&EPA
United States
Environmental Protection
Agency
Office of Research and
Development
Cincinnati, Ohio 45268
EPA/625/R-95/004
September 1995
Publication
National Conference on
Environmental
Problem-Solving with
Geographic Information
Systems
Cincinnati, Ohio
September 21-23,1994
-------
EPA/625/R-95/004
September 1995
Seminar Publication
National Conference on Environmental Problem-Solving
with Geographic Information Systems
September 21-23, 1994
Cincinnati, Ohio
U.S. Environmental Protection Agency
Office of Research and Development
National Risk Management Research Laboratory
Center for Environmental Research Information
Cincinnati, Ohio
-------
Notice
This document has been subjected to EPA's peer and administrative review and has been approved
for publication as an EPA document. The views expressed are those of the authors and do not
necessarily reflect those of the Agency. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
When an NTIS number is cited in a reference, that document is available from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
703-487-4650
Please note that the Soil Conservation Service of the U.S. Department of Agriculture is now the
Natural Resources Conservation Service; "Soil Conservation Service" is used throughout this
document because this was the name in use at the time of the conference.
-------
Contents
Page
GIS Concepts
GIS Uncertainty and Policy: Where Do We Draw the 25-Inch Line?
James E. Mitchell 3
Data Quality Issues Affecting GIS Use for Environmental Problem-Solving
Carol B. Griffin 15
You Can't Do That With These Data! Or: Uses and Abuses of Tap Water Monitoring Analyses
Michael R. Schock and Jonathan A. Clement 31
Ground-Water Applications
Using GIS/GPS in the Design and Operation of Minnesota's Ground Water
Monitoring and Assessment Program
Tom Clark, Yuan-Ming Hsu, Jennifer Schlotthauer, Don Jakes, and Georgianna Myers. 45
Use of GIS in Modeling Ground-Water Flow in the Memphis, Tennessee, Area
James Outlaw and Michael Clay Brown 50
MODRISI: A PC Approach to GIS and Ground-Water Modeling
Randall R. Ross and Milovan S. Beljin 60
GIS in Statewide Ground-Water Vulnerability Evaluation to Pollution Potential
Navulur Kumar and Bernard A. Engel 66
Verification of Contaminant Flow Estimation With GIS and Aerial Photography
Thomas M. Williams 74
Geology of Will and Southern Cook Counties, Illinois
Edward Caldwell Smith 81
Watershed Applications
The Watershed Assessment Project: Tools for Regional Problem Area Identification
Christine Adamus 97
-------
Contents (continued)
Page
Watershed Stressors and Environmental Monitoring and Assessment Program
Estuarine Indicators for South Shore Rhode Island
John F. Paul and George E. Morrison 101
CIS Watershed Applications in the Analysis of Nonpoint Source Pollution
Thomas H. Cahill, Wesley R. Homer, and Joel S. McGuire 110
Using CIS To Examine Linkages Between Landscapes and Stream Ecosystems
Carl Richards, Lucinda Johnson, and George Host 131
Nonpoint Source Water Quality Impacts in an Urbanizing Watershed
Peter Coffin, Andrea Dorlester, and Julius Fabos 142
A CIS for the Ohio River Basin
Walter M. Grayman, Sudhir R. Kshirsagar, Richard M. Males, James A. Goodrich,
and Jason P. Heath 151
Nonpoint Source Pesticide Pollution of the Pequa Creek Watershed, Lancaster County, Pennsylvania:
An Approach Linking Probabilistic Transport Modeling and CIS
Robert T. Paulsen and Allan Moose 156
Integration of CIS With the Agricultural Nonpoint Source Pollution Model:
The Effect of Resolution and Soils Data Sources on Model Input and Output
Suzanne R. Perlitsh 164
XGRCWP, a Knowledge- and CIS-Based System for Selection, Evaluation, and
Design of Water Quality Control Practices in Agricultural Watersheds
Runxuan Zhao, Michael A. Foster, Paul D. Robillard, and David W. Lehning 175
Integration of EPA Mainframe Graphics and CIS in a UNIX Workstation Environment
To Solve Environmental Problems
William B. Samuels, Phillip Taylor, Paul Evenhouse, and Robert King 183
Wetlands Applications
Wetlands Mapping and Assessment in Coastal North Carolina:
A CIS-Based Approach
Lori Sutter and James Wuenscher 199
Decision Support System for Multiobjective Riparian/Wetland Corridor Planning
Margaret A. Fast and Tina K. Rajala 213
Design of CIS Analysis To Compare Wetland Impacts on Runoff
in Upstream Basins of the Mississippi and Volga Rivers
Tatiana B. Nawrocki 218
IV
-------
Contents (continued)
Page
Water Quality Applications
Vulnerability Assessment of Missouri Drinking Water to Chemical Contamination
Christopher J. Barnett, Steven J. Vance, and Christopher L Fulcher 235
Reach File 3 Hydrologic Network and the Development of CIS Water Quality Tools
Stephen Bevington 238
EPA's Reach Indexing Project: Using CIS To Improve Water Quality Assessment
Jack Clifford, William D. Wheaton, and Ross J. Curry 242
Environmental Management Applications
Ecological Land Units, CIS, and Remote Sensing: Gap Analysis in the Central Appalachians
Ree Brannon, Charles B. Yuill, and Sue A. Perry. 255
A CIS Strategy for Lake Management Issues
Michael F. Troge 261
A Watershed-Oriented Database for Regional Cumulative Impact Assessment and Land Use Planning
Steven J. Stichter 266
A CIS Demonstration for Greenbelt Land Use Analysis
Joanna J. Becker 273
CIS as a Tool for Predicting Urban Growth Patterns and Risks
From Accidental Release of Industrial Toxins
Samuel V. Noe 278
Integration of CIS and Hydrologic Models for Nutrient Management Planning
Clyde W. Fraisse, Kenneth L Campbell, James W. Jones, William G. Boggess,
and Babak Negahban 283
Other GIS Applications
Expedition of Water-Surface-Profile Computations Using GIS
Ralph J. Haefner, K. Scott Jackson, and James M. Sherwood 295
Small Is Beautiful: GIS and Small Native American Reservations
Approach, Problems, Pitfalls, and Advantages
Jeff Besougloff 299
A CIS-Based Approach to Characterizing Chemical Compounds
in Soil and Modeling of Remedial System Design
Leslie L Chau, Charles R. Comstock, and R. Frank Keyser 302
-------
Contents (continued)
Page
Polygon Development Improvement Techniques for Hazardous Waste Environmental Impact Analysis
David A. Padgett 308
Comparing Experiences in the British and U.S. Virgin Islands in Implementing CIS
for Environmental Problem-Solving
Louis Potter and Bruce Potter 313
Application of CIS for Environmental Impact Analysis in a Traffic Relief Study
Bruce Stauffer and Xinhao Wang 322
VI
-------
A cknowledgments
The success of the conference and this document is due largely to the talents and efforts of many
individuals. Gratitude goes to each person involved.
Authors
A special thanks goes to the many authors of the papers presented in this document. Their efforts
in preparing papers made this document possible and led to the overall success of the conference.
Peer Reviewers
The following individuals peer reviewed this document:
Jim Goodrich, U.S. Environmental Protection Agency, National Risk Management Research
Laboratory, Cincinnati, Ohio.
Nancy Phillips, Hollis, New Hampshire.
John Shafer, Earth Sciences & Resources Institute, University of South Carolina, Columbia,
South Carolina.
Editorial Review and Document Production
Heidi Schultz, Eastern Research Group, Inc., Lexington, Massachusetts, directed the editorial
review and production of this document.
Technical Direction and Coordination
Daniel Murray and Susan Schock, U.S. Environmental Protection Agency, Office of Research and
Development, National Risk Management Research Laboratory, Center for Environmental Re-
search Information, Cincinnati, Ohio, coordinated the preparation of this document and provided
technical direction throughout its development.
Special Thanks
Special thanks goes to the following individuals for their support:
Tom Davenport, U.S. Environmental Protection Agency, Region 5, Chicago, Illinois.
Mike Forrester, Urban and Regional Information Systems Association, Washington, DC.
Bill French, American Society for Photogrammetry and Remote Sensing, Bethesda, Maryland.
Carol Griffin, Idaho Falls, Idaho.
Lyn Kirschner, Conservation Technology Information Center, West Lafayette, Indiana.
Kevin Klug, Association of American Geographers, Washington, DC.
VII
-------
John Lyon, Ohio State University/American Society of Photogrammetry and Remote Sensing,
Columbus, Ohio.
Mark Monmonier, Department of Geography, Syracuse University, Syracuse, New York.
Sam Noe, Joint Center for CIS & Spatial Analysis, University of Cincinnati, Cincinnati, Ohio.
Nancy Phillips, Hollis, New Hampshire.
Javier Ruis, U.S. Department of Agriculture, Soil Conservation Service, Fort Worth, Texas.
Don Schregardus, Ohio Environmental Protection Agency, Columbus, Ohio.
VIM
-------
Introduction
The National Conference on Environmental Problem-
Solving with Geographic Information Systems was held
in Cincinnati, Ohio, September 21 to 23, 1994. The
conference was a forum for over 450 environmental
professionals to exchange information and approaches
on how to use geographic information systems (CIS) to
define, assess, and solve environmental problems.
Cross-media pollutant transport and watershed-based
decision-making have made the process of solving en-
vironmental problems more complex. The application of
CIS to environmental problem-solving has greatly in-
creased our ability to manipulate and analyze relational
and spatial data, providing environmental decision-mak-
ers with a powerful tool for analyzing multimedia envi-
ronmental data over increasingly broader areas (e.g.,
watersheds, states, regions). While the approach to
using CIS varies from application to application, a com-
mon, technically sound framework for applying CIS to
environmental problems should be developed and im-
plemented. This conference was an initial step in defin-
ing this framework by examining the following areas:
Problem identification and definition.
Data requirements (e.g., coverage, scale), availabil-
ity, documentation, reliability, and acquisition.
Approaches considered and selected for problem-
solving.
Unique challenges and pitfalls encountered.
Interpretation of results, including level of confidence
achieved based on data quality and approach taken.
Presenters were requested to address one or more of
these areas in papers and posters that focused on
applications of CIS to specific environmental problems.
This document presents peer-reviewed papers from the
conference. The papers have been organized by gen-
eral topic area as follows:
CIS Concepts
Ground-Water Applications
Watershed Applications
Wetlands Applications
Water Quality Applications
Environmental Management Applications
Other CIS Applications
The purpose of this document is to share the information
presented at the conference with individuals who were
unable to attend. This document will be useful to indi-
viduals who are currently applying CIS to environmental
situations or considering CIS for application in environ-
mental problem-solving. These individuals include envi-
ronmental regulatory personnel at the federal, state and
local level; university professors, researchers, and stu-
dents; private sector personnel, including industry rep-
resentatives and environmental consultants; and other
interested persons. The goal of sharing this information
with a broader audience is to help users apply CIS to
environmental problem-solving with a greater aware-
ness of the power and limitations of this very useful tool.
IX
-------
GIS Concepts
-------
GIS Uncertainty and Policy: Where Do We Draw the 25-Inch Line?
James E. Mitchell
Institute for Environmental Studies, Louisiana State University, Baton Rouge, Louisiana
Abstract
The growing availability of improved hardware and soft-
ware for geographic information systems (GIS) has out-
stripped most users' ability to identify and represent
uncertainty in the available data. In practice, the prolif-
eration and compounding of errors and uncertainty in-
crease as information becomes more easily handled
and combined from different sources.
Various stages of GIS database development and analy-
sis generate different forms and amounts of error and
uncertainty. In most cases, inherent uncertainty within
source data is simply ignored and its nature eventually
lost through subsequent processing. Both the location
of features and their attributes can include error and
uncertainty. By the time decision-makers receive mapped
information, it is typically represented as correctly lo-
cated and attributed.
The use of weather and climate information provided by
the National Climatic Data Center (NCDC) is a common
example of this scenario. Weatherstation locations pro-
vided by NCDC are reported to the nearest truncated
degree-minute. A minute is one-sixtieth of a degree of
arc. In the center of the continental United States, 1
minute of latitude averages approximately 6,000 feet
and 1 minute of longitude averages approximately 4,800
feet. Thus, the station location is only known to lie within
a box of approximately 1 square mile. Map repre-
sentations of these data should reflect this uncertainty.
Under the Municipal Solid Waste Landfill (MSWLF) Cri-
teria, the U.S. Environmental Protection Agency has
dictated that the 25-inch precipitation contour line be
used as a regulatory boundary for the level of protection
required at municipal landfill sites. The way in which
these lines are created and interpreted has important
policy implications. Indeed, the cost and practicality of a
given location must take this into account. If the 25-inch
precipitation figure is critical, characterizing its uncer-
tainty is also important.
In this work, uncertainty is considered a property of the
data (1). A Monte Carlo procedure is used to represent
the stochastic character of contour lines generated from
point data with known locational uncertainty. The 30-
year normal precipitation data for Kansas are used as
an example. The results of this study are compared with
the 25-inch contour used for regulatory purposes in
Kansas. This study demonstrates that the method of
interpolation greatly influences the resulting contours. In
addition, locational uncertainty changes the results un-
predictably using four different contouring methods. Fi-
nally, the differences have potentially significant policy
implications. The nature and origin of these factors are
discussed.
Problem Statement
The increasing power of geographic information sys-
tems (GIS) and the availability of digital data have en-
abled users and decision-makers to perform complex
spatial analyses for a great variety of environmental
applications (2). The rapid expansion of GIS has re-
sulted in a parallel growing concern about the quality of
data (3).
An understanding of error and uncertainty is critical for
proper use of spatial information. For the purposes of
this discussion, error is defined as a deviation between
the GIS representation of a feature and its true value (4).
For a location, this might arise from rounding or truncat-
ing digits. Attribute error can involve misclassification of
a feature or some other form of incorrectly accounting
for its nature. Error is a measurable value quantifying
these differences.
Furthermore, uncertainty shall refer to a characteristic
for which the exact location and/or quantity cannot be
calculated (5) or an attribute whose value represents a
distribution or some other ensemble (composite) meas-
ure. Locational uncertainty often arises when inappro-
priate measurement systems are used. An example of
this is the use of a Public Land Survey designation (often
-------
referred to as a legal location) to specify a point location.
This system is designed to represent a tract of land (an
area). It is not accurate for locating points (1).
The uncertainty associated with an attribute is an impor-
tant characteristic of that feature. It quantifies the preci-
sion of a stochastic quantity; that is, one that is not
accurately represented by a single value. Annual pre-
cipitation is represented by a single number, typically the
30-year mean annual precipitation (30-year normal).
This number varies each year, however, and that uncer-
tainty can be quantified by the variance or other statis-
tical measures. In this sense, uncertainty is a known or
calculable value that can be used in spatial analyses.
Unreliable CIS data and products may lead to adverse
environmental and legal consequences. The National
Center for Geographic Information Systems and Analy-
sis (NCGISA) chose data quality as the first initiative on
its CIS research agenda (6). Many efforts have been
made priorto this, and since, to understand and manage
error and uncertainty in CIS applications.
CIS analyses are inherently subject to propagation of
error and uncertainty (4, 7). No data set can represent
every spatial reality of a geographically dispersed phe-
nomenon. Monmonier (8) points out that as long as the
three-dimensional earth's surface is transformed to a
two-dimensional plane, error and uncertainty of various
forms will be produced. Goodchild and Min-Hua (9) point
out two issues that are important when dealing with error
and uncertainty:
Minimization of error in the creation of CIS products.
Measurement and presentation of error and uncer-
tainty in a useful fashion.
CIS technology introduces error and uncertainty through
two major sources: (1) inherent error and (2) operational
error. Inherent error is the error present in source data.
It is generated when the data are collected. Operational
error is generated during data entry and manipulation
(7, 10-13). Examples include locational shifts due to
projection or combining information from different
source scales.
Most error and uncertainty contained in CIS data cannot
be eliminated. Instead, they are actually created, accen-
tuated, and propagated through CIS manipulation pro-
cedures (14-16). Most operational errors are difficult to
estimate.
The selection by the U.S. Environmental Protection
Agency (EPA) of a 25-inch per year local precipitation
limit as one of the criteria to determine whether small
municipal solid waste landfills (MSWLF) are subject to
the provisions of Subtitle D provides an excellent exam-
ple of how uncertainty and errors enter into a CIS analy-
sis and its subsequent products. It demonstrates all of
the major forms and purveyors of error and uncertainty:
Spatial (locational) error
Statistical (sampling) uncertainty
Temporal (time domain) error
Error proliferation (processing error)
Analytical (choice of methodology) error
Cartographic representation error
Many of these are avoidable; some are known and
understood, yet they remain largely ignored by users of
CIS technology. This work presents each of these fac-
tors, discusses their origins, and shows how CIS could
have been used to better serve the policy and regulatory
processes. The Kansas example demonstrates that ig-
noring the factors influencing error and uncertainty can
result in incorrect conclusions and inappropriate policy
decisions.
Data Requirements and Sources
To perform an analysis of precipitation, data are typically
obtained from the National Climatic Data Center
(NCDC), located in Asheville, North Carolina. This is the
national repository for such data. These data are also
available through state or regional climate centers. The
Kansas Weather Library at Kansas State University pro-
vided data for this study. The 1990 "normal precipitation"
data (17) and locations were obtained and generated
into an ARC/INFO point coverage. Figure 1 displays the
locations of the precipitation stations used in this study.
Normal precipitation is defined as the average annual
precipitation for a three-decade (30 years) period at
each station for which reliable data are available. To
avoid "edge effects" (processing anomalies due to a lack
of data along edges of an area), all stations in Kansas
and some from neighboring states were used. A total of
380 stations compose this data set. In addition, precipi-
tation contours from the "Availability of Ground Water in
Kansas Map" (18) were digitized from a [paper] source
map. The Geohydrology Section of the Kansas Geologi-
cal Survey provided base map coverages of carto-
graphic features.
All data represent the best available information from the
source institutions noted above. Those organizations
use the data in their analytical and cartographic re-
search and production operations.
Methodology
To examine the influence of locational uncertainty on the
representation of three-dimensional, natural phenom-
ena, a Monte Carlo approach was adopted (1). Using
this technique, random realizations of point locations are
generated for each rain gauge, in each of 50 separate
simulations. From this, 50 possible representations of
the unknown locations of each gauge are used to create
-------
*
.
\
.'
Figure 1. Location of rain gauges used in this study.
50 different sets of contours. All Monte Carlo calcula-
tions and data generation were performed using Statis-
tical Analysis System (SAS) (19, 20).
These 50 simulations were sequentially processed us-
ing the four different contouring methods available within
ARC/INFO. This provided a means to examine analyti-
cal error propagation. The first of the four methods is
kriging (21, 22). This is referred to in the paper as the
UK method, for its use of linear universal kriging (23).
The other three are manipulations of the triangular-
irregular network (TIN) contouring algorithm available in
ARC/INFO. These differ by the number of interpolation
points used along the edges of the elements in the TIN
data structure (24). The first used the default 1, the
second used 5, and the third used 10 (the largest value
available). These are labeled D1, D5, and D10, respec-
tively.
CIS operations used in this work include overlay analy-
sis, areal calculation, and arc intersection. ARC/INFO
was used for all CIS and cartographic production in
this work.
Identifying the Sources of Uncertainty
Spatial Error
Data obtained from NCDC is provided with the knowl-
edge that weather station locations are reported using
truncated degrees and minutes of longitude and latitude.
NCDC cannot provide any better locational accuracy at
this time. Because each location is reported with error,
this clearly has the potential to affect any contours or
other three-dimensional features interpolated from the
data. The magnitude and nature of this influence are
unknown and unpredictable (1).
In addition to the poorly defined station locations, exami-
nation of the data revealed other anomalies. The loca-
tions in the publication reporting normal precipitation
(17) were not identical to those identified by the Kansas
State climatologist and NCDC. Some of the discrepan-
cies were quite large. These anomalies were brought to
the attention of all parties involved. No resolution was
provided to this investigator's satisfaction, however.
The contours digitized from the "Availability of Ground
Water in Kansas Map" are stated to originate from the
1960 normals (18). No documentation exists, however,
concerning the way the lines were derived or the number
of rain gauges used. Presumably, they were contoured
by hand.
Statistical Uncertainty
This is a sampling consideration based on the size of
the data, the nature of the process being sampled, and
its variability. Unfortunately, precipitation is a particularly
"patchy" phenomenon. That is, rain falls in a discon-
tinuous fashion, and adjacent gauges can depict very
different patterns. This is confounded by the fact that
most contouring algorithms and other approaches to
-------
represent three-dimensional surfaces assume a rela-
tively smooth (locally) and continuous process.
Areal processes are almost always sampled as point
information. Most contouring algorithms require a regu-
lar grid from which to interpolate surface features. In
practice, rain gauges, as well as other environmental
sampling programs, are irregularly distributed. Place-
ment often depends on factors other than grid sampling
(e.g., convenience, access to communications, fi-
nances, Congressional districts). This creates a "nonex-
perimental sampling" design (25). Nonexperimental
sampling can contribute to uncertainty (26, 27).
Temporal Error
The normals are recalculated each decade and can
change drastically in local areas. These changes arise
for various reasons. First, some stations enter and drop
from the database. Stations are deleted due to changes
in location or extended periods of data collection prob-
lems. On occasion, new stations are added. Thus, the
size and areal coverage of the data set changes with time.
In addition, weather patterns change with time. Ex-
tended periods of drought or excess rain or snow alter
measured precipitation. In turn, three-dimensional rep-
resentations change unevenly.
Error Proliferation
Once an error enters into the database and is included
in CIS operations, spatial analysis, or spatial interpola-
tion, its effect passes into the next stage of processing.
In the 1990 normal precipitation data for Kansas, two
stations are reported in Garden City. Despite the fact
that the two are only a few miles apart, their annual total
precipitation differs by 2 inches! In consultation with the
state climatologist (Mary Knapp, Kansas State Univer-
sity), one was eliminated from the analysis. This process
was repeated for an additional six stations where re-
ported values appeared to be anomalous compared with
nearby stations or the previous normal precipitation
(1951 to 1980).
Errors can also proliferate through the normal handling
of data. With geographic data, this often occurs while
converting data from raster to vector and vector to raster
forms (28). Some CIS operations are best accomplished
in one form or another. As a result, transformations are
often "hidden" from the user. Commonly, features
"move" slightly after each step in an analysis.
Analytical Error
Different techniques have been developed for perform-
ing spatial interpolation, and an abundance of software
is available for this purpose. All these methods have
strengths and weaknesses. Each is based on a specific
set of assumptions about the form and nature of the
data. Some are more robust (less sensitive to data
anomalies) than others. Most importantly, some provide
additional information useful in data analysis.
Unfortunately, users often "take the defaults" when using
sophisticated techniques and ignore the assumptions
behind the method. Parameters can be varied and their
effect evaluated, as in a sensitivity analysis (29). Often,
the best approach is to try several methods and evaluate
their joint performance (30, 31).
Another difficulty is the need to assign values to areas.
By definition, polygons in a CIS are considered to be
homogeneous. In reality, they bound areas that are a
gradation from one characteristic to another. On the
other hand, contours are commonly used to depict sur-
face gradients but are useless (within a CIS) for analyti-
cal or modeling purposes. Ultimately, data sampling is
accomplished as a point process (except, perhaps, in
remote sensing), while many forms of data analysis and
processing require areal information.
Cartographic Representation Error
Communicating the uncertainty of map features is not a
trivial endeavor. Maps can be produced in two basic
forms: as a raster (e.g., orthophotoquads or satellite
images) or a composition of vectors (e.g., contour
maps). The printing process, however, often reduces all
of this to a raster representation at a very fine pixel size.
Each method poses its own problems in depicting un-
certainty.
Rasters can be used effectively in conjunction with color
information theory to produce a continuum of shading
within a thematic map layer (32). The choice of colors,
however, can influence the interpretation of the data,
and no universal scheme exists for depicting thematic
variability. For example, blue shades often represent
water or cold, while yellow and/or red often represent
temperature or heat.
Vectors present a different suite of problems. Contouring
is the primary technique for using vectors to depict areal
variation. By definition, however, contour lines represent
an exact isoline or single value along its length. Uncer-
tainty cannot be represented in a line. Rather, a com-
posite of lines can be displayed that represents a set of
possible interpretations of the data. This is not a practi-
cal solution for mapping, however, as it can create a
jumble of intersecting lines that makes interpretation
difficult and is not an aesthetic means of presentation.
Challenges Encountered in This Study
This work attempts to discover and account for sources
of error and uncertainty in CIS analysis. Given this
information, the challenges are to find the best way to
incorporate it into the analysis and to represent it in a
useful manner. Another challenge is finding ways to use
-------
CIS uncertainty to support policy and management de-
cisions. Addressing these manifold problems starts with
identifying the sources of error and uncertainty, the way
they enter the analysis, and the manner in which they
are propagated through the use of CIS.
This study includes a number of known sources of un-
certainty. In practice, this is not always the case. Users
of CIS data and technology should always assume that
the sources of uncertainty discussed in this paper are
present and attempt to determine their nature. Uncer-
tainty should be considered a property of the data and
appropriately represented (1). This is the approach
taken in this work.
After examining these factors, a Monte Carlo simulation
was deemed the most appropriate approach to capture
the nature of the locational uncertainty. Four different
methods of contouring were used to examine analytical
uncertainty (uncertainty due to the choice of a contour-
ing algorithm). In addition, the contribution of statistical
(sampling) uncertainty could have been addressed
through incorporating information about the standard
error of the point precipitation measurements (normals)
used as the base data. Time limitations precluded ex-
amining this dimension of the question. Comparing the
contours resulting from the 1960 and 1990 precipitation
normals demonstrates the effect of temporal variation.
The greatest challenge is communicating the uncer-
tainty in a manner useful to decision-makers. This paper
presents a series of maps, figures, and tables aimed at
addressing this problem. Some of the maps (see Fig-
ures 2 through 5) show the uncertainty resulting from
each of the contouring methods. Figure 6 depicts the
union (overlay) of the four approaches and displays their
correspondence. The pie chart in Figure 7 is a nonspa-
tial representation of this correspondence and the rela-
tive area represented within the different combinations
of overlapping regions of uncertainty. Tables 1 through
3 further compare these quantities. Figure 8 is the map
that the Kansas Department of Health and Environment
(KDHE) chose to define the regulatory boundary (the
25-inch contour). The contours resulting from this study
can be seen in Figure 9. Finally, the map in Figure 10 is
a cartographic comparison of the differences between
the contours used by KDHE (based on the 1960 nor-
mals) and those generated by the currently available
data (the 1990 normals).
There is no single best approach for meeting these
challenges, and there may never be one. The real chal-
lenge to address is how to educate technical CIS pro-
fessionals and the users of their work to look for
uncertainty and consider its influence on their decision-
making process.
Results
The zones of uncertainty defined by the results from the
four contouring methods used in this study are displayed
in Figures 2 through 5. For each method, these zones
represent the areal extent of the overlain contour lines
produced in the 50 simulations. Each region is bounded
by the furthest west or east contour generated along any
length of the region. Table 1 shows the relative area
falling within each of these zones as they traverse the
state of Kansas. Clear differences exist between the
total areas of uncertainty. It is their placement and rela-
tive location, however, that have policy and manage-
ment implications. CIS is required to examine these
questions.
Table 1. Comparison of Absolute and Relative Area of
Uncertainty Arising From Four Methods of
Determining the 25-Inch Precipitation Contour
Method3
UK
D1
D5
D10
Total Area
(square
miles)
289.33
494.38
602.13
631.11
Difference
Between
This and
UK Method
(square
miles)
205.05
312.80
341 .78
Percentage
of UK
Method
170.87
208.11
218.13
Percentage
of Combined
Areab of
Uncertainty
19.65
33.58
40.90
42.87
UK = universal kriging with linear drift, D1 = TIN interpolation with 1
subdivision, D5 = TIN interpolation with 5 subdivisions, D10 = TIN
interpolation with 10 subdivisions (23, 24).
bA union (overlay) of all four sets of regions of uncertainty creates a
combined area of 1,472.07 square miles. This includes the zones of
uncertainty for each method of contouring and areas not included
within any of the four regions of uncertainty (gaps between them).
Figures 2 through 5 clearly show differences in both the
extent of uncertainty in the 25-inch contour line and its
positional interpretation. Each method has a slightly
different bend or twist. Islands (isolated regions where
the 25-inch line appears as a closed loop) are mani-
fested differently depending on the interpolation
scheme. It is interesting to note the relative correspon-
dence between the general shape of the D1 and UK
methods. In the south-central border region, D1 and UK
represent the local uncertainty as a bulge, while D5 and
D10 depict it as an island of lower precipitation.
Areal correspondence and difference are depicted in
Figure 6 and Table 2. Figure 7 is a pie chart visualizing
the information in Table 2. These results are somewhat
surprising in that areas where none of the four methods
located the 25-inch line ("None Present") represent the
second largest composite area. Because of the large
number of "sliver polygons," the graphic representation
of the overlay is somewhat difficult to interpret. Table 2
clarifies these interrelationships by breaking down the
various categories. The average area per polygon value
-------
Figure 2. Regions of uncertainty produced by the UK method of contouring.
Figure 3. Regions of uncertainty produced by the D1 method of contouring.
-------
Figure 4. Regions of uncertainty produced by the D5 method of contouring.
Figure 5. Regions of uncertainty produced by the D10 method of contouring.
-------
UK
D1
D5
D10
Not Present
Figure 6. Union of the regions of uncertainty from all four methods of contouring. The numerous "sliver" polygons make this a
difficult presentation to interpret at this scale. The black areas appear prominently, however. These represent areas where
no method placed contour lines.
adds to the interpretation of the relative areas by incor-
porating the number of polygons in each category. An
inspection of this column makes those categories with
a multitude of very small polygons stand out. It also
displays a number of large jumps in magnitude. As this
number increases, the significance of the correspondence
increases.
Table 2 highlights the correspondence between the D5
and D10 approaches and the D1 and UK methods of
interpolation. This relationship is interesting because the
algorithms used by the UK and D1 methods both are
forms of linear interpolation. The D5 and D10 algorithms
are designed to provide more "smoothing" and appear
to create increasingly more "bull's eyes." Only D5 and
D10 generate these features. Of the 50 simulations, a
particular bull's eye appears west of the 25-inch contour
(see Figures 3 and 4) four times using D5 and 39 times
using D10. The size and location of these anomalies
also vary with the input data. Polygons containing con-
tours from all four methods rank ninth in total area and
seventh in average area (out of 17 categories). This
supports the conclusion that the four chosen methods
have a relatively low spatial correspondence.
Table 3 breaks down the area of uncertainty by county.
Although the zones of uncertainty appearto be relatively
small when displayed on a statewide basis, they have
important impacts in local areas. In particular, combining
this information with soils, topography, ground water,
and other information can clearly indicate whether a
specific location is suitable for a landfill. Often, informa-
tion developed at one scale is used in another. In this
case, statewide information is being used fora site-specific
application.
Figure 8 is a copy of the map the KDHE used to deline-
ate the 25-inch precipitation contour. The results from
LRGflty
12.13%
D1,03. acd DIG
11.11%
Ml Oiling;
Norn Praianl
22.13%
Figure 7. Breakdown of the total area in each category result-
ing from the union (overlay) of the regions of uncer-
tainty from all four contouring methods. The "None
Present" category represents a surprisingly large
proportion among the 17 possible combinations.
10
-------
Table 2. Comparison of Absolute and
Relative Area
of
Table 3.
Uncertainty Arising From Four Methods of
Determining the
Methods of
Contouring
Found Within
Area
D5 and D10
None present
D1 only
UK only
D1, D5, and D10
D10 only
UK and D1
D5 only
Common to all
UK, D5, and D10
D1 and D5
D1 and D10
UK and D5
UK, D1, and D5
UK, D1, and D10
UK and D10
Combined total
N
28
28
37
27
24
69
19
73
10
14
38
38
19
12
12
13
461
25-Inch Precipitation Contour
Total Area
(square
miles)
350.13
326.06
221 .02
178.60
163.61
70.07
65.82
34.68
23.45
15.47
10.68
6.50
2.21
1.89
1.41
0.47
1472.07
Percentage
of Total
Combined
Area
23.78
22.15
15.01
12.13
11.11
4.76
4.47
2.36
1.59
1.05
0.73
0.44
0.15
0.13
0.10
0.03
100.00
Average
Area per
Polygon3
(square
miles)
12.50
11.64
5.97
6.61
6.82
1.02
3.46
0.48
2.35
1.11
0.28
0.17
0.12
0.15
0.12
0.04
County
Barber
Barton
Clark
Area of Uncertainty for Each County Arising From
Four Methods of
Determining the 25-Inch
Precipitation Contour
UK Area
(square
miles)
36.38
28.47
12.64
Comanche 44.30
Edwards
Ellis
Jewell
Kiowa
Osborne
Pawnee
Pratt
Rush
Russell
Smith
Stafford
Total
aA dash
method
a
15.15
43.87
18.84
0.04
27.17
27.59
34.88
289.3
D1 Area
(square
miles)
24.29
96.84
11.49
52.94
6.14
6.65
14.32
99.82
11.30
8.51
22.71
68.47
70.90
494.4
indicates that no contours appeared
specified.
D5 Area
(square
miles)
28.05
56.16
12.97
51.43
9.27
23.05
17.81
168.2
0.11
11.56
31.32
28.30
67.38
96.49
602.1
D10
Area
(square
miles)
28.31
59.56
13.02
52.23
9.44
8.90
24.26
31.12
172.6
0.30
11.43
29.51
28.00
66.34
96.05
631.1
in that county for the
Average Area per Polygon =
to compare the relative size
(Total Area) / N. This a useful measure
of each polygon in each classification.
the UK method were selected as the best available
representation of normal precipitation across Kansas
(see Figure 9). The figure displays the undipped contour
lines generated from the data. This is done to point out
the importance of "edge effect." Note the incoherent
behavior of the contour lines at their termini. If a smaller
window of data points were used, interpolation problems
would have lain across the region of interest. When
present, these features require more handling and time
for analysis. They often introduce additional error and
uncertainty.
The policy implications of this example are demon-
strated in Figure 10. Here, the map shows the combina-
tion of the "official" KDHE map and the data interpreted
in this study. The pattern of noncorrespondence is note-
worthy. The lightest areas are regions that currently
experience higher annual precipitation than forecast by
the 1960 normals (from the KDHE map). Black areas
are expected to have lower precipitation under current
climatic conditions. Therefore, large areas of Kansas
that should be under regulation according to the MSWLF
regulations are not.
In summary, the figures and tables clearly show that
locational uncertainty of data measured as points is
propagated into contour lines. The nature and magni-
tude of that uncertainty varies with location and method
of interpolation and shows no regular (predictable) pat-
tern. Perhaps most importantly, uncertainty that appears
small at one scale can be relatively more significant at
another. In addition, seemingly small geographic feature
and uncertainty can be an important factor in decision-
making.
Discussion and Conclusions
CIS is an established and accepted technology, espe-
cially in applications related to natural resource and
environmental management. Despite the widespread
proliferation of CIS into these areas, the available data
are not always appropriate for the intended application.
Furthermore, adequate documentation is not always
available to determine whether the data are adequate
for a given use. The development of metadata standards
will play an important role in addressing this problem.
Errors and uncertainty will always be present in CIS
data. Recognizing their presence, incorporating them
into the analysis, and representing them in CIS products
will remain a constant challenge.
This study demonstrates the influence that various
sources of CIS uncertainty can leverage on the results
of an analysis. The example of the 25-inch precipitation
11
-------
Yield of Greater Than
500 Gallons per Minute
* Precipitation Contours in Inches Per Year
Yield of 100 to 500 Gallons
of Water per Minute
Yield of Less Than 100
Gallons of Water per Minute
100 Miles
Figure 8. The map that the KDHE selected as the definitive source for the location of the 25-inch precipitation contour (18).
Figure 9. Map of the precipitation contours resulting from applying the UK method with linear drift to the 1990 normals (17). Here,
the contour interval is 5 inches. Note the incoherent behavior of the contours around the margins of the map. This is
referred to as "edge effect."
12
-------
D
Higher Precipitation in
1990 Data
Lower Precipitation in
1990 Data
Figure 10. This map represents the union (overlay) of the information in Figures 8 and 9. The lighter regions represent areas
exhibiting higher precipitation in the 1990 normals (1961 to 1990) than was apparent in the 1960 normals (1931 to 1960).
The darker areas show the opposite relationship.
line in Kansas is a clear example of how the use of
inappropriate data can have far-reaching effects on pol-
icy and management. The regulatory agency, KDHE,
chose the wrong map upon which to base its regulatory
authority. As a result, numerous potential sites for small
municipal solid waste landfills will be considered that are
in violation of the letter and intent of the law.
Ultimately, the responsibility for proper use of CIS tech-
nology lies in the hands of practitioners. Technical staff
performing CIS analysis must be knowledgeable about
sources of error and uncertainty and ensure that users
of their work are aware of their influence on CIS output.
The problems demonstrated in the Kansas example
could have been avoided simply by investigating the
appropriateness of the data. Instead, a convenient
source was chosen without seeking any other sources
of "better" information. Indeed, familiarity with the nature
of the data (30-year normals) should have led the policy
analyst to select the most current data and not data that
are 30 years out of date! An understanding that contour
lines represent a generalization of the point precipitation
measurements should also have led to the conclusion
that locations near the boundary line ought to be moni-
tored for compliance. Both the temporal and spatial
characteristics of climate can change, as exemplified by
the difference in the 1960 and 1990 normals. The "Dust
Bowl" periods of the 1930s and 1950s significantly
influenced the 1960 normals (33). As a result, they are
not appropriate for this application.
Although a powerful tool, CIS does not hold all the
answers. The technical community and policy-makers
must work together to ensure its proper use. In reality,
no 25-inch precipitation line floats over Kansas. It is
merely the interpretation of scientists and policy-makers
who select its location. The only way to arrive at a
reasonable answer is to gather the best available infor-
mation and allow all parties to scrutinize it. CIS can be
a wonderful tool to do this.
Acknowledgments
The author wishes to express his gratitude to the Kan-
sas Geological Survey for its computing support before,
during, and after his move to Louisiana. Also, the author
wishes to thank Mr. M. Schouest, who made the move
both possible and tolerable. Mr. C. Johnson of Johnson
Controls in Baton Rouge, Louisiana, produced the ex-
cellent presentation graphics. Finally, this work could not
have been performed without the Internet.
References
1. Mitchell, J.E. 1993. The influence of (x,y) uncertainty on predic-
tion error and contour lines from a three-dimensional surface.
Water Resources Bull. 29(5):863-870.
13
-------
2. Northeast Waste Management Officials Association. 1992. Cata-
logue and directory of New England states and regional CIS
organizations and activities, and an assessment of their future
CIS needs. EPA/G-D001456/90-2.
3. Campbell, W.G., and D.C. Mortenson. 1989. Ensuring the quality
of geographic information system data: A practical application of
quality control. EPA/600/J-89/306.
4. Chrisman, N.R. 1991. The error component in spatial data. In:
Maguire, Goodchild, and Rhine, eds. Principles of geographic
information systems, Vol. 1. pp. 164-174.
5. Isaaks, E.H., and R.M. Srivastava. 1989. An introduction to applied
geostatistics. New York, NY: Oxford University Press.
6. National Center for Geographic Information and Analysis. 1989.
The research plan of the National Center for Geographic Infor-
mation and Analysis. Int. J. Geo. Info. Sys. 3(2):117-136.
7. Vitek, J.D., S.J. Walsh, and M.S. Gregory. 1984. Accuracy in
geographic information systems: An assessment of inherent and
operational errors. Proceedings of the PECORA IX Symposium.
pp. 296-302.
8. Monmonier, M.S. 1977. Maps, distortion, and meaning. Associa-
tion of American Geographers Resource Paper Number 75-4.
9. Goodchild, M.F., and W Min-Hua. 1988. Modeling error in ras-
ter-based spatial data. Proceedings of the Third International
Symposium on Spatial Data Handling, Sydney, Australia (August
17-19). pp. 97-106.
10. Walsh, S.J. 1989. Inherent and operational error within a CIS:
Evaluation from a user perspective. In: Goodchild, M.F., ed. Ac-
curacy of spatial databases, p. 290.
11. Burrough, PA. 1986. Principles of geographic information sys-
tems for land resources assessment. Oxford, England: Clarendon
Press.
12. Marble, D.F., and D.J. Peuquet. 1983. Geographic information
systems and remote sensing: Manual of remote sensing, 2nd ed.
Falls Church, VA: American Society of Photogrammetry and Re-
mote Sensing, pp. 923-958.
13. Mead, D.A. 1982. Assessing data quality in geographic informa-
tion systems. In: Johannsen, C.J., and J.L. Sanders, eds. Remote
sensing for resource management. Ankeny, OH: Soil Conserva-
tion Society of America, pp. 51-59.
14. Heuvelink, G.B.M., and PA. Burrough. 1989. Propagation of
errors in spatial modeling with GIS. Int. J. Geo. Info. Sys.
3(4):303-322.
15. Dunn, R., A.R. Harrison, and J.C. White. 1990. Positional accu-
racy and measurement error in digital databases of land use: An
empirical study. Int. J. Geo. Info. Sys. 4(4):385-398.
16. Walsh, S.J., D.R. Lightfoot, and D.R. Butler. 1987. Recognition
and assessment of error in geographic information systems. Pho-
togrammetric Eng. and Remote Sensing 53(10):1,423-1,430.
17. Owenby, J.R., and D.S. Ezell. 1992. Climatography of the United
States: Monthly station normals of temperature, precipitation, and
heating and cooling; degree days 1961-1990, Kansas. No. 81.
U.S. Department of Commerce, National Oceanic and Atmos-
pheric Administration, National Climatic Data Center, Asheville,
NC (January).
18. Bayne, C.K. 1975. General availability of ground water and nor-
mal annual precipitation in Kansas (1:500,000 map). Kansas
Geological Survey Map Series M4-A.
19. SAS Institute, Inc. SAS/GRAPH Software: Reference, Version 6,
Vol. 1, 1sted. Gary, NC: Statistical Analysis System Institute, Inc.
1990.
20. SAS Institute, Inc. SAS/GRAPH Software: Reference, Version 6,
Vol. 2, 1st ed. Gary, NC: Statistical Analysis System Institute, Inc.
1990.
21. Olea, R.A. 1992. Understanding allays intimidation. Geobyte
7(5):12-17.
22. Bastin, G., B. Lorent, C. Duque, and M. Gevers. 1984. Optimal
estimation of the average areal rainfall and optimal selection of
rain gauge locations. Water Resour. Res. 20(4):463-470.
23. ESRI. 1991. Cell-based modeling with GRID. Redlands, CA: En-
vironmental Systems Research Institute, Inc.
24. ESRI. 1991. Surface modeling with TIN, surface analysis, and
display. Redlands, CA: Environmental Systems Research Insti-
tute, Inc.
25. Davis, J.C. 1982. Statistics and data analysis in geology, 2nd ed.
New York, NY: John Wiley and Sons.
26. Mitchell, J.E. 1991. A regionalization approach based on empiri-
cal Bayes techniques with applications to hydrologic forecasting,
Ph.D. dissertation. Duke University, Durham, NC.
27. Mitchell, J.E., H. Sun, and S. Chakrabarti. 1992. Predicting the
form of a surface from random, nonexperimental samples using
neural networks. Proceedings of the First Midwest Electro-Tech-
nology Conference, Iowa State University, Ames, IA (April 10-11).
28. Congalton, R.G., and D.M. Schallert. 1992. Exploring the effects
of vector to raster and raster to vector conversion. EPA/600/
R-92/166.
29. Mitchell, J.E., and H. Sun. 1992. Generalization in higher order
neural networks when predicting polynomial surfaces, implica-
tions for the free-form surface problem. In: Intelligent engineering
systems through neural networks, Vol. 2. Proceedings of the
Artificial Neural Networks in Engineering (ANNIE '92) Confer-
ence, St. Louis, MO (November 15-18).
30. Bates, J.M., and C.W.J. Granger. 1969. The combination of fore-
casts. Operational Res. Quarterly 20:451-468.
31. Granger, C.J., and R. Ramanathan. 1984. Improved methods of
combining forecasts. J. Forecasting 3:197-204.
32. Carr, D.B., A.R. Olsen, and D. White. 1994. Hexagon mosaic
maps for display of univariate and bivariate geographical data.
EPA/68-C8/0006.
33. Rule, G.K. 1960. Dust bowl. In: The world book encyclopedia,
Vol. 4. pp. 315-316.
14
-------
Data Quality Issues Affecting GIS Use for Environmental Problem-Solving
Carol B. Griffin
Henry's Fork Foundation, Island Park, Idaho
"Abandon hope, all ye who enter here." Dante's quote
might well be the advice that experienced geographic
information system (GIS) users give to nonusers about
to confront data quality issues associated with GIS
use. Indeed, after reading this paper, some decision-
makers might abandon attempts to use a GIS because
of the error associated with it. Others may want to spend
an inordinate amount of time and money trying to elimi-
nate all error associated with GIS use. Neither option is
prudent.
Data quality is important because it affects how reliable
CIS-generated information is in the decision-making
process. Too often, the availability of inexpensive digital
data overshadows data quality concerns; people fre-
quently use digital data because they are available, not
because they have the necessary accuracy.
A GIS can help decision-makers use spatial informa-
tion more fully than manual methods allow, but some-
times data quality issues cause concern about using
CIS-generated outputs. Making environmental deci-
sions without adequate consideration to data quality
may lead to an erroneous decision, erode public confi-
dence, or cause an agency to incur liability. This paper
attempts to encourage decision-makers to become
more aware of data quality issues, including the sources
and magnitude of error.
GIS error research has necessarily progressed in a
linear fashion, beginning with identifying and classifying
sources of error. This paper discusses both inherent
(source) error and the error that GIS operations intro-
duce (operational error) during data input, storage,
analysis/manipulation, and output (1). Strategies for
coping with error and research into error reduction tech-
niques have only recently received attention. Unfortu-
nately, the answers to error management questions
such as, "How will the error affect decision-making?" are
not clear. The end of this paper covers several error
management suggestions and anticipated software im-
provements designed to reduce errors, however.
Data Quality Concepts and Their
Importance
Data quality is a major issue for CIS-generated maps,
much more so than it is for paper maps. In part, this is
because a GIS can perform operations on spatial data
that would be nearly impossible without a GIS because
of scale, complexity, and generalization issues (2). Car-
tographers adjust for these problems when they manu-
ally manipulate and instantly combine paper maps by
adhering to long-standing cartographic principles, but
GIS personnel may not be fully trained in these princi-
ples. A GIS enables an analyst, whether trained in car-
tographic principles or not, to combine or manipulate
data in appropriate or in inappropriate, illogical, and
erroneous ways. Lack of training coupled with the speed
of spatial data manipulation can have serious conse-
quences for an agency whose personnel produce and
use CIS-generated maps.
Limited scientific understanding, limited ability to meas-
ure data, sampling error, inherent variability, and inade-
quacy of mathematical representations all contribute to
uncertainties associated with spatial data. Uncertainty
about spatial data consists of two parts: ignorance and
variability. Ignorance means that variables have a "true"
value, but it is unknown to us, whereas variability means
one value cannot represent the variables.
Data quality defies a simple definition. For this paper,
data quality can roughly mean how "good" the data are
for a given purpose. People usually think of data quality
in terms of error, but the term is broader and encom-
passes the six components outlined in the next section.
Error can mean the difference between the observed
values and the "true" value. The "true" value of a variable
is usually unknown and unknowable, but for this paper's
purposes, "true" could be the known value or the value
one would obtain from field measurements (the discus-
sion of data collection tries to dispel the notion that there
is one "true" value for many variables, such as soil type
in a given area orwatertemperature in a lake). Imperfect
15
-------
equipment or observers and environmental effects
cause spatial error. According to Thapa and Bossier (3),
errors fall into three categories:
Gross errors and blunders (people or equipment).
Systematic errors (which introduce bias).
Random errors (due to imperfect instruments and
observers).
In addition, another view divides spatial error into two
different components: accuracy and precision. Accuracy
means how close a value is to the "true" value or a
known standard (absence of bias). Precision can have
two definitions: it can be a measure of dispersion (stand-
ard deviation) of observations about a mean, or it can
refer to the number of decimal digits used to represent
a value (4). In the first definition of precision, a meas-
urement of 6 feet plus or minus 1 foot is more precise
than one of 6 feet plus or minus 3 feet. In the second
definition, a value of 6.1794 feet is more precise than
one of 6.1 feet. Figure 1 provides a graphic explanation
of the difference between error, accuracy, and precision.
Error
Accuracy
Precision
True Value
4 1 Mean 2 3
Observations
Figure 1. Relationship between error, accuracy, and precision.
Data are not accurate or inaccurate. Instead, data accu-
racy exists on a continuum, ranging from low to high
accuracy. Although people strive for accurate (error-
free) data, obtaining 100-percent accurate data is im-
practical. The list below provides some of the reasons
why total accuracy is not obtainable (5):
Objects to be measured are often vaguely defined.
Some phenomena are variable in nature.
Classification schemes are imprecise.
Measurements are inherently imprecise.
Gross errors of a nonstatistical nature can occur during
measurement.
Attributes encoded on an ordinal scale (high, me-
dium, low) are approximate.
Data represent a past state of reality.
Users of geographic data should strive for data that are
only as accurate as they need. A variety of factors, of
course, can determine need:
Intended use of the data
Budget constraints
Time constraints
Data storage considerations
Potential liability
The main barrier to highly accurate data is lack of funds.
Male (6) suggests that rather than abandoning a CIS
project because funds are not sufficient to achieve the
desired accuracy, an agency should collect data at the
desired accuracy from smaller areas, such as areas
being developed or redeveloped. Overtime, data collec-
tion at the desired accuracy can expand to include areas
that lacked data due to budgetary constraints. Smith and
Honeycutt (7) outline the use of a value of information
approach in determining the need for more data (or
more accurate data) based on the expected costs and
benefits associated with data collection. If the benefits
of increased data accuracy are greater than the ex-
pected costs, additional funds should be allocated to
obtain more accurate data.
The intended use of data affects the type of data, as well
as the data quality needed. Beard (8) divides CIS appli-
cations into six types (see Table 1). The specific type of
data quality one needs (e.g., positional accuracy, attrib-
ute accuracy) also varies with the intended application.
Analysts with inventory applications such as agricultural
production are less concerned about positional accu-
racy than with an accurate assessment of anticipated
crop yields (attribute accuracy). Decision-makers must
Table 1. Types of GIS Applications (8)
Application Example
Siting Finding optimal location (fire station, waste site)
Logistic Movement or distribution through space
(emergency response, military movement)
Routing Optimal movement through a known network (mail,
school bus)
Navigation Way finding; may or may not involve a known
network (ground, sea, air)
Inventory Count and location of objects for a given time
(census, tax rolls)
Monitoring/ Examining processes over space and time
Analysis (ecological, zoological, geological, epidemiological
studies)
16
-------
decide which data quality component is the most impor-
tant for their use because optimizing all six components
can be very expensive (9). An obvious conflict arises
when local and state governments must meet multiple
application needs simultaneously and thus feel forced to
try to optimize several data quality components.
The nature of the decision may also help decision-makers
determine the data quality they need. Beard (8) lists sev-
eral of these factors (see Table 2). A political, high-risk
decision requires higher quality data than a nonpolitical,
low-risk decision because more public attention focuses
on the former decision.
Table 2. Factors That May Affect the Data Quality Needed for
Decision-Making (8)
Lower Data Quality
Possibly Needed
Higher Data Quality
Possibly Needed
Routine
Nonpolitical
Minimal risk
Noncontroversial
Indefinite
Local implication
Nonroutine
Political
High risk
Controversial
Immediate
Global implication
Components of Data Quality
The National Committee for Digital Cartographic Data
Standards (9) identifies six components of digital carto-
graphic data quality. This section discusses each of
these components:
Lineage
Positional accuracy
Attribute accuracy
Logical consistency
Completeness
Temporal accuracy
Most components of data quality apply to both source
and operational error.
Lineage
Because uses and users of data change, those at the
national level have noted a recent push to include docu-
mentation when disseminating spatial data. Data line-
age, also known as metadata or a data dictionary, is data
about data. Metadata consists of information about the
source data such as:
Date of collection
Short definition
Data type, field length, and format
Control points used
Collection method, field notes, and maps
Data processing steps
Assessment of the reliability of source data
Data quality reports
Access to this information can help CIS personnel de-
termine if the data are appropriate for their use, thereby
minimizing risks associated with using the wrong data
or using data inappropriately. According to Chrisman
(10), the only ethical and probably best legal strategy for
those who produce spatial data is to reveal more infor-
mation about the data (metadata) so that users can
make informed decisions. Eagan and Ventura's article
(11) contains a sample of a generic environmental data
lineage report. The U.S. Environmental Protection
Agency's (EPA's) new locational data policy requires
contractors to estimate data accuracy and provide infor-
mation about the lineage of the data (12).
Positional A ecu racy
Anyone who has used a map has probably come across
features that are not located where the map says they
should be located and has experienced low positional
accuracy. (Undoubtedly, they have also detected fea-
tures that were not on the map, but that is a different
issue.) Positional accuracy, frequently referred to as
horizontal error, is how close a location on a map is to
its "true" ground position. Features may be located inac-
curately on maps for many reasons, including (13):
Poor field work.
Distortion of the original paper map (temperature,
humidity).
Poor conversion from raster to vector or vector to
raster data.
Data layers are collected at different times.
Natural variability in data (tides, vegetation, soil).
Human-induced changes (altering reservoir water
levels).
Movement of features (due to scale of the map and
printing constraints) so they can be easily discerned
by the map reader.
Combining maps with different scales.
Combining maps with different projection and coordi-
nate systems.
Different national horizontal datum in source materials.
Different minimum mapping units.
Positional accuracy has two components: bias and pre-
cision. Bias reflects the average positional error of the
17
-------
sample points and indicates a systematic discrepancy
(e.g., all locations are 7 feet east of where they should
be). Estimating precision entails calculating the stand-
ard deviation of the dispersion of the positional errors.
Usually, root mean square error (RMSE) is reported as
the measure of positional accuracy, but it does not dis-
tinguish bias from precision (14). RMSE is frequently
monitored during digitizing to minimize the introduction
of additional positional error into the CIS.
To determine positional accuracy, one must compare the
location of spatial data with an independent source of
higher accuracy. Federal agencies that collect data and
produce maps adhere to National Map Accuracy Stand-
ards (NMAS) for positional accuracy. Maps such as
United States Geological Survey (USGS) topographic
maps that conform to NMAS carry an explicit statement
on them. Other groups also have developed standards
for large-scale mapping (15).
NMAS for positional accuracy require that not more than
10 percent of well-defined points can be in error by more
than one-thirtieth of an inch for maps at a scale of
1:20,000 or larger. For smaller scale maps, not more
than 10 percent of well-defined points can be in error by
more than one-fiftieth of an inch (16). Thus, less than 10
percent of the well-defined locations on a USGS
1:24,000 map can stand more than 40 feet from their
"true" location; the other 90 percent of the well-defined
points must stand less than 40 feet from their "true"
location. Table 3 shows the acceptable positional accu-
racy for commonly used maps. Note that as scale de-
creases from 1:1,200 to 1:100,000, positional accuracy
decreases.
Several important issues relate to NMAS. First, not all
maps adhere to NMAS, which means their positional
accuracy may be lowerthan NMAS or may be unknown.
Second, NMAS do not indicate the location of points in
error. Third, 10 percent of the well-defined points can
have a positional error greater than the standards allow,
but neither the location nor the magnitude of these
errors are known. Fourth, NMAS apply to well-defined
points; therefore, areas that are not well defined may
Table 3. NMAS Horizontal (Positional) Accuracy
Scale
1:1,200
1 :2,400
1 :4,800
1:12,000
1 :24,000
1:63,360
1:100,000
1 Inch = x Feet
100
200
400
1,000
2,000
5,280
8,333
Horizontal
Accuracy +/- Feet
3.33
6.67
13.33
33.33
40.00
105.60
166.67
have even lower positional accuracy. The implication of
these errors in location is that users should use caution
in making decisions that require high positional accu-
racy. Positional accuracy issues are particularly trouble-
some for CIS operations on small-scale maps or when
combining large-scale maps (1:1,200) with small-scale
maps (1:100,000).
Recently, global positioning systems (GPS), which the
U.S. military developed, have helped to obtain more
accurate feature locations. GPS is not without error,
however. The list below notes some of the possible
sources of error associated with GPS use, some of
which can be controlled while others cannot (17):
Errors in orbital information.
Errors in the satellite clocks.
Errors in the receiver clocks.
Ionospheric or tropospheric refraction.
Deliberate degrading of the satellite signal.
Obstructions that block the signal.
Reflection of the GPS signal off buildings, water, or
metal.
Human error.
The importance of positional accuracy depends on the
intended use of the data. In an urban area, a posi-
tional error of 1 foot on a tax map may be unaccept-
able because 1 foot may be worth millions of dollars.
In a rural area, however, tax boundaries mapped
within 10 feet of their surveyed location may be accu-
rate enough (6). Somers (18) reports that positional
accuracy of 10 to 20 feet may be sufficient for envi-
ronmental analysis. She says the cost of increasing
accuracy to 5 feet could increase the cost of data col-
lection by a factor of 10. The decision-maker must de-
termine the needed positional accuracy.
A ttribute A ccuracy
Attribute accuracy refers to how well the description of
a characteristic of spatial data matches what actually
exists on the ground. For some spatial data, the location
does not change overtime, but the value of the attribute
does (e.g., the location of a census tract does not
change, but the population within a census tract
changes). Attribute accuracy is reported differently for
continuous data (i.e., elevation, which has an infinite
number of values) or discrete data (i.e., gender, which
has a finite number of values).
NMAS exist for elevation contour lines on topographic
maps. NMAS for vertical accuracy state that not more
than 10 percent of the points tested shall be in error by
more than one-half of the contour interval (16). A well-
defined point on a USGS topographic map with a 10-foot
18
-------
contour interval could vary by 10 feet because the actual
elevation could be 5 feet higher or lower than the map
indicates. The implications of these errors are similar to
the ones for positional accuracy. In addition, errors in
elevation are important because small changes in ele-
vation may significantly affect some CIS analysis opera-
tions such as the determination of aspect, slope,
viewshed, and watershed boundaries.
NMAS do not exist for discrete variables such as land
use derived from satellite imagery. Instead, a classifica-
tion matrix reports attribute accuracy. Field checking or
checking a portion of the classified image against a map
of higher accuracy determines the accuracy of the land
use classification. The result of the comparison is a table
from which to calculate overall, producer's, and user's
accuracy. Table 4 is an example of a classification accu-
racy matrix.
Table 4. Example of a Classification Accuracy Matrix (19)
Reference Data ("Ground Truth")
Number of Cells
Classified Data
(Satellite Image)
Number of Cells
Forest
Water
Urban
Total
Forest
28
1
1
30
Water
14
15
1
30
Urban
15
5
20
40
Total
57
21
22
100
Overall Accuracy (sum of the main diagonal)
63
100
= 63%
Producer's Accuracy
(column total)
OQ
Forest = ^ = 93%
oU
User's Accuracy
(row total)
28
Forest = - =
49%
Water = ^ = 50%
oU
Urban = ^ = 50%
Water = ^
20
Urban =
71%
91%
Overall accuracy is the percentage of correctly classified
cells calculated as the sum of the main diagonal (19).
Producer's accuracy is the total number of correct pixels
in a category divided by the total number of pixels of that
category as derived from the reference data (column
total). It corresponds to how well the person classifying
the image (the "producer") can correctly classify or map
an area on the earth. In this example, the producer most
accurately classified forested land (93 percent). User's
accuracy describes the probability that a sample from
the classified area actually represents that category on
the ground. The map "user" is concerned about the
map's reliability. In this example, the most accurately
classified land use from the user's perspective is urban
(91 percent).
The significance of overall, producer's, and user's accu-
racy depends on the intended use of the data. As an
example, Chrisman (20) says that the error in distin-
guishing wetland from pasture may not matter to some-
one estimating open space, but the difference is critical
if the person is estimating the amount of wildlife habitat
available. Story and Congalton (19) provide an example
of how to interpret a classification matrix. A forester
looks at the classification matrix and sees that forest
classification is 93 percent accurate (producer's ac-
curacy); therefore, the analyst did not identify only 7
percent of the forest on the ground. Once the forester
field checks the supposed forested area, she finds that
only 49 percent (28 cells) of the sites mapped as forest
are actually forest; the rest are water (14 cells) or urban
(15 cells) areas.
A report of overall, producer's, and user's accuracy can
help decision-makers determine the appropriateness of
the classified image for their use by identifying potential
errors in classification. This can help direct field work,
which can improve the classification of the image and
perhaps subsequent images. Because CIS analysis fre-
quently uses land use, decision-makers need to know
that significant variability can result when several ana-
lysts classify the same image. Bell and Pucherelli (21)
found that consistency in classification can improve by
having one person classify the entire image. McGwire (22)
even found significant differences between analysts in
unsupervised classification of Landsat imagery. Com-
puters primarily perform unsupervised classification,
which implies that different analysts would classify the
same image in the same way.
Logical Consistency
Logical consistency focuses on flaws in the logical rela-
tionships among data elements. For example, a vector
CIS should label all polygons with only one label per
polygon, and all polygons should be closed. Logical
inconsistency can also occur by collecting data layers at
different times or from different scale maps with different
positional accuracies. For example, the edge of a lake
on the hydrology data layer should coincide with the
edge of land in the land use data layer. If data on the
lake were collected during a wet year rather than a dry
year, the lake's volume would be higher than normal,
affecting its location on the map. If land use data for the
same area were collected during a dry year, the bound-
ary of the lake on the two layers would not be the same.
Logical inconsistencies usually do not appear until the
two maps are overlaid and the boundaries do not coin-
cide (see Figure 2). The user must determine the "cor-
rect" location of the feature that appears misaligned on
one or more data layers. The inconsistency between the
19
-------
Land Use
Hydrology
Overlay
Figure 2. Logical inconsistency in lake and forest location.
location of the two layers resolves through a process
called conflation. All maps are adjusted so that the
feature on each data layer lines up with the same feature
on the base map.
Completeness
Completeness focuses on the adequacy of data collec-
tion procedures. Robinson and Frank (5) discuss two
kinds of uncertainty associated with collecting spatial
data that can lead to error. One type of uncertainty is the
inability to measure or predict an inherently exact char-
acteristic or event with certainty. Examples of this are
blunders in data collection or measurement error, nei-
ther of which can be accurately predicted. The other kind
of uncertainty is associated with concepts that are inher-
ently ambiguous. Crisp data sets, such as property
boundaries, have little ambiguity; the only issue related
to error is the positional accuracy in measuring the
boundary. Because land use data are not crisp data
sets, the challenge is to accurately represent an inher-
ently inexact concept.
Although we know spatial data are variable, our classi-
fication systems generally ignore the second type of
uncertainty. Analysts map data as though all variables
had exact boundaries and all polygons consisted of
homogeneous data. Burrough (4) reports that spatial
variation of natural phenomena is "not just a local noise
function or inaccuracy that can be removed by collecting
more data or by increasing the precision of measure-
ment, but is often a fundamental aspect of nature that
occurs at all scales. . . ."
Mapping spatial data is a function of how humans ag-
gregate and disaggregate data either in space, catego-
ries, quantities, or time; spatial data seldom exist in
nature the way maps depict them (23). Data and rela-
tionships between data are sensitive to the scale and
the zoning system in which the data are reported (24,
25). The modifiable area unit problem occurs because
an analyst can recombine a given set of units or zones
into the same total number of units producing very dif-
ferent results (see Figure 3).
The scale problem occurs because an analyst can com-
bine a set of small units into a smaller number of larger
units, which can change the inferences that can be
Figure 3. Modifiable area unit. (Number of units is constant;
location of units changes.)
made from the data. In Figure 4, the area containing the
highest values changes from the southwest corner in the
first picture to the northern half in the second picture.
For example, water quality data are scale-dependent
because they vary based on the size and location of the
collection area (e.g., adjacent to a point source dis-
charge, a stream segment, the entire river, or the lake
the river discharges into).
Kennedy (25) reports on a similar problem known as the
small number problem. This problem occurs when cal-
culations use a percentage, ratio, or rate for a geo-
graphic area for which the population of interest
(denominator) is sparse or the numerator is a rare event
(1 case of cancer per 1 million people). CIS-generated
maps may highlight a statistically insignificant change in
rare events. Small, random fluctuations in the numerator
may cause large fluctuations in the resulting percent-
age, ratio, or rate. If policy-makers use these maps,
priorities for public health policy may change because
of the erroneous belief that an area is experiencing more
unwanted rare events.
Data can be collected using a tag- or count-based sys-
tem, which affects their usefulness. The tag approach
categorizes items based on the dominant or average
attribute and is ideal for planners who want only one
value for each area. For example, each polygon in a
county soil survey is tagged with one soil type. Accord-
ing to soil taxonomy rules, however, only about 35 per-
cent of a delimited area on a soil survey must match its
classification, and up to 10 percent may be a radically
different soil (26). Although the text in the soil survey
sets limits on data accuracy by listing major impurities
found with each soil type, the CIS seldom carries that
information because analysts only digitize soil bounda-
ries and label data with the dominant attribute. This
Figure 4. Scale problem (number of units changes).
20
-------
leads to the depiction of apparently homogeneous soil
units although the text specifies that the data are not
homogeneous (27).
Some soil or land cover phenomena, even though pre-
sent in small quantities and thus not mapped, may have
great significance for hydrologic models, which makes
the tag approach to data collection troublesome. Data
collected using the count system allow the analyst to
tabulate the frequency of occurrence or areal extent of
a particular phenomenon. Environmental modelers pre-
fer count data but are usually forced to use tag data,
which can introduce error into their models (26). The
new digital soils databases, STATSGO and SSURGO,
are collected and depicted using a count format, which
will help experienced analysts use the data more fully.
Figure 5 shows the difference between tag and count
methods of data collection.
Data are seldom complete because analysts use clas-
sification rules to indicate how homogeneous an area
must be before it is classified a particular way (e.g.,
more than 50 percent, more than 75 percent). Another
decision an analyst must make is where to draw the
boundary between two different areas; it is seldom clear
where a forest leaves off and a rural development be-
gins. Analysts must also decide how or if to show inclu-
sions (e.g., a forested area in the middle of agricultural
land uses).
Temporal A ecu racy
Collecting data at different times introduces error be-
cause the variable may have changed since data collec-
tion. The effect of time, reported as the date of the
source material, depends on the intended use of the
data. Some natural resource data have daily, weekly,
seasonal, or annual cycles that are important to con-
sider. For example, obtaining land use data from re-
motely sensed imagery in November for North Dakota
produces a very different land use map than data ana-
lysts obtain during the July growing season.
In addition, demographic and land use information
changes quickly in a rapidly urbanizing area. Data col-
lected at several times can produce logical inconsis-
tency between data layers, forcing the analyst to adjust
the location of features to coincide with the base map.
Another problem with collecting data at different times
Soil A
Soil B
Soil B 55%'
Soil C 45%
Soil A 30%
Soil B 25%
Soil C 20%
Soil D 25%
30%
Tag Count
Figure 5. Tag and count methods of data collection.
is that data may be collected using different standards,
which may not be apparent to the user (4).
Source Errors in a GIS
Source (or inherent) error derives from errors in data
collection. The amount of error present in collected data
is a function of the assumptions, methods, and proce-
dures used to create the source map (28). Primary data
refers to data collected from field sampling or remote
sensing. Causes of the errors associated with this data
are (3, 4, 8, 14,29):
Environmental conditions (e.g., temperature, humidity).
Sampling system (e.g., incomplete or biased data
collection).
Time constraints.
Map projection.
Map construction techniques.
Map design specifications.
Symbolization of data.
Natural variability.
Imprecision due to vagueness (e.g., classifying a forest).
Measurement error from unreliable, inaccurate, or
biased observers.
Measurement error from unreliable, inaccurate, or
biased equipment.
Lab errors (e.g., reproducibility between lab proce-
dures and between labs).
The process of converting primary data to secondary
data (usually a map) introduces additional error. Many
of the data layers that a GIS analyst acquires are sec-
ondary data. Some of the errors associated with map-
making are (3):
Error in plotting control points.
Compilation error.
Error introduced in drawing.
Error due to map generalization.
Error in map reproduction.
Error in color registration.
Deformation of the material (temperature, humidity).
Error introduced due to using a uniform scale.
Uncertainty in the definition of a feature (boundary
between two land uses).
Error due to feature exaggeration.
Error in digitization or scanning.
21
-------
Converting paper maps to digital data for entry into a
CIS (tertiary data) introduces still more error (the errors
generated from converting paper maps into a digital
format are discussed in the section on input error), in
part because the purpose for which the data was col-
lected differs from the intended use of the data.
Many types of error are associated with data collection:
Data for the entire area may be incomplete.
Data may be collected and mapped at inappropriate
scales.
Data may not be relevant for the intended application.
Data may not be accessible because use is restricted.
Resolution of the data may not be sufficient.
Density of observations may not be sufficient.
The following discussion explains these types of errors.
Data for the Entire Area May Be Incomplete
An incomplete data record may be due to mechanical
problems that interrupt recording devices, cloud cover
or other types of interference, or financial constraints.
Possible solutions to this problem include collecting ad-
ditional data for the incomplete area, using information
from a similar area, generalizing existing large-scale
maps to match the less detailed data needed, or con-
verting existing small-scale maps to large-scale maps to
obtain data at the desired scale. Collecting additional
data may not be a feasible solution because of time or
money constraints. Extrapolating data from the surro-
gate area to the desired area can cause problems be-
cause the areas are not identical and the scale,
accuracy, or resolution of the surrogate area data
may be inappropriate for the intended use. The sec-
tion on analysis/manipulation of data within a CIS
covers the effect of generalization on data quality
as well as the effect of converting small-scale maps
to large-scale maps.
Data May Be Collected and Mapped at a Scale
That Is Inappropriate for the Application
A variety of guidelines suggest the appropriate map
scale to use for various applications (see Table 5). Also,
Table 5. Relationship Between Map Scale and Map Use (6)
Map Scale Map Use
1:600 or larger
1:720 to 1:1,200
1:2,400 to 1:4,800
1:6,000 and smaller
Engineering design
Engineering planning
General planning
Regional planning
some maps and digital databases suggest the type of
application for which they are appropriate (e.g., the
STATSGO digital soil database is suitable for state and
regional planning, whereas SSURGO is suitable for lo-
cal level planning). Tosta (30) cites an example of com-
bining wetland data with parcel boundaries to determine
ownership of the land containing a wetland. If wetland
mapping was done to plus or minus 100 feet positional
accuracy and parcels are 40 feet wide, then the scale of
the wetland map is inappropriate for determining if a
wetland is located on a specific parcel.
Identifying the optimal scale of the necessary data is
crucial because at some point, the cost of collection and
storage exceeds the benefits of increasing the map
scale. Lewis Carroll (1893) summed up the quest for
data mapped at an ever larger scale and the problems
associated with large-scale maps:
"What do you consider the largest map that would
be really useful?"
"About six inches to the mile."
"Only six inches!" exclaimed Mein Herr. "We very
soon got to six yards to the mile. Then we tried a
hundred yards to the mile. And then came the
grandest idea of all! We actually made a map of the
country, on the scale of a mile to the mile!"
"Have you used it much?" I enquired.
"It has never been spread out, yet," said Mein Herr.
"The farmers objected: they said it would cover the
whole country, and shut out the sunlight! So now we
use the country itself, as its own map, and I assure
you it does nearly as well."
Data Collected May Not Be Relevant for the
Intended Application
Frequently, using surrogate data is quicker or cheaper
than collecting needed data (e.g., Landsat imagery
rather than data field collection used to determine land
use) (4). The accuracy and classification scheme used
in collecting the data depends on the intended use of
the data, which may not coincide with the analyst's
purpose. For instance, soil maps were developed to aid
farmers in determining what crops they should plant and
for estimating crop yield. Soil maps, however, see wide
use for very different purposes (e.g., hydrologic and
other environmental models). In addition, STORETdata,
collected at points, are typically extrapolated to repre-
sent water quality in an entire stream stretch.
Data May Not Be Accessible Because
Use Is Restricted
An example of restricted data is Census data on individ-
ual households. An agency may not want to release data
that reveal the location of endangered species. Another
22
-------
example is that people may not even want the informa-
tion mapped. For example, some cavers do not want to
reveal the location of caves to the U.S. Forest Service,
which is charged under the federal Cave Resources
Protection Act with protecting caves, because they think
the best way to protect the caves is to not map them
(31). The National Park Service is putting the location of
petroglyphs in the Petroglyph National Monument into a
CIS. Making their location known to the public, however,
is troublesome because this may, in fact, encourage
their vandalism (32). Other problems in obtaining data
include difficulty in acquisition even if access is not
restricted, expensive collection or input, or unsuitable
format (4, 14).
Resolution of the A vailable Data
May Not Be Sufficient
Spatial resolution is the minimum distance needed be-
tween two objects for the equipment to record the ob-
jects as two entities; that is, resolution is the smallest
unit a map represents. To obtain an approximation of a
map's resolution, divide the denominator of the map
scale by 2,000 to get resolution in meters; for instance,
a 1:24,000-scale map has a resolution of approximately
12 meters (33).
Resolution relates to accuracy in that different map
scales conform to different accuracy standards. Two air
photos shot from the same camera at the same distance
above the ground have the same scale. If one photo has
finer grain film, however, smaller details are evident on
it, and this photo produces a map with higher resolution
(34). According to Csillag (33), analysts cannot simulta-
neously optimize attribute accuracy and spatial resolu-
tion. As spatial resolution increases, attribute complexity
increases (35). Also, the finer the spatial resolution, the
greater the probability that random error significantly
affects a data value.
Resolution of the data is not necessarily the same as
the size of a raster cell in a database. Statistical sam-
pling theory suggests using a raster cell size that is half
the length (one-fourth of the area) of the smallest feature
an analyst wishes to record. Raster data have a fixed
spatial resolution that depends on the size of the cell
employed, but a CIS analyst can divide or aggregate
cells to achieve a different cell size. Frequently, an
analyst transforms data collected at one level of resolu-
tion to a higher level of resolution than existed in the
original source material. According to Everett and Si-
monett (23), "Geographic analysis, however, can be no
better than that of the smallest bit of data which the
system is capable of detecting." Vector data are limited
by the resolution of input/output devices, limits on data
storage, and the accuracy of the digitized location for
individual points (36). The spatial resolution of the data-
base and the processes that operate on it should be
reduced to a level consistent with the data's accuracy.
The spatial resolution needed depends on the intended
use of the data, cost, and data storage considerations.
As resolution increases, so does the cost of collection
and storage. Resolution sufficient to detect an object
means that an analyst can reveal the presence of some-
thing. Identification, the ability to identify the object or
feature, requires three times the spatial resolution of
detection. Analysis, a finer level of identification, re-
quires 10 to 100 times the resolution that identification
needs (23). Increasing resolution increases the amount
of data for storage, with storage requirements increas-
ing by the square of the resolution of the data. For
example, if the resolution of the data needs to change
from 10-meter to 1-meter pixels, file size increases by 102
or 100 times (14).
Density of Observations May Be Insufficient
The density of observations serves as a general indica-
tor of data reliability (4). Users need to know if sampling
was done at the optimum density to resolve the pattern.
Burrough determined that boulder clay in The Nether-
lands could be resolved by sampling at 20-meter inter-
vals or less, whereas coversand showed little variation
in sampling from 20- to 200-meter intervals.
Some strategies for reducing data collection errors are to:
Adhere to professional standards
Allocate enough time and money
Use a rigorous sampling design
Standardize data collection procedures
Document data collection procedures
Calibrate data collection instruments
Use more accurate instruments
Perform blunder checks to detect gross errors
Documenting data collection procedures and distribut-
ing them along with data allows potential users to deter-
mine if the data are suitable for their purposes. By not
documenting procedures, errors in the source material
are essentially "lost" by inputting the data to a CIS, and
the errors become largely undetectable in subsequent
CIS procedures. The result is that agencies that make
decisions based on the CIS-generated map assume the
source data are accurate, only to discover later that the
map contains substantial errors in part due to errors in
the source material.
Operational Errors in a GIS
Data input, storage, analysis/manipulation, and output
can introduce operational errors. Digital maps, unlike
23
-------
paper maps, can accumulate new operational errors
through CIS operations (8). Even if the input data were
totally error-free, which the last section demonstrated is
not the case, CIS operations can produce positional and
attribute errors. The CIS operation itself determines to
a large extent the types of errors that result.
Input Errors
The process of inputting spatial and attribute data can
introduce error. The major sources of input error are
manual entry of attribute features and scanning or dig-
itizing spatial features. Manual entry errors include in-
complete entry of attribute data, entering the wrong
attribute data, or entering the right attribute data at the
wrong location. Digitizing errors originate from equip-
ment, personnel, or the source material (see Table 6).
Digitizing errors, such as under- and overshoot of lines
and polygons that are not closed, can introduce error
(see Figure 6). CIS software can "snap" lines together
that really do not connect. Depending on the tolerance
Table 6. Types of Digitizing Errors (4, 14, 37)
Personnel Errors
Changes in the origin
Incorrect registration of the map on the digitizing table
Creation of over- and undershoots
Creation of polygons that are not closed
Incomplete spatial data when data are not entered
Duplication of spatial data when lines are digitized twice
Line-following error (inability to trace map lines perfectly with the
cursor)
Line-sampling error (selection of points used to represent the map)
Physiological error (involuntary muscle spasms)
Equipment Errors
Digitizing table (center has higher positional accuracy than the
edges)
Resolution of the digitizer
Differential accuracy depending on cursor orientation
Errors in Source Material
Distortion because source maps have not been scale-corrected
Distortion due to changes in temperature and humidity
Necessity of digitizing sharp boundary lines when they are gradual
transitions
Width of map boundaries (0.4 mm) digitized with a 0.02-mm accu-
racy digitizer
Undershoot Overshoot Polygon Not Closed
Figure 6. Common digitizing errors.
selected, this can result in the movement of both lines,
which can decrease the accuracy of the resultant map.
Despite the long list of personnel errors associated with
digitizing, a good operator probably contributes the least
error in the entire digitizing process (38). Giovachino
discusses methods that can help determine equipment
accuracy, including checking the repeatability, stability,
and effect of cursor rotation. Digitizing accuracy var-
ies based on the width, complexity, and density of the
feature being digitized but typically varies from 0.01
to 0.003 (3).
One problem with digitized data is that the data can
imply a false sense of precision. Boundaries on paper
maps are frequently 0.4 mm wide but are digitized with
0.02-mm accuracy. The result is that the lines are stored
with 0.02-mm accuracy, implying a level of precision that
far exceeds the original data.
Minimizing digitizing errors is important because the
errors can affect subsequent CIS analysis. Campbell
and Mortenson (39) provide a list of procedures they
used to reduce errors associated with digitizing and
labeling:
Use log sheets to ensure consistency and account-
ability, and to provide documentation.
Check for completeness in digitizing all lines and
polygons.
Check for complete and accurate polygon labeling.
Set an acceptable RMSE term for digitizing (usually
0.003).
Always overshoot rather than undershoot when
digitizing.
Overlay a plot of the digitized data with the source
map to check lines and polygons. If light passes be-
tween the digitized line segment and source map,
redigitize it.
Check digitized work immediately to provide feed-
back to the digitizer operator and to help identify and
correct systematic errors.
Limit digitizing to less than 4 hours a day.
Involve people in doing CIS-related jobs other than
digitizing to decrease turnover and increase the level
of experience.
Storage Errors
Data storage in a CIS usually involves two main types
of errors. First, many CIS systems have insufficient
numerical precision, which can introduce error due to
rounding. Integers are stored as 16 or 32 bits, which
have four significant figures. Real numbers are stored
as floating point numbers either in single precision (32 bit,
24
-------
7 significant figures) or double precision (64 bit, 15 or 16
significant figures). If the data in a CIS range from
fractions of a meter to full UTM coordinates, typical
32-bit CIS systems cannot store all the numbers. Using
double precision (64 bits) reduces this problem but in-
creases storage requirements.
Second, CIS processing and storage usually ignore
significant digits (data precision). As a result, the preci-
sion of CIS processing frequently exceeds the accuracy
of the data (40). When a CIS converts a temperature
recorded and entered as 70 degrees Fahrenheit (near-
est degree) to centigrade, the CIS stores the tempera-
ture as 21.111 degrees rather than 21 degrees, which
the significant figures in the original temperature meas-
urement would dictate. Using the accuracy of the data,
not the precision of floating point arithmetic, partially
resolves this but requires the user to make a special
effort because the CIS does not automatically track
significant figures.
Analysis/Manipulation Errors
CIS analysis/manipulation functions, designed to trans-
form or combine data sets, also can introduce errors.
These errors originate from the measurement scale
used or during data conversion (vector to raster and
rasterto vector), map overlay, generalization, converting
small-scale to large-scale maps, slope, viewshed, and
other analysis functions. One of the biggest problems
associated with CIS use is that data in digital form are
subject to different uses than data in paper form be-
cause the user has access to multiple data layers.
Measurement Scale
Four measurement scales can depict spatial data: nomi-
nal, ordinal, interval, or ratio scales. A name or letter
describes nominal data (e.g., land use type, hydrologic
soil group C). Performing mathematical operations such
as addition and subtraction on nominal data is meaning-
less. Ordinal or ranked data have an order to them such
as low, medium, and high. Interval data have a known
distance between the intervals such as 0, 1 to 5, 6 to 9,
more than 9. Ratio data are similar to interval data
except ratio data have a meaningful zero (e.g., tempera-
ture on the Kelvin scale).
Often during CIS operations, analysts convert interval
or ratio data into nominal data (e.g., low slope is 0 to
3 percent, medium slope is 4 to 10 percent), resulting in
a loss of information. Analysts should preserve the origi-
nal slope values in the CIS in case the user later wants
to modify the classification scheme. Robinson and
Frank (5) describe the tradeoff between information con-
tent and the meaning that can be derived from it, which
partly helps explain why interval data are frequently
converted to nominal data. The authors identify a con-
tinuum progressing from nominal data at one end that is
highly subjective, has low information content, and high
meaning (low slope means something to the average
user) to ratio data that has low subjectivity, high infor-
mation content, and low meaning (a slope of 7 percent
may not mean much to the average user).
Data Conversion
Errors can occur in converting a vector map to a raster
map or a raster map to a vector map. For instance,
remotely sensed data are collected using a raster-based
system. Using a vector CIS, however, requires conver-
sion from raster to vector data. The size of the error
depends on the conversion algorithm, complexity of fea-
tures, and grid cell size and orientation (13).
A line on a vector map converted to a raster map has
lower accuracy in the raster representation because
vector data structures store data more accurately than
raster ones. When polygons in a vector CIS are con-
verted to a raster CIS, the coding rule usually used
assigns the value that covers the largest area within the
cell of a categorical map to the entire cell (see Figure 7).
For example, when placing a grid over a vector map with
an urban land polygon adjacent to an agricultural poly-
gon, the cell placement can include part of both poly-
gons. If the resultant cell comprises 51 percent urban
and 49 percent agricultural land, the cell is assigned 100
percent urban. Converting a numerical map between
raster and vector systems requires spatial interpolation
procedures. CIS software packages use different inter-
polation methods that can produce a different output
even when using the same input data.
Vector Raster
Figure 7. Polygon conversion from vector to raster data.
Map Overlay
Map overlay, used extensively in planning and natural
resource management, is the combining of two or more
data layers to create new information. In a vector CIS,
slivers or spurious polygons can result from overlaying
two data layers to produce a new map (slivers cannot
be formed in a raster-based CIS). When combining the
data layers, lines do not coincide, resulting in the crea-
tion of a new polygon or sliver that did not exist on either
layer (see Figure 8). Unfortunately, as accuracy in digit-
izing increases, so does the number of slivers (41).
Positional error in the boundaries can occur because of
25
-------
o
o
Sliver
Figure 8. Sliver example.
mistakes in measuring or converting the data to digital
form, incremental expansion or recession of a real world
boundary over time, or the fact that certain boundaries
are difficult to determine and thus are generalized
differently (42).
The number of map layers, accuracy of each map layer,
and the coincidence of errors at the same position from
several map layers all determine the accuracy of the
map overlay procedures (43). Using probability theory,
Newcomer and Szajgin determined that the highest ac-
curacy to expect from a map overlay is equal to the
accuracy of the least accurate map layer. The lowest
accuracy in map overlay occurs when errors in each
map occur at unique points.
In the quest for more accurate results, CIS modelers
have increased the complexity of their models and
therefore have increased the number of data layers
needed. Guptill (44) states, "Conventional wisdom
would say that as you add more data to the solution of
a problem, the likelihood of getting an accurate solution
increases. However, if each additional data layer de-
grades the quality of the combined data set, and hence
the accuracy of the solution, then additional data sets
may be counterproductive."
Generalization
Monmonier (45) provides an extensive discussion of
geometric and content generalization procedures used
in map-making. Table 7 lists common types of generali-
zation. Generalizing data on a map helps to focus the
user's attention on one or two types of information and
to filter out irrelevant details. Generalizing is performed
by reducing the scale of the data; a 1:24,000-scale map
can be generalized to a 1:100,000-scale map so that all
data layers have the same scale. With generalizing,
areas on a large-scale map become point or line
features on a small-scale map (35). Obtaining some
measurements from small-scale maps, however, re-
quires caution. For example, a map may depict a
40-foot wide road as a single line one-fiftieth of an inch
wide. On a 1:100,000 map, one-fiftieth of an inch
translates into a 160-foot wide roadfour times the
actual width of the road.
Several studies have pointed to errors that can result
from generalization. Wehde (46) compared soil maps
generated from 0.017-acre grid cells and 11 progres-
sively increasing grid cell sizes. He found that as grid
Table 7. Generalization Operations (45)
Geometric Generalization
Generalizing a Line
Simplification
Displacement
Smoothing
Enhancement
Selection
Generalizing a Point
Selection
Displacement
Graphic association
Abbreviation
Aggregation
Area conversion
Generalizing an Area
Selection
Simplification
Displacement
Smoothing
Enhancement
Aggregation
Dissolution
Segmentation
Point conversion
Line conversion
Content Generalization
Selection
Classification
cell size increased, map accuracy decreased. More re-
cently, Stoms (47) found that generalizing a habitat map
from 1 to 25, 100, and 500 hectares decreased the
number of habitat types and the number of species
predicted.
Transforming Small-Scale Maps to
Large-Scale Maps
Converting small-scale maps (1:250,000) to large-scale
maps (1:24,000) is advisable only if the analyst fully
appreciates the effect of this procedure on map quality.
Data mapped at a small scale are subject to different
accuracy standards than data mapped at a large scale.
Connin (48) reports, "Problems with accuracy arise
when positions are reported to decimal parts of a foot or
meter, but the method of data capture may cause the
positional error to be as much as hundreds of feet or
meters." Yet when converting the data from small- to
large-scale, the data appear to have the accuracy of the
large-scale map. Theoretically, data should not be trans-
formed and used at a scale larger than the scale of the
document from which the data are derived (3).
26
-------
Slope and Viewshed
CIS software packages use a variety of algorithms to
calculate slope and viewsheds and can produce very
different results. Algorithms are an unambiguous set of
rules or a finite sequence of operations used to carry out
a procedure. Smith, Prisley, and Weih (49) used six
different CIS algorithms to determine slope on 5,905
acres of land in order to calculate the amount of land
deemed unsuitable for timber harvesting. They found
that unsuitable land varied from 175 to 1,735 acres,
indicating that different algorithms produce very different
results. Felleman and Griffin (50) found that CIS pack-
ages with different algorithms generate alternate view-
sheds (the area that can be seen from a point).
Output Errors
A variety of errors are associated with data output:
Output devices create error.
Paper shrinks and swells.
Line implies certainty that may not exist because
boundaries are gradual.
A cell or polygon implies homogeneity.
Scale can be modified to imply higher accuracy than
exists in the source data.
Precision can be modified to imply higher precision
than exists in the source data.
Depiction of symbols and colors may not follow
convention.
An important problem associated with CIS-generated
maps is that users make informal assessments about
data quality, partially based on how they perceive the
quality of the output. A hand-drawn map connotes a
lower level of accuracy than a five-color, CIS-produced
map complete with scale and agency logo. Another
problem with output is that distinguishing highly accu-
rate data from less accurate data is impossible on a
CIS-generated map. Users want the output from a CIS
to look like maps they usually see, perpetuating the
notion that lines mark exact boundaries and that poly-
gons or cells are homogeneous. Maps that federal map-
ping agencies produce frequently follow NMAS, but
CIS-generated maps seldom adhere to published map
accuracy standards. An agency could require that CIS
map products meet NMAS, which would establish and
maintain data standards from data collection to output.
A pen stroke of one-fiftieth of an inch on an output device
translates to an error of 40 feet on the ground for a
1:24,000-scale map (6). Small changes in paper maps
due to changes in temperature and humidity can repre-
sent several feet on the ground. As previously noted,
analysts can modify the scale of CIS maps to whatever
they desire. The basic rule of informational integrity is
that the implied precision of data output should not
exceed the precision (spatial, temporal, or mathemati-
cal) of the least precise input variable (26).
CIS-generated maps probably do not differ significantly
from paper maps in their implication that lines and poly-
gons on the map represent certainty and homogeneity.
CIS-generated maps, however, may not depict standard
symbols, sizes, shapes, colors, and orientation. For ex-
ample, paper geological maps use dashed lines to show
inferred, rather than actual, field collected data, but geo-
logical maps in a CIS may not follow the same conven-
tion (27). Cartographers conventionally use blue lines to
indicate water, but a CIS map-maker can show water as
red rather than blue.
Even more troublesome are the color schemes that
some analysts use in depicting model output. Analysts
often give little thought to assigning the colors to model
results depicted as ordinal rankings. For example, areas
of high erosion might be blue, medium erosion might be
red, and low erosion might be green. This selection of
colors ignores the intuitive meaning that people assign
to colors. It has been suggested that the color ordering
used in stop lights might provide a better option. In that
case, areas of high erosion would be red, medium ero-
sion would be yellow, and low erosion would be green.
Error Reduction Techniques
Although CIS users and researchers develop error re-
duction strategies, ultimately users must rely on CIS
software developers to implement new error reduction
techniques in CIS packages. Error reduction techniques
range from simple software warnings to prohibiting a
user from performing selected CIS procedures. Dutton
(51) predicts that future CIS programs will automate
data manipulation (i.e., size, format, and placement
of feature labels on maps) in keeping with standard
cartographic principles. Dutton (51) and Beard (8) also
predict that future CIS packages will enforce metadata-
based constraints such as operations that are illegal or
illogical (e.g., determining the average value of nominal
data such as land use), or are inadvisable (e.g., over-
laying maps with widely different scales).
Another change Dutton anticipates is that software ven-
dors will include information in manuals that explains
how executing a specific command may affect the data-
base. Graphic techniques to depict error are being de-
veloped for nonexpert users while experts tend to use
spatial statistics. Felleman (52) and Berry (53) present
an interesting graphic portrayal of an error map that may
indicate the future of error maps. Additional research
must determine what effect errors will have on decision-
making.
27
-------
Error Management
Ultimately, the decision-maker must determine what to
do with the information in this paper. A decision-maker
has a variety of possible courses of action, ranging from
prudent steps that attempt to minimize error and the
effect it has on decisions, to other less useful options.
Possible actions are to:
Abandon use of a CIS.
Ignore the error associated with CIS use.
Attempt to collect "error-free" data.
Determine if the data are accurate enough for the
intended purpose.
Develop and use data quality procedures.
Obtain and use an error report with CIS-generated
output.
Ask that CIS-generated maps show potential errors.
Continually educate users about the appropriate use
of spatial data.
First, the decision-maker could abandon any attempt to
use a CIS because of the errors associated with its use.
At times, this may be the appropriate strategy, but this
approach ignores the potential benefits associated with
CIS use.
Second, the decision-maker could ignore the error associ-
ated with CIS use and continue to use the CIS for deci-
sion-making. This type of "head in the sand" approach is
not advisable because of the potential liability associated
with making decisions based on inaccurate data.
Third, the decision-maker could engage in an expensive
and time-consuming effort to collect highly accurate er-
ror in hopes that error becomes a nonissue. Depending
on the intended use of the data, the cost of collecting
more accurate data may exceed the benefit.
Fourth, the decision-maker could assess whether the
information available is accurate enough for the in-
tended purpose. If data quality is too low, the decision-
maker may opt to collect new data at the desired quality.
If collecting additional data is not possible, the decision-
maker can explore what types of decisions are possible
given the attainable data quality. For instance, Hunter
and Goodchild (54) found that the data they were using
were suitable only for initial screening rather than for
regulatory and land-purchasing decisions.
Fifth, procedures to ensure high quality data could be
developed and used in the data collection, input, and
manipulation stages of building a CIS database.
Sixth, the decision-maker could require a quantitative or
at least a qualitative report on the sources, magnitude,
and effects of errors. The absence of an error report
does not mean the map is error-free (36). Dutton (51)
predicts that in the near future users of geographic data
will demand error reports, confidence limits, and sensi-
tivity analyses with CIS-generated output.
Seventh, the decision-maker could ask for CIS-generated
maps that adequately portray the error in the final map. For
example, areas where the uncertainty is high could appear
in red on maps. Another option is to place a buffer around
lines to indicate the relative positional accuracy of a line or
to show transition zones. Finally, an analyst can present
the output in ways other than a dichotomous yes or no;
instead, the analyst may use yes, maybe, or no depictions
or even more gradations.
Finally, Beard (8) introduced the concept of directing
efforts toward educating users about use error. She
defines use error as the misinterpretation of maps or
misapplication of maps to tasks for which they are not
appropriate. "We can't assume that CIS will automat-
ically be less susceptible to misuse than traditional
maps, and it may, in fact, exacerbate the problem by
expanding access to mapped information." Beard ar-
gues that money directed to reducing source and opera-
tional error, while important, may not matter if use error
is large.
Conclusions
CIS is a powerful tool for analyzing spatial data. Every-
one who uses CIS-generated output, however, must be
aware of source errors and operational errors intro-
duced during data input, storage, analysis/manipulation,
and output. Increased awareness of the sources and
magnitude of error can help decision-makers determine
if data are appropriate for their use. Decision-makers
cannot leave data quality concerns to CIS analysts be-
cause efforts to improve data quality are not without
cost, and the decision-makers typically control funding.
Decision-makers must not get caught up in the glamour
of the spatial analyses and outputs that a CIS can
produce. These attributes may lead decision-makers to
ignore issues associated with uncertainty, error, accu-
racy, and precision. Inexpensive digital data can make
analysts and decision-makers ignore data quality. If sub-
sequent management decisions are made based on
poor quality data, the resultant decisions may turn out
wrong. This would give decision-makers a jaded view of
the usefulness of CIS. An adequate understanding of
data quality issues can help decision-makers ask the
right questions of analysts and avoid making decisions
that are inappropriate given the data quality.
References
1. Walsh, S.J., D.R. Lightfoot, and D.R. Butler. 1987. Recognition
and assessment of error in geographic information systems. Pho-
togrammetric Eng. and Remote Sensing 53(10):1,423-1,430.
28
-------
2. Openshaw, S. 1989. Learning to live with errors in spatial data-
bases. In: Goodchild, M., and S. Gopal, eds. The accuracy of
spatial databases. Bristol, PA: Taylor and Francis, pp. 263-276.
3. Thapa, K., and J. Bossier. 1992. Accuracy of spatial data used
in geographic information systems. Photogrammetric Eng. and
Remote Sensing 58(6):835-841.
4. Burrough, PA. 1986. Principles of geographical information sys-
tems for land resources assessment. New York, NY: Oxford Uni-
versity Press.
5. Robinson, V.B., and A.U. Frank. 1985. About different kinds of
uncertainty in collections of spatial data. In: Proceedings of Digital
Representations of Spatial Knowledge, Washington, DC, Auto-
Carto 7. Falls Church, VA: ASPRS and ACSM. pp. 440-449.
6. Nale, O.K. 1992. Do traditional map standards conflict with a CIS
landbase? CIS World 5(7):50-53.
7. Smith, G.R., and D.M. Honeycutt. 1987. Geographic data uncer-
tainty, decision-making, and the value of information. In: Proceed-
ings of CIS'87, San Francisco, CA (October 26-30). Falls Church,
VA: ASPRS and ACSM. pp. 300-312.
8. Beard, K. 1989. Dimensions of use and value of geographic
information. In: Onsrud, H., H.W Calkins, and N.J. Obermeyer,
eds. In: Proceedings of use and value of geographic information:
Initiative four specialist meeting, Tenants Harbor, ME (August).
Vol. 89-7. Santa Barbara, CA: NCGIA.
9. National Committee for Digital Cartographic Data Standards.
1988. The proposed standards for digital cartographic data. Amer.
Cartographer 15(1):11-140.
10. Chrisman, N. 1994. A vision of digital libraries for geographic
information, or how I stopped trying to find the on-ramps to the
information superhighway. Geo. Info. Sys. 4(4):21-24.
11. Eagan, P.O., and S.J. Ventura. 1993. Enhancing value of envi-
ronmental data: Data lineage reporting. J. Environ. Eng. 119(1):5-
16.
12. U.S. EPA. 1990. Information resources management policy man-
ual. Washington, DC.
13. Lunetta, R.S., R.C. Congalton, L.K. Fenstermaker, J.R. Jenson,
K.C. McGwire, and L.R. Tinney. 1991. Remote sensing and geo-
graphic information system data integration: Error sources and
research issues. Photogrammetric Eng. and Remote Sensing
57(6):677-688.
14. Aronoff, S. 1989. Geographic information systems: A manage-
ment perspective. Ottawa, Ontario: WDL Publications.
15. ASPRS. 1990. ASPRS accuracy standards for large-scale maps.
Photogrammetric Eng. and Remote Sensing 56(7):1,068-1,070.
16. Thompson, M.M. 1987. Maps for America: Cartographic products
of the U.S. Geological Survey and others, 3rd ed. Washington,
DC: U.S. Geological Survey.
17. Ferguson, J. 1993. What is it we obtain from GPS? Professional
Surveyor 13(4):3-4.
18. Somers, R. 1994. Is a CIS department the answer? Geo. Info.
Sys. 4(6):29-32.
19. Story, M., and R.G. Congalton. 1986. Accuracy assessment: A
user's perspective. Photogrammetric Eng. and Remote Sensing
52(3):397-399.
20. Chrisman, N.R. 1987. The accuracy of map overlays: A reassess-
ment. Landscape and Urban Planning 14:427-439.
21. Bell, A.D., and M.J. Pucherelli. 1992. Audubon Lake island ero-
sion study using aerial photography and geographic information
systems. Report No. R-92-12. Denver, CO: U.S. Department of
the Interior, Bureau of Reclamation.
22. McGwire, K.C. 1992. Analyst variability in labeling of unsuper-
vised classifications. Photogrammetric Eng. and Remote Sensing
58(12):1,673-1,677.
23. Everett, J., and D.S. Simonett. 1976. Principles, concepts, and
philosophical problems in remote sensing. In: Lintz, J., and D.S.
Simonett, eds. Remote sensing of the environment. Reading, MA:
Addison-Wesley.
24. Fortheringham, A.S. 1989. Scale-independent spatial analysis. In:
Goodchild, M., and S. Gopal, eds. The accuracy of spatial data-
bases. Bristol, PA: Taylor and Francis, pp. 221-228.
25. Kennedy, S. 1989. The small number problem and the accuracy
of spatial databases. In: Goodchild, M., and S. Gopal, eds. The
accuracy of spatial databases. Bristol, PA: Taylor and Francis, pp.
187-196.
26. Brown, D.A., and PJ. Gersmehl. 1987. Maintaining relational ac-
curacy of geocoded data in environmental modeling. In: Proceed-
ings of CIS '87, San Francisco, CA (October 26-30). Falls Church,
VA: ASPRS and ACSM. pp. 266-275.
27. Fisher, PF. 1987. The nature of soil data in GISerror or uncer-
tainty. In: Proceedings of the International Geographic Informa-
tion Systems (IGIS) Symposium, Arlington, VA (November 15-18).
Vol. Ill, pp. 307-318. Arlington, VA: Association of American Ge-
ographers.
28. Walsh, S.J. 1989. User considerations in landscape charac-
terization. In: Goodchild, M., and S. Gopal, eds. The accuracy of
spatial databases. Bristol, PA: Taylor and Francis, pp. 35-43.
29. Stoms, D. 1987. Reasoning with uncertainty in intelligent geo-
graphic information systems. In: Proceedings of GIS '87, San
Francisco, CA (October 26-30). Falls Church, VA: ASPRS and
ACSM. pp. 693-700.
30. Tosta. 1993. The data wars: Part II. Geo. Info. Sys. 3(2):22-26.
31. Anonymous. 1994. Between a rock and a hard place. Backpacker
22(8):16-18.
32. Fletcher, M., and D. Sanchez. 1994. Etched in stone: Recovering
Native American rock art. GPS World 5(10):20-29.
33. Csillag, F. 1991. Resolution revisited. In: 1991 ACSM-ASPRS
Annual Convention, Baltimore, MD (March 25-28). Auto-Carto 10,
Volume 6. Bethesda, MD: ACSM and ASPRS. pp. 15-28.
34. Star, J.L., and J.E. Estes. 1990. Geographic information systems:
An introduction. Englewood Cliffs, NJ: Prentice Hall.
35. Goodchild, M. 1993. The state of GIS for environmental problem-
solving. In: Goodchild, M.R., B.O. Parks, and L.T Steyaert, eds.
Environmental modeling with GIS. New York, NY: Oxford Univer-
sity Press, pp. 8-15.
36. Veregin, H. 1987. Error modeling in geographic information sys-
tems: A review. Unpublished manuscript.
37. Keefer, B.J., J.L. Smith, and T.G. Gregoire. 1991. Modeling and
evaluating the effects of stream mode digitizing errors on map
variables. Photogrammetric Eng. and Remote Sensing 57(7):957-
963.
38. Giovachino, D. 1993. How to determine the accuracy of your
graphic digitizer. Geo. Info. Sys. 3(3):50-53.
29
-------
39. Campbell, W.G., and D.C. Mortenson. 1989. Ensuring the quality
of geographic information system data: A practical application of
quality control. Photogrammetric Eng. and Remote Sensing
40. Goodchild, M., and S. Gopal. 1989. Preface. In: Goodchild, M.,
and S. Gopal, eds. The accuracy of spatial databases. Bristol,
PA: Taylor and Francis.
41 . Goodchild, M. 1 993. Data models and data quality: Problems and
prospects. In: Goodchild, M.F., B.O. Parks, and L.T Steyaert, eds.
Environmental modeling with GIS. New York, NY: Oxford Univer-
sity Press, pp. 94-103.
42. Lester, M., and N. Chrisman. 1991. Not all slivers are skinny: A
comparison of two methods for detecting positional error in cate-
gorical maps. In: Proceedings of GIS/LIS '91, Atlanta, GA (Octo-
ber 28-November 1). Bethesda, MD: ASPRS. pp. 648-658.
43. Newcomer, J.A., and J. Szajgin. 1984. Accumulation of thematic
map errors in digital overlay analysis. Amer. Cartographer
44. Guptill, S.C. 1989. Inclusion of accuracy data in a feature-based,
object-oriented data model. In: Goodchild, M., and S. Gopal, eds.
The accuracy of spatial databases. Bristol, PA: Taylor and Fran-
cis. pp. 91-98.
45. Monmonier, M. 1991. How to lie with maps. Chicago, IL: Univer-
sity of Chicago Press.
46. Wehde, M. 1982. Grid cell size in relation to errors in maps and
inventories produced by computerized map processing. Photo-
grammetric Eng. and Remote Sensing 48(8): 1, 289-1 ,298.
47. Stoms, D.M. 1992. Effects of habitat map generalization in bio-
diversity assessment. Photogrammetric Eng. and Remote Sens-
ing 58(11):1,587-1,591.
48. Connin, R. 1994. The trouble with data. Professional Surveyor
14(1):24-25.
49. Smith, J.L., S.P. Prisley, and R.C. Weih. 1991. Considering the
effect of spatial data variability on the outcomes of forest man-
agement decisions. In: Proceedings of GIS/LIS '91, Atlanta, GA
(October 28-November 1). Bethesda, MD: ASPRS. pp. 286-292.
50. Felleman, J., and C.B. Griffin. 1990. The role of error in CIS-
based viewshed determination: A problem analysis. Report No.
IEPP-90-2. Syracuse, NY: Institute of Environmental Policy and
Planning, State University of New York, College of Environmental
Science and Forestry.
51. Dutton, G. 1991. Improving spatial analysis in GIS environments.
In: 1991 ACSM-ASPRS Annual Convention, Baltimore, MD
(March 25-28). Auto-Carto 10, Volume 6. Bethesda, MD: ACSM
and ASPRS. pp. 168-185.
52. Felleman, J. 1992. Letter to the editor. Geo. Info. Sys. 2(8):12-13.
53. Berry, J.K. 1994. The this, that, there rule. GIS World 7(7):22.
54. Hunter, G.J., and M.F. Goodchild. 1993. Managing uncertainty in
spatial databases: Putting theory into practice. In: Proceedings
of the annual meeting of the Urban and Regional Information
Systems Association, Atlanta, GA (July 25-29). Washington, DC:
Urban and Regional Information Systems Association, pp. 1-15.
30
-------
You Can't Do That With These Data! Or: Uses and Abuses of
Tap Water Monitoring Analyses
Michael R. Schock
Drinking Water Research Division, U.S. Environmental Protection Agency, Cincinnati, Ohio
Jonathan A. Clement
Black & Veatch, Cambridge, Massachusetts
Introduction
Linkage between human health and drinking water qual-
ity has been an area of interest in the United States for
many years. Over the past approximately 10 years,
drinking water monitoring requirements have expanded
rapidly under the Safe Drinking Water Act (SDWA).
Growing public and governmental interest in this envi-
ronmental area makes the aggregation and consolida-
tion of data on the occurrence and distribution of many
organic and inorganic contaminants and background
constituents of drinking water an important process.
These data can then be made available for systemiza-
tion and visualization to regulators, municipalities, water
utilities, public interest groups, health researchers, con-
sulting engineers, and water treatment scientists.
Given a sufficient number of data points and a conven-
ient computerized database/mapping platform, a wide
variety of maps can be generated to use in research and
decision-making processes. The validity of doing so,
however, rests inseparably upon the basis of the sam-
pling plan and protocols, as well as the precision and
accuracy of the analytical methods used forthe constitu-
ents of interest. The well-known problem of matching the
proper scale of the source data to that employed in the
maps for interpretation is a critical problem with drinking
water sampling, where many unappreciated small-scale
variations render many, if not most, attempts to make
generalizations inaccurate or meaningless.
This paper introduces and describes many concepts re-
lated to what generates or controls the concentrations
of metals and other constituents in drinking water, ways
in which the sampling protocol affects apparent levels of
constituents, and the magnitude of temporal and spatial
variability present in both municipal and private water
supplies. Illustrations from water quality studies show in
practical terms how generalizations must be kept to a
minimum and how the data input into a geographic
information system (CIS) for interpretation and evalu-
ation must be carefully analyzed and screened to deter-
mine the appropriateness for various well-intended
purposes. The discussion and examples show how
many apparently significant trends and assessments of
exposures or occurrences turn out to be merely artifacts
of critical (yet subtle) inconsistencies or errors in the
planning and execution of the sample collection proc-
ess, or inconsistencies caused by the fact that regula-
tory (and not research) requirements govern the origin
of the data.
The concepts this paper covers are equally valid in
many other disciplines using or contemplating the use
of CIS for interpretation of all kinds of "field" data.
Why Maps Are Useful for Drinking
Water Studies
Maps and CIS databases could have wide applicability
to drinking water studies. For example, they could pro-
vide the basis for investigating the occurrence of regu-
latory contaminants or related constituents, either to
estimate the costs of compliance with a regulation or to
estimate human health effects. Mapping could be useful
to utilities and consultants investigating process changes
for a utility or determining the effectiveness of some
existing treatment such as corrosion control orchlorina-
tion. Use of CIS could also assist in assessing the
feasibility and impact of system expansion. Another
promising application would be CIS assistance in devel-
oping and implementing wellhead protection plans.
Many other areas of application may be possible now,
or will be discovered in the future, as CIS technology
and regulatory requirements continue to develop.
31
-------
Sampling Protocols for Data Usable in GIS
Several SDWA regulations have resulted or will result in
the collection of geographically diverse drinking water
quality data that may interest mappers. The Lead and
Copper Rule, the Surface Water Treatment Rule, the
proposed Information Collection Rule, and the Disinfec-
tion/Disinfection Byproduct Rule are but four examples.
Many states have their own variations on federal drink-
ing water regulations, so their data collection require-
ments may differ somewhat. Considerable data may
also be collected for specific research studies of either
academic or purely practical nature.
Chemical Factors in Constituent Behavior
For the purposes of this discussion, chemical constitu-
ents in drinking water may be classified as being gener-
ally reactive or nonreactive. Reactive constituents may
change concentrations or chemical form for a variety of
reasons, such as:
A result of interaction with the background composi-
tion of the drinking water.
By precipitation or dissolution reactions with pipe ma-
terial used for the distribution system.
By chemical reactions with disinfectants added at
water treatment plants.
By slow chemical reactions started at water treatment
plants.
Nonreactive constituents may play an important role by
providing a chemical background that indirectly influ-
ences the speed or extent of other chemical reactions
and transformations. Table 1 gives a summary of many
common constituents of drinking water and identifies
whether they function essentially as reactive or nonre-
active constituents.
Reactive Constituents
Clearly, chemical species or compounds that can
change in concentration or transform into other species
or compounds during distribution make mapping on very
large scales difficult to justify. Reactive constituents may
also change concentration in the same place overtime,
such as water standing overnight in a home, school, or
building, which is discussed in a later section. Some
examples of reactions during water distribution follow:
During lime softening processes at some central
water treatment plants, a supersaturated state is
used for the compound calcium carbonate to remove
calcium (and sometimes magnesium) ions from the
water. This condition is sometimes maintained into
the distribution system as well to assist in maintaining
chemical conditions useful for corrosion control of
lead and copper. Thus, calcium levels, pH, and car-
Table 1. General Reactivity Trends for Common Drinking
Water Constituents
Constituent
General Reactivity Tendency
PH
Dissolved oxygen
Calcium
Magnesium
Total carbonate
Total alkalinity
Chlorine residual
Temperature
Iron
Copper
Lead
Zinc
Silica
Sulfate
Orthophosphate
Polyphosphate
Total phosphate
Nitrate
Chloride
Fluoride
Trihalomethanes
Haloacetic acids
Highly reactive
Reactive
Nonreactive (reactive when cementitious
pipe linings are present)
Nonreactive
Nonreactive
Reactive, particularly with pH changes
Reactive
Either
Reactive
Reactive
Reactive
Reactive
Nonreactive
Nonreactive
Reactive
Reactive
Reactive
Nonreactive
Nonreactive
Nonreactive
Reactive
Reactive
bonate concentrations (and consequently, alkalinity)
drop as water passes away from the plant (1, 2).
The metals in pipe materials, such as iron, copper,
zinc (in galvanized pipe), and lead, are oxidized by
oxygen, free chlorine, chloramines, ozone, and other
disinfectants, which renders them into a form that
water can transport, unless other chemical conditions
are such that a highly insoluble scale deposits on the
pipe, immobilizing the metal (1, 3).
Prolonged contact with chlorine disinfectant species
converts a fraction of natural organic matter present
in many distributed waters into regulated "disinfection
byproduct" compounds, such as trihalomethanes,
chloroform, and haloacetic acids (4, 5).
Following the addition of chlorine or after increasing
pH to enable some corrosion control for copper and
lead, iron present in well waters in dissolved ferrous
(Fe2+) form oxidizes into Fe3+ form, which is much
less soluble. Obnoxious "red water" results, as ferric
oxyhydroxide precipitate forms and clouds the water.
Polyphosphate chemicals added to "sequester" iron
or manganese in well waters break down into simpler
32
-------
polyphosphate forms of shorter chain lengths, plus
orthophosphate. The orthophosphate frequently be-
comes present at high enough concentration to aid
in controlling lead or copper (1, 6-8).
Water passes through newly installed cement mor-
tar-lined pipes, or aggressive water passes through
older asbestos-cement pipes. Because of the particu-
lar chemical nature of the water, calcium carbonate
and calcium hydroxide in the cement dissolve, raising
the pH and hardness of the water (1).
Free chlorine is added to disinfect water and is such
a strong oxidant that it is unstable in water at normal
concentrations. Additionally, it reacts with miles of
unlined cast iron pipe, accelerating the decomposi-
tion of hypochlorous acid or hypochlorite ion to chlo-
ride. Consequently, the overall redox potential of the
water supply and the effectiveness of disinfection
decrease.
A concentration of 1 milligram per liter (as PO4) phos-
phoric acid is added to a distributed water at pH 7.5
to control lead corrosion. The orthophosphate reacts
with exposed iron in the distribution main, however,
and the residual concentration of orthophosphate
decreases throughout distribution passage to the
point where the level is no longer adequate to create
the lead orthophosphate passivating film needed (1,
6, 8, 9).
Unless a constituent is known to be nonreactive, maps
may be falsely generated under the premise that the
concentration of a constituent is essentially a constant
over some geographic area. Following the changes in
concentration or chemical form of reactive constituents
would also seem to be a useful application of CIS tech-
nology. One major restriction applies to the viability of
that approach, however. Presuming that the analytical
techniques used can adequately quantify the concentra-
tion and concentration changes observed, the scale of
the variability or concentration change relative to the
scale of the mapping perspective becomes critical to
accurate mapping. A latersection ofthis paper considers
this critical factor in more specific detail.
Nonreactive Constituents
Almost no inorganic constituents in natural or drinking
water are purely chemically inert. Under some condi-
tions, and at some concentrations, significant reactions
can occur. Some constituents that are actually reactive
may act as if they are nonreactive constituents, how-
ever, because they are present in high enough concen-
trations relative to the extent of chemical reactions
taking place that no discernible change in their concen-
tration results. An obvious example is the dissolved
inorganic carbonate (DIG = hteCOs* + HCOs- + COs2-)
concentration (1, 10). Complexation and formation of
passivating basic carbonate solid films of lead and cop-
per by carbonate and bicarbonate ion dominate the
corrosion control chemistry of copper(ll) and lead(ll) (1,
11). The concentration of DIG in water on either a molar
or weight basis, however, is normally a factor of 500 to
10,000 higher than the lead or copper concentrations.
Hence, changes in the DIG content from these reactions
normally are analytically undetectable.
Another example is fluoride ion, which is often used as
a distribution system water flow "tracer" because of its
relative inertness. Actually, fluoride ion can form strong
complexes with aluminum left in water following coagu-
lation treatment with alum. The solubility of fluoride-
containing solids with other major drinking water
components (such as calcium and sodium) is very high,
however, and fluoride reacts only weakly with metallic
plumbing materials in the distribution system. Therefore,
total fluoride concentrations tend to remain constant.
Relatively accurate maps of the occurrence and distri-
bution of nonreactive constituents can be made, but
their usefulness depends on the scale of the mapping
relative to their occurrence and the particular question
under investigation. All of this supports the need to
ensure that the question asked can be answered cor-
rectly at the map scale.
Scale of Drinking Water Constituent
Sources
More than 59,000 public water suppliers exist in the
United States (12). Of these, approximately 660 are
considered large water systems, which serve over
50,000 in population. These municipal systems use
source water supplies that can be ground-water wells,
"surface" waters (i.e., rivers, reservoirs, lakes), or a
combination of both. Some water suppliers perform
minimal water treatment of their own and purchase
water from another water system or systems to satisfy
their needs.
Surface Water Sources
Many water utilities use a single water treatment plant
to treat surface waters, which could satisfy the entire
water demand of the community all year. In many cases,
however, utilities combine several surface water
sources and use a different treatment plant to treat each
water source. The water plants usually discharge into
the distribution system at different points, and system
hydraulics dictate the areas of the system in which
waters mix. This is important because the water quality
characteristics, which often differ among treatment
plants, influence the corrosivity of the waters to various
plumbing materials in the distribution system. Different
water constituents also may affect the disinfection effec-
tiveness of the treatment and the formation of unwanted
disinfection byproducts.
33
-------
For surface water systems, the chemical composition of
the water depends on the upstream or watershed geo-
chemistry, the seasonal nature of the water body used
as the source, and the characteristics that the treatment
imparts, such as coagulation with ferric sulfate or alum
(aluminum sulfate), lime softening, filtration, pH adjust-
ment, corrosion control treatment, chlorination, etc.
The scale of the source water chemical data, therefore,
is large, driven by the geology, soil nature, land use, and
climate. The chemical nature of the treated water, how-
ever, may differ significantly from that of its source.
Ground-Water Supplies
Many water utilities use multiple ground-water wells. A
water supply of medium to large size usually uses mul-
tiple wells, instead of or in addition to the surface water
supplies. Wells number from only two or three to more
than 100 for very large water systems. Wells normally
operate in different patterns, and only rarely do all wells
operate at the same time. The yield of the wells and their
water quality dictates the combination and number of
wells used at a particular moment. The wells may or may
not be from the same aquifer, and even if they are, local
inhomogeneities frequently exist in water composition
(especially with iron and manganese) that limit the use-
fulness of certain wells without substantial treatment.
Historically, utilities have treated some (but not neces-
sarily all) wells with a chemical such as a polyphos-
phate or sodium silicate to sequester the iron and
manganese from wells. Some utilities install physical
removal processes such as ion-exchange softeners, re-
verse-osmosis plants, aeration systems for iron re-
moval, air stripping towers for volatile organic compound
or radon removal, or "greensand" filters for the removal
of iron and manganese. These facilities sometimes exist
at only certain well sites or at some point where water
from multiple wells is combined.
The scale of chemical controls on ground-water sup-
plies, therefore, becomes only hundreds of feet. Con-
taminants of raw waters, such as arsenic, nitrate, or
chromium, are geologically and geochemically control-
led. Therefore, their occurrence is geographically vari-
able on even a small scale, and the variability exists
vertically in the subsurface as well as horizontally. A
municipality may use wells of different depths into differ-
ent aquifers, or even approximately the same depth
spread out over hundreds of feet to many miles in the
same aquifer or a variety of geologic units.
The variability of individual ground-water wells overtime
(such as seasonally) is usually less apparent than with
surface water sources, but the fact that many wells are
frequently used in different combinations and for differ-
ent lengths of time (hours to days, usually) makes char-
acterizing "influent" water quality complicated. The
same observation applies to water systems that allow
different amounts of water to bypass treatment proc-
esses (e.g., ion-exchange, reverse osmosis) depending
on the levels of targeted undesirable contaminants (e.g.,
nitrate, sulfate, arsenic).
These characteristics of the nature of chemical compo-
sition, use, and treatment of ground-water supplies
clearly show that generalizations over areas such as
states or geographic regions (e.g., New England, Upper
Midwest) are at least very gross and uncertain and at
worst, entirely misleading when decisions are to be
made about risk and health assessments, or estimates
of the necessity for certain treatments or economic im-
pacts of different potential drinking water regulations.
Combination Systems
Some municipalities combine the use of surface water
supplies and ground-water wells. Therefore, general
water chemical characteristics vary throughout the sys-
tem in a regular manner in response to the location and
use of different sources, as well as relative amounts of
water that the different sources produce and deliver.
Distribution System Mains
The next lower level of scale is the distribution system
network of pipes and storage. Common materials used
for distribution system piping include cast iron, ductile
iron, cement mortar-lined iron, iron with organic coat-
ings, asbestos-cement (A-C), and various forms of plas-
tic. Pipe diameters range from about 4 inches to many
feet, depending on size of the water utility and commu-
nity, size of the neighborhood fed by the line, and dis-
tance of travel for the water. Here, because of the large
volume of water involved relative to the pipe diameter,
the major chemical interactions involve such constitu-
ents as hardness (calcium and magnesium) ions, pH,
iron, bicarbonate and carbonate ions, and chlorine re-
sidual species, and possibly microbiological parameters
such as total plate counts, heterotrophic plate counts,
and assimilable organic carbon. Disinfection byproducts
(DBFs) may change in concentration and type because
of the time involved in the water traveling through the
piping from the treatment plant. Trace metal contamina-
tion, such as lead and copper, is usually negligible from
this source, unless it is present when distributed from
the wells or water treatment plants.
Depending on prevalent economics and construction
practices during periods of water system growth, the
materials will not be either randomly or uniformly distrib-
uted geographically within system boundaries. Water
flow often varies greatly within the distribution system,
and water lines sometimes terminate in dead-end areas
with minimal flow rates. Water quality often differs sub-
stantially in these dead ends from that in the fully flowing
parts of the distribution system.
34
-------
Household Service Lines
Service lines represent the connection between the
house or building and the distribution main. Sometimes,
the service lines are joined to the mains by a flexible,
approximately 2- to 3-foot long pipe called a "goose-
neck" or "pig-tail." Historically, this connector was often
made of lead. Recently, copper has been the most
widely used material, with plastic gaining in acceptance.
Service lines for homes are usually 0.75 to 1 inch in
diameter, with service lines for many commercial build-
ings or multifamily dwellings ranging in size from 1.5 to
3 inches in diameter. Service line for homes and build-
ings have usually been made of lead, brass, copper,
galvanized steel, or plastic. The material used depends
on the age of the water connection and the construction
practices of the area involved. A recent report estimated
that approximately 6.4 million lead connections ("goose-
necks") still exist in the United States, and about 3.3
million lead service lines still exist (13). In many commu-
nities, old lead service lines remain a major source of
lead in drinking water.
Like distribution system materials, service line materials
may vary greatly within a distribution system by space
and time. For instance, in large eastern cities, very old
neighborhoods may have many (or even mostly) lead
service lines. New neighborhoods likely have copper or
plastic service lines. Galvanized steel or copper pipes
may have been installed between the era when lead was
used and modern times. With the exception of Chicago,
where lead service lines were occasionally installed into
the 1980s, the use of lead for service lines generally
stopped in the late 1940s or early 1950s. An example of
nonuniform distribution of service line materials is
shown by Figure 1, a map indicating Cincinnati sampling
sites for Lead and Copper Rule (14-17) monitoring.
Erratic clustering of different service line materials is
evident.
Rehabilitation of old houses or replacement of failed
piping results in a mixture of new and old material in
areas where houses are predominantly old. Following
completion of the construction, maps of service line
material would show many clusters representing preva-
lent plumbing codes and economics.
Interior Plumbing
Interior plumbing of buildings and houses reflects even
more variability than service lines. This is the dominant
contributor to lead and copper levels at most sites cov-
ered under the Lead and Copper Rule (14-17). Interior
plumbing consists of piping, plus a large number of
valves, connectors, fixtures, and perhaps soldered joints
and a water meter. Any or all of these components are
replaced at varying intervals as a result of failures or
remodeling. Therefore, even generalizations within a
small neighborhood are risky, unless the neighborhood
is very new and uniformly constructed. When attempting
to survey the composition of plumbing materials that
might be the source of drinking water contamination,
merely asking for the age of the house or building is
insufficient. Questions must be asked to obtain the nec-
essary precise information on the age and type of
plumbing materials and components in the building.
Typical interior plumbing materials include lead, galva-
nized steel, copper, and different plastics for pipes.
Some brass and black steel have been used for short
times in some areas. Faucets are almost always made
with either brass or plastic internal parts, which differ in
composition from the exteriors, which are usually plated
with chrome or other metal. Interior faucet volumes typi-
cally range from about 30 milliliters to 120 milliliters,
depending upon design. Valves and meters are also
frequently made of brass or bronze, which are cop-
per/zinc alloys usually containing 2 percent to 6 percent
lead. Until recently, solders used to join copper drinking
water pipe sections were usually a tin and lead combi-
nation, containing 40 percent to 60 percent lead. Occa-
sionally, connector lines to fixtures include copper,
stainless steel, aluminum, or flexible plastic sections.
Private Water Systems
The many possible designs of domestic water systems
originating from wells or cisterns are too numerous to
illustrate. Figure 2 gives an example of one such system
layout. Private systems share many features with do-
mestic systems supplied by water utilities, however.
Interior plumbing shares most of the same configura-
tions and materials. For private water systems, addi-
tional plumbing that could cause contamination or water
chemistry changes includes well casing material, sub-
mersible pump casing and fittings, pressure tank feed
and control plumbing, and nonsubmersible pump inte-
rior materials. Therefore, problems with determining the
frequency and distribution of levels of potential contami-
nants include those present for domestic situations in
general, plus those complications arising from cycling of
the pumps, pressure tank system, or both.
Water Samples Representing Distances
One of the most important fundamentals of under-
standing drinking water sampling is that volumes of
water (e.g., 1-liter samples, 250-milliliter samples) rep-
resent the linear distance of plumbing material in contact
with the water sampled. Because of water mixing and
flow during use or sampling, they are also integrated
samples of that volume. This understanding is at the
heart of designing accurate water sampling programs
and making viable interpretations of existing monitoring
data that may be contained in (or mappable by) a CIS.
Table 2 summarizes some interesting and important
relationships between pipes of different inside diameters
35
-------
-4 ii
S_ wL -
-Tr ' i£s .,
""*j '
-V . : - .' Jf%£. '- .
:- : ..?";.
>--.:-" '
-
. Lead Site
» Copper Site
W.Q.P. Site
C.W.W. Facility
* Bolton D.S.
*** California D.S.
Not C.W.W. D.S.
Service Limits
California Water
Treatment Complex
Figure 1. Cincinnati Water Works Lead and Copper Rule compliance monitoring, July to December 1992.
Figure 2. Distribution system.
(IDs) and the volumes of water they contain per unit of
length (7). Much domestic interior plumbing has an ID of
approximately 0.5 inches, depending upon the material.
Figure 3 shows schematically what parts of a plumbing
system would likely be represented by samples of differ-
ent volumes taken after water was allowed to stand in
the pipe for many hours. Faucets, bubblers, and other
terminating fixtures vary widely in volume. Kitchen-type
fixtures usually contain from 60 to 120 milliliters of water.
Bathroom-type fixtures may contain only about 30 to 60
milliliters of water. Bubblers, such as those frequently
found on school or office drinking fountains, are smaller
still. As can be seen schematically in Figure 3a, a small
volume such as 125 milliliters captures the faucet and a
short distance of pipe immediately leading to it. In many
plumbing systems, this volume catches water in contact
with numerous soldered joints. On the other hand, if a
single 1-liter first-draw sample is taken, the water in the
bottle represents a much longer distance back into the
plumbing system. In a situation where the source of lead
in drinking water is a new brass faucet, or soldered joints
of lead-tin solder, this larger volume usually gives a
lower lead concentration than the smaller volume be-
cause more water in the sample is not in intimate contact
with materials containing lead.
Other sampling schemes logically follow. For instance,
if examining copper pipe corrosion, discarding the first
36
-------
Table 2. Interrelationships Among Pipe Length, ID, and
Internal Volume for Selected Common Plumbing
Materials and Pipe Sizes
Material
Copper
tubing
Copper
tubing
Copper pipe
Galvanized
steel pipe
Lead pipe
or tube
Lead pipe
or tube
PVCor
CPVC pipe
Length
for 1,000
Identification/ True True Milliliters
Type ID OD (Feet)
0.5-inch, type 0.545 0.625 22
L, annealed
0.5-inch, type 0.545 0.625 22
L, drawn
0.5-inch, 0.622 0.840 17
schedule 40
0.5-inch, 0.616 0.840 17
schedule 40
0.5-inch ID, 0.50 1.00 26
0.25-inch wall
0.75-inch ID, 0.75 1.25 11.5
0.25-inch wall
0.5-inch, 0.546 0.840 22
schedule 80
Milliliters
per Foot
46
46
60
59
39
87
46
Plumbing Represented by Samples
a. 1,000 Milliliters = 22 Feet @ 0.5-Inch ID Cu Type L
Faucet
1,000-Milliliter Sample
Main
Service Line
125 or 250 milliliters of water is likely to give more
accurate information because it minimizes the effects of
the faucet material as well as piping that connects the
faucet to the interior line. Often, this connecting piping
is not copper. If examining the corrosivity of the water to
lead service lines, wasting a volume of water corre-
sponding to the distance from the outlet to the service
line better estimates the effect, although not without
uncertainty (18). Many other sampling schemes are
possible and useful, but users must be aware that the
sampling protocol may have as much or more influence
on the observed metal concentration than water quality
or other variables. Hence, incorporation of monitoring
data into a CIS database must be done only when the
source represents equivalent samples.
Because of turbulent mixing during flow, local high con-
centrations of lead (or other contaminant) may become
broadened and diluted by the time the water to be
sampled reaches the sample collection bottle (18). In
many cases, therefore, numerous small-volume se-
quential samples can be taken and used to profile a
plumbing system to locate brass valves, connectors,
soldered joints, etc. Figure 4 illustrates sequential sam-
pling results for one room of a building. Peaks in the
distribution of samples physically correspond to the lo-
cation of a chrome-plated brass faucet and to a later
concentration of fresh Sn:Pb soldered joints. Unfortu-
nately, even small-volume sequential tap water samples
must pass over other potentially contaminating or alter-
ing surfaces on the way through the sampling tap.
Kuch and Wagner have shown how water can dissolve
large amounts of lead simply by traveling through long
distances in lead pipes with small IDs (9, 19). Although
this study specifically examined lead, the principle ap-
plies to other metallic piping materials. This phenome-
non is inseparable from the aspect of time, which is the
next subject.
Faucet
b. 125 Milliliters = 2.75 Feet @ 0.5-Inch ID Cu Type L
Household
Plumbing
T
Soldered
Joints
125-Milliliter Sample
Main
Meter
1UF-
Service Line
Figure 3. Schematic diagram of plumbing materials repre-
sented by sample volumes of a) 1 liter and b) 125
milliliters.
300 -,
| 250 -
I 150-
o
1, 100 -
1 50-
Samples
1-2 60 Milliliters
3-12 125 Milliliters
1 2 3 4 5 6 7 8 9 10 11 12
Sample Sequence, #
Figure 4. Sequential sampling results from a room on the
ground floor of a building.
37
-------
Effect of Time
While some chemical reactions are instantaneous,
many dissolution and precipitation reaction steps that
are important in controlling metal levels in water take
many hours to many days to reach equilibrium. In fact,
passivation films and scales on pipes that inhibit corro-
sion and reduce leaching of trace metals may take
months to decades to develop substantially. Some other
chemical transformations, such as creation of triha-
lomethanes from chlorination of natural organic matter
during disinfection, or processes such as inactivation of
pathogens, may occur over hours (20, 21).
Many steps in an overall chemical reaction process
could be rate-limiting. Figure 5 shows how lead levels
increase in 0.5-inch ID pipe given two different assump-
tions. The top curve shows how lead increases and
levels off after about 8 to 12 hours (9, 19). This curve is
closely applicable to any metal, as long as the limiting
factor on the rate of metal migration into the water is the
radial diffusion of the soluble metal species away from
the pipe surface. The second curve shows, schemati-
cally, the effect of a diffusion barrier film (e.g., calcium
carbonate, adsorbed iron hydroxide mixed with organic
matter, aluminosilicate mineral deposits) or inhibition of
metal oxidation rates on lead migration into water after
different amounts of time.
In some water systems, significant chemical changes
can occur while the water is standing that can drastically
affect the oxidation or solubility of the plumbing material.
For example, dissolved oxygen and free chlorine react
quickly in new copper pipe or brass. If the water stands
0.01
Stagnation Level Controlled by
Metal Ion Diffusion
Stagnation in Presence of Diffusion
or Oxidation Barrier Film
10
20
30 40 50
Time, Hours
60
70
Figure 5. Comparison of lead concentrations that would be ob-
served after water stands different amounts of time
given different controlling chemistry factors.
sufficiently long, their concentrations may become neg-
ligible, which would significantly alter the redox condi-
tions governing metal solubility. In the absence of
oxygen or chlorine species, the dominant form of copper
in water and on plumbing material then becomes cop-
per(l) instead of copper(ll), resulting in different solubility
characteristics after consumption of the oxidant than at
initiation of the standing period (11).
Seemingly identical water samples collected from the
same taps in houses, schools, or other buildings yield
different metal concentrations, depending on the time
the water was in contact with the faucet, solder, or piping
material. Similarly, samples taken for disinfection bypro-
ducts after different chlorine contact times may produce
different concentrations and different speciation (e.g.,
trihalomethanes, haloacetic acids). This factor causes
considerable confusion in many investigations of con-
tamination of school or building drinking water taps and
water coolers and complicates estimating human expo-
sure for health-effects studies.
Interconnectedness of Distance and Time
In innumerable situations, the effects of distance and
time are impossible to separate. Some generalizations
and examples follow.
Dead ends and slow rate areas produce long residence
times for the distributed water. This results in long con-
tact times with pipe materials, so reactive constituents
can change considerably in concentration. The process
is totally interactive, in that concentration changes of
reactive constituents are in response to contact with the
pipe materials, and in turn, the materials respond to the
water composition. Water may take hours to days to
reach a particular home or building and may traverse
many miles of distribution system piping of the same or
differing composition. Water thoroughly run through a
household faucet for 5 or 10 minutes to purge the lines
is "fresh" from the resident's perspective, but may be
"old" from the distribution system perspective.
The profile of the water line shown in Figure 4 was made
on the basis of filling small volume (60 or 125 milliliters)
sample bottles, one after another, without wasting any
water. If the objective were to capture only the highest
risk of lead contamination from a lead service line after
some hours of stagnation, then the sampling process
would be different. Instead of collecting all water be-
tween the tap and the service line, the water can be run
until either a target volume is wasted (representing the
linear plumbing distance to the service line) or can be
run at a given rate until, after the appropriate length of
time passes, the sample bottle can "intercept" the slug
of water residing in the service line. Beware that differ-
ences may exist in the peak concentration of the con-
taminant and the "width" of the slug of elevated
38
-------
contaminant level, depending upon the rate of waterflow
before and during sampling (18, 22, 23).
Other Sources of Variability in Water
Samples
Variability in water samples can stem from many sources
aside from those discussed in this paper (18). The na-
ture of the errors and their likely magnitude may vary
with each episode of sampling and analysis and is far
beyond the present scope of discussion. A brief listing
to consider, however, when drawing conclusions from
"field" data includes:
Analytical imprecision or bias.
Flow rate of the water during sampling.
Temperature.
Particulate erosion from plumbing materials during
sampling.
Effect of container material.
Effect of air contact or other handling effects during
sample collection and shipment.
A Case Study of Easy Misinterpretation
Interpretations of water quality problems based on ag-
gregate monitoring data can be very misleading unless
analysis is performed at the appropriate scale. The situ-
ation of one utility described below provides a good
example of how using a CIS approach could have
helped solve the problem but also highlights how care-
fully data would need to be matched and consolidated
only at the proper scale if CIS were to be employed for
evaluating some kinds of water quality problems.
The utility at Hopkinton, Massachusetts, found very high
lead and copper levels exceeding the regulatory action
levels under the Lead and Copper Rule (14-17). The
90th percentile copper levels even exceeded 6 milli-
grams per liter, compared with an action level require-
ment of only 1.3 milligrams per liter. Some sites with lead
service lines are present in the system. Figure 6 shows
a schematic representation of the distribution system for
this utility. Five wells feed the system, and four (1, 2, 4,
and 5) are used regularly.
The background hardness, alkalinity, and carbonate
concentrations are fairly similar for all wells. The pH of
the ground water from the wells is usually slightly above
6. Chlorine solution is dosed for disinfection. High iron
levels are present in wells 1, 2, and 3, and high manga-
nese is also present in well 3. Generally, high dosages
of a polyphosphate chemical were added to wells 1 and
2 to respond to consumer complaints about the "red
water" that results from iron oxidation and precipitation.
Water from different wells mixes in the distribution sys-
tem, but the water tends to partition into two zones as
WelM
Well 4
Well 5
Figure 6. Schematic representation of the utility's distribution
system.
Figure 6 indicates. The lead and copper levels tended
to be distinctly lower in the section where the polyphos-
phate was dosed, marked as the "treated" part of the
system. The ongoing research study has employed ap-
proximately 22 monitoring sites.
From the information presented thus far, the system
clearly cannot be characterized by a discrete value for
lead or copper contamination, as well as the chemical
background of water throughout the system. Hence,
putting data at the "whole system" scale into a statewide
or countrywide data system would be tempting, but it
could be very misleading in solving the treatment prob-
lem. Having accurate spatially distributed data for back-
ground water qualities, monitoring site characteristics,
and metal levels at the subsystem scale, such as that
which could be integrated into CIS, would have been
extremely convenient, however. Yet, even more informa-
tion at a smaller scale is necessary to understand and
solve the whole treatment problem.
The utility initially observed that because the lowest lead
and copper levels coincided with the area of the system
fed by the polyphosphate chemical, that chemical likely
caused the corrosion inhibition. Median lead levels, for
example, were between about 200 and 300 milligrams
per liter in the "untreated" section, compared with about
10 to 15 milligrams per liter in the "treated" section.
Median copper levels were approximately 4 to 5 milli-
grams per liter in the untreated section, but only about
0.3 to 0.5 milligrams per liter in the treated section. The
utility and the researchers wondered whether the poly-
phosphate chemical should also be added to the other
wells. This is a matter of significant concern because
some studies indicate polyphosphate chemicals can en-
hance lead corrosion (1, 6) and the subject has rarely
been studied under statistically valid controlled conditions.
Additional site-by-site investigation, however, first re-
vealed that the sites with lead service lines all lay in the
untreated area of the distribution system. Because the
39
-------
research sampling program included two successive
1-liter samples, the additional contamination from the
service lines was confirmed by higher lead levels in the
second 1-liter sample than in the first in many cases.
Therefore, physical reasons, in addition to chemical
ones, explained the discrepancy in the lead levels. Fur-
ther, when considering only the treated system sites, the
lead levels were still high enough to be of concern.
Focusing on the copper sites resulted in the collection
of more important and interesting small-scale informa-
tion. Figure 7 shows the difference between average
copper levels in the two sections of the system. Almost
all sites in both parts of the system had copper interior
plumbing with 50:50 or 60:40 Sn:Pb soldered joints and
faucets with brass-containing internal materials. Though
the chemical added for iron control was ostensibly a
polyphosphate chemical, it also contained an initially
present fraction of orthophosphate and also tended
to partially break down to orthophosphate in the pres-
ence of iron and calcium (as most polyphosphates do).
Figure 8 shows the orthophosphate concentrations in
the two different parts of the system. While the levels of
orthophosphate present in the treated section would be
far too low to significantly inhibit lead leaching at the
background pH (1, 6, 7, 9, 22), the orthophosphate
plausibly may significantly inhibit copper dissolution, in
concordance with recent research projections (11).
Having determined through detailed small-scale sam-
pling and analysis that the chemistry affecting metal
levels in the system is generally consistent with modern
knowledge, a new treatment plan is being implemented
to control copper and lead levels through pH adjustment
in conjunction with iron control through a compatible
sodium silicate/oxidation treatment. Incorporation of
system and monitoring site physical characteristic data,
plus monitoring results, into CIS could have saved con-
siderable investigatory effort. The importance of this
case history, however, is that the data must be of the
appropriate scale and highly documented to be useful in
problem-solving. Failure to use data meeting these re-
Untreated Average
---- Treated Average
Oct.
Nov.
Dec.
Mar.
Apr.
0.6 -,
0.5-
0.4-
0.3-
0.2-
0.1 -
J
/
**
^
*»^
*<
/
t j_
Untreated Average
Treated Average
\
T/ > I
1 T .
!-!--
Sept. Oct.
Nov. Dec. Mar. Apr.
Figure 7. Average copper levels in treated and untreated sec-
tions of the system.
Figure 8. Average orthophosphate levels in treated and un-
treated sections of the system.
quirements, as well as overgeneralization to a large
mapping scale, can lead to ineffective if not damaging
water treatment choices that could adversely affect pub-
lic health.
Conclusions
The examples and discussion above lead to several
general conclusions about the use of CIS with drinking
water monitoring data:
Temporal and spatial variability stems from many
causes, down to a very small scale.
Sampling protocols must be keyed to the precise
questions under investigation.
Regulatory sampling, whose results are generally
readily available, is usually inappropriate to assess
human exposure to trace metals or other parameters
of interest (such as DBPs).
Generalizations on a large scale are often impossible
because of the geology and water chemistry variations.
Additionally, some considerations apply to the types of
mapping that could be employed by CIS. For example,
a mapping technique such as contouring may be espe-
cially inappropriate for use with drinking water data.
Major problems could result from:
Discrete, small-scale (such as within an individual
house) variability in distributions of certain contami-
nants, such as lead and copper.
Physical constraints of the distribution system network.
The small number of monitoring sites in relation to
the size of the distribution network.
Different chemical or hydraulic zones in the distribu-
tion system.
Employing CIS could be very useful in solving a variety
of drinking water problems. Users must be extremely
conscious of the nature of the source information, how-
ever, to avoid abusive extrapolations and generalizations.
40
-------
Acknowledgments
Some of the examples cited were investigated as part
of a U.S. Environmental Protection Agency cooperative
agreement (Darren A. Lytle, Project Officer) with the
New England Waterworks Association. Jack DeMarco
of the Cincinnati Water Works kindly provided Figure 1
on distribution of sampling sites for lead and copper
monitoring.
References
1. Schock, M.R. 1990. Internal corrosion and deposition control. In:
Water quality and treatment: A handbook of community water
supplies, 4th ed. New York, NY: McGraw Hill.
2. Benefield, L.D., and J.S. Morgan. 1990. Chemical precipitation.
In: Water quality and treatment: A handbook of community water
supplies, 4th ed. New York, NY: McGraw Hill.
3. Glaze, W.H. 1990. Chemical oxidation. In: Water quality and treat-
ment: A handbook of community water supplies, 4th ed. New
York, NY: McGraw Hill.
4. Cotruvo, J.A., and C.D. Vogt. 1990. Rationale for water quality
standards and goals. In: Water quality and treatment: A handbook
of community water supplies, 4th ed. New York, NY: McGraw Hill.
5. Tate, C.H., and K.F. Arnold. 1990. Health and aesthetic aspects
of water quality. In: Water quality and treatment: A handbook of
community water supplies, 4th ed. New York, NY: McGraw Hill.
6. Schock, M.R. 1989. Understanding corrosion control strategies
for lead. JAWWA 81:7:88.
7. AWWARF. 1990. Lead control strategies. Denver, CO: American
Water Works Association Research Foundation and American
Water Works Association.
8. AWWARF. 1994. Corrosion control in water distribution systems,
2nd ed. Denver, CO: American Water Works Association Re-
search Foundation/Engler Bunte Institute.
9. Schock, M.R., and I. Wagner. 1985. The corrosion and solubility
of lead in drinking water. In: Internal corrosion of water distribution
systems. Denver, CO: American Water Works Association Re-
search Foundation/DVGW Forschungsstelle.
10. Pankow, J.F. 1991. Aquatic chemistry concepts. Chelsea, Ml:
Lewis Publishers, Inc.
11. U.S. EPA. 1995. Effect of pH, DIG, orthophosphate and sulfate
on drinking water cuprosolvency. Cincinnati, OH: Office of Re-
search and Development. In press.
12. AWWA. 1993. Initial monitoring experiences of large water utilities
under U.S. EPA's Lead and Copper Rule. Denver, CO: Water
Industry Technical Action Fund/American Water Works
Association.
13. AWWA. 1990. Lead service line replacement: Benefit to cost
analysis. Denver, CO: Water Industry Technical Action
Fund/American Water Works Association.
14. AWWA. 1991. Lead and copper: A working explanation of the
Lead and Copper Rule. Denver, CO: American Water Works
Association.
15. Federal Register. 1991. Lead and copper: Final rule correction.
Fed. Reg. 56:135:32,112 (July 15).
16. Federal Register. 1991. Lead and copper: Final rule. Fed. Reg.
56:110:26,460 (June 7).
17. Federal Register. 1992. Lead and copper: Final rule correction.
Fed. Reg. 57:125:28,785 (June 29).
18. Schock, M.R. 1990. Causes of temporal variability of lead in
domestic plumbing systems. Environ. Monit. Assess. 15:59.
19. Kuch, A., and I. Wagner. 1983. Mass transfer model to describe
lead concentrations in drinking water. Water Res. 17:10:1,303.
20. Ireland, J.C. 1993. Alternatives to chlorine for disinfection of drink-
ing water. In: Strategies and technologies for meeting SDWA
requirements. Lancaster, PA: Technomic.
21. Miltner, R.J. 1993. Pilot scale treatment for control of disinfection
byproducts. In: Strategies and technologies for meeting SDWA
requirements. Lancaster, PA: Technomic.
22. Sheiham, I., and P.J. Jackson. 1981. Scientific basis for control
of lead in drinking water by water treatment. J. Inst. Water Engrs.
and Scientists 35:6:491.
23. Heumann, D.W. 1989. Los Angeles Department of Water and
Power: Solid lead gooseneck slug dispersion in consumer plumb-
ing system. Proceedings of the AWWA Water Quality Technology
Conference, Philadelphia, PA.
41
-------
Ground-Water Applications
-------
Using GIS/GPS in the Design and Operation of Minnesota's Ground Water
Monitoring and Assessment Program
Tom Clark, Yuan-Ming Hsu, Jennifer Schlotthauer, and Don Jakes
Minnesota Pollution Control Agency, St. Paul, Minnesota
Georgianna Myers
Water Management Consultants, Denver, Colorado
Abstract
Minnesota's Ground Water Monitoring and Assessment
Program (GWMAP) is administered by the Minnesota
Pollution Control Agency (MPCA) to evaluate baseline
ground-water quality conditions regionally and state-
wide. The program uses a systematic sampling design
to maintain uniform geographic distribution of randomly
selected monitoring stations (wells) for ground-water
sampling and data analysis. In 1993, geographic infor-
mation system (CIS) and global positioning system
(GPS) technologies were integrated into GWMAP, auto-
mating the selection of wells and the field determination
of well locations.
GWMAP consists of three components: the statewide
baseline network, regional monitoring cooperatives,
and a trends analysis component. In the statewide
baseline network, Minnesota is divided into over 700
121-square-mile grid cells, each with a centralized,
9-square-mile sampling region. Wthin each target area,
single-aquifer, cased and grouted wells are sampled for
about 125 metals, organic compounds, and major cat-
ions and anions. We are currently finishing the second
year of a 5-year program to establish the statewide grid.
When complete, the statewide baseline component will
consist of about 1,600 wells representing Minnesota's
14 major aquifers.
In 1993, approximately 4,000 well construction records
were selected for geologic and hydrologic review, using
a CIS overlay, from a database of 200,000 water well
records maintained in the state's County Well Index
(CWI). Using GPS, 364 wells were sampled and field
located. The semiautomatic well selection process uses
existing electronic coverage of public land survey (PLS)
data maintained in CWI in conjunction with the digitized
systematic sampling grid. CIS has greatly reduced the
time needed for selecting sampling stations. With the
combination of CIS and GPS, program costs have de-
creased, allowing more resources to be applied toward sam-
pling, while efficiency and quality of data have improved.
Introduction
Quantitative assessment of ground-water quality condi-
tions requires a highly organized data collection pro-
gram that includes statistical evaluation of monitoring
results (1, 2). States have difficulty providing the staff
and financial resources necessary to generate state-
wide quantitative ground-water information. Wth the
use of geographic information system (CIS) and global
positioning system (GPS) technologies, however, states
have the potential to improve the quality of environ-
mental monitoring programs and to reduce the amount
of staff time needed to collect and evaluate data, thus
decreasing costs. The degree to which states realize
these potential benefits depends largely on how effec-
tively the technology can be incorporated into the design
of the monitoring program. This paper describes how
CIS and GPS technologies are being integrated into the
design and operation of Minnesota's Ground Water
Monitoring and Assessment Program (GWMAP) to im-
prove overall effectiveness.
The Minnesota Pollution Control Agency (MPCA) has
sampled and analyzed ambient ground-water quality in
the state's 14 principal aquifers since 1978. In 1990, the
MPCA began a redesign of its ground-water monitoring
program to better assess water quality conditions state-
wide (3). Three program components resulted from the
redesign: a statewide baseline network for complete
geographic coverage, a trends analysis component for
intensive studies of how ground-water quality in specific
areas changes with time, and a regional monitoring
cooperative link to governmental units such as counties
45
-------
to meet specific local ground-water assessment needs.
This paper describes the design and operation of the
statewide baseline network.
The design of the statewide network is geographically
and statistically based to automate well selection and
data interpretation. In 1993, the MPCA began integrat-
ing CIS and GPS technologies into this part of the
program. The implementation of CIS and GPS sur-
passed our expectations by reducing staff time re-
quired to select wells and evaluate analytical results
(see Table 1). In addition, through the elimination of
previously uncontrollable variables, the use of CIS and
GPS has increased the accuracy of GWMAP data.
Monitoring Program Description
Since 1992, GWMAP has selected 150 to 250 existing
water supplies yearly for ground-water sampling and
analysis of about 125 parameters, including major cat-
ions and anions, metals, and volatile organic com-
pounds. Well selection is a fundamental element of
GWMAP that, if efficiently performed, supports the pro-
gram objectives by upholding the quality of the monitor-
ing data and minimizing the operating costs.
A key to the interpretation of monitoring data is the
technique used to select wells for sampling (2, 4, 5).
Minnesota has over 200,000 active water wells with
approximately 10,000 new installations annually. For
each well selected for GWMAP monitoring, a hydrologist
must individually review many well construction records.
An automated prescreening mechanism to facilitate well
selection can result in considerable time (and therefore
cost) savings. GWMAP chose CIS as the best tool for
this task. CIS enables the program to combine a sys-
tematic sampling technique with hydrogeologic criteria
to ensure an efficient and consistent selection process.
As Table 1 shows, CIS allowed us to more than triple
our geographic coverage and wells initially selected,
while dramatically reducing the records that must be
individually reviewed. We realized a time savings of 2
months compared with the time required before CIS
implementation.
In general, systematic sampling techniques use a ran-
domly generated uniform grid to determine sampling
locations in space and/or time (5). Systematic sampling
was initially implemented in GWMAP in 1991 using a
manually generated spatial grid defined by the public
land survey (PLS) (3). Although the PLS is not 100
Table 1. Well Selection in 1992 and 1993
percent geographically uniform, it was selected for the
grid to expedite well selection from existing digital data-
bases in which wells are organized by PLS location.
Systematic Sample Site Selection
Using GIS
Systematic sample site selection is a three-step process.
First, a database search of Minnesota's County Well Index
(CWI) (6), containing nearly 200,000 driller's records, is
conducted to include all available water wells in the
region of interest. Second, the candidate pool is reduced
to those wells located within regularly spaced grid cells.
Third, further wells are eliminated from the candidate
pool by applying geologic and well construction criteria
mandated in the GWMAP design (7).
Generating the Sampling Grid
The statewide sampling grid was generated from a randomly
selected origin (8). This grid consists of approximately 700
square cells, 11 miles on a side (see Figure 1). The
centroid of each cell is consecutively numbered and was
extracted to produce the origin of the sampling zone.
Figure 1. Statewide baseline network sampling grid.
Year
1992
1993
Area
Covered
9 counties
26 counties
PLS Sections
Selected
500
1,659
Well Logs
Selected
3,000
11,000
Well Logs
Reviewed
3,000
834
Wells
Sampled
158
206
Time
Spent
6 months
4 months
46
-------
Establishing the Sampling Zone
Each sampling zone consists of a 3- by 3-mile box from
which potential sampling sites are selected. It is gener-
ated by computing the coordinates of the four corners of
the box using the grid cell's centroid as the origin. To link
the sampling zone and grid cell, both are identified with
the same numerical code.
These sampling "target" zones, a series of regularly
spaced, 9-square-mile boxes, are then made into a CIS
coverage and overlaid on top of the PLS coverage to
extract those sections that are associated with each of
the sampling zones. Ideally, each sampling zone should
cover exactly nine PLS sections (3). Due to irregularities
in the PLS system, however, portions of 16 to 20 sec-
tions usually fall within the sampling zone of each cell
(see Figure 2).
Watonwan
Legend
/ \ / PLS Boundary
Sample Zone
Sample Grid
County Border
Figure 2. PLS and the sampling grid, Watonwan County.
Selection of PLS Sections
The PLS coverage was derived from the Minnesota
Land Management Information Center (LMIC) "GISMO"
file. It was originally created in 1979 by digitizing every
section corner in Minnesota from the U.S. Geological
Survey (USGS) 7.5-minute quadrangle map series.
The PLS section information is necessary in the well
selection process because the original well construction
logs, maintained by the Minnesota Geological Survey
(MGS), are organized by PLS. Although most of the well
selection process can be automated, manual file
searches for well records are still necessary and require
the PLS information.
Well Selection
After identifying the PLS sections within the sampling
grid, the statewide well database is imported as a point
coverage and overlaid with the selected PLS section
coverage. Thus, all wells that fall within the 16 to 20
sections are selected as potential candidates. The ac-
curacy of the well locations in CWI varies; most of the
point locations are approximated to four quarters
(2.5 acres). The CWI does not contain all well construc-
tion information, however, requiring that copies of
driller's logs be made for GWMAP files.
The final well selection is done after applying the
9-square-mile sampling zone over the potential pool of
candidates. For wells that fall within the zone, the well
construction records are pulled from MGS files, copied,
and submitted for hydrologist review. Depending on the
target cell location, the number of candidate wells requir-
ing review may range from a few to more than 100. For
newly installed water wells whose records have not yet
been digitized by LMIC, the PLS locations of the wells
are manually plotted onto a map to confirm whether they
fall into a sampling grid cell. Typically, from 5 percent to
as many as 20 percent of selected wells that meet the
location criteria are sampled. This accounts for the hydro-
geologic and well construction criteria and the coopera-
tion of well owners participating in the program.
Currently, interest in ground-water protection programs
runs high in rural Minnesota, with an acceptance rate of
up to 80 percent.
The implementation of CIS in well selection helped
GWMAP excel in two major areas. First, the develop-
ment of the statewide CIS grid eliminated previously
uncontrolled variables by removing the PLS spatial in-
consistencies from the systematic grid. Second, the CIS
reduced the manual workload with the automation of two
important steps in the well selection process: the gen-
eration of PLS section information to facilitate the data-
base search, and the identification of wells that meet the
geographic location criteria. The success of GWMAP
relies largely on the ability to use existing CIS cover-
ages. In using coverages created by other entities, this
program identified the need for a uniform standard for
data conversion and transfer.
Application of Global Positioning
Systems in Ground-Water Sampling
In 1991, the U.S. Environmental Protection Agency
(EPA) established a policy that all new data collected
after 1992 should meet an accuracy goal of 25 meters
or better (9). The purpose of EPA's Locational Data
Policy (LDP) is to establish principles for collecting and
47
-------
documenting consistently formatted locational data to
facilitate cross-programmatic, multimedia analyses. Ac-
curate geographic information is important to the spatial
analysis of well sampling results. Any uncertainty in
sample location can compromise hydrogeologic analy-
sis (10). GPS is an easy, cost-effective solution.
Global Positioning System Field Application
Beginning in October 1992, GWMAP employed GPS in
the field to assist in locating sample sites. Applying GPS
in the field has proven to be quite easy. The program
uses a multichannel C/Acode receiver with internal data
logging capability. Typically, the receiver is placed di-
rectly on top of the wellhead and logs 100 to 150 GPS
readings into the receiver's internal memory in approxi-
mately 5 minutes.
The GPS is also used for navigation in the field to locate
sampling sites. Because sampling sites are predeter-
mined, their locations can be extracted from a topo-
graphic map. The approximate coordinates can then be
loaded into a GPS receiver. In most cases, the receiver
successfully led the field team within visual range of the
sampling site.
Because of the inherent selective availability (SA) of the
GPS, raw field data must go through a differential cor-
rection process to achieve the goal of 25-meter accu-
racy (9, 11).
Data Management and Processing
Once the GPS receiver is brought back from the field,
data are downloaded to a personal computer (I486 proc-
essor at a speed of 50 MHz) and differentially corrected
(11). The average or mean of the 100 or more readings
collected onsite is calculated and reported as the site
location.
The MPCAdoes not operate a GPS base station for the
purpose of differential correction. The base station data
are obtained through a computer network (Internet) from
the Minnesota Department of Health (MDH) base station
located in Minneapolis.
To facilitate future data integration and document data
accuracy for secondary application, GWMAP proposed
quality assurance codes for GPS data collected by the
MPCA. The value of the accuracy proposed is a nominal
value rather than an absolute number (see Table 2).
Each of the seven processing methods is assigned a
separate code.
In the field experience of GWMAP, a nominal accuracy
of 2 to 5 meters has been consistently achieved after
the postdifferential correction and averaging have been
applied to the data. This technology is suitable for any
program that is designed to conduct either large-area or
intensive monitoring activities. It helps to cut costs by
Table 2. Proposed Nominal Accuracy Reference Table
Type of GPS
Receiver Used
Processing Method
Used To Correct Data
Nominal
Accuracy
(meters)
Navigational
quality C/A code
receiver
Navigational
quality with carrier
aid receiver
Survey quality
receiver (dual or
single frequency)
Postdifferential corrected 2-5
Real-time differential
corrected (RTCM) 2-5
Autonomous mode (no
correction) 15-100
Postdifferential corrected < 1
Real-time differential
corrected (RTCM) < 1
Autonomous mode (no 15-100
correction)
Postdifferential corrected <0.1
increasing efficiency and accuracy of the data. The data
collected by GWMAP can be used not only in a regional
study but could be used directly in a site-specific inves-
tigation as well.
GWMAP also found that GPS can be used most effi-
ciently by separating the two roles of field operator and
data manager. The field operators receive only the brief
instructions necessary to operate a GPS receiver before
going into the field. The data manager handles the data
processing details. The field operators can then concen-
trate their efforts on obtaining ground-water samples
and conducting the hydrogeologic investigation.
Conclusions
CIS and GPS technologies made it possible for the
MPCA to implement the statewide GWMAP project by
optimizing the available funding and staff time. CIS mini-
mized staff time spent on identifying sampling areas,
manipulating the sampling grid, and selecting monitor-
ing sites. In addition, CIS enabled GWMAP to integrate
a variety of databases and maps of different scales.
Using GPS to locate sampling sites enabled GWMAP to
efficiently obtain accurate geographic locational data
with relative ease. This eliminated the degree of uncer-
tainty that previously might have compromised the sta-
tistical evaluation of the hydrogeologic data.
GWMAP's success in integrating existing digital data to
automate the well selection process clearly demon-
strated the importance of the ability to share information
with others and the great need for a broadly applied
standard for data conversion and transfer.
48
-------
Acknowledgments
The authors wish to thank Renee Johnson of the Min-
nesota Department of Natural Resources for her work
to convert the PLS data layerto CIS coverage. Susanne
Maeder of LMIC supplied the statewide CWI coverage,
and Susan Schreifels of MPCA conducted research on
the LDP and made valuable suggestions on implement-
ing GPS.
References
1. Nelson, J.D., and R.C. Ward. 1981. Statistical considerations and
sampling techniques for ground-water quality monitoring. Ground
Water 19(6):617-625.
2. Ward, R.C. 1989. Water quality monitoringA systems approach
to design. Presented at the International Symposium on the De-
sign of Water Quality Information Systems, Colorado State Uni-
versity, Ft. Collins, CO.
3. Myers, G., S. Magdalene, D. Jakes, and E. Porcher. 1992. The
redesign of the ambient ground water monitoring program. St.
Paul, MN: Minnesota Pollution Control Agency.
4. Olea, R.A. 1984. Systematic sampling of spatial functions. Series
on Spatial Analysis, No. 7. Kansas Geological Survey, University
of Kansas, Lawrence, KS.
5. Gilbert, R.0.1987. Statistical methods for environmental pollution
monitoring. New York, NY: Van Nostrand Reinhold.
6. Wahl, I.E., and R.G. Tipping. 1991. Ground-water data manage-
mentThe county well index. Report to the Legislative Commis-
sion on Minnesota Resources. Minnesota Geological Survey,
University of Minnesota, St. Paul, MN.
7. Clark, T, Y Hsu, J. Schlotthauer, and D. Jakes. 1994. Ground-
water monitoring and assessment programAnnual report. St.
Paul, MN: Minnesota Pollution Control Agency.
8. ESRI. 1993. ARC/INFO version 6.1, ARCPLOT command refer-
ences. Redlands, CA: Environmental Systems Research Insti-
tute, Inc.
9. U.S. EPA. 1992. Global positioning systems technology and its
application in environmental programs. CIS Technical Memoran-
dum 3. EPA/600/R-92/036. Washington, DC.
10. Mitchell, J.E. 1993. A characterization of the influence of (x,y)
uncertainty on predicting the form of three-dimensional surfaces.
Proceedings of the AWRA Spring Symposium on Geographic
Information Systems and Water Resources, Mobile, AL. pp. 559-
567.
11. Trimble Navigation, Ltd. 1993. GPS Pathfinder System, general
reference. Sunnyvale, CA: Trimble Navigation, Ltd.
49
-------
Use ofGIS in Modeling Ground-Water Flow in the Memphis, Tennessee, Area
James Outlaw and Michael Clay Brown
University of Memphis, Memphis, Tennessee
Abstract
Memphis, Tennessee relies solely on ground water for
its municipal and industrial water supply. Memphis Light,
Gas, and Water (MLGW) Division owns and operates
over 160 water wells in 10 production fields throughout
Shelby County. MLGW produces an average of approxi-
mately 200 million gallons per day, excluding much of
the industrial demand. The city obtains its water from a
thick, prolific aquifer known as the Memphis Sand,
which was thought to be separated from a surficial
aquifer by a thick confining layer. In recent years, evi-
dence of leakage from the surficial aquifer to the Mem-
phis Sand has been found.
The University of Memphis Ground Water Institute
(GWI) is developing a hydrogeologic database of the
Memphis area to study the aquifer. The database serves
as the basis for several ground-water flow models that
have been created as well as part of the wellhead
protection programs currently being developed for Mem-
phis and other municipalities in Shelby County. A geo-
logic database was developed and is constantly being
updated from borehole geophysical logs made in the
area. Well locations are being field verified using a
global positioning system (GPS).
Use of the database has allowed the development of a
three-dimensional model of the Memphis area subsur-
face. The database also contains locations of and infor-
mation on both private and public production and
monitoring wells, Superfund sites, underground storage
tanks, city and county zoning, land use, and other per-
tinent information. Procedures for linking the database
to ground-water flow and solute transport models have
been developed. The data visualization capabilities and
the ability to link information to geographic features
make geographic information systems (CIS) an ideal
medium for solving ground-water problems.
An example of CIS use in ground-water flow modeling
is the study of the Justin J. Davis Wellfield. The water
quality parameters of alkalinity, hardness, sulfate, and
barium have significantly increased over the past 10
years at this facility. To understand why these changes
are occurring, MLGW, the GWI, and the U.S. Geological
Survey (USGS) participated in a joint investigation of the
wellfield.
In the spring of 1992, a series of 12 monitoring wells was
drilled into the surficial aquifer nearthe production wells.
Geophysical logging and split-spoon sampling revealed
an absence of the confining layer, referred to as a
window, at one of the monitoring wells. All other wells
penetrated various thicknesses of clay. This window in
the confining layer suggests that the water quality
changes could be due to leakage from the surficial
aquifer to the Memphis Sand.
The CIS database was used to construct a flow model
of the Davis area. Also, using the surface modeling
capabilities of CIS, the extent of the confining layer
window was estimated and used to calculate leakage
between the two aquifers. The results of these analyses
also indicate that further subsurface exploration is
needed to more accurately define the extent of the con-
fining layer window.
Introduction
Memphis, Tennessee, relies solely on ground water for
its municipal and industrial water supply. The Memphis
Light, Gas, and Water (MLGW) Division owns and op-
erates over 160 water wells in 10 production fields
throughout Shelby County, as shown in Figure 1. MLGW
produces an average of approximately 200 million gal-
lons per day, excluding much of the industrial demand.
The city obtains its water from a thick, prolific aquifer
known as the Memphis Sand, which was thought to be
separated from the surficial aquifer by a thick confining
layer. In recent years, evidence of leakage from the
surficial aquifer to the Memphis Sand has been found.
The University of Memphis Ground Water Institute
(GWI) is developing a hydrogeologic database of the
Memphis area to study the aquifer. Several ground-water
50
-------
Mississippi Alluvia] Plain
N
A
0 25 50
Scale in Thousands of Feet
Figure 1. Physiographic description and MLGW Wellfields in
Shelby County.
flow models have been developed using the database.
Also, the database is an integral part of wellhead pro-
tection programs being developed for Memphis and
other municipalities in Shelby County. A geologic data-
base was developed and is constantly being updated
from borehole geophysical logs made in the area. Well
locations are being field verified using a global position-
ing system (GPS).
Use of the database has allowed the development of a
three-dimensional model of the subsurface of the Mem-
phis area. The database also contains locations of and
information on both private and public production and
monitoring wells; Superfund sites; underground storage
tanks; city and county zoning; land use; and other per-
tinent information. Water quality measurements for
every MLGW production well have been obtained, and
a history of water quality for the Memphis Sand is being
developed. Procedures have been developed for linking
the database to ground-water flow and solute transport
models (1). The data visualization capabilities and the
ability to link information to geographic features make
geographic information systems (CIS) an ideal medium
for solving ground-water problems.
GIS Database
The GWI has developed and is continuing to update a
hydrogeologic database for the Memphis area.
ARC/INFO (marketed by Environmental Systems Re-
search Institute, Redlands, California) is the GIS pro-
gram that the GWI is using. The program runs on a
network of 10 SUN SPARC stations. The capabilities of
ARC/INFO and the computational speed of the SPARC
stations allow very sophisticated ground-water analyses
to be performed and have allowed the development of
an extensive electronic database.
The basic unit of data storage in ARC/INFO is a cover-
age. A coverage is a digital representation of a single
type of geographic feature (e.g., points may represent
wells, lines may represent streets or equipotential lines,
and polygons may represent political boundaries or zon-
ing classifications). Information may be associated with
an individual geographic feature in a feature attribute
table. This information may then be queried and used in
analyses. ARC/INFO also has its own macro language
that allows the customization and automation of many
ARC/INFO procedures.
A relatively new feature of ARC/INFO is address match-
ing. This procedure compares a file containing the street
address of a particular feature with an address cover-
age. This coverage is basically a library of addresses
that are linked to a geographic coordinate. As the ad-
dresses from the input file are compared with the ad-
dress coverage, the matching points are written to a
second coverage. Any addresses in the input file that do
not match an address in the address coverage are
written to a "rejects" file. These can be matched by hand
on a one-by-one basis.
Address matching serves as an alternative to digitizing,
as long as a good address coverage for a specific area
exists. The GWI has used this capability extensively and
has developed a coverage of underground storage tank
(UST) locations inside Shelby County. A database of
private and monitoring wells is also being developed
and updated using address matching. The raw informa-
tion was obtained in an ASCII format from the appropri-
ate regulating agencies (i.e., the Tennessee Department
of Environment and Conservation Division of Under-
ground Storage Tanks and the Memphis/Shelby County
Health Department). The ASCII information was im-
ported into ARC/INFO and address matched. The crea-
tion of a suitable address coverage and completion of
the address matching of the UST file has taken almost
2 years. The private well coverage is currently being
updated from historical information provided by regulat-
ing agencies and local well drilling companies.
An important part of the database is the geologic infor-
mation obtained from geophysical logs in the area.
Gamma logs, resistivity logs, and spontaneous potential
(SP) logs are three major types of electric geophysical
logs. Gamma logs measure naturally occurring radiation
emitted from soil in the borehole. Clays and shales emit
gamma rays. A high gamma count indicates the pres-
ence of clay or shale, and a low gamma count implies
that little or no clay is present. Sand layers that contain
fresh water are located using resistivity logs. Maximum
values of resistivity indicate the possibility of a sand
51
-------
layer. Clays and sands that contain salt water may
exhibit similar resistivities. SP logs are used to differen-
tiate between the two (2).
Corroborating data such as formation logs, geologic
studies, and available material samples should be con-
sulted when reading and interpreting geophysical logs.
The accuracy and reliability of an application based on
well logs is completely dependent on a realistic interpre-
tation of the geophysical data. A sample interpretation
of a set of geophysical logs is shown in Figure 2. The
results of interpretations like this are entered into the
point attribute file of a well coverage.
The Triangulated Irregular Network (TIN) module of
ARC/INFO is used to create a three-dimensional sur-
face from information stored in a coverage. TIN creates
a surface from a set of nonoverlapping triangles defined
by a set of irregularly or regularly spaced points. In this
study, the points defining the triangular TIN surfaces are
the locations of the wells in the model area. TIN uses
various interpolation routines to estimate surface val-
ues. Once the surfaces have been developed, two-di-
mensional profiles can be made that show the relative
thicknesses of the various soil strata, as shown in Figure
3. These profiles aid in selecting boundary conditions
and defining layers in ground-water flow models (3, 4).
In addition to the creation of profiles, a process that
extracts surface values for use in a ground-water flow
model has been developed. The GWI uses the United
States Geological Survey (USGS) flow model, MOD-
FLOW (5). Being a cell-based model, MODFLOW re-
Gamma
SP
Resistivity
Figure 2. Example interpretation of geophysical logs.
quires a value for each hydrogeologic parameter for
each cell in the model grid. A series of FORTRAN pro-
grams and arc macro language (AMI) programs were
coupled to extract the required hydrogeologic data from
surface models. For example, piezometric surface val-
ues are required to set initial conditions for the model. A
coverage of the piezometric surface of the Memphis
Sand was created, converted to a TIN surface, and the
required values for each cell in the model were extracted
using the procedure described above.
The results and hydrogeologic data from the calibrated
model can be read back into the database and con-
verted into coverages. This allows piezometric contours
to be developed and displayed with other information in
the database to aid in decision-making. Also, capture
zones forthe wells can be brought into the database and
compared with surface features like industries, landfills,
Superfund sites, LIST locations, or other sites that may
have an impact. This has proved especially helpful in
developing wellhead protection programs where a com-
plete contaminant source inventory must be performed
for the capture zone of each well and within a fixed
radius around the well. The procedure used to develop
model data from the CIS database is summarized in
Figure 4.
Me Cord Wellfield Wellhead
Protection Program
MLGWand the GWI performed a demonstration project
funded by the U.S. Environmental Protection Agency
(EPA) for the C.M. McCord Wellfield Wellhead Protec-
tion Program (6). This wellfield was selected because of
multijurisdictional problems that will be encountered dur-
ing plan implementation. The City of Memphis owns all
the wells, but many of them, and all future well lots, are
located within the city limits of Bartlett. A wellhead pro-
tection plan will have to involve the cooperation of both
municipalities. The existing wellfield is shown in Figure 5.
Tennessee wellhead protection regulations require the
delineation of two zones of protection for a city the size
of Memphis: a 750-foot radius around the wellhead and
a 10-year capture zone for the well. The 10-year capture
zone (called the Zone 2 area) was delineated using two
flow models and information obtained from the GWI
database. Results were imported into the CIS database
and compared with existing information. The Zone 1
area was delineated by buffering each well point in the
coverage by the appropriate radius. Each well location
was verified using a Trimble GPS unit and is accurate
to within 2 meters. A contaminant source inventory was
performed using the coverages developed by address
matching. The primary potential sources of contamina-
tion in this area are USTs. A windshield survey located
other potential sources, such as dry cleaners. These
locations were entered into the database also by using
52
-------
Surficial Aquifer
Confining Layer/Clay Lens
Transitional Layer
Memphis Sand
Figure 3. Two-dimensional profile constructed from surface models.
Study Area Defined and
Model Grid Developed
Within ARC/INFO
«s*
*e
Model Input Files
Developed
V
Model Calibration
and Verification
the address-matching capabilities of ARC/INFO. The
Zone 2 areas for present wells and future wells, along
with the potential sources of contamination, are shown
in Figure 6.
Davis Wellfield Study
An example of CIS use in ground-water flow modeling
is the study of the Justin J. Davis Wellfield. The Davis
Wellfield is one of 10 producing fields operated by
MLGW It is located in the southwestern corner of
Shelby County and consists of 14 wells, as shown in
Figure 1. Production at the Davis Wellfield began in
1971, and an estimated 13 million gallons per day are
currently withdrawn from the Memphis Sand aquifer.
Since 1972, MLGW has collected water quality data
from the wells at the Davis Wellfield, including values for
alkalinity, hardness, chloride, sulfate, iron, and barium.
Water quality parameters of alkalinity, hardness, sulfate,
and barium have significantly increased in the past 10
years.1 A possible explanation for the change in water
quality is water leakage from the upper aquifer through
the confining unit to the Memphis Sand aquifer.1 The
water chemistry from the two aquifers is noticeably dif-
ferent. The surficial aquifers generally have a higher
total dissolved solids concentration, hardness, and alka-
linity than water from the Memphis Sand.1
Figure 4. Procedure for integrating GIS and flow model.
Webb, J. 1992. Memphis Light, Gas, and Water Division. Personal
interview.
53
-------
Shelby |
County
Well Location
Pumping
Station
Municipal Boundary
Figure 5. Existing McCord Wellfield.
MLGW, the GWI, and the USGS participated in a joint
investigation of the wellfield to determine why the water
quality changes are occurring. In the spring of 1992, 12
monitoring wells were drilled into the surficial aquifer
near the production wells, as shown in Figure 7.
Geophysical logging and split-spoon sampling revealed
an absence of the confining layer at one monitoring well,
GWI-3. All other wells penetrated various thicknesses of
clay. This "window" in the confining layer suggests that
the water quality changes could be due to leakage from
the surficial aquifer to the Memphis Sand. The logs for
these monitoring wells were combined with an existing
geophysical log coverage. The extent of the confining
layer window was estimated using CIS surface model-
ing capabilities, as shown in Figure 8.
Two-dimensional profiles were created to further show
the extent of the confining layer window. The locations
of the profiles in relation to various surface features are
shown in Figure 9. Profiles 1 and 2 were taken across
the river bluff, and Profiles 2 and 3 were taken across
the window. The profiles are shown in Figure 10.
Many important features of this area's geology can be
inferred by looking at the profiles. A connection of the
alluvial and fluvial aquifers is shown in Profile 2. Else-
where along the bluff, the connection of the two aquifers
is less prominent, as shown in Profile 1. The connection
of the two aquifers in Profile 2 may be the cause of a
peculiar mounding effect in the water table of the alluvial
aquifer in that area. The thinning of the top soil in Profiles
1 and 3 may indicate a local recharge area for the
54
-------
0 5 10
Scale in Thousands of Feet
10 -Year Capture Zones
Developed Wellfield
Fully Developed Wellfield
Underground Storage Tank
Aboveground Storage
Figure 6. Underground storage tanks and aboveground storage locations.
alluvial aquifer. Profiles 2 and 3 show the confining layer
window that suggests a connection between the surficial
aquifer and the Memphis Sand.
Following the convention of the USGS, a "window" is
defined as any area where the aggregate clay thickness
is less than 10 feet (8). Asurface of the thickness of the
confining layer was generated from the geophysical log
coverage. The surface model was converted to a con-
tour line coverage on a 5-foot interval. Using the ARC-
EDIT module of ARC/INFO, the contour line coverage
was converted to a polygon coverage. The area
bounded by the 10-foot contour of the surface model
was calculated to be 840,000 square feet (about 19
acres). The area was calculated by adding the areas
between the 10- and 5-foot contours and the area within
the 5-foot contour.
A flow model of the area was developed based on the
hydrogeologic data contained in the database. A steady
state model was calibrated to hydraulic conditions re-
corded during fall 1992 by the USGS and the GWI; the
root mean square (RMS) error for this model was 1.76
feet. A second steady state model was developed to
simulate conditions recorded during spring 1993 (a pe-
riod of high water levels in area lakes). The RMS error
for this simulation was 5.19 feet. This higher error may
indicate that the high water levels in surface water bod-
ies are not realistic for steady state boundary conditions.
Realistically, monitoring wells that are relatively far from
a surface water body are affected more by average
water levels over time ratherthan relatively short periods
of highs and lows.
Using average values of hi, h2 (head in upper and lower
aquifers), / (vertical flow distance), and VCONT (a pa-
rameter used in MODFLOWto allow for vertical conduc-
tance), the estimated flow rate through the window for
fall 1992 may be computed as:
k = VCONT 1 = 1.76e-3 x 199.2 = 0.351
ft
day
h!-h2 186.5-156.1
I
199.2
= 0.153
Q = kAi = 0.351 x 840,000x0.153 = 45,111
day
45,111 -- x 7.48
day
= 0.34 MGD
The flow rate calculated from average spring 1993 (a
period of high water levels in the surficial aquifers)
model results was computed as:
55
-------
Horn Lake Cutoff
Sewanee Road
Horn Lake Cutoff
Raines Road
Shelby Drive
Legend
A GWI Well
mi Water Body
NUU^UU^ BlUff
Roads
01234
Scale in Thousands of Feet
Figure 7. Location of GWI monitor wells.
k = VCONT I = 176e-3 x 199.2 = 0.351 -^-
day
. h!-h2 196.4-156.8
I
199.2
Q = kAi = 0.351 x 840,000 x 0.199 = 58,673
58,673 -ft- x 7.48 gall°ns= 438,874 - _,
day ff day
day
Using the GIS-delineated window, an estimated 0.34 to
0.44 million gallons per day flow from the alluvial aquifer
to the Memphis Sand aquifer. This variation in the flow
rate was due to seasonal variations of water level in the
alluvial aquifer.
Since the wellfield pumps approximately 13 million gal-
lons per day, the effect of the window on the entire Davis
Wellfield may not be significant. The window lies within
the 30-year capture zone of two wells in the field, however.
Raines Road
Shelby Drive
Legend
Water Body
Bluff
A
GWI Well
Roads
Scale in Thousands of Feet
Thickness of Confining Unit
< 5 Feet
5 to 10 Feet
10 to 20 Feet
20 to 30 Feet
30 to 40 Feet
> 40 Feet
Figure 8. Location and extent of window in confining unit.
The flow rate through the window is approximately 20
percent of the total production of these two wells. Addi-
tionally, since the wells would probably not operate si-
multaneously, the flow rate through the window may
account for approximately 40 percent of the flow at
either well.
Particle tracking in the Memphis Sand was developed
using MODPATH (9) from MODFLOW results. Particles
were placed at model well screens and tracked back-
ward for 30 years. The output from MODPATH was read
into the CIS database for comparison with other data,
as shown in Figure 11. The hole in the confining unit lies
56
-------
Legend
Bluff Roads ~, profile
0 5 10 15 20
Scale in Thousands of Feet
Water Body
Figure 9. Location of selected profiles.
within the 30-year capture zone of MLGW wells 418 and
419. Historically, 419 was the first well that experienced
water quality changes. The water quality is becoming
similar to water found in the alluvial aquifer.
A change in the water quality in well 418 is not as
immediately noticeable as in well 419. This inconsis-
tency in data may indicate that the window does not
extend northward from GWI-3, as the TIN model pre-
dicted. To determine which capture zone (418 or 419)
encompasses GWI-3, particles were tracked backward
for 40 years, as shown in Figure 12. GWI-3 lies on the
edge of the capture zone for 419. The flow lines from
418 and 419 move toward the northwest and southwest
in the Memphis Sand, pass up through the window, and
emerge in the upper aquifer.
Conclusions
The explanation of the Davis Wellfield investigation ad-
dressed some limitations of the database. The utility that
this hydrogeologic database provides greatly outweighs
the disadvantages, however. Without the ability to map
and define hydrogeologic features, this project may not
have been completed in the allotted time frame or may
not have been completed in the same level of detail. CIS
greatly enhances the development and evaluation of
ground-water flow models.
Specific conclusions that can be drawn from the analy-
sis performed in this project are:
The delineation of a window in the confining layer
using a CIS database is possible.
Based on the CIS-generated window, an estimated
0.34 to 0.44 million gallons per day flow from the
upper aquifer to the Memphis Sand, which may
account for as much as 40 percent of the flow at
either well 418 or 419.
The drilling of more monitoring wells north, south,
east, and west of GWI-3 may provide for a more
accurate delineation of the window.
57
-------
East
Profile 1
East
West
Profile 2
Profile 3
01234
Top Soil
Horizontal Scale in Thousands of Feet r^. gg^ an(j Qrave|
g|| Confining Clay
Q1234
Vertical Scale in Hundreds of Feet
Memphis Sand Aquifer
Figure 10. Selected subsurface cross sections in the Davis
Wellfield.
Sewanee Road
N
A
Legend
* MLGW Production Well -
U Window in Confining Unit
01234 __
Flow Path
Water Body
- Bluff
General conclusions that may be drawn from this dis-
cussion are:
CIS provides a convenient method of viewing flow
model input and output.
A flow model may be developed and evaluated in a
relatively short time using CIS.
CIS provides a convenient means of compiling and
managing the information required to develop a well-
head protection program.
Some CIS disadvantages that have been noted are:
The time required to develop a database and learn
to apply the CIS program in a particular situation may
be prohibitive.
CIS-generated results from a limited database may
be misleading and should be corroborated with other
analysis methods.
Acknowledgments
The authors would like to acknowledge the efforts of the
professors and students of the GWI, both present and
past, for their contributions to the database and this
Sewanee Road
Particles From Well
418 Emerging in ^
the Upper Aquifer >|
GWI-3
Legend
*
Sc:
GWI Well
Window in Confining Unit
n 1 > n a
Flow Path
Flow Path (#41 9)
Figure 11. Backward tracking for 30 years.
Figure 12. Backward tracking for 40 years.
58
-------
project, especially: Dr. John W. Smith, Director, and Dr.
Charles V. Camp for their patience and guidance in
interpreting the model results and hydrogeologic condi-
tions of the area; and David W. Kenley, Brian A.
Waldron, and Robert B. Braun for their help in creating
the well log database.
References
1. Camp, C.V., J.E. Outlaw, and M.C. Brown. 1994. CIS-based
ground-water modeling. Microcomputers in Civil Eng. 9:281-293.
2. Driscoll, F.G. 1986. Groundwater and wells. St. Paul, MN: Johnson
Division.
3. Camp, C.V., and M.C. Brown. 1993. A CIS procedure for develop-
ing a three-dimensional subsurface profile. J. Computing in Civil
Eng. (July).
4. Camp, C.V., and J.E. Outlaw. 1993. Constructing subsurface pro-
files using CIS. Adv. in Eng. Software. 18:211-218.
5. McDonald, M., and A. Harbaugh. 1988. A modular three-dimen-
sional finite-difference ground-water flow model. Open File Report
83-875. U.S. Geological Survey.
6. Palazolo, P.J., J.W. Smith, J.L. Anderson, and C.V. Camp. 1994.
C.M. McCord wellhead protection demonstration grant. Final
report. Herff College of Engineering Ground Water Institute,
University of Memphis, Memphis, TN.
7. Richardson, G. 1989. A study of potential sources of leakage into
the Memphis sand aquifer beneath the Davis well field in Memphis,
Tennessee. M.S. thesis. Memphis State University, Memphis, TN.
8. Parks, W.S. 1990. Hydrology and preliminary assessment of the
potential for contamination of the Memphis aquifer in the Memphis
area, Tennessee. Water Resources Investigation Report 90-4092.
U.S. Geological Survey.
9. Pollock, D. 1989. Documentation of computer programs to com-
pute and display pathlines using results from the U.S. Geological
Survey modular three-dimensional finite difference ground-water
flow model. Open File Report 89-381. U.S. Geological Survey.
59
-------
MODRISI: A PC Approach to GIS and Ground-Water Modeling
Randall R. Ross
Robert S. Kerr Environmental Research Laboratory, U.S. Environmental Protection Agency,
Ada, Oklahoma
Milovan S. Beljin
University of Cincinnati, Cincinnati, Ohio
Abstract
It is widely accepted that ground-water contamination
problems cannot be adequately defined or addressed
until the governing physical, chemical, and biological
processes affecting the transport and fate of contami-
nants are adequately characterized. Recent research
has led to a better understanding of these complex
processes and their effect on the movement of contami-
nants in the subsurface. The compilation and application
of such information has yet to be accomplished at many
hazardous waste sites, however. Too often, copious
quantities of data are collected, only to be stored, ig-
nored, or misplaced, rather than used for problem-solving.
Geographic information systems (GIS) are computer-
based tools that are relatively new to many environ-
mental professionals. GIS allows the manipulation,
analysis, interpretation, and visualization of spatially re-
lated data (e.g., hydraulic head, ground-water velocity,
and contaminant concentration). GIS is more than a
cartographic utility program, however. The analytical ca-
pabilities of GIS allow users to display, overlay, merge,
and identify spatial data, thereby providing the basis for
effective environmental decision-making.
IDRISI is a widely used PC-based raster GIS system
that provides numerous analytical capabilities that are
directly applicable to hydrogeologic studies. Raster sys-
tems are particularly well suited for analysis of continu-
ous data such as elevation (e.g., water table, land and
bedrock surfaces), precipitation, recharge, or contami-
nant concentrations and may be readily integrated with
finite-difference ground-water models. Because the for-
mats for IDRISI and ground-water model input data sets
are different, a need exists for a program to integrate
these two types of robust tools.
MODRISI is a collection of utility programs that allows
easy manipulation and transfer of data files between
IDRISI and ground-water models (e.g., MODFLOW,
ASM, MOC). In addition, MODRISI integrates other
widely used commercial and private domain software
packages, such as SURFER, Geopack, GeoEas, Auto-
CAD, CorelDraw, and various spreadsheet programs.
Two-dimensional arrays of models' input data sets can
easily be created from IDRISI image files. AutoCAD
vector files obtained by digitizing model boundaries, well
locations, rivers and streams, or U.S. Geological Survey
digital elevation model (DEM) files can also be trans-
lated into model input file formats. MODRISI can proc-
ess model output files and prepare GIS image files that
can be displayed and manipulated within IDRISI. Thus,
MODRISI is more than a pre- and postprocessor for
ground-water models; it is a complete GIS/ground-water
modeling interface that is accessible to most ground-
water hydrologists.
Introduction
Hydrogeologists collect and analyze large volumes of
data during a ground-water modeling process. These
data are stored and presented in many different forms
such as maps, graphs, tables, computer databases, or
spreadsheets. To most hydrogeologists, geographic in-
formation systems (GIS) are relatively new tools. They
have been developed and applied in other natural and
social science fields for over two decades, however, and
can also be used in the ground-water modeling process.
GIS represents a new, powerful set of tools that can
significantly improve the usefulness of results obtained
during the ground-water modeling process. Bridging the
disciplines of ground-water modeling, computer graph-
ics, cartography, and data management, GIS represents
a computer-based set of tools to display and analyze
spatial data (e.g., water level elevations, ground-water
quality data, modeling results, ground-water pollution
potential). Efficient use of increasingly large volumes of
60
-------
data can be achieved only with powerful systems capa-
ble of acquiring information from a variety of sources,
scales, and resolutions.
CIS can be defined as a computer-assisted system for
the efficient acquisition, storage, retrieval, analysis, and
representation of spatial data. Most CIS platforms con-
sist of numerous subsystems that perform the listed
tasks. The subsystems have the ability to query spatially
related information and incorporate statistical analyses
and modeling of relations and their temporal changes
within the database. More than just a mapping system,
CIS allows the user to analyze spatially related data and
visualize results in either paper map form or graphically
on screen. The data to be analyzed are a collection of
spatial information represented by points, lines, and
polygons and their associated attributes (characteristics
of the features such as elevation or concentration). The
cartographic tools of CIS allow the analyst to display,
overlay, measure, merge, and identify the data to sup-
port a particular analysis. By allowing spatial data analy-
sis and display, CIS provides the means necessary for
effective environmental decision-making and implemen-
tation of environmental management plans.
CIS uses two basic map representation techniques:
vector and raster. Vector representations describe fea-
tures with a number of connected points. Raster repre-
sentations subdivide a study area into a mesh of grid
cells, each cell containing either a quantitative attribute
value or feature identifier. Raster systems are well suited
for analysis of continuous data (e.g., water level eleva-
tions, infiltration and recharge rates). This makes raster-
based systems ideal for integration with ground-water
models that use regularly spaced nodes. The objective
of this paper is to illustrate such an integration of MOD-
FLOW (1), a widely used U.S. Geological Survey
(USGS) finite-difference ground-water flow model, and
IDRISI (2), a raster-based CIS.
Previous Studies
To enhance understanding of a hydrogeologic system,
and also to develop a credible ground-water model of
the system, hydrogeologic features such as lithological
logs, recharge and withdraw rates, estimates of spatial
distribution of hydraulic conductivity, or specific storage
can be plotted using CIS capabilities of data retrieval
and overlay options to interactively define an area of
interest (3-6). Two previous studies that combined CIS
and ground-water modeling are briefly described.
Torak and McFadden (7) used CIS to facilitate finite-element
modeling of ground-water flow. Complex aquifer geometry
and irregularly distributed aquifer-system characteristics
that influence ground-water flow affect the design of the
finite-element mesh. CIS systems represent the com-
plex arrangement of nodes and elements and the dis-
tribution of aquifer properties to provide input to the
flow model. Point-data coverages of pertinent aquifer
characteristics are rated from a relational database and
are displayed using CIS.
Contoured surfaces based on point-data coverages pro-
duce triangulated irregular networks (TINs) that are su-
perposed on the finite-element mesh to delineate zones
of elements having similar aquifer properties. Zone
boundaries are identified using the contoured TIN sur-
face and by manually determining where boundaries
align with the element sides. The allocation of well
pumping rates to nodes in the finite-element mesh is
performed efficiently with CIS for model input. Well
pumping rates are accumulated by element from the
combined coverages of the pumping data and the mesh,
and element data are distributed to the node points for
input. CIS is also used to prepare data for model input
and to assess the adequacy of the data priorto simulation.
Three-dimensional perspectives showing TIN cover-
ages of aquifer-property data are used to analyze and
interpret complexities within the flow system before
zonation. Additionally, CIS is used to display computed
hydraulic heads over the finite-element mesh to produce
contour maps of the simulated potentiometric surface.
Because the node points in the finite-element mesh are
not arranged in an orthogonal fashion, such as a finite-
difference grid, a map display of the computed values of
hydraulic head at the nodes is prepared for efficient and
accurate interpretation of simulation results.
Harris et al. (8) conducted the Remedial Investiga-
tion/Feasibility Study (RI/FS) of the San Gabriel basin.
Vast amounts of hydrogeologic data have been gath-
ered, and a comprehensive systematized CIS database
has been developed. The identified hydrologic bounda-
ries, recharge basins, stream locations, well locations,
and contaminant distributions are some of the features
considered in developing a base map. The CIS-gener-
ated base map has allowed development of a finite-ele-
ment grid for the basin. For each finite element, the initial
estimates of the hydraulic conductivity, specific yield,
recharge rates, and other input parameters were provided.
Using simple interfacing programs, the retrieval CIS
nodal and elemental data were converted to required
formats for the input files of the Couple Fluid, Energy,
and Solute Transport (CFEST) code. Simulated ground-
water levels were compared with the CIS-generated
potentiometric surfaces. In areas of wide variations be-
tween simulated and observed data, the zonal distribu-
tion of controlling parameters was reevaluated,
analyzed, and updated. Data processing, development
of input files for computerized analysis of ground-water
flow, and analysis of simulation results with different
alternative conceptualizations is time consuming and
tedious. Efficient use of CIS and CFEST not only eased
the burden of conducting multiple simulations but
61
-------
reduced the probability of errors as well as the amount
of time and effort required for each simulation.
IDRISI
IDRISI is a grid-based geographic information and im-
age processing system developed by the Graduate
School of Geography at Clark University and supported
by the United Nations Institute for Training and Re-
search (UNITAR) and the United Nations Environment
Programme Global Resource Information Database
(UNEP/GRID) (9). IDRISI is a collection of over 100
program modules that are linked through a menu
system. These programs are organized into several
groups:
The core modules provide data entry and database
management capabilities.
The geographic analysis modules provide tools for
database analysis.
The statistical analysis modules allow statistical char-
acterization of images.
The peripheral modules provide a series of utilities.
IDRISI and other raster-based systems divide data sets
into map layers; each layer contains data for a single
attribute. For the example of a ground-water model,
these layers could correspond to the MODFLOW two-
dimensional arrays (e.g., initial water levels, transmis-
sivity distribution, IBOUND arrays, computed hydraulic
heads). IDRISI provides many analytical tools that are
useful in hydrogeologic studies.
Three of the most important categories of these tools
are database query, map algebra, and context operator.
Asemihypothetical case described below illustrates the
use of these analytical tools. IDRISI provides an exten-
sive set of tools for image processing, geographic and
statistical analysis, spatial decision support, time series
analysis, data display, and import/export and conver-
sion. In addition, as a set of independent program mod-
ules linked to a broad set of simple data structures, the
system is designed such that researchers may readily
integrate into the system their own modules, written in
any programming language.
IDRISI uses three types of data files: image, vector, and
attribute. Image files contain rasterized information
relating to a spatial variable. Vector files contain the
coordinates of points, lines, and polygonal features. An
attribute file lists the identifiers of features and the asso-
ciated attribute values. Values files can be extracted
from the existing image files, or image files can be
created from existing values files. The values files can
be combined and stored in a dBASE format. Each image,
vector, or attribute file has a corresponding documentation
file that contains information about the data file (e.g.,
title, number of rows and columns).
MODRISI: MODFLOW/IDRISI Interface
MODRISI is a set of utility programs that allows the
transfer of data files between MODFLOW, IDRISI, Gold-
en Software SURFER, GeoEas, and other software.
Preparation of two-dimensional arrays for the MOD-
FLOW input files is generally tedious and time consum-
ing. The arrays can be created easily from the IDRISI
image files, however. Thus, MODRISI serves as a pre-
processor for MODFLOW. For example, when the val-
ues of a variable are available only for irregularly spaced
points, interpolation routines in SURFER or GeoEas
may be used to estimate the values of the variable on
regularly spaced grid-nodes. MODRISI translates
SURFER or GeoEas files into IDRISI image files for
manipulation, analysis, and display. IDRISI recognizes
either latitude and longitude geodetic coordinates or
arbitrary Cartesian plane coordinates. IDRISI assigns
the lower left grid-block of a raster image as a zero-row,
zero-column block.
Vector files, such as model boundaries, well locations,
and rivers, may be created within IDRISI and translated
into a MODFLOW input file format. For example, the
location of a river may be digitized on screen in IDRISI.
The vector-to-raster function may be invoked, assigning
all blocks through which the river passes as river nodes.
Similarly, the positions of wells may be digitized and
translated into the row-column positions and saved as a
MODFLOW input file for the well package. Once the
MODFLOW input files are prepared, MODFLOW simu-
lations may be initiated. The MODFLOW hydraulic head
output files may be read by MODRISI and modified to
create IDRISI image files. Again, the image files may be
displayed and evaluated within IDRISI. Thus, MODRISI
is used as a postprocessor for MODFLOW.
Case Study
The utility of MODRISI was demonstrated at a hazard-
ous waste site. Previous investigations provided site-
specific information, including water level and bedrock
and land surface elevations, which was then analyzed
using GeoEas, a public domain geostatistical software
program. These data were kriged to produce a grid of
regularly spaced data. These data were imported into
MODRISI and converted to IDRISI image files. IDRISI
was used to visualize the surfaces that GeoEas generated.
Several prominent features are obvious upon inspection
of the kriged bedrock topography (see Figure 1). Abed-
rock ridge trending northwest to southeast is flanked by
a minor trough to the east and a major trough to the
southwest. The outline of the site is visible on all figures.
The kriged water level elevation map (see Figure 2)
illustrates the general hydraulic gradient to the west.
Land surface elevations (see Figure 3) range from
greater than 200 feet in the northeastern portion of the
site to a low of 166 feet on the western boundary.
62
-------
70-80
80-90
90-100
100-110
110-120.
120-130
130-140
140-150
150-160
(Feet)
-- - - fcfeMss&SSt: 310 Feet
Figure 1. Bedrock elevation contour map derived from kriged data and transformed by MODRISI into IDRISI image file format.
,.-,=.- - _f#>*
&^- "
168
1 69
1 70
-|y-|
-| 72
175 -
176 :
(Feet)
iiialK 310 Feet
Figure 2. Water level contour map derived from kriged water level data and transformed by MODRISI into IDRISI image file format.
63
-------
165-170 i
170-175
175-180
180-185 :
185-190 .
190-195
195-200
(Feet)
310 Feet
Figure 3. Land elevation contour map generated from kriged data and transformed by MODRISI into IDRISI image file format.
Additionally, the value (e.g., elevation) and (x,y) coordi-
nates may be queried for any point, line, or area of an
image file.
The analytical capabilities of IDRISI are illustrated in the
following example. The saturated thickness was deter-
mined by subtracting the bedrock surface from the water
level surface. The results illustrate the spatial variability
of saturated thickness of the overburden aquifer (see
Figure 4). The results correspond favorably with the
general bedrock topography, as would be expected
given the relatively low hydraulic gradient across the
site. This OVERLAY (subtract) function may also readily
be used to evaluate the adequacy of model calibration.
For example, predicted values may be compared with
observed values graphically, allowing the modeler to
quickly visualize and identify areas of the model domain
requiring additional consideration and manipulation.
Conclusions
One of the most tedious tasks in a ground-water mod-
eling process is preparing input data and postprocess-
ing model results. CIS allows rapid incorporation and
evaluation of new site characterization information. The
proposed combination of IDRISI, a raster-based CIS
system, and MODRISI, a set of utility programs, could
significantly reduce the amount of time necessary for
entering data in required array formats. The visualization
capabilities of IDRISI in conjunction with MODRISI and
MODFLOW allow project managers to better under-
stand the three-dimensional nature of subsurface envi-
ronmental problems.
Acknowledgments
The authors wish to thank Chet Janowski, U.S. Environ-
mental Protection Agency Region I, for his support on
this project.
References
1. McDonald, M., and A. Harbough. 1988. A modular three-dimen-
sional finite-difference ground-water flow model: USGS techniques
of water resources, Book 6, Chapter A1.
2. Eastman, J.R. 1992. IDRISI: Technical reference. Clark University,
Worcester, MA (March).
3. Van Metre, P.C. 1990. Structure and application of an interface
program between a geographic-information system and a ground-
water flow model. USGS Open File Report 90-165. Denver, CO:
U.S. Geological Survey.
4. Hinaman, K.C. 1993. Use of a geographic information system to
assemble input-data sets for a finite-difference model of ground-
water flow. In: Proceedings of the AWRA Symposium on Geo-
graphic Information Systems and Water Resources, Mobile, AL,
(March 14-17). pp. 405-412.
5. Rifai, H.S., L.A. Hendricks, K. Kilborn, and P.B. Bedient. 1993. A
geographic information system (CIS) user interface for delineating
wellhead protection areas. Ground Water 31 (3):480-488.
6. Orzol, L.L., and T.S. McGrath. 1992. Modifications of the USGS
modular three-dimensional finite-difference, ground-water flow
model. U.S. Geologic Survey Open File Report 92-50.
64
-------
< 20
20-30 &:.
30-40 :
40-50 r:v.;
50-60
60-70
70-80
80-90
90-100 53
100-110
(Feet)
lisis&s&BliifisiiSifi -, -----;--' :."':- : " .".-.' -.- - :=;, -.6.;"i'':.;--,; " 310 Feet
Figure 4. Saturated thickness contour map developed by subtracting bedrock image file from water level image file within IDRISI.
7. Torak, L.J., and K.W. McFadden. 1988. Application of a geographic
information system to finite-element modeling of ground-water
flow. In: Proceedings of the CIS Symposium Integrating Technol-
ogy and Geoscience Applications, Denver, CO (September 26-30).
8. Harris, J., S. Gupta, G. Woodside, and N. Ziemba. 1993. Integrated
use of CIS and three-dimensional, finite-element model: San
Gabriel basin ground-water flow analyses. In: Goodchild, M., B.
Parks, and L. Steyaert, eds. Environmental modeling with CIS.
New York, NY: Oxford University Press.
9. Eastman, J.R., PA. Kyem, J. Tolendano, and W. Jin. 1993. CIS
and decision making. Explorations in geographic information sys-
tems technology, Vol. 4. United Nations Institute for Training and
Research, Geneva, Switzerland.
65
-------
GIS in Statewide Ground-Water Vulnerability Evaluation to Pollution Potential
Navulur Kumar and Bernard A. Engel
Department of Agricultural Engineering, Purdue University, West Lafayette, Indiana
Abstract
The ground-water vulnerability of Indiana to pollution
potential was evaluated using a geographic information
systems (GIS) environment. The Geographic Re-
sources Analysis Support System (GRASS) and the
GRID submodule of ARC/INFO were used to conduct
the analysis and to identify and display the areas sensi-
tive to ground-water pollution potential. The state soils
geographic (STATSGO) database was employed to re-
trieve statewide soils information required for the analy-
sis. The information from the STATSGO database was
used in two models, DRASTIC (acronym representing
the following hydrogeologic settings: Depth to water
table, aquifer Recharge, Aquifer media, Soil media,
Topography, Impact of vadose zone, and hydraulic
Conductivity of the aquifer) and SEEPAGE (System for
Early Evaluation of Pollution Potential of Agriculture
Ground-Water Environments). These models employ a
numerical ranking system and consider various hydro-
geologic settings that affect the ground-water quality of
a region. Ground-water vulnerability maps were pre-
pared for the state of Indiana based on DRASTIC and
SEEPAGE results. Continuing work is planned to deter-
mine the accuracy of the results by comparing the ex-
isting well-water quality data. The DRASTIC Index and
SEEPAGE Index number (SIN) maps show great poten-
tial as screening tools for policy decision-making in
ground-water management.
Introduction
Ground-water contamination due to fertilizer and pesti-
cide use in agricultural management systems is of wide
concern. In 1989, reports of ground-water contamination
in New York wells led the U.S. Environmental Protection
Agency (EPA) to conduct a nationwide survey on well
contamination in the United States. These wells were
tested for presence of nitrate, pesticides, and pesticide
breakdown products (1). Statistically, the wells selected
represent more than 94,600 wells in approximately
38,300 community water systems. Over 52 percent of
the community water systems and 57 percent of the
rural domestic wells tested contained nitrates (2).
Indiana has abundant ground-water systems providing
drinking water for 60 percent of its population. A study
on well-water quality detected pesticides in 4 percent of
wells tested in Indiana. Also, 10 percent of private wells
and 2 percent of noncommunity wells contained exces-
sive nitrate levels (3).
Statewide maps showing the areas vulnerable to
ground-water contamination have many potential uses
such as implementation of ground-water management
strategies to prevent degradation of ground-water qual-
ity and monitoring of ground-water systems. These
maps will be helpful in evaluating the existing and po-
tential policies for ground-water protection. Ground-
water models such as SEEPAGE (System for Early
Evaluation of Pollution Potential of Agriculture Ground-
Water Environments) and DRASTIC (acronym repre-
senting the following hydrogeologic settings: Depth to
water table, aquifer Recharge, Aquifer media, Soil
media, Topography, Impact of vadose zone, and hy-
draulic Conductivity of the aquifer) can be applied on a
regional scale to develop such maps.
The data layers required forthese models are commonly
available data such as pH and organic matter content.
For most states, the statewide ground-water vulnerabil-
ity maps generated using DRASTIC were produced
from 1:2,000,000-scale data (4). EPA (2) found that
these maps did not correlate well with the water quality
analysis performed for the national survey of pesticides
in drinking water wells. States need more detailed and
accurate maps to implement ground-water management
programs. The state soils geographic (STATSGO)
database at the 1:250,000-scale might be useful for
studies at a larger scale.
The geographic information systems (GIS) environment
is widely applied for diverse applications in resources
management and other areas. It offers the facilities to
store, manipulate, and analyze data in different formats
66
-------
and at different scales. The DRASTIC and SEEPAGE
models can be integrated within the CIS environment to
produce the final ground-water vulnerability maps.
Objectives
The purpose of the study was to prepare maps showing
areas in Indiana vulnerable to ground-water pollution.
This goal was accomplished by considering hydro-
geologic factors in each region that affect the mobility
and leaching of the contaminant reaching the aquifer.
The prime objectives of this research were to:
Evaluate Indiana's ground-water vulnerability to
pollution potential using the DRASTIC and SEEP-
AGE models:
- Integrate and evaluate the models in a CIS envi-
ronment (Geographic Resources Analysis Support
System [GRASS] ARC/INFO).
- Develop a graphic user interface (GUI) in ARC/INFO
to conduct the analyses.
Compare the pollution potential map from the
DRASTIC model with the map developed using the
SEEPAGE Index number (SIN).
Validate the accuracy of the present approach by
comparing the vulnerability maps with the existing
well-water quality data sampled across the state.
DRASTIC
DRASTIC is a ground-water quality model for evaluating
the pollution potential of large areas using the hydro-
geologic settings of the region (4-6). EPA developed this
model in the 1980s. DRASTIC includes different hydro-
geologic settings that influence a region's pollution po-
tential. A hydrogeologic setting is a mappable unit with
common hydrogeologic characteristics. This model em-
ploys a numerical ranking system that assigns relative
weights to parameters that help evaluate relative
ground-water vulnerability to contamination.
The hydrogeologic settings that make up the acronym
DRASTIC are:
[D] Depth to water table: Compared with deep water
tables, shallow water tables pose a greater chance
for the contaminant to reach the ground-water surface.
[R] Recharge (net): Net recharge is the amount of
water per unit area of soil that percolates to the aqui-
fer. This is the principal vehicle that transports the
contaminant to the ground water. Higher recharges
increase the chances of the contaminant being trans-
ported to the ground-water table.
[A] Aquifer media: The material of the aquifer deter-
mines the mobility of the contaminant traveling
through it. An increase in travel time of the pollutant
through the aquifer increases contaminant attenuation.
[S] Soil media: Soil media is the uppermost portion
of the unsaturated zone/vadose zone characterized
by significant biologic activity. This, in addition to the
aquifer media, determines the amount of water per-
colating to the ground-water surface. Soils with clays
and silts have larger water holding capacity and thus
increase the travel time of the contaminant through
the root zone.
[T] Topography (slope): The higher the slope, the
lower the pollution potential due to higher runoff and
erosion rates, which include pollutants that infiltrate
the soil.
[I] Impact of vadose zone: The unsaturated zone
above the water table is referred to as the vadose
zone. The texture of the vadose zone determines the
travel time of the contaminant. Authors of this model
suggest using the layer that most restricts water flow.
[C] Conductivity (hydraulic): Hydraulic conductivity of
the soil media determines the amount of water per-
colating through the aquifer to the ground water. For
highly permeable soils, the travel time of the pollutant
is decreased within the aquifer.
The major assumptions outlined in DRASTIC are:
The contaminant is introduced at the surface.
The contaminant reaches ground water by precipitation.
The contaminant has the mobility of water.
The area of the study site is more than 100 acres.
DRASTIC evaluates pollution potential based on the
seven hydrogeologic settings listed above. Each factor
is assigned a weight based on its relative significance in
affecting the pollution potential. Each factor is also as-
signed a rating for different ranges of the values. Typical
ratings range from 1 to 10, and weights range from 1 to
5. The DRASTIC Index, a measure of pollution potential,
is computed by summation of the products of rating and
weights of each factor as follows:
DRASTIC Index =
DrDw + RrRw + ArAw + SrSw + TrTw + Irlw + CrCw
where:
Dr = Ratings for the depth to water table
Dw = Weights for the depth to water table
Rr = Ratings for different ranges of aquifer recharge
Rw = Weights for the aquifer recharge
Ar = Ratings for the aquifer media
Aw = Weights for the aquifer media
Sr = Ratings for soil media
Sw = Weights for soil media
Tr = Ratings for topography (slope)
Tw = Weights for topography
Ir = Ratings for the vadose zone
67
-------
Iw = Weights for the vadose zone
Cr = Ratings for different rates of hydraulic conductivity
Cw = Weights for hydraulic conductivity
DRASTIC assigns two different weights depending upon
the type of contaminant. Pesticides are given different
weights than general contaminants. In assigning the
weights, DRASTIC considers the different properties of
pesticides as they travel through the vadose zone and
root zone of the soil media.
The higher the DRASTIC Index, the greater the relative
pollution potential. The DRASTIC Index is divided into
four categories: low, moderate, high, and very high. The
sites with high and very high categories are more vul-
nerable to contaminations and hence should be re-
viewed by the site specialist. These weights are relative,
however. Low pollution potential does not necessarily
indicate that a site is free from ground-water contami-
nation. It indicates only that the site is less susceptible
to contamination than sites with high or very high
DRASTIC ratings.
SEEPAGE
The SEEPAGE model is a combination of three models
adapted to meet the Soil Conservation Service's
(SCS's) need to assist field personnel (7, 8). SEEPAGE
considers hydrogeologic settings and physical proper-
ties of the soil that affect ground-water vulnerability to
pollution potential. SEEPAGE is also a numerical rank-
ing model that considers contamination from both con-
centrated and dispersed sources.
The SEEPAGE model considers the following parameters:
Soil slope
Depth to water table
Vadose zone material
Aquifer material
Soil depth
Attenuation potential
The attenuation potential further considers the following
factors:
Texture of surface soil
Texture of subsoil
Surface layer pH
Organic matter content of the surface
Soil drainage class
Soil permeability (least permeable layer)
Each factor is assigned a numerical weight ranging from
1 to 50 based on its relative significance, with the pa-
rameter that has the most significant effect on water
quality assigned a weight of 50 and the least significant
assigned a weight of 1. The weights are different for
concentrated or site-specific sources, and dispersed or
nonspecific sources.
Similar to DRASTIC, each factor can be divided into
ranges and ratings, varying from 1 to 50. The ratings of
the aquifer media and vadose zone are subjective and
can be changed for a particular region. Once the scores
of the six factors are obtained, they are summed to
obtain the SIN. These values represent pollution poten-
tial, where a high SIN implies relatively more vulnerabil-
ity of the ground-water system to contamination. The
SIN values are arranged into four categories of pollution
potential: low, moderate, high, and very high. A high or
very high SIN category indicates that the site has signifi-
cant constraints for ground-water quality management (7).
GIS
CIS has been widely used for natural resources man-
agement and planning, primarily during the past decade.
A GIS can be combined with a ground-water quality
model to identify and rank the areas vulnerable to pol-
lution potential for different scenarios and land use prac-
tices. Many GIS software packages are available.
GRASS is a raster-based public domain software devel-
oped by the U.S. Army Construction Engineers Re-
search Laboratory (9). This software can assign different
weights to, or reclass, the data layers and combine map
layers, and is suitable for implementing the DRASTIC
and SEEPAGE models. ARC/INFO is a GIS software
developed by Environmental Systems Research Insti-
tute (ESRI) in Redlands, California. The GRID sub-
module of ARC/INFO facilitates the handling of raster
data. Also, the capability to develop a menu-based GUI
helps users easily implement the models. The GRID
submodule also can reclass and manipulate the map
layers suitable for conducting the analyses.
Methodology
Developing the Data Layers in GRASS and
ARC/INFO
The STATSGO database from SCS comes at a scale of
1:250,000 and is distributed in different data formats.
This study used the STATSGO database in the
ARC/INFO format. The database is organized into map
units that have up to 21 components. These map com-
ponents have information assigned to layers of soil ho-
rizons. Each layer is attributed various soil properties
such as pH or organic matter content (10). Each prop-
erty is assigned a high and a low value for a map unit.
The STATSGO map for Indiana is available in the vector
format. This map was exported to GRASS as a vector
coverage (11) and was converted into a raster coverage
within the GRASS GIS environment. This was used as
68
-------
the base map for the DRASTIC and SEEPAGE analy-
ses. The hydrogeologic parameters required for the
models were identified from the corresponding INFO
data tables and were exported into an ASCII file. Code
was developed to generate a GRASS reclass file as-
signing the weighted values of the parameters to the
corresponding map units in the base map. The STATSGO
base map imported into GRASS was reclassed for each
hydrogeologic setting (e.g., topography, pH) to create the
data layers required for DRASTIC and SEEPAGE analyses.
The map layers of the hydrogeologic parameters in
GRASS were then exported to ARC/INFO as raster
coverages. A GRASS command was developed that
allows the output ASCII file from GRASS to be imported
into ARC/INFO directly without further modifications to
the header in the ASCII file.
Developing a Graphic User Interface in
ARC/INFO
The dynamic form-menu option (12) was used to develop
a GUI for both DRASTIC and SEEPAGE analyses (see
Figures 1 and 2). Because ratings for some parameters
are subjective, the GUI provided an option to change the
weights assigned to hydrogeologic settings. The cover-
ages must already be assigned ratings before using the
interface, however. The interface also allows users to
reclassify the final vulnerability maps qualitatively (13) into
four categories (low, moderate, high, and very high) after
viewing the range of DRASTIC Index or SIN values.
Conducting the Analyses
The data layers were developed separately for the high
and low values of the hydrogeologic settings. Once all
the data layers were compiled, the corresponding rat-
ings and weights were assigned and the analyses were
conducted using the GUI. The data layers aquifer re-
charge, aquifer media, and vadose zone media were not
available, so the analyses were conducted without these
base maps. The SEEPAGE analysis was performed for
concentrated/point sources of pollution. The final vulner-
ability indexes from the analyses were classified into
four categories (low, medium, high, and very high) (see
Table 1) to generate the final statewide vulnerability maps.
Table 1. Pollution Potential Categories Using SEEPAGE and
DRASTIC Indexes
Range of DRASTIC/SEEPAGE Index
Analysis
Low
Moderate High
Very High
SEEPAGE
DRASTIC
1-24
30-70
25-48
71-100
49-70
101-110
>70
> 110
Validating the Accuracy of the
Vulnerability Maps
The ground-water vulnerability maps produced by
DRASTIC analysis were compared with those generated
using the SEEPAGE model. The final statewide ground-
water vulnerability maps from either approach were
compared with the well-water quality data sampled from
over 2,500 wells (see Figure 3), and the number of wells
falling into each vulnerability category was tabulated.
Results and Discussion
Statewide analysis of ground-water vulnerability to pol-
lution potential was conducted using the DRASTIC and
SEEPAGE analyses at a scale of 1:250,000 in the raster
format. The analyses were conducted for both the high
and low values of the hydrogeologic settings, and the
final vulnerability maps were prepared for the state of
Indiana (see Figures 4 and 5). The vulnerability maps
from both approaches were compared in the GRASS
environment.
In both analyses, the low values of hydrogeologic set-
tings resulted in more areas being classified as high and
very high categories, compared with the high values of
hydrogeologic settings in a map unit. The DRASTIC
analysis placed more areas in the very high vulnerability
category, compared with the SEEPAGE analysis, which
categorized the same areas as high vulnerability. The
nitrate-nitrogen concentrations observed in the wells
were compared with the final vulnerability maps, and the
number of wells falling into each of the four vulnerability
categories (low, moderate, high, and very high) were sum-
marized (see Tables 2 and 3).
Approximately 80 percent of the wells with concentra-
tions less than 5 parts per million are classified under
the moderate vulnerability category in SEEPAGE analy-
sis. Overall, the results from the analyses did not corre-
late satisfactorily with the observed well-water quality
data. Unavailability of the data layers aquifer media,
aquifer recharge, and vadose zone media may have
caused these results. The well-water quality data of
nitrate-nitrogen contaminations was considered only for
testing map accuracy, whereas the analyses do not
account for the type of contaminant, its severity, and its
volume in the generation of vulnerability maps. Other
limitations of these approaches, including that the fac-
tors influencing aquifer contamination (e.g., direction of
water flow, land use, population at risk, point sources of
pollution) are not considered in the ground-water vulner-
ability evaluation, might also have led to the observed
results.
The DRASTIC and SEEPAGE analyses can be improved
by incorporating data layers such as land use and nitrate
loadings in computing the DRASTIC and SEEPAGE
Indexes. The STATSGO database can be used for
69
-------
Figure 1. GUI for DRASTIC analysis.
70
-------
"* fff VS r t . tf f f f rr,
,' fff'f f f ff/'sf ,f",,, ', '$ '{ f" f x"x*. ' .,. f , f f f'ff f, ixxxi ;
*- t*$ \tr ^ fff ^,«' «- »> sv ^; , , w f fit ff ff
* jf f A-J: ' *sf . fffff j-j-f fSff jff ^ t v. f f . fj? ff ^ fff *& fffe
-'& ,^r~ -^'/;-^/ ,
T* ' «.'% **
e* 4^.< . .
frfft AV ^
^ %*'$ ^V*^^.;|^r^ *" f *H ' ^"
'* 'x,,''^ i'«^Si*t'^ry'fl«ifeu- 1T'» '"'
v f $z&f*~. M5^^f ^^ ^"" ''"
5 ^ ^ /A' f, "f,f ,%.,
",' » vsxj-x ^,xx ,. -,; ^
B # ~, ^ 'xi*' x ,"ff ff ;;_ -5, ^ ';,
\^~*:*.::*z'i *:.\ *.
Figure 2. GUI for SEEPAGE analysis.
71
-------
Figure 3. Sampling sites (wells) for water quality data.
Low
Moderate
High
Very High
^ Low
1H Moderate
High
Very High
Figure 5. Ground-water vulnerability map using SEEPAGE
analysis.
Table 2. Comparison of SEEPAGE Results With Observed
Nitrate-Nitrogen Concentrations in Wells
Nitrate-Nitrogen Levels
Figure 4. Ground-water vulnerability map using DRASTIC yerv hjqh
analysis.
Category
Low
Moderate
High
Very high
Table 3.
Category
Low
Moderate
High
Very high
0-5 5-10
7 2
1 ,322 76
249 11
194 2
>10
8
384
1
64
Comparison of DRASTIC Results With Observed
Nitrate-Nitrogen Concentrations in Wells
Nitrate-Nitrogen Levels
0-5 5-10
11 2
541 41
720 39
500 9
> 10
9
290
213
65
72
-------
developing most of the data layers required for the
analyses. CIS is a useful tool for integrating ground-
water quality models and facilitates testing the models for
different scenarios. The GUI helps users easily conduct
analyses and facilitates changing the weights for subjec-
tive hydrogeologic settings. DRASTIC and SEEPAGE ap-
proaches show great potential as screening tools for policy
decision-making in ground-water management.
Summary
Ground-water pollution from agricultural management
systems is of wide concern. Few models address
ground-water vulnerability on a regional scale. The
DRASTIC and SEEPAGE models are numerical ranking
models that consider various hydrogeologic settings af-
fecting the contamination of a region. The data required
for these models are commonly available data, and the
STATSGO database at 1:250,000-scale was used in this
study. These models were integrated in the CIS environ-
ment of GRASS and ARC/INFO in the raster format. A
menu-based GUI was developed in ARC/INFO for con-
ducting the analyses. The vulnerability maps generated
from DRASTIC and SEEPAGE analyses were com-
pared. The statewide vulnerability maps also were com-
pared with the well-water quality data to validate the
accuracy of the models.
References
1. U.S. EPA. 1990. National pesticide survey: Phase I report. Wash-
ington, DC.
2. U.S. EPA. 1992. Another look: National survey of pesticides in
drinking water wells, Phase II report. EPA/579/09-91/020. Wash-
ington, DC.
3. Department of Environmental Management, Groundwater Sec-
tion. 1989. Indiana ground-water protection: A guide book (June).
4. U.S. EPA. 1987. DRASTIC: A standardized system for evaluating
ground-water pollution potential using hydrogeologic settings.
EPA/600/2-87/035. Washington, DC.
5. U.S. EPA. 1985. DRASTIC: A standardized system for evaluating
ground-water pollution potential using hydrogeologic settings.
EPA/600/2-85/0108. U.S. EPA, Robert S. Kerr Environmental Re-
search Laboratory, Ada, OK.
6. Deichert, L.A., and J.M. Hamlet. 1992. Nonpoint ground-water
pollution potential in Pennsylvania. ASAE Paper No. 922531.
American Society of Agricultural Engineers International Winter
Meeting, Nashville, TN.
7. Richert, S.E., S.E. Young, and C. Johnson. 1992. SEEPAGE: A
GIS model for ground-water pollution potential. ASAE Paper No.
922592. American Society of Agricultural Engineers International
Winter Meeting, Nashville, TN.
8. Engel, B.A., and D.D. Jones. 1992. Technique for developing
ground-water vulnerability to nitrate maps for large areas: Re-
search proposal 12-1992.
9. U.S. CERL. 1990. GRASS: Geographical Resources Analysis
Supporting System user's manual. Champaign, IL: U.S. Army
Construction Engineers Research Laboratory.
10. SCS. 1992. State Soils Geographic Database (STATSGO) user's
guide. Soil Conservation Service Publication No. 1492.
11. ESRI. 1992. ARC/INFO command references: Arc command ref-
erences. Redlands, CA: Environmental Systems Research Institute.
12. ESRI. 1992. ARC/INFO user's guide: AML user's guide. Red-
lands, CA: Environmental Systems Research Institute.
13. ESRI. 1992. ARC/INFO user's guide: Cell-based modeling with
GRID. Redlands, CA: Environmental Systems Research Institute.
73
-------
Verification of Contaminant Flow Estimation With GIS and Aerial Photography
Thomas M. Williams
Clemson University, Georgetown, South Carolina
Abstract
Estimation of contaminant movement in ground water
requires interpolation of data from sampling wells that
represent a very small sample of aquifer volume. Spatial
statistics and kriging provide the best unbiased estima-
tor of interpolated concentrations. Hurricane Hugo
provided an opportunity to compare these estimators
with actual forest mortality caused by saltwater inunda-
tion associated with the tidal surge. During the 9- to
15- month period after the hurricane, salt from the tidal
surge moved within the shallow water table aquifer,
causing widespread tree mortality on Hobcaw Forest in
eastern Georgetown County, South Carolina. A small
watershed (12 acres) was instrumented with 24 multi-
level sampling wells. Piezometric potential and samples
for salt concentration were collected for 12 months
(months 18 to 30 after the tidal surge). These data
produced three-dimensional estimations of flow direc-
tions and two-dimensional maps of chloride concentra-
tion. These maps led to the identification of important
heterogeneities in the water table aquifer. Apparently,
the infiltrated salt water moved to the bottom of the
aquifer (15 feet) and emerged, killing the forest, where
aquifer heterogeneity resulted in upward movements of
ground water.
Georgetown County implemented a geographic infor-
mation system (GIS) for tax mapping in 1988 and pre-
pared 1:400-scale orthophotographs of the entire county
with true ground accuracy of less than 5 feet. Color
infrared aerial photographs were taken from a Cessna 150
platform annually after the hurricane. ERDAS GIS soft-
ware and the accurate photo base allowed removal of
scale irregularities and distortion that resulted from using
a small aircraft. Scanned images, using a 10-square-foot
pixel, were compared with kriged chloride concentration
maps, also using a 10-foot cell size. Grid cells with
estimated chloride concentration of more than 500 mil-
ligrams per liter also exhibited low reflectance in the
infrared-enhanced color band, indicating tree mortality.
Here, a small number of sampling wells accurately pre-
dicted ground-water movement of a contaminant (NaCI),
and GIS and remote sensing verified this movement.
Introduction
Estimating contaminant flow in ground water is difficult
because we cannot "see" the aquifer. We know that
aquifers comprise sediments that vary from place to
place, that changes in hydraulic conductivity determine
the rate of water movement, and that the spatial variabil-
ity of the aquifer sediment determines the hydraulic
conductivity. Our inability to accurately represent spatial
variability of the aquifer limits our ability to predict
ground-water flow and, thereby, contaminant transport.
A large variety of prediction models are available (1-3),
and stochastic methods of estimating spatial heteroge-
neity have been developed (4, 5) and tested (6, 7). On
well-characterized field sites, these techniques can pro-
duce predictions of tracer movements that accurately
predict experimental plumes in terms of mass behavior.
Even at these research sites, the spatial distribution of
hydraulic conductivity is not known well enough to pre-
dict behavior at any particular point.
Ground-water measurements generally derive from
wells that are single points. To understand movement of
an entire plume, these single point samples must be
extended to represent areas. The geostatistical ap-
proach allows quantitative estimation of the spatial vari-
ation of point estimates (8). Kriging is a technique that
uses spatial covariance to estimate values at points
where no measurement exists (9). It produces the best
linear unbiased estimator of nonmeasured points (8).
Following Hurricane Hugo, these techniques were used
to study saltwater movement in the water table aquifer
in forested stands of eastern Georgetown County, South
Carolina. Clemson University received a grant from the
U.S. Forest Service to examine forest mortality and
regeneration success within the forest zone covered by
salt water during the tidal surge. In this study, we used
a small sample of ground water to estimate the direction
74
-------
and concentration of salt moving in the aquifer. Geo-
graphic information systems (CIS) proved to be a useful
tool to verify conclusions based on the small sample
size. Onsite sampling, aerial photography, vector and
raster CIS, and spatial statistics were combined into one
analysis system. The system estimated and verified
directions of salt movement within the aquifer. CIS and
remote sensing of forest mortality produced an inde-
pendent indicator of salt movement that could be com-
pared with the geostatistical technique.
Problem Statement
The main goal of the research project was to evaluate
problems for forest regeneration in areas covered by salt
water during the hurricane. In many of these areas, the
mature trees died during the summer following the hur-
ricane. These areas have very low elevation, little relief,
and abundant rainfall, causing the water table to remain
near the soil surface. The hypothesis was that salt
movement within the aquifer killed the mature trees and
could limit regeneration success. We divided the prob-
lem into three tasks: to determine if salt concentrations
in the aquifer were high in areas where mature trees
died, to determine pathways of salt movement within the
aquifer that could explain high salt concentrations, and
to predict regeneration success from the pattern of salt
movement.
CIS contributed both to testing the initial hypothesis and
to extending predictions to areas not initially studied.
CIS has been used primarily to store and display spatial
data in a way that preserves and presents the spatial
relationships as well as the data. For this project, we
collected two dissimilar data types. To determine salt
movement, we measured salt concentrations and pie-
zometric pressures in a series of wells. CIS had to
represent the well data and the domain of the kriging
procedures in a coordinate system compatible with the
mortality data. We determined forest mortality data from
infrared-enhanced color aerial photography. CIS also
had to allow separation of the infrared signature of the
photography, transform the signature into data that were
comparable with the well data, and ensure that the
coordinate systems of the well data and the mortality
data represented the same true ground positions.
To use the CIS ability for this project, we needed to
choose several CIS parameters. In this case, we esti-
mated ground-water chloride concentration using
kriging, which produced data on a grid comparable with
mortality interpreted from photographs. A raster CIS
representation could be compared with individual grid
chloride values. Each grid cell was 10 square feet so
that each cell would be within a single tree crown.
Methods
Study Location
The study was located on 12 acres of a small watershed
located on the eastern side of Hobcaw Forest, an ex-
perimental forest managed by Clemson University, De-
partment of Forest Resources. Hobcaw Forest is
located on the end of a peninsula between the Winyah
Bay and the Atlantic Ocean in eastern Georgetown
County, South Carolina. The study watershed is located
immediately west of the salt marsh and barrier island
separating the forest from the Atlantic Ocean and is in
Pleistocene-aged beach sediment. Watershed divides
were created by former low dune lines, and the stream
is within a small depression between these former dune
lines. Divides are from 7 to 8 feet above sea level and
the stream from 4 to 6 feet above sea level.
The study watershed is 50 miles northeast of Char-
leston, South Carolina, where the eye of Hurricane
Hugo struck the U.S. coastline. Along this portion of the
South Carolina coast, the tidal surge was approximately
10 feet above mean sea level (10), covering the entire
watershed. Afterthe hurricane, shallow auger holes con-
tained water with sodium concentrations of 4,000 milli-
grams per liter (11). The hurricane winds did little
damage to the watershed forest, but 25 percent of the
large oaks were windthrown (12). Beginning in the
spring of 1990, however, many hardwoods and pines
began dying. By the winter of 1990 and 1991, a large
portion of the forest on the watershed had died. Tree
mortality did not correspond with high salinity measured
by the initial auger-hole method, suggesting movement
in the water table aquifer.
Well Installation
The water table aquifer is about 20 feet thick, consisting
of fine sand similar to the present beach, with thin beds
of shells 10 feet beneath the stream. The bottom of the
aquifer is a bed of clay up to 3 feet thick over a leaky
artesian aquifer composed of shell and sand. Local
rainfall recharges the water table aquifer. Recharge for
the lower aquifer is provided by leakage from the water
table aquifer beneath the center of the peninsula, about
2 miles west of the watershed, where land elevations
are 15 to 25 feet above sea level. Piezometric potential
in the lower aquifer is generally a few inches above the
water table aquifer, making it only weakly artesian (13).
We installed 24 multilevel ground-water samplers (14)
in the water table aquifer. Five samplers were located
in regeneration measurement plots (15) placed within
the stream. Two samplers, one at each edge of the
hardwood wetland, formed a line perpendicular to the
stream at the regeneration plot. Two more samplers
were located along these lines near the watershed di-
vides on each side of the watershed. The 24 samplers
75
-------
formed five transects across the stream (see Figure 1).
Piezometric potential and ground-water chloride con-
centrations were measured from these samplers from
March 31, 1991, through April 1, 1991. Williams (16)
provides a complete description of samplers, sampling
procedures, and laboratory analysis.
GIS Implementation and Measures
A GIS system for Hobcaw Forest management was
developed in 1987 using Environmental Systems Re-
search Institute's PC ARC/INFO software (17). The in-
itial system consisted of forest stand boundaries
digitized onto 1:100,000 digital line graphs (DLGs) pur-
chased from the U.S. Geological Survey. These rela-
tively crude maps were combined with stand records
and used for management decisions that did not require
exact locations of stand boundaries. Later, management
of the endangered red-cockaded woodpecker's habitat
required that mapped stand lines be closer to true
ground locations than the original DIG data scale al-
lowed. A program of ground surveys and aerial photog-
raphy was conducted in the late 1980s to locate stand
boundaries more accurately (18).
In 1988, Georgetown County began a program to con-
vert county tax mapping to computer-based systems.
The first step was to acquire survey grade orthopho-
tography. Copies of 1:400-scale orthophotographs, with
guaranteed ground accuracy of plus or minus 5 feet,
Location of Sample Points
554740
554576 -
554412 _|
D)
C
!c
t
554248
554084 -
553920
Transect 5
Transect 4
Transect 3
Transect 2
Transect 1
2548470 2548634 2548798 2548962
Easting (feet) SC State Plane 1983 Datum
Figure 1. Position of sampling wells in rectangle defined for
estimation of salinity movements.
became available in 1990. Roads and stand boundaries
were digitized from these photographs into the PC
ARC/INFO database. The new, accurate map was com-
bined with stand records of previous coverages to create
stand record coverages on a map that was true to the
ground within 5 feet, plus or minus 0.3 percent.
In 1991, the GIS programs ERDAS VGA and LIVE LINK
were obtained. ERDAS VGA programs allow image
processing and raster GIS to be done on personal com-
puters with VGA and some Super VGA monitor adapt-
ers. The LIVE LINK program allows display of both
ARC/INFO and ERDAS images on the same monitor
screen. Orthophotographs were scanned using 5-
square-foot pixels and rectified with less than 1-pixel
mean error, giving accurate ground locations of plus or
minus 8 feet.
Ground surveys from a nearby benchmark provided accu-
rate locations of the sampling wells. PC-TRAVERSE per-
formed coordinate geometry from the survey notes, which
was then plotted on the ARC/INFO forest stand data-
base. Accuracy was checked by plotting the recogniz-
able points on the survey with the scanned
orthophotograph using the LIVE LINK software.
In February 1991, the Hobcaw Forest was photo-
graphed with infrared-enhanced color film. In this film,
the red layer is sensitive to near infrared radiation that
is strongly reflected by chlorophyll. Red colors in result-
ing prints indicate living vegetation. The color photogra-
phy was not corrected for scale variation (from small
fluctuation in aircraft altitude) or for distortion (caused by
slight variations in the aircraft attitude). One photograph
(1:1,320 scale) covering the study watershed was
scanned into the ERDAS program. This image was
rectified to the 1988 orthophotograph image using con-
trol points visible on both. A 10-square-foot pixel was
used to sample individual tree crowns. The mean error
of rectification was 1.5 pixels for a ground location, plus
or minus 15 feet.
Ground-water chloride values at any one point varied
over three orders of magnitude in all three dimensions
and over two orders of magnitude with time. Annual
averages of piezometric potential and chloride concen-
tration, however, yielded interpretable results that were
also statistically significant (16). Averaged values could
then be combined with the surveyed sampler locations
in the ARC/INFO system. The GIS also calculated the
corner coordinates for a rectangle that would include all
the sampler locations.
Geostatistical calculations were performed using the
GS+ software. The data input to this program is an ASCII
file of sample point locations in x,y coordinates and data
values. The program allows calculation of semivario-
grams with various combinations of active and maxi-
mum lag distance and fitting of various model types to
76
-------
best fit the semivariogram (19). An active lag of 65 feet
and a Gausian model produced the best fit:
i(h) = 0.001 + 1.337 (1 - exp [-h2/20736]), r2 = 0.616
where i(h) is the semivariance at lag distance h.
This best fit model was then used in a block kriging (9)
procedure. The procedure used eight nearest neighbors
and calculated average values for 10- by 10-square-foot
blocks. Values were calculated within the rectangle de-
fined by the corner coordinates from the ARC/INFO
procedure described above.
Finally, the rectangle defined around the sampler posi-
tion formed the region of comparison between the rate
of mortality, as sampled by infrared reflection, and esti-
mated average chloride concentrations. The first com-
parison involved mapping reflection in the red band of
the aerial photograph as a gray-scale map and compar-
ing it with the contoured map of chloride concentration.
The rectangle coordinates were used in the ERDAS
software to create a subset of the scanned aerial pho-
tograph that included only the red band in the 5,145
pixels defined by the rectangle surrounding the sam-
plers. In this subset, the infrared reflection was scaled
as a gray-scale value between 0 and 255 for each
10- by 10-square-foot block defined in the concentration
map. In addition to mapping, a regression of chloride
concentration to gray-scale value was performed using
the individual blocks.
Results
Ground Water
Ground-water chloride data reflected a consistent expla-
nation of salt movement. Initial auger-hole data col-
lected within a month of the hurricane indicated most
salt was near the surface of the pine ridges, where,
presumably, salt water had filled the aquifer to the soil
surface during the hurricane. Data collected 30 months
after the hurricane indicated the bulk of the salt had
moved to the bottom of the aquifer underthe pine ridges.
Figures 2 and 3 represent the most significant results
interpreted from the piezometric potential and chloride
concentration measurements.
Figure 2 represents a cross section of the chloride con-
centrations and directions of ground-water flow in tran-
sect 4, the second most northern transect. The common
information represented in this cross section was the
west to east movement of ground water, representing
the regional flow toward the forest edge. Also, there is
an area of upwelling just east of the stream at the bottom
of the aquifer, representing a leaky spot in the underlying
clay layer. Upwelling causes the west to east stream-
lines to rise toward the surface along the western edge
of the wetland. Chloride concentrations indicate large
Transect 4
Pine
Chloride Concentration
[] < 50 Milligrams per Liter 500 to 750 Milligrams per Liter
PI 50 to 100 Milligrams pel
f Liter
100 to 250 Milligrams per Liter
750 to 1,000 Milligrams per Liter
1,000 to 2,000 Milligrams per Liter
[>] 250 to 500 Milligrams per Liter M > 2,000 Milligrams per Liter
Figure 2. Cross section of aquifer at transect 4; gray scale rep-
resents chloride concentration, boxes at sampler po-
sitions represent 95-percent confidence limits of
average chloride concentration in same gray scale,
and arrows are perpendicular to contours of pie-
zometric potential and represent two-dimensional
vectors of streamlines.
reservoirs of salt beneath each of the pine ridges and
small pockets of fresher water near the surface, prob-
ably the result of rain infiltration during the 30 months
since the hurricane. The water upwelling from the arte-
sian aquifer was consistently fresh. Flow passing
through the concentrated zone beneath the western
ridge was pushed to the surface beneath the stream,
where evaporation caused chloride concentrations to
average more than 1,000 milligrams per liter.
Figure 3 represents two plan views of the site at depths
of 4 and 12 feet below the surface. The 12-foot plan view
indicated that the east to west flow in transect 4 is only
the east vector of a southeast flow. Other areas of
upwelling exist beneath the stream. Chloride concentra-
tions are highest beneath both ridges and lowest in the
stream center. At the 4-foot depth, the east and south-
east flows are also obvious. Also, as deeper flows from
the western ridge are turned to the surface, high con-
centrations of chloride are present near the surface.
High concentrations within the wetland result from water
being carried to the surface due to upwelling within the
wetland.
GIS Evaluations
The ground-water interpretations show a consistent ex-
planation of salt movement. These interpretations are
77
-------
Plan View 12-Foot Depth
Plan View 4-Foot Depth
Figure 3. Plan views of aquifer at depths of 4 and 12 feet below the surface with same chloride scale and two-dimensional plan
vectors of streamlines.
based on only 24 sample points. Interpolation was linear
using the nearest neighbor. The samples removed from
the aquifer represent only 0.000014 percent of the aqui-
fer volume. Interpretation of a three-dimensional flow
regimen from such small sampling does not produce
great confidence in the validity of the interpretation.
Salt at the 4-foot depth is most likely to interact with tree
roots, and concentrations at this depth were used for
kriging. Kriged results, mapped in the same manner as
the aerial photography, show general agreement of high
mortality and predicted chloride concentrations over 500
milligrams per liter (see Figure 4). A regression of chlo-
ride concentration (chloride) to gray-scale value (G) for
the individual points yielded a significant negative cor-
relation. The regression line G = 169-0.12 (chloride)
explained only 27 percent of the variation in gray-scale
value, however.
Conclusions
CIS was successfully used to verify interpretations of
ground-water flow and salt movements in a shallow
water table aquifer. A variety of computer software com-
bined to create a system of analysis that allowed inte-
gration of field sampling, aerial photography, vector and
raster CIS, and spatial statistics. Using this system, we
compared chloride movement, measured by subsurface
samplers, with remotely sensed tree mortality caused by
soil salinity. The overall pattern of mortality was pre-
dicted by a 500-milligram-per-liter chloride contour esti-
mated by kriging averaged concentrations. Estimation
of mortality on a single tree basis was less successful,
with a regression of chloride to infrared reflection ex-
plaining only 27 percent of the variation in reflection. The
regression did not fit values of high reflection well but
did predict reflection values of 100 or below (regions of
high mortality) for concentrations above 500 milligrams
per liter.
The most important factor in the success of this project
was the availability of large-scale orthophotography.
Georgetown County's investment in accurate mapping
allowed creation of a map base that made scale correc-
tion of less costly aerial photography possible. Without
assurance that pixels on the aerial photograph corre-
sponded to the same locations as the subsurface sam-
plers, correlations would have been meaningless.
Another factor that contributed greatly to the research
was the fact that most of the computer software ex-
ported or imported data from simple ASCII files. The
standard (x coordinate, y coordinate, data value) format
in ASCII allowed files to be manipulated with spreadsheets
78
-------
Kriged Chloride Concentrations
Forest Mortality (Dark)
554740
554576 -
554412
554248
554084 -
553920
2548470 2548634 2548798 2548962
Easting (feet)
554740
554576
554412
554248
554084
553920
2548470 2548634 2548798 2548962
Easting (feet)
<200
<500
<700
<1,000
<1,200
Figure 4. Plan view of chloride concentrations from krig analysis and gray scale of infrared reflection from aerial photograph. Lighter
tones represent greater infrared reflection and less forest mortality.
or word processors. Creation of headers, positioning of
columns, or changing order of rows or columns could be
done for import into the next program. Although more
difficult than point and click file transfers of the modern
software, the simple standard format creates freedom to
use the software in ways not anticipated by the software
developers.
Finally, a clear problem statement aided in selecting the
most applicable CIS techniques. CIS software allows
several methods of data representation. In this example,
we chose a raster representation with a cell the size of
a tree crown. Criteria for choosing these parameters
included physical dimensions of the phenomenon of
interest, dimensions of CIS accuracy, and a desire for
automated determination of values for individual com-
parisons. A careful review of the problem to be solved,
data available, and capabilities of the CIS software are
all necessary ingredients for a useful problem statement.
References
1. Van Genutchen, M.T., and W.J. Alves. 1982. Analytical solutions
of the one-dimensional convective-dispersive solute transport
equation. U.S. Department of Agriculture Technical Bulletin 1661.
2. Wang, H.F., and M.P. Anderson. 1982. Introduction to ground-
water modeling: Finite difference and finite element methods. San
Francisco, CA: W.H. Freeman.
3. Bear, J., and A. Verruijt. 1987. Modeling ground-water flow and
pollution. Dordrecht, The Netherlands: D. Reidel Publishing Co.
4. Gelhar, L.W. 1993. Stochastic subsurface hydrology. Englewood
Cliffs, NJ: Prentice Hall.
5. Dagan, G. 1987. Theory of solute transport by ground water. Ann.
Rev. Fluid Mech. 19:83-215.
6. Garabedian, S.P., D.R. LeBlanc, L.W. Gelhar, and M.A. Celia.
1991. Large-scale natural gradient tracer test in sand and gravel,
Cape Cod, Massachusetts, 2. Analysis of spatial moments for a
nonreactive tracer. Water Resour. Res. 19(1): 1,387-1,397.
7. Rehfield, K.R., J.M. Boggs, and L.W. Gelhar. 1992. Field study
of dispersion in a heterogeneous aquifer, 3. Geostatistical analy-
sis of hydraulic conductivity. Water Resour. Res. 28(12):3,309-
3,324.
8. de Marsily, G. 1986. Quantitative hydrogeology. San Diego, CA:
Academic Press.
9. Krige, D.G. 1966. Two-dimensional weighted moving average
trend surfaces for ore-evaluation. J. South African Instit. Mining
and Metallurgy 66:13-38.
10. Coch, N.K., and M.P.Wolfe. 1991. Effect of Hurricane Hugo storm
surge in coastal South Carolina. J. Coastal Res. SI 8:201-208.
11. Gardner, L.R., W.K Michner, E.R. Blood, T.M. Williams, D.J. Lip-
scomb, and W.H. Jefferson. 1991. Ecological impact of Hurricane
Hugosalinization of a coastal forest. J. Coastal Res. SI 8:301-318.
12. Gresham, C.A., T.M. Williams, and D.J. Lipscomb. 1991. Hurri-
cane Hugo wind damage to southeastern U.S. coastal tree spe-
cies. Biotropica 23:420-426.
13. Nittrouer, PL. 1988. Ground-water flow patterns in a forested
beach ridge-scale system adjacent to a South Carolina salt
marsh. Master's thesis, Department of Geological Sciences, Uni-
versity of South Carolina.
79
-------
14. Williams, T.M., and J.C. McCarthy. 1991. Field-scale tests of col-
loid-facilitated transport. In: National Research and Development
Conference on the Control of Hazardous Materials, Anaheim,
California. Greenbelt, MD: Hazardous Materials Control Re-
search Institute.
15. Conner, W.H. 1993. Artificial regeneration of bald cypress in three
South Carolina forested wetland areas after Hurricane Hugo. In:
Brissette, J.C., ed. Proceedings of the Seventh Southern Silvicul-
tural Research Conference. U.S. Department of Agriculture For-
est Service, General Technical Report SO-93. Southern Forest
Experiment Station, Mobile, AL.
16. Williams, T.M. 1993. Salt water movement within the water table
aquifer following Hurricane Hugo. In: Brissette, J.C., ed. Proceed-
ings of the Seventh Southern Silvicultural Research Conference.
U.S. Department of Agriculture Forest Service, General Technical
Report SO-93. Southern Forest Experiment Station, Mobile, AL.
17. Lipscomb D.J., and T.M. Williams. 1988. A low-cost, map-based
data retrieval system for forest managers. In: Proceedings of the
American Congress of Surveying and Mapping/American Society
for Photogrammetry and Remote Sensing, Fall Convention. Falls
Church, VA: American Congress of Surveying and Mapping.
18. Lipscomb, D.J., and T.M. Williams. 1990. Developing a CIS for
forest management in the 1990s. In: Resource technology 90:
Second international symposium on advanced technology in
natural resource management. Bethesda, MD: American Society
of Photogrammetry and Remote Sensing.
19. Rossi, R.E., D.J. Mulla, A.G. Journel, and E.H. Franz. 1992.
Geostatistical tools for modeling and interpreting ecological spa-
tial dependence. Ecol. Monographs 62:277-314.
80
-------
Geology of Will and Southern Cook Counties, Illinois
Edward Caldwell Smith
Illinois State Geological Survey, Champaign, Illinois
Introduction
The Silurian dolomite aquifer is the primary source of
ground water in northeastern Illinois. It is overlain by
glacially derived sands and gravels or tills. The sands
and gravels within the glacial drift hydrologically interact
with the fractured and creviced dolomite bedrock.
The purpose of this study was to define the extent of
major glacial drift aquifers and their relationship to the
shallow bedrock aquifer surface. The study succeeded
in identifying two principal sand and gravel aquifers: an
"upper" drift aquifer within the glacial tills and a "basal"
drift aquifer overlying the bedrock. Bedrock topography,
drift thickness, thickness of the Silurian dolomite, and
thickness of major sand and gravel units were mapped
to help define the geologic and hydrologic system and
the interaction of the upper bedrock aquifer and the drift
aquifers.
The data collected to create the various maps came
from well records, engineering borings, oil and gas tests,
and structure tests on file at the Illinois State Geological
Survey (ISGS). Reviewing published reports, manu-
scripts, and unpublished reports on open file at the ISGS
provided an overall perspective of the geology of the
study area. Previously, no detailed studies of the hydro-
geology of the entire area had been conducted. Incor-
porating water well and other data into a computer
database greatly facilitated map construction. Prelimi-
nary maps were developed using Interactive Surface
Modeling (ISM) software and a geographic information
system (CIS).
Past regional geologic studies of the northeastern Illi-
nois area that have encompassed this study area in-
clude Thwaites (1), Bretz (2), Bergstrom et al. (3), Bretz
(4), Suter et al. (5), Hughes et al. (6), and Willman (7).
Bogner (8) and Larsen (9) included interpretive maps of
the surficial geology of the area as a part of planning
studies for northeastern Illinois.
Map Construction
Creating the database used in the construction of the
maps for this project entailed inputting information from
well driller's logs into a PC-based computerized spread-
sheet (Quattro Pro). Well logs were primarily from water
wells and engineering borings. Data items input into the
spreadsheet included:
Well identification (ID) number
Owner name
Location of well
Thickness of drift
Depth to top and bottom of the bedrock
Depth to top and bottom of each sand unit
The ground surface elevation of each well was interpo-
lated from United States Geological Survey (USGS)
7.5-minute quadrangles. Elevations of the top of bed-
rock and top and bottom of sand bodies were calculated
based on the interpolated elevations. Locations were
verified wherever possible using plat books by matching
either landowner names or the address location from the
well log. After compilation, the data were converted to
ASCII text and transferred into an ARC/INFO (Versions
5.0.1 and 6.0) database on a SUN SPARC workstation.
ARC/INFO is a product of Environmental Systems Re-
search Institute, Inc., of Redlands, California.
Of the more than 10,000 records reviewed for this pro-
ject, over 5,100 were input into the database. Sub-
sequently, numerous data quality checks ensured that
duplicate well ID numbers were corrected, locations
were corrected, thicknesses were checked so that the
sand thickness data reported did not exceed drift thick-
ness, and elevations were checked so that elevation of
a sand body was not below the bedrock surface. After
running the data quality checks and removing questionable
data from the database, approximately 5,000 records
remained.
ISM, a contouring package from Dynamic Graphics,
Inc., of Alameda, California, helped to create two di-
mensional grid representations of:
Surface topography
Drift thickness
81
-------
Bedrock topography
Bedrock isopach
Intermediate sand body isopach
Basal sand isopach
ISM also allowed for the creation of contoured output of
the grids. Grids are regularly spaced rectangular arrays
of data points (nodes) that allow for efficient mathemati-
cal calculations and contouring. ISM uses a minimum
tension gridding technique, allowing for the curvature
(change in slope) of the surface to be spread throughout
the surface rather than being concentrated at the input
data points. The ISM program uses a biharmonic in-
verse cubic spline function (algorithm) to assign data
values to grid nodes. This function assumes that for any
grid node assignment, input data points farther away
from the node being evaluated have less influence on
that node's value than nearer data points. To determine
each grid node value, ISM calculates an average value
from the surrounding scattered input data (up to 15 input
data points) and finds the standard deviation. ISM con-
tinues to refine the values of the grid nodes until the
standard deviation is minimized (10).
Several grid spacings were reviewed to determine which
would best represent the density of the data. The grids
that ISM uses, as described above, determine the fine-
ness to which the data control the resultant contours.
Experimentation was necessary to determine a grid
spacing that adequately represented the data. Too fine
a grid spacing can exaggerate or overly weight individ-
ual points, causing the resultant contours to be overly
jagged. With too large of a grid spacing, the contours
can become overgeneralized and become much less
data dependent because the calculated grids are
ove rave raged.
The two-dimensional grid of the land surface topography
was based on surface topography lines and spot eleva-
tions digitized from USGS 7.5-minute quadrangles. The
linework for each quadrangle was converted to ASCII
files of data points. The ASCII files contained x and y
coordinates and the elevation value of each data
point. After inputting the ASCII files into ISM, a
two-dimensional grid for each quadrangle was created.
ISM also generated contour lines from each grid. Com-
paring plots of the generated lines with USGS 7.5-minute
topographic maps allowed for the correction of errors and
ensured that the grid elevation values were within 10
feet of the elevations shown on the USGS maps. An ISM
two-dimensional grid of the entire area's surface topog-
raphy was created by combining the grids. After creating
a contoured surface of the grid, an ARC/INFO coverage
of the output was produced. ARC/INFO was used to edit
the coverage and produce the final map.
The two-dimensional grid of the bedrock surface topog-
raphy was based on data from water well and engineer-
ing boring logs, ISGS field observations of outcrop
locations, and previous ISGS mapping (9). An ASCII file
of x and y coordinates and the elevation of each bedrock
top was input into ISM. Subtracting the bedrock topog-
raphy grid from the land surface grid produced a grid of
the drift thickness. A contoured output of the grid was
produced, and an ARC/INFO coverage of the output was
created. Again, ARC/INFO was used to edit the cover-
age and produce the final map.
Creating the isopach maps entailed subtracting the top
and bottom elevations of each unit to calculate the
thickness of each unit. ASCII files of the xand y coordi-
nates and the thickness values for each data point were
input into ISM. ISM then created two-dimensional grids
of each isopach. Contoured output of each grid was
produced, which allowed for the creation of ARC/INFO
cove rages of the output. ARC/INFO was used to edit the
coverages and produce the final maps.
Bedrock Geology of the Study Area
All the sedimentary bedrock units are of the Paleozoic
Era. The Paleozoic bedrock comprises sequences of
sandstones, dolomites, limestones, and shales. The
stratigraphic column of Figure 1 illustrates the vertical
succession of the bedrock. Major tectonic activity of the
area includes the formation of the Kankakee Arch in
Ordovician time (11) and faulting along the Sandwich
Fault Zone. Faulting along the Sandwich Fault Zone
(see Figure 2) may have occurred coincidentally with the
formation of the Lasalle Anticlinorum in early Pennsyl-
vanian time (12). No further faulting has been noted
since deposition of glacial sediments. Bedrock units
gently dip to the east (7). The majority of the area lies
on the Niagara cuesta, a south and west facing scarp
that comprises the resistant Silurian strata that have an
eastward dip of roughly 15 feet per mile (13). The Silu-
rian strata are absent west of the Kankakee River as
well as in an area west of the Des Plaines River in
west-central Will County (see Figure 2). This study re-
lates to the hydrogeology of the Silurian strata and the
drift materials, and details only the uppermost bedrock
units. The report, however, does briefly summarize units
below the Maquoketa Group using information from
Hughes et al. (6) and Visockey et al. (14).
Precambrian Bedrock
Granites or granitic rock compose the Precambrian
basement of northern Illinois. Few details about the
nature of the basement rocks are known because few
wells have completely penetrated the sedimentary bed-
rock of the region. The elevation of the top of the Pre-
cambrian basement probably stands at 4,000 feet below
mean sea level in the study area.
82
-------
SYSTEM
QUATER-
NARY
PENNSYL-
VANIAN
z
<
(r
^
to
ORDOVICIAN
CAMBRIAN
PHE-
CAMBRIAN
SERIES
PLEISTOCENE
DESMOINESIAN
NIAGAHAN
ALEXANDRIAN
CINCINNATIAN
CHAMPLAINIAN
CANADIAN
CROIXAN
GROUP OR
FORMATION
Spoon and
Carbondale
Racine
Sugar Run
Joliet
Kankakee
Elwood
Wilhelmi
Maquoketa
Galena
Platteville
Glenwood
St. Peter
Shakopee
New
Richmond
Oneota
Gunter
Eminence
Potosi
Franconia
Iron ton
Galesville
Eau Claire
Elmhurst
Member
Mt. Simon
AQUIFER
Sands
and
Gravels
Silurian
Galena-
Platteville
Glenwood-
St. Peter
Prairie du
Chien
Eminence
Potosi
Franconia
Ironton-
Galesville
omite aquifer system
8
S
I
\
\
\
\
\
brian-Ordovician aquifer system
$
Elmhurst-
Mt. Simon
aquifer
system
LOG
<.:,'.>: x
S&t
a O "o D". »
&S&S
*^"V
A
'£
WW
rWi
f
i\
/
/
/
/.
/
/
f
/-
~T
7-L
-AZ
TA
-7
7
L / T^fl /
:£^sn/»
^\f"i
/-
/
^^«/
17 _zr
i ~i '
/ /
/ / i
. . A. A,...
'/"VV
7 /
< i "V
1 1 A' 1 '
* / /
':s:':''-~^/:.
^ '''
/ -'
/ .
YA
/ : .
/ /
/. .
. / *
/ '
/ ...
/ .
/ -A-
ilfcA-
4%^
fr
i
/
f~
/A/
/
/ /
/
/:
/
/
/ ..
/
7-'-
/
|r-h'%s'^.-
W&&
£!TiiT* *-~ * r\ "*"
-/. SY* -/*--{
/ ~~/1m /, ~ /,'t
* * * _ *f fj^t
t * * *^^ ~
^^^\
x^/X^/
THICKNESS
(FT)
0-260
0- 110
- reef
0 350
0-100
80-250
310-380
125-600
0-410
0-280
110- 160
390 - 570
2200
DESCRIPTION
Unconsolidated glacial deposits-pebbly clay
(till), silt, sand and gravel
Alluvial silts and sands along streams
Shale, sandstone, clay, limestone, and coal
Dolomite, very pure to argillaceous, silly,
cherty; reefs in upper part
Dolomite, slightly argillaceous and stlty
Dolomite, very pure to shaly and shale,
dolomitic; white, light gray, green, pink,
maroon
Dolomite, pure lop 1 ' - 2', thin green shale
partings, base glauconitic
Dolomite, slightly argillaceous, abundant
layered white chert
Dolomite, gray, argillaceous and becomes
dolomitic shale at base
Shale, red to maroon, oolites
Shale, silty, dolomitic, greenish gray, weak
(Upper unit)
Dolomite and limestone, white, light gray,
interbedded shale (Middle unit)
Shale, dolomitic, brown, gray (Lower unit)
Dolomite, and/or limestone, cherty (Lower
part)
Dolomite, shale partings, speckled
Dolomite and/or limestone, cherty, sandy
at base
Sandstone, fine anH coarse-grained; little
dolomite; shale at top
Sandstone, fine to medium-grained; locally
cherty red shale at base
Dolomite, sandy, cherty (oolitic); sandstone
Sandstone interbedded with dolomite
Dolomite, white to pink, coarse-grained
cherty (oolitic)
Sandstone, medium-grained, slightly dolomitic
Dolomite, light colored, sandy, thin sandstones
Dolomite, fine-grained gray to brown,
drusy quartz
Dolomite, sandstone and shale, glauconitic,
green to red, micaceous
Sandstone, fine to coerse- grained, well
sorted; upper part dolomitic
Shale and siltstone, dolomitic, glauconitic;
sandstone, dclomitic. glauconitic
Sandstone, coarse-grained, white, red in lower
half; lenses of shale and siltstone, red,
micaceous
Granitic rocks
Figure 1. Generalized stratigraphic column of rock units and aquifers in northern Illinois (prepared by M.L. Sargent, ISGS).
83
-------
Cambrian
The Elmhurst-Mt. Simon Sandstone comprises the old-
est sedimentary units in Illinois and consists of medium-
grained sandstones. It has a total thickness of
approximately 2,500 feet. The upper part of this unit has
acted as an aquifer in the Chicago region in the past;
ground-water mining of the aquifer (a nonreplenished
lowering of the static water level), however, has led to a
discontinuation of its use for that purpose. The Eau
Claire Formation, the Basal Sandstone Confining Unit
(14), consists of dolomitic shale and siltstone with thin
beds of sandstone. It has a thickness of 300 feet to 400
feet and separates the Elmhurst-Mt. Simon aquifer from
the Ironton-Galesville Sandstones. The Ironton-Galesville
Sandstones have a thickness of 150 feet to 250 feet and
serve as a source of ground water in northern Illinois (6).
The Galesville Sandstone is fine-grained, while the Iron-
ton Sandstone is coarser grained and contains more
dolomite. The Knox Megagroup, the Middle Confining
Unit (14), comprises all the bedrock units between the
Ironton-Galesville Sandstones and the Ancell Group. It
includes the:
Cambrian Franconia Formation
Potosi Dolomite
Eminence Formation
Jordan Sandstone
Ordovician Prairie du Chien Group
The Knox Megagroup is primarily dolomitic in compo-
sition, though it contains thin sandstones. Its thick-
ness ranges from 400 feet in the northern portion of
the study area to about 700 feet in the southernmost
tip of Will County. The sandstones tend to be some-
what discontinuous and, where present, offer a local-
ized source of ground water. The group as a whole
acts as a confining unit between the Ironton Sand-
stone and the Ancell Group.
Ordovician
The Ancell Group, which contains the St. Peter Sand-
stone and Glenwood Sandstone, has a thickness of
roughly 200 feet throughout the study area except in
north-central Will County where it is over 400 feet. The
thickness of the Ancell Group varies considerably in
northern Illinois because it rests on an erosion surface.
The Ancell Group is the shallowest aquifer present in
this area below the Silurian dolomite aquifer. The eleva-
tions of the top of the Ancell Group range from just over
sea level in the northwest corner of Will County to 500
feet below mean sea level in the southwestern corner.
The Galena and Platteville Groups provide a sequence
of carbonate rocks that are primarily dolomitic in com-
position. The Platteville Group conformably overlies the
Ancell Group. The two units have a combined thickness
of 350 feet throughout this part of the state. The Galena
and Platteville Groups, combined with the overlying
Maquoketa Shale Group, act as an aquitard between the
Ancell aquifer and the Silurian dolomite aquifer.
Maquoketa Shale Group
The study area has three subaerially exposed bedrock
units. The oldest of these that this report details are
Ordovician-aged strata comprising the Cincinnatian Se-
ries Maquoketa Shale Group. The thickness of the
Maquoketa Group ranges from 260 feet in eastern Will
County to 120 feet in the northwestern corner of Will
County and is unconformably overlain by Silurian strata
(15). The Maquoketa Group comprises four formations:
Scales Shale
Fort Atkinson Limestone
Brainard Shale
Neda Formation
The Scales Shale forms the lowermost unit and consists
of gray to brown dolomitic shale. Thin layers with phos-
phatic nodules and pyritic fossils occur near the top and
base of the unit. The Scales Shale may attain a thickness
of up to 120 feet in this region (15). The Fort Atkinson
Limestone, a coarse-grained crinoidal limestone to fine-
grained dolomite, may range up to 60 feet thick (15). The
Brainard Shale comprises greenish gray dolomitic shale
and has a thickness of generally less than 100 feet (16).
The Neda Formation, the youngest formation in the
Maquoketa Group, is relatively thin with a thickness of
usually less than 10 feet. In some places, it may attain a
maximum thickness of 15 feet. The Neda is exposed along
the Kankakee River, and the Silurian-aged Kankakee For-
mation typically overlies it. The Neda Formation consists
mostly of red and green shale with interbedded goethite
and hematite oolite beds (7, 16).
Silurian System
Silurian-aged rocks consist almost solely of dolomites
and dolomitic limestones. The Silurian is divided into the
Alexandrian and Niagaran Series. The Alexandrian Se-
ries is about 25 feet thick and is represented by the
Kankakee, Elwood, and Wilhelmi Formations. These for-
mations are a fine- to medium-grained, white, gray to
pinkish gray dolomite. The Kankakee Formation is exposed
along the Kankakee River in southern Will County (17).
The Niagaran Series comprises much of the bedrock
surface of this area and includes three formations. The
Joliet Formation has a lower member of dolomite with
interbedded red and green shale, and two upper mem-
bers with an increasing purity of dolomite toward the
top of the formation (7). The Sugar Run Formation,
formerly termed the Waukesha Formation (17), is an
argillaceous, fine-grained, medium- to thick-bedded,
84
-------
brownish gray dolomite (7). The Racine Formation is the
thickest unit in the Niagaran Series, attaining a thick-
ness of as much as 300 feet (17). The Racine Formation
contains large reefs that are as high as 100 feet and
consist of vugular gray dolomite. The inter-reef rock
consists of dense, cherty gray dolomite. The Racine
Formation is exposed in the bluffs along the Des Plaines
River from Joliet to Blue Island, Illinois (17).
Figure 2 is an isopach of the Silurian dolomite indicat-
ing the thickness of the unit in the study area and the
boundary of the Silurian rocks. The Silurian dolomite
aquifer has a maximum thickness of just over 500 feet
in the southeast corner of Will County and becomes
thicker to the east and south. It rapidly increases in
thickness from its margin along the western border of
Will County, where it has eroded. The contact be-
tween the Silurian dolomite and the underlying
Maquoketa Shale Group has relatively little relief.
Thus, the major differences in thickness of the unit
result from erosion of the bedrock surface. Joints and
fracture patterns within the upper bedrock have a domi-
nantly northwest-southeast and northeast-southwest
orientation (18).
Pennsylvanian System
Pennsylvanian-aged bedrock is found in the southeast-
ern portion of Will County west of the Kankakee River
with an outcropping at the confluence of the Des Plaines
and Kankakee Rivers. The lowermost unit, the Spoon
Formation, is very thin and consists of clay beds with
scattered occurrences of coal formed in channel-like
depressions (19). The Spoon Formation overlies the
Maquoketa Shale Group. The overlying Carbondale
Formation may attain a thickness of over 100 feet in the
southwestern corner of Will County. The Carbondale
Formation consists of shale with thin limestone beds.
The lowermost unit, the Colchester (Number 2) Coal
Member, outcrops in this area and attains a thickness of
up to 3 feet. It has been extensively mined along the
Will-Grundy-Kankakee County border where large ar-
eas of strip-mined land are evident. Most of the available
coal has been mined out, and numerous gob piles exist
R9E
- -
R14E
R15E
Scale of Miles
5 10
Legend
300"' Thickness in Feet
Contour Interval 50 Feet
. _-" Western Boundary of
Silurian Dolomite
***-* Fault
it
Figure 2. Thickness of the Silurian dolomite.
85
-------
in the area of Braidwood. The Francis Creek Shale
Member, which overlies the Number 2 coal, constitutes
the remainder of Pennsylvanian units in the study area.
The Francis Creek Shale is gray with numerous flat-
tened concretions that contain the Mazon Creek flora of
Pennsylvanian-aged fossils (19). Weathering of the
mine slag materials may have exposed fossiliferous
concretions in the gob piles (7).
Bedrock Topography
The highest bedrock elevations are in east-central Will
County where the bedrock rises to over 700 feet above
mean sea level (see Figure 3). Bedrock uplands occur
as a broken curved ridge from the southeast to the
northwest with bedrock elevations consistently rising
over 650 feet above mean sea level. The bedrock sur-
face slopes from the bedrock upland high westward to
the Des Plaines River. It also has a regional downward
slope to the south into Kankakee County, Illinois, to the
northeast into the Lake Michigan basin, and to the east
into Indiana. West of the Des Plaines River, the bedrock
surface rises to over 650 feet above mean sea level in
northeastern Kendall County. Elsewhere, the surface
has relatively low relief.
The dominant features of the bedrock surface are the
river valleys. The Des Plaines River valley is better
expressed than the Kankakee River valley. This is true,
in part, because it is older and acted as a drainageway
for glacial meltwater where it may have become en-
trenched in the present valley. The Kankakee River
valley may be less expressed partly because of the
amount of scouring that occurred over a large area
during the Kankakee flood event such that the river is
not entrenched in most places. Also, smoothing of the
study's contour maps has generalized some of the de-
tail.
The buried Hadley Bedrock valley, described initially by
Horberg and Emery (20), probably existed prior to
glaciation and concurrently with the preglacial Des
Plaines River. The valley may have acted as a drainage-
way for glacial meltwaters until the time that glacial
debris buried it. Glacial scouring was originally believed
to have formed the valley, but evidence presented by
McConnel indicated a fluvial origin of the valley (21).
Also, the base of the Hadley valley does not overhang
or lie much below the Des Plaines valley but rather joins
it at a smooth juncture.
The bedrock surface contains a number of sinkholes
or closed depressions that are expressions of karst
Legend
^o*" Bedrock Elevation in Feet
Above Mean Sea Level;
^ Contour Interval 25 Feet
^- Western Boundary of
x^ Silurian Dolomite Aquifer
s\ Bedrock Quarry
Figure 3. Topography of the bedrock in the study area.
86
-------
development that formed prior to continental glaciation.
Karst, a terrain developed on limestone or dolomite by
solution or dissolving of the rock, is characterized by
closed depressions and cavity development along joints
and fractures. Fischer (22) first noted karst features in
the Joliet area where early Pennsylvanian sediments of
shale and clay filled cavities in the upper bedrock.
Buschbach and Heim (23) indicated closed depressions
in the Silurian dolomite surface in their bedrock topog-
raphy map for the Chicago region. They speculated
these depressions were expressions of karst develop-
ment. McConnel (21) demonstrated the existence of
sinkholes in the area of the buried Hadley Bedrock
valley northeast of Joliet by using seismic refraction
survey data.
Glacial Geology
The sediments overlying the bedrock comprise tills,
sands and gravels, lacustrine deposits from glacial
lakes, and surficial eolian deposits of loess and sand.
The unconsolidated deposits are over 150 feet thick
along the crest of the Valparaiso Morainic System.
Figure 4, adapted from Willman (24), indicates the
principal moraines. In the area where the Hadley Bed-
rock valley is present, the deposits attain a thickness of
over 175 feet. Bedrock is mainly exposed along the Des
Plaines River valley and its tributaries. It is also exposed
in isolated areas in southeastern Cook County. The drift
thickness map (see Figure 5) indicates the distribution
of the earth materials overlying the bedrock and the
locations of bedrock outcrops. The bedrock outcrop in-
formation for this map was derived from Piskin (25) and
Berg and Kempton (26).
Erosion of the glacial sediments was a major factor in
controlling the drift thickness of the area. Succeeding
glaciers scraped off previously deposited sediments, but
glacial meltwaters, which came from the east and north
along the river channels, caused much of the erosion.
Both the Kankakee River and Des Plaines River acted
as meltwater channels as the glaciers melted. The Du
Page River acted as a minor drainageway and was most
active during large-scale flooding events. The thickness
of the drift varies in the area also because of the topo-
graphic control that the bedrock on the overlying sedi-
ments exercises. The crest of the Valparaiso moraines
coincides with the topographic high in the bedrock
End Moraines
Scale in Miles
0 10
Figure 4. End moraines (late Wisconsin) in Will and southern Cook Counties, Illinois.
87
-------
j
Legend
Drift Thickness in Feet; Contour
Interval 25 Feet Between 0 to 50
Feet and 50 Feet Between 50 to
150 Feet
^Sy Bedrock Outcrop Area
A Isolated Bedrock Outcrop
2^1 Strip-Mined Areas
Figure 5. Thickness of the glacial drift in the study area and bedrock outcrop information (25, 26).
surface. The cross sections in Figure 6 also show this.
The bedrock high may have caused late Woodfordian
glaciers to stall repeatedly in the same area, causing
moraines to build atop one another sequentially (27).
Descriptions by well drillers note few variations in the
character of the unconsolidated sediments; therefore,
we did not attempt to correlate these deposits. The drift
materials present in the study area are late Wisconsinan
or younger. Though this region experienced glaciating
repeatedly prior to the Wisconsinan glaciation, no Illi-
noisan or pre-lllinoisan deposits have been identified
(28). The drift units divide into three main units (29):
The Lemont Drift
The Yorkville Drift
The Wadsworth Drift
The three drift units are all part of the Wedron Formation
of Wisconsinan age. The Lemont Drift has a dolomitic
character because the source material for the diamicton
was glacially eroded Silurian dolomite. The Lemont Drift
is the oldest of the three units and is found only under-
lying the Wadsworth Drift. The Yorkville Drift is the only
drift unit present west of the Valparaiso Morainic System
boundary within the study area. It overlies the bedrock
surface wherever the basal sand unit is not present. The
Wadsworth Drift comprises silty and clayey diamictons
and is the youngest of the drifts (29). It overlies the
Lemont Drift and the upper sand unit. In the cross
sections (see Figure 6), where the upper sand unit is
present, it roughly indicates the boundary between the
Lemont and Wadsworth Drifts. The gradation between
the different drift units at the Valparaiso System bound-
ary is not well defined. The Wadsworth Drift appears to
grade into the Yorkville Drift because they are very
similar in composition near the boundary (9).
The large Kankakee flood left extensive deposits of
sand and gravel and lacustrine sediments along the
Kankakee River and Des Plaines River. The flood oc-
curred as glacial meltwaters built up behind a constric-
tion at the Marseilles Morainic System to the west (30).
Large glacial lakes, which developed during the flood,
subsequently emptied into the Illinois River valley after
a breach in the moraines developed. The force of the
flood waters eroded the glacial deposits along the river
valleys, flattened the surface of the drift, and, in places,
88
-------
North
South
800-1
Legend
Drift, Undifferentiated
Sand
Silurian Dolomite
Potentiometric Profile of
' Silurian Dolomite Aquifer
Vertical Exaggeration: 167x
450 J
-450
West
800! D R10E
rSOO
-750
Figure 6. Geologic cross sections of the glacial drift and potentiometric profile of the Silurian dolomite aquifer.
89
-------
exposed the underlying bedrock. The flood event formed
thin, dispersed lake plain deposits of silt, clay, and sand
in southwestern Will County. Some lacustrine deposits
lie between morainic ridges in southern Cook County
where small glacial lakes developed as the Valparaiso
Moraines were being deposited (8).
Figure 7 shows the locations of some of the surficial
materials. Sands and gravels were also deposited along
tributary creeks and in abandoned channels that once
connected the Du Page River and Des Plaines River
north of their present juncture. Wind has reworked the
surficial sand deposits forming low dunes along the
Kankakee River in southern Will County. Masters (31)
classified the sand and gravel deposits of the area by
their origin, indicating that most of the deposits present
in the valley of the Des Plaines River formed as well-
sorted valley train deposits. In the Kankakee River val-
ley, the sands and gravels were primarily deposited as
riverine sediments during the Kankakee flood event.
Sand and Gravel Isopachs
The sand and gravel isopach maps (see Figures 8 and
9) indicate the variations in thickness of the upper and
basal sand and gravel units. The most extensive depos-
its of both exist throughout the area overlain by the
Valparaiso Morainic System. This may be associated
with bedrock control on the formation of the moraines
and associated deposits referred to earlier.
The thickest deposits lie in the buried Hadley Bedrock
valley where thicknesses of both units can exceed over
100 feet. The upper sand unit may be found in the glacial
drift within a wide range of elevations. For mapping
purposes, we defined the upper sand unit as a sand unit
greater than 1 foot thick that occurs between two fine-
grained layers. The basal sand unit includes all coarse-
grained materials that overlie the bedrock surface. Most
of the basal sands present west of the Des Plaines River
were formed as valley train deposits along the river
T
32
N
Legend
Surficial Glacial Geology
Scale of Miles
5
10
n Equality Fm - Glacial Lake I
Plain Deposits I
pv-pa Erosional Meltwater Channel [
K&-F-I Cut Through Equality Fm
ffff, :j Henry Fm - Glacial Outwash f
*'«'-'* Sand and Gravel Deposits
Geologic Cross Section
'A*
Wendron Fm - Wadsworth Drift
- Tinley Moraine
Wedron Fm -Wadsworth Drift
- Valparaiso Morainic System
Wedron Fm - Yorkville Drift
- Minooka Morainic System
Silurian Dolomite
Strip-Mined Areas
Figure 7. Surficial glacial geology (23).
90
-------
R8E R 1QE
R13E RUE
°%N
-*v-
R15E
Scale of Miles
5 10
Legend
Thickness of Upper Sand
0 Isolated areas > 10 Feet
10 to 50 Feet
50 to 100 Feet
> 100 Feet
Figure 8. Thickness of the upper sand unit.
channels as the glaciers melted back. The origin of the
extensive deposits underlying the Valparaiso Moraines
is not clear. They may have been formed during early
Wisconsinan glacial events as outwash plain deposits
or they may have been deposited subglacially. The
cross sections (see Figure 6) can reveal the variability
and complexity of the sand and gravel layers as they
occur within the drift. The sand and gravel deposits very
seldomly act as aquifers in this region because almost
all wells are completed in the Silurian dolomite aquifer.
Clearly, Figures 8 and 9 indicate that some ground-
water resource potential may exist within these depos-
its.
References
1. Thwaites, FT. 1927. Stratigraphy and geologic structure of north-
ern Illinois. Illinois State Geological Survey Report of Investiga-
tions 13.
2. Bretz, J.H. 1939. Geology of the Chicago region, Part I: General.
Illinois State Geological Survey Bulletin 65.
3. Bergstrom, R.E., J.W. Foster, L.F. Selkregg, and W.A. Pryor.
1955. Ground-water possibilities in northeastern Illinois. Illinois
State Geological Survey Circular 198.
4. Bretz, J.H. 1955. Geology of the Chicago region, Part II: The
Pleistocene. Illinois State Geological Survey Bulletin 65.
5. Suter, M., R.E. Bergstrom, H.F. Smith, G.H. Emerich, W.C. Wal-
ton, and T.E. Larson. 1959. Preliminary report on the ground-
water resources of the Chicago region, Illinois. Illinois State
Geological Survey and Illinois State Water Survey Cooperative
Ground-Water Report 1.
6. Hughes, G.M., P. Kraatz, and R.A. Landon. 1966. Bedrock aqui-
fers of northeastern Illinois. Illinois State Geological Survey Cir-
cular 406.
7. Willman, H.B. 1971. Summary of the geology of the Chicago
area. Illinois State Geological Survey Circular 460.
8. Bogner, J.E. 1976. Geology for planning in northeastern Illinois,
Vol. V: Cook County. Illinois State Geological Survey Open-File
Report.
9. Larsen, J.I. 1976. Geology for planning in northeastern Illinois,
Part VI: Will County. Illinois State Geological Survey Open-File
Report.
91
-------
R9E
R10E
R15E
Scale of Miles
0 5 10
Legend
Basal Sand Thickness, Feet
OtolO
10 to 50
HI 50 to 100
I >100
Figure 9. Thickness of the basal sand unit.
10. Dynamic Graphics, Inc. 1991. Interactive surface modeling users
guide. Alameda, CA.
11. Ekblaw, G.E. 1938. Kankakee Arch in Illinois. Illinois State Geo-
logical Survey Circular 40.
12. Kolata, D.R., T.C. Buschbach, and J.D. Treworgy. 1978. The
Sandwich Fault Zone of northern Illinois. Illinois State Geological
Survey Circular 500.
13. Horberg, L. 1950. Bedrock topography in Illinois. Illinois State
Geological Survey Bulletin 73.
14. Visocky, A.P., M.G. Sherrill, and K. Cartwright. 1985. Geology,
hydrology, and water quality of the Cambrian and Ordovician
systems in northern Illinois. Illinois Department of Energy and
Natural Resources Cooperative Groundwater Report 10.
15. Kolata, D.R., and A.M. Graese. 1983. Lithostatigraphy and depo-
sitional environments of the Maquoketa group (Ordovician) in
northern Illinois. Illinois State Geological Survey Circular 528.
16. Willman, H.B., et al. 1975. Handbook of Illinois stratigraphy. Illi-
nois State Geological Survey Bulletin 95.
17. Willman, H.B. 1973. Rock stratigraphy of the Silurian system in
northeastern and northwestern Illinois. Illinois State Geological
Survey Circular 479.
18. Foote, G.R. 1982. Fracture analysis in northeastern Illinois and
northern Indiana. Master's thesis, University of Illinois.
19. Smith, W.H. 1968. Strippable coal reserves of Illinois, Part 6: La
Salle, Livingston, Grundy, Kankakee, Will, Putnam, and parts of
Bureau and Marshall Counties. Illinois State Geological Survey
Circular 419.
20. Horberg, L., and K.O. Emery. 1943. Buried bedrock valleys east
of Joliet and their relation to water supply. Illinois State Geological
Survey Circular 95.
21. McConnel, D.R. 1972. Bedrock topography and paleogeomor-
phology northeast of Joliet, Illinois. Master's thesis, University of
Illinois.
22. Fischer, D.J. 1925. Geology and mineral resources of the Joliet
quadrangle. Illinois State Geological Survey Bulletin 51.
23. Buschbach, T.C., and G.E. Heim. 1972. Preliminary geologic in-
vestigations of rock tunnel sites for flood and pollution control in
the greater Chicago area. Illinois State Geological Survey Envi-
ronmental Geology Notes 52.
24. Wilman, H.B., and J.C. Frye. 1970. Pleistocene stratigraphy of
Illinois. Illinois State Geological Survey Bulletin 94.
25. Piskin, K. 1975. Thickness of glacial drift in Illinois. Illinois State
Geological Survey 1:500,000 scale map.
92
-------
26. Berg, R.C., and J.P. Kempton. 1988. Stack-unit mapping of geo-
logic materials in Illinois to a depth of 15 meters. Illinois State
Geological Survey Circular 542.
27. Hansel, A.K., and W.H. Johnson. 1987. Ice marginal sedimenta-
tion in a late Wisconsinan end moraine complex, northeastern
Illinois, USA. In: van der Meer, J.J.M., ed. Tills and glaciotecton-
ics. Proceedings of an INQUA Symposium on genesis and lithol-
ogy of glacial deposits.
28. Horberg, L. 1953. Pleistocene deposits below the Wisconsinan
drift in northeastern Illinois. Illinois State Geological Survey Re-
port of Investigations 165.
29. Johnson, W.H., and A.K. Hansel. 1985. The Lemont section. In:
Johnson, W.H., A.K. Hansel, B.J. Socha, L.R. Follmer, and J.M.
Masters, eds. Depositional environments and correlation prob-
lems of the Wedron Formation (Wisconsinan) in northeastern
Illinois. Illinois State Geological Survey Guidebook 16.
30. Gross, D.L., and R.C. Berg. 1981. Geology of the Kankakee
River system in Kankakee County, Illinois. Illinois State Geologi-
cal Survey Environmental Geology Notes 92.
31. Masters, J.M. 1978. Sand and gravel resources in northeastern
Illinois. Illinois State Geological Survey Circular 503.
93
-------
Watershed Applications
-------
The Watershed Assessment Project: Tools for Regional
Problem Area Identification
Christine Adamus
St. Johns River Water Management District, Palatka, Florida
The St. Johns River Water Management District of Flor-
ida recently completed a major water resources plan-
ning effort. As part of this planning effort, the St. Johns
River Water Management District created a geographic
information systems (CIS) project called the Watershed
Assessment, which included a nonpoint source pollution
load model. This paper introduces the planning project
and the Watershed Assessment, and describes how the
results of the model are being used to guide water
management activities in northeast Florida.
Background
The St. Johns River Water Management District (Dis-
trict), one of five water management districts in Florida,
covers 12,600 square miles (see Figure 1). The St.
Johns River starts at the southern end of the District and
flows north; it enters the Atlantic Ocean east of the city
District Boundary
Major Drainage
Basin Boundaries
Figure 1. St. Johns River Water Management District, Florida.
of Jacksonville. The cities of Orlando, Daytona Beach,
and Jacksonville are partially or entirely within the Dis-
trict boundaries. Ad valorem taxes provide primary fund-
ing for the District.
The District boundaries are somewhat irregularly
shaped because Florida water management districts are
organized on hydrologic, not political, boundaries, which
greatly improves the District's ability to manage the
resources. On the north, the District shares the St.
Mary's River with the state of Georgia and on the south,
shares the Indian River Lagoon with another water man-
agement district. Most of the water bodies the District
manages, however, have drainage basins that are en-
tirely contained within the District's boundaries.
Water management districts in Florida have amassed
extensive CIS libraries, which they share with local and
statewide agencies. These libraries include basic data
layers such as detailed land use, soils, and drainage
basins. Districts also coordinate data collection and
management to ensure data compatibility.
District Water Management Plan
All activities and programs of the water management
districts are related to one or more of the following
responsibilities: water supply, flood protection, water
quality management, and natural systems management.
Each water management district recently completed a
district water management plan (Plan). The main pur-
pose of these Plans is to provide long-range guidance
for the resolution of water management issues. The
Florida Department of Environmental Protection will use
these five Plans as the basis for a state water manage-
ment plan. Each water management district used the
same format, which comprised the following components:
Resource assessment: What are the problems and
issues related to each of the four responsibilities
listed above?
97
-------
Options evaluation: What options are available for
addressing the problems?
Water management policies: What existing District
policies influence the decisions that must be made?
Implementation strategy: What is the best plan for
addressing the problems?
The Watershed Assessment Project
The District created the Watershed Assessment project
as part of is resource assessment. This CIS project
examines the entire District to identify problems related
to flood protection, ecosystems protection, and surface
water quality.
The flood protection component is the only part of the
Watershed Assessment that is not complete. It will in-
volve simple overlays of floodplain boundaries with ex-
isting and future land use. Floodplain boundaries are
defined as Federal Emergency Management Agency
(FEMA) flood insurance rate map 100-year flood hazard
areas. In many areas, these designations are not very
accurate, yet we decided to proceed with their use
because they are the best available information for
many parts of the District. In areas where little hydrologic
information is available and where the District has not
conducted any related studies, the FEMA data are a
helpful starting point. This echoes a theme of the Water-
shed Assessment project: the assessment is primarily
intended to fill in gaps where we have not performed
previous resource assessments, not to supplant existing
information.
The ecosystems protection component of the Water-
shed Assessment is based heavily on a project identify-
ing priority habitat in Florida, conducted by the Florida
Game and Fresh Water Fish Commission (1). It is similar
to gap analyses that the U.S. Fish and Wildlife Service
currently is conducting in many parts of the country. For
the Watershed Assessment, we modified the data some-
what and examined ways to protect the habitat in coop-
eration with local agencies.
The surface water quality component of the Watershed
Assessment has two main parts. The first uses water
quality data from stations that have been spatially refer-
enced so that we can map them and combine the infor-
mation with other information, such as the second part
of the water quality component. This second part is a
nonpoint source pollution load model, which is dis-
cussed in more detail below.
The Pollution Load Screening Model
The nonpoint source pollution load model is the Pollu-
tion Load Screening Model (PLSM), a commonly used
screening tool in Florida. It is an empirical model that
estimates annual loads to surface waters from storm-
water runoff. Our goal in designing this model was to
identify pollution load "problem areas" for examination in
the Plan.
In these types of models, annual pollutant loads are a
function of runoff volume and mean pollutant concentra-
tions commonly found in runoff. Runoff volume varies
with soil and land use, while pollutant concentrations
vary with land use. For the PLSM, pollutant concentra-
tions were derived from studies conducted solely in Flor-
ida. A report describing the model in detail is available (2).
Usually, this kind of model combines CIS with a spread-
sheet: the CIS supplies important spatial information
that is input into a spreadsheet where the actual calcu-
lations are made. The PLSM is different, however, be-
cause we programmed it entirely within CIS. The
District's CIS software is ARC/INFO, and the model
employs an ARC/INFO module called GRID, which uses
cell-based processing and has analytical capabilities
(3). All the model calculations are done in the CIS
software, resulting in a more flexible model with useful
display capabilities.
Model input consists of grids, or data layers, with a
relatively small cell size (less than 1/2 acre). We chose
this cell size based on the minimum mapping unit of the
most detailed input data layer (land use) and the need
to retain the major road features. The model has four
input grids: land use, soils, rainfall, and watershed
boundaries. For any given cell, the model first calculates
potential annual runoff based on the land use, soil, and
rainfall in that cell. It then calculates annual loads by
applying land-use-dependent pollutant concentrations
to the runoff.
For this model:
Land use is from 1:24,000-scale aerial photography
flown in 1988 and 1989. The model incorporates 13
land use categories.
Soils are the Soil Conservation Service (SCS)
SSURGO database, which corresponds to the county
soil surveys. The PLSM uses the hydrologic group
designation of each soil type.
Rainfall was taken from a network of long-term rainfall
stations located throughout the District.
Watersheds were delineated by the United States
Geological Survey (USGS) on 1:24,000-scale,
7.5-minute maps and digitized.
Model output consists of a runoff grid and six pollutant
load grids. We calculated loads for total phosphorus,
total nitrogen, suspended solids, biochemical oxygen
demand, lead, and zinc. We chose these pollutants
because reliable data were available and because they
characterize a broad range of nonpoint pollution-gener-
ating land uses, from urban to agricultural. The model
98
-------
calculates runoff and loads for any point in space, allow-
ing the user to see the spatial distribution of loads. An
example of a total phosphorus load grid for one sub-
basin in the Jacksonville, Florida vicinity is shown in
Figure 2.
The grids themselves provide a detailed view of model
output. Model results can also be summarized by water-
shed, using the watershed boundary grid, and the infor-
mation can be examined from a basinwide perspective.
We have applied PLSM results in other useful ways at
the District. For example, District staff felt that previous
sediment sampling sites were not appropriately located,
so the District water quality network manager used
model results to locate new sampling sites, focusing on
problem areas as well as areas where we expect to see
little or no nonpoint impact.
Figure 2. Distribution of total phosphorus loads, Ortega River
subbasin (darker areas represent higher loads).
Application of Model Results in the Plan
Because the goal of the model was to identify potential
stormwater runoff problem areas, we needed to simplify,
or categorize, the model results for use in the Plan. We
calculated the per acre watershed load for each pollut-
ant and defined "potential stormwater runoff problem
areas" as those individual watersheds with the highest
loads for all pollutants. Problem areas for one major
basin in the District, the lower St. Johns River basin, are
depicted in Figure 3.
We also ran the model with future land use data ob-
tained from county comprehensive plans. Because the
Potential Stormwater
Runoff Problem Areas
Figure 3. Potential stormwater runoff problem areas, lower St.
Johns River basin.
county maps are guides to future development, and not
predictions of actual development, we exercised caution
when using the results. Problem areas were defined as
those watersheds with projected loads greater than or
equal to existing problem areas. Also, District planners
combined model results with information about individ-
ual counties' regulations and policies to evaluate where
problems are most likely to occur.
Prior to compiling the Plan, the District conducted work-
shops in each county in the District, in which problem
areas identified by the PLSM were discussed with local
agency staff, officials, and the public. We provided large,
hard copy maps depicting stormwater runoff problem
areas combined with results of a separate water quality
analysis on county-based maps. These maps proved to
be powerful tools for initiating discussions and gathering
feedback.
99
-------
In the Plan, stormwater runoff problem areas were re-
ported for each of the 10 major drainage basins in the
District. The information was also repackaged in a
county-based format to create a quick reference for local
agencies. District planners recommended strategies for
addressing problems; these strategies vary as appropri-
ate for each county. Examples include the need to as-
sess compliance with existing stormwater permits,
encourage stormwater reuse during the stormwater and
consumptive use permitting processes, coordinate with
municipalities that are implementing stormwater man-
agement plans, encourage and assist significantly af-
fected municipalities to create stormwater utilities, and
improve monitoring in problem areas that do not have
sufficient water quality data.
In conclusion, the Watershed Assessment CIS project
has proved to be useful not only to the St. Johns River
Water Management District, but also to local govern-
ments. Large projects such as this could not be com-
pleted in a reasonable time without the use of CIS. Also,
for ARC/INFO users who have been restricted to vector
processing, the cell-based processing available in GRID
is a powerful modeling tool.
References
1. Cox, J., R. Kautz, M. MacLaughlin, and T. Gilbert. 1994. Closing
the gaps in Florida's wildlife habitat conservation system. Florida
Game and Fresh Water Fish Commission, Office of Environmental
Services, Tallahassee, FL.
2. Adamus, C.L., and M.J. Bergman. 1993. Development of a non-
point source pollution load screening model. Technical Memoran-
dum No. 1. Department of Surface Water Programs, St. Johns
River Water Management District, Palatka, FL.
3. ESRI. 1992. Cell-based modeling with GRID. Redlands, CA: En-
vironmental Systems Research Institute, Inc.
100
-------
Watershed Stressors and Environmental Monitoring and Assessment Program
Estuarine Indicators for South Shore Rhode Island
John F. Paul and George E. Morrison
Environmental Research Laboratory, U.S. Environmental Protection Agency,
Narragansett, Rhode Island
Abstract
The U.S. Environmental Protection Agency has initiated
the Environmental Monitoring and Assessment Program
(EMAP), a nationwide ecological research, monitoring,
and assessment program whose goal is to report on the
condition of the nation's ecological resources. During the
summers of 1990 through 1993, data were collected from
approximately 450 sampling locations in estuarine waters
of the Virginian Biogeographic Province (mouth of the
Chesapeake Bay to Cape Cod). During this period, sam-
pling stations were located in the coastal ponds and
coastal area of south shore Rhode Island.
One objective of EMAP is to explore associations be-
tween indicators of estuarine condition and stressors in
the watersheds of the sampled systems. Extensive wa-
tershed information for south shore Rhode Island is
available in geographic information system (CIS) for-
mat. Watershed stressors along south shore Rhode
Island were compared with EMAP indicators of estu-
arine conditions using CIS analysis tools. The indicator
values for coastal EMAP stations (those offshore from
coastal ponds) were associated with all of the aggre-
gated south shore watershed stressors. The coastal
pond indicator values were associated with stressors in
the individual coastal pond watersheds. For the total south
shore watershed, the major land use categories are resi-
dential and forest/brush land, followed by agriculture.
Closer to the coast, residential land use is more preva-
lent, while further from the coast, forests/brush lands
dominate. All coastal EMAP stations, with one exception,
exhibited unimpacted benthic conditions, indicating no
widespread problems. For the individual watersheds, the
major land use categories are residential and forest/brush
land. The population density (persons per square mile)
shows an increasing trend from west to east. Impacted
benthic conditions were observed at EMAP sampling sites
in two coastal ponds. These two impacted benthic sites
appear to be organically enriched.
Introduction
Since its inception in 1970, the U.S. Environmental Pro-
tection Agency (EPA) has had the responsibility for regu-
lating, on a national scale, the use of individual and
complex mixtures of pollutants entering our air, land, and
water. The Agency's focus during this period centered
primarily on environmental problems attributable to the
use of individual toxic chemicals. Regulatory policy,
while continuing to control new and historical sources of
individual chemicals (i.e., "end of the pipe") and remedi-
ate existing pollution problems, will have to address the
cumulative impacts from multiple stresses over large
spatial and temporal scales.
In this decade, the focus of environmental problems, or
"scale of concern," has shifted from point-source and
local scales to regional and global scales. Concurrently,
the focus has shifted from chemical to nonchemical
stressors. The threat posed by nonchemical stresses
(e.g., land use, habitat alteration and fragmentation,
species loss and introduction) presents a substantial
risk to the integrity of both specific populations and
ecosystems, and entire watersheds and landscapes.
The shift in the scale of concern for environmental prob-
lems presents a unique challenge for environmental
decision-making. Traditionally, environmental informa-
tion has been collected over local spatial and short
temporal scales, focused on addressing specific prob-
lems, limited in the number of parameters measured,
and collected with a variety of sampling designs that
were neither systematic nor probabilistic. It is not sur-
prising, then, that several scientific reviews concluded
that the information needed to assess, protect, and man-
age marine and estuarine resources was either insuffi-
cient or unavailable and recommended a national
101
-------
network of regional monitoring programs (1, 2). Two key
recommendations resulted from these reviews: (1) the
need for a national monitoring program designed to
determine the status and trends of ecological resources,
and (2) the need for an assessment framework for syn-
thesizing and interpreting the information being pro-
duced in a timely manner and in a form that the public
can understand and decision-makers can use. EPA's
response to these recommendations was to institute a
long-term monitoring program, the Environmental Moni-
toring and Assessment Program (EMAP), and to adopt
a risk-based strategy for decision-making.
EMAP is a nationwide ecological research, monitoring,
and assessment program whose goal is to report on the
condition of the nation's ecological resources. During
the summers of 1990 through 1993, data were collected
from approximately 450 sampling locations in estuarine
waters of the Virginian Biogeographic Province (mouth
of the Chesapeake Bay to Cape Cod) (3-5). During this
period, some of the sampling stations were located in
the coastal ponds and coastal area of south shore
Rhode Island. One objective of EMAP is to explore
associations between indicators of estuarine condition
and stressors in the watersheds of the sampled sys-
tems. Extensive watershed information for south shore
Rhode Island is available in geographic information sys-
tems (CIS) format.
The intent of this paper is to compare watershed stres-
sors with EMAP indicators of estuarine condition along
south shore Rhode Island using CIS analysis tools. The
indicator values for coastal EMAP stations (those off-
shore from coastal ponds) are associated with all of the
aggregated south shore watershed stressors. The
coastal ponds indicator values are associated with
stressors in the individual coastal pond watersheds.
The project reported on in this paper served as a pilot
for integrating watershed information with wide-scale
ecological data collected to assess condition of estu-
arine waters.
Ecological Risk Assessment Context
Robert Huggett, EPA's Assistant Administrator for Re-
search and Development, is using the risk assessment-
risk management paradigm as a framework to
reorganize the EPA research laboratories (6). Huggett is
also reorienting the research that EPA conducts to be
risk based (both human and ecological). The major
thrust of the research to be conducted in the EPA labo-
ratories will be directed toward reducing the uncertain-
ties in the risk assessment process. In this way, the risk
assessment context provides the "why" for the research
conducted.
Ecological risk assessment is defined as a process for
evaluating the likelihood that adverse ecological effects
have occurred, are occurring, or will occur as a result of
exposure to one or more stressors (7). The value of the
risk assessment framework lies in its utility as a proc-
ess for ordering and analyzing exposure and effects
information, and in its flexibility for describing past, pre-
sent, and future risks.
One way of depicting the ecological risk assessment
process is shown in Figure 1 (8). The key points are that
the process is continuous; the process can be oriented
in either direction, dependent upon the form of the ques-
tion or issue being addressed; and monitoring is at the
hub, providing information to all activities. The end result
of the effort is to provide better information for making
environmental management decisions.
Figure 1. Ecological risk assessment framework (8).
Overview of EMAP and Estuarine Results
EMAP has been described as an approach to ecological
research, monitoring, and assessment (9). It is not the
only approach but is an approach that is driven by its
goal to monitor and assess the condition of the nation's
ecological resources. The objectives of the program to
address this goal are to:
Estimate the current status, trends, and changes in
selected indicators of the condition of the nation's
ecological resources on a regional basis with known
confidence.
Estimate the geographic coverage and extent of the
nation's ecological resources with known confidence.
Seek associations among selected indicators of natu-
ral and anthropogenic stress and indicators of eco-
logical condition.
Provide annual statistical summaries and periodic as-
sessments of the nation's ecological resources.
102
-------
The approach used by the program to meet its objec-
tives and address its goal includes:
Use of a large, regional scope that encompasses the
entire county but provides information on the scale that
federal and regional environmental managers require.
Emphasis on ecological indicators to provide the in-
formation to assess condition (i.e., collect information
on the ecological systems themselves to determine
their condition or "health").
A probability-based sampling design to produce sta-
tistically unbiased estimates on condition and to pro-
vide uncertainty bounds for these estimates.
A vision of the program as long-term, continuing into
the next century, which is consistent with the large,
regional spatial scale being addressed.
Development through partnerships with other agencies
that have natural resource stewardship responsibility.
The estuarine component of EMAP was initiated in
1990, with monitoring in the estuarine waters of the
Virginian Biogeographic Province (mouth of
Chesapeake Bay northward to Cape Cod) (10). Figure 2
depicts the biogeographic provinces of estuarine re-
sources of the country. These provinces have been deline-
ated based upon major climatic zones and the prevailing
offshore currents (11). This is comparable with the
ecoregion approach used to describe the distribution of
terrestrial ecosystems (12). The biogeographic province
is the comparable approach for coastal ecosystems.
Monitoring in the Virginian Province continued through
Columbian
1993; monitoring was conducted in the Louisianian
Province from 1991 to 1994; monitoring was initiated in
the Carolinian Province in 1994; and monitoring will be
initiated in the \Afest Indian Province in 1995.
A suite of measurements was collected at each of the
EMAP-Estuaries sampling sites that were selected with
a probability-based sampling design (13, 14). As indi-
cated above, the measurements emphasized ecological
conditions indicators, which included biotic indicators
such as benthicand fish abundance, biomass, diversity,
and composition, and also included abiotic indicators
such as dissolved oxygen, sediment contaminant con-
centration, and sediment toxicity (15).
In the Virginian Province, approximately 450 probability-
based sampling sites were visited during the summer
periods in 1990 through 1993 using consistent indica-
tors and collection and analysis procedures. An example
of the results is shown in Figure 3, which presents the
condition of benthic resources (16). The benthic condi-
tion is reported using a benthic index, which is an ag-
gregate of individual benthic measurements that were
combined using discriminant analysis to differentiate
impacted from unimpacted sites (3, 17). The figure pre-
sents results for values of the benthic index that were
determined to be impacted. The bar chart is the stand-
ard EMAP format for results: province-scale results with
95-percent confidence intervals about estimates. The
large, small, and tidal categories refer to the strata used
in the probability-based sampling design: large systems
are the broad expanses of water such as in Chesapeake
Bay, Delaware Bay, and Long Island Sound; small
Acadian
Californian
Lousianian
West Indian
Figure 2. Biogeographic provinces used by EMAP-Estuaries (13).
103
-------
Impacted EMAP Site
Philadelphia
Baltimore
Washington
1
,**
l£**~ H< ^
$T V^p
\ yf
jS&&*>*~ ^x-,J v_ l^jP*
')
ia jf*
V*%&
Province Large
Small
Tidal
Figure 3. Condition of benthic communities in Virginian Province.
systems include the bays and harbors along the edges
of the major systems and embayments along the coast;
and large tidal rivers include the Potomac, James, Rap-
pahannock, Delaware, and Hudson Rivers.
The results indicate that 24 percent + 4 percent of the
estuarine waters of the Virginian Province have im-
pacted benthic communities. The small and tidal river
systems have proportionately more impacted area than
the large systems.
All the EMAP data are geographically referenced; there-
fore, the data can be spatially displayed to explore
patterns. The spatial display of the impacted benthic
community information is a simple spatial analysis of the
EMAP data. This analysis shows that the impacted ben-
thic resources are distributed across the entire province,
with more impacted sites in the vicinity of the major
metropolitan areas.
In addition to analyzing the EMAP results at the regional
scale, analyses have been conducted at the watershed
scale (see Figure 4). The probability-based sampling
design permits the data to be aggregated (poststratified)
in ways other than the way the original design was
stratified. The only restriction to the aggregation is the
number of available sample sites for the aggregation; a
small number of sites leads to large uncertainties in the
results. Figure 4 shows the aggregation for four major
watersheds: Chesapeake Bay, Delaware Bay, Hudson-
Raritan system, and Long Island Sound. This watershed
scale is close to the practical scale at which environ-
mental management decisions are implemented. The
data need to be analyzed at smaller scales, however, to
focus on environmental management of the smaller wa-
tersheds (e.g., contaminated sediments). This leads into
the need to conduct the pilot project addressing water-
shed information.
South Shore Rhode Island Pilot Project
EMAP's third objective relates to exploring associations
between indicators of estuarine condition and water-
shed stressors. Note that the word "watershed" was
added. One way to address environmental management
remediation strategies is to look at the watershed activi-
ties that could possibly be modified or changed to im-
prove estuarine conditions.
Watershed stressors for the estuarine environment in-
clude land-based sources of pollution, such as point
sources of pollution, and land use activities (i.e., how the
land is actually used, including landscape patterns).
104
-------
\
Figure 4.
LIS
Condition of benthic communities in major watersheds in Virginian Province (Chesapeake Bay, Delaware Bay, Hudson-
Raritan system, and Long Island Sound).
Which of these stressors is more important for a particu-
lar situation depends on the types of estuarine impact
(localized or systemwide) and the management ques-
tion that is being addressed.
The specific objective of the south shore Rhode Island
pilot project was to compare watershed stressors with
EMAP indicators of estuarine condition using CIS analy-
sis tools. This project was not intended to be a definitive
study by itself of south shore Rhode Island butto explore
the process necessary to undertake the comparisons, to
investigate the feasibility of pulling the necessary infor-
mation together, and to identify potential problems before
undertaking this comparison on a much larger scale.
The south shore Rhode Island study area is depicted in
Figure 5. This coastal area drains into the coastal waters
of Block Island Sound. The project was intentionally
restricted to a limited geographic area to avoid being
overwhelmed with the tremendous volumes of data that
could have been encountered.
All data sources used in this project were available
electronically. Digitized U.S. Geological Survey quad
maps were available from the Rhode Island Geographic
Information System (RIGIS) at the University of Rhode
Island. The National Pollutant Discharge Elimination
System (NPDES) was available for major dischargers
from the National Oceanic and Atmospheric Administra-
tion's National Coastal Pollution Discharge Inventory
(18). The 1990 census was also available from RIGIS.
The EMAP 1990 through 1993 estuarine data were
available from the EMAP-Estuaries Information System
at the EPA Environmental Research Laboratory in Nar-
ragansett, Rhode Island. The RIGIS data were already
available as ARC/INFO coverages. The NPDES and
EMAP data had to be converted to ARC/INFO point
coverages.
Two approaches were used to conduct spatial analyses.
Buffer zones at 1, 3, 5, 10, and 20 kilometers from the
south coast of Rhode Island were created and used to
clip the south shore area coverages (e.g., land use,
population, point sources). The watershed boundaries
of three south shore coastal ponds were manually de-
lineated and used to clip the south shore area cover-
ages. The ponds were Quonochontaug, Ninigret, and
Point Judith (west to east).
105
-------
Point Judith Pond . Vji
Ninigret Pond
Quonochontaug Pond
Block Island Sound
1
Figure 5. South shore Rhode Island study area.
South Shore Rhode Island Pilot
Project Results
The results for land use by distance from the coast are
presented in Figure 6. These results give the broad-scale
coastal perspective. For the total south coast watershed,
the major land use categories are residential and for-
est/brush land, followed by agriculture. Closerto the coast,
residential land use is more prevalent, while farther from
the coast, forests/brush lands dominate. Population (see
Figure 7) increases with distance from the coast, but popu-
lation density does not appear to be a function of distance
from the coast. Only one out of five coastal EMAP stations
exhibited impacted benthic conditions, indicating no wide-
spread benthic problems in the coastal waters. The one
station that was classified as impacted was dominated by
an extremely high number of individuals of one species.
A smaller scale view can be gained by looking at the
results for the individual watersheds. This view also
provides an east-west perspective compared with the
south-north perspective with the distance from the
coast. Again, for the individual watersheds, the major
land use categories are residential and forest/brush land
(see Figure 8). The population increases from west to
east, and population density shows an increasing trend
from west to east (see Figure 9). Impacted benthic
conditions were observed at the EMAP stations in
Quonochontaug and Point Judith Ponds. These stations
exhibited organic enrichment (total organic carbon in the
sediments exceeding 2 percent), possibly from histori-
cally improperly treated sewage. No benthic data were
available for the Ninigret Pond station; however, dis-
solved oxygen was observed to be low at this station.
No major NPDES point sources are located in the coastal
pond watersheds, although two are located on the eastern
edge of the Point Judith Pond watershed boundary.
Discussion
A pilot project was conducted for south shore Rhode
Island to compare watershed stressors with EMAP indi-
cators of estuarine condition. The results indicate that
such a comparison can be accomplished, with the wa-
tershed information providing a qualitative link to the
estuarine conditions observed. One potential problem is
the need to delineate the watershed boundaries for all
106
-------
CO
(U
CO
5
-------
2,500 -
2,000
1,500
in
-------
4. U.S. EPA. 1994. Statistical summary: EMAP-Estuaries Virginian
Province (1991). EPA/620/R-94/005. Narragansett, Rl.
5. U.S. EPA. 1994. Statistical summary: EMAP-Estuaries Virginian
Province (1992). EPA/620/R-94/019. Narragansett, Rl.
6. Newman, A. 1995. A new direction for EPA's Office of Research
and Development. Environ. Sci. and Technol. 29(3): 126A-129A.
7. U.S. EPA. 1992. Framework for ecological risk assessment.
EPA/630/R-92/001. Washington, DC.
8. Gentile, J.H., J.F. Paul, K.J. Scott, and R.A. Linthurst. 1995. The
role of regional monitoring in ecological risk assessment: An
EMAP example for watersheds. Manuscript in review.
9. Messer, J.J., R. Linthurst, and W. Overton. 1991. An EPA program
for monitoring ecological status and trends. Environ. Monitor, and
Assessment 17(1):67-78.
10. Paul, J.F., K.J. Scott, A.F. Holland, S.B. Weisberg, J.K. Summers,
and A. Robertson. 1992. The estuarine component of the U.S.
EPA's Environmental Monitoring and Assessment Program.
Chemistry and Ecology 7:93-116.
11. Bailey, R.G. 1983. Delineation of ecosystem regions. Environ.
Mgmt. 7:365-373.
12. Omernik, J.M. 1987. Ecoregions of the conterminous United
States. Annals of the Assoc. of Amer. Geographers 77:118-125.
13. U.S. EPA. 1990. Near coastal program plan for 1990: Estuaries.
EPA/600/4-90/033. Narragansett, Rl.
14. U.S. EPA. 1991. Design report of the Environmental Monitoring
and Assessment Program. EPA/600/3-91/053. Chervils, OR.
15. U.S. EPA. 1990. Environmental Monitoring and Assessment Pro-
gram: Ecological indicators. EPA/600/3-90/060. Washington, DC.
16. U.S. EPA. EMAP-Virginian Province four-year assessment (1990-
1993). EPA report in review.
17. Engle, V.D., J.K. Summers, and G.R. Gaston. 1994. A benthic
index of environmental condition of Gulf of Mexico estuaries.
Estuaries 17(2):372-384.
18. NOAA. 1993. Point-source methods document. NOAA technical
report. Silver Spring, MD: Strategic Environmental Assessments
Division, National Ocean Service, National Oceanic and Atmos-
pheric Administration.
19. Comeleo, R.L., J.F. Paul, P.V August, J. Copeland, C. Baker, S.
Hale, and R.W Latimer. 1995. Relationships between landscape
stressors and sediment contamination in Chesapeake Bay estu-
aries. Manuscript in review.
109
-------
GIS Watershed Applications in the Analysis of Nonpoint Source Pollution
Thomas H. Cahill, Wesley R. Horner, and Joel S. McGuire
Cahill Associates, West Chester, Pennsylvania
Introduction
Geographic information systems (GIS) have been used
to evaluate the impact of nonpoint source (NPS) pollu-
tion in a variety of watersheds and drainage systems
over the past 20 years (1-6). During that period, our
understanding of the sources and hydrologic transport
mechanisms of NPS pollutants, both in particulate and
soluble forms, has greatly increased (7-9). Our ability to
create and manipulate land resource data, however, has
advanced at a far more dramatic rate. Whereas 20 years
ago, both computer system capabilities and peripheral
hardware limited the process of encoding, storing, and
displaying spatial data, today we can encode land re-
source data, analyze it, and produce stunning visual
displays at a relatively low cost.
The question is: what has this experience told us regard-
ing the yet unresolved problem of water quality degra-
dation from NPS pollution in our streams, lakes, and
coastal waters (10, 11)?
The purpose of this paper is to report on several recent
studies of this nature that created a GIS as a tool to
analyze NPS pollution. This paper will not cover all
aspects of these studies; detailed reports on each pro-
ject are available from the authors or respective clients.
The objects of these studies were:
A medium-sized lake draining a fairly small watershed
A riverine system with multiple use impoundments
A 100-mile stretch of Atlantic coastal estuary
These water bodies all have one common ingredient:
NPS pollution significantly affects them. While the pri-
mary focus of these studies was to understand the
dynamics of surface water quality, and specifically the
NPS component, the further objective was to document
the causal link between identified water resource prob-
lems and the watershed-wide management actions
needed for their remediation. Thus, GIS serves not only
as a mechanism for analysis of NPS pollution sources
but also as the tool by which to evaluate alternative
methods that would reduce or prevent this pollution.
Study Concepts
These three studies illustrate different approaches to
both aspects of this problem. In the 93-square-mile Up-
per Perkiomen Creek watershed (UPW) study, the ob-
jective was to develop a management program that
would reduce nutrient load in a system of reservoirs at
the base of the watershed. An essential element in the
analysis underlying GIS design (ARC/CAD) was to be
able to differentiate and evaluate pollution sources in the
watershed, while providing the technical basis for an
innovative and far-reaching management program on all
levels of government; that is, GIS was used not only to
analyze the problem but to help formulate the solution.
In the more focused Neshaminy Creek study, Cahill
Associates (CA) designed a detailed pixel/raster format
for GIS to support detailed hydrologic modeling (12) and
NPS loading analysis. This study, carried out under
Pennsylvania's Act 167 stormwater management pro-
gram, was under a legal requirement to translate tech-
nical findings into subdivision regulations that all 30
watershed municipalities would adopt. This mandate
required much more geographically specific rigor in the
GIS approach and in the management recommenda-
tions the law stipulated.
These two projects (see Figure 1), when taken together,
illustrate the critical relationship between understanding
the appropriate level of detail in GIS system design, GIS
development with modeling and other analytical require-
ments, and ultimately, the proposed management ac-
tions for watershed-wide implementation.
In the New Jersey Atlantic Coastal Drainage (ACD)
study, the objective was to document more completely
the magnitude and sources of NPS pollutants, espe-
cially nutrients, entering New Jersey estuarine coastal
waters. The GIS design placed special attention on the
role of urban or developed land uses situated along the
coastal fringe, particularly the maintained or landscaped
portions of developed sites. Most previous studies have
largely ignored this factor. Instead, they have focused
water quality analysis typically on NPS loadings as a
110
-------
Upper Perkiomen
Watershed
Pennsylvania
N. HAMPTON
-x ^ Vl
LEHIGH
Perfiomen Creek Watershed^"y
Figure 1. Regional location of Upper Perkiomen and Neshaminy
basins.
function of impervious area coverage, with the assump-
tion that loadings increase as imperviousness increases.
On the contrary, the CA thesis states that certain pollut-
ant loadings, such as nutrients, maximize in areas with
relatively moderate densities (1/2- to 1-acre lots) and
percentage impervious cover but with large maintained
lawnscapes. Because sandy soils allow soluble NPS
pollutants to pass as interflow to points of surface dis-
charge with surprising ease, they exacerbate the prob-
lem of nutrient applications in typical coastal drainage
areas. CIS application in this case enabled estimation
of the nutrient loading to coastal waters. Existing fertil-
ized lawn areas were calculated to be a significant
source of nutrient pollution, with loadings from new land
development posited as an even more serious problem
for New Jersey's coastal waters. CIS was then applied
to evaluate the suitability of various best management
practices (BMPs), based on the physical and chemical
properties of the soil mantle and the existing and antici-
pated land use.
The Upper Perkiomen Creek
Watershed Study
Background
The UPWin southeastern Pennsylvania is a tributary of
the Schuylkill River in the Delaware River basin (see
Figure 2). Serious eutrophication problems occurring in
the system of reservoirs lying at the base of this rela-
tively rural watershed prompted the study. The study
New Jersey
HUNTERDON
Upper Perkiomen Creek Water:
-\,.-
SCHUTLKIU.
Scale in Miles
Figure 2. The Perkiomen Creek watershed in the Delaware River
basin.
effort evolved from concerns on the part of the Delaware
Riverkeeper, a private nonprofit environmental organi-
zation dedicated to promoting the environmental well-
being of the Delaware River watershed. The Upper
Perkiomen Creek has experienced various water quality
problems, especially the eutrophication of Green Lane
Reservoir, a large raw water supply storage reservoir
(see Figure 3). Green Lane's highly eutrophic condition
has been a constant since shortly after initial construc-
tion over 35 years ago, but the relative importance of
NPS inputs has dramatically increased. Whereas 10
years ago point source input was the major source of
phosphorus, elimination of some point sources and ad-
vanced waste treatment for others has greatly reduced
that component of pollutant loading, while NPS sources
have remained constant or increased. Current analysis
indicates that NPS pollution constitutes over 80 percent
of the annual load of phosphorus (see Figure 4) into the
Green Lane Reservoir and is well in excess of the
desired loading to restore water quality (see Figure 5).
Nonpoint Source Analysis
Calculating the NPS load was an essential ingredient in
the study and relied on developing accurate measure-
ment of NPS transport during stormwater runoff periods.
Certain pollutants, specifically those associated with
sediment and particulate transport such as phosphorus,
have produced a "chemograph" that parallels but does
not exactly follow the traditional form of the hydrograph
(see Figure 6). The pollutant mass transport associated
with this runoff flux frequently constitutes the major frac-
tion of NPS discharge in a given watershed (8, 13). In
111
-------
T3 ^
CD &1
si ro
(/) T3
a i
-'a i
i
':
f
:/ ' ^
.."V
" »
o
Z
D)
0)
'8
(D
Q.
W
OJ
CL
'Jb-
......
S
1 v
£
ro
T3
CD
X
Figure 3. Green Lane Reservoir in the Upper Perkiomen watershed, 814 acres, 4.3 BG.
112
-------
Dustfall
2.50%
500 Pounds
per Year
Waterfowl
2.50%
500 Pounds
per Year
Nonpoint Sources
Dry Flow
5,604 Pounds
perYi
Nonpoint Source
Total = 84%
Point Sources
Direct 2,486
Pounds per Year
Point Sources
to Tributaries 592
Pounds per Year
Direct Drainage
1,172 Pounds
Per Year
Nonpoint Sources
Storm Flow 8,036
Pounds per Year
Total Load = 18,889 Pounds per Year
Figure 4. Sources of total phosphorus mass transport into the
Green Lane Reservoir from the Upper Perkiomen wa-
tershed (71 square miles) in an average flow yearin
pounds per year.
20,000
18,000
Average Year, Riverkeeper-
1993 = 18,889 Pounds per Year
Eutrophic = 8,416 Pounds per Year
Oligotrophic = 4,208 Pounds per Year
Riverkeeper -1993 Eutrophic
Oligotrophic
the UPW study, operating continuous sampling stations
at two key gage locations above the reservoir allowed
the measurement of stormwater chemistry of this type
and produced estimates of wet weather transport of phos-
phorus and sediment. Surprisingly, the NPS transport
during dry weather, calculated by subtracting the point
sources, was also significant and is attributed to live-
stock discharges and septage drainage.
But the wet weather proportion of NPS pollution still
dominates lake water quality. Many have said that water
quality in a given watershed is a function of land use,
but that statement is as unsatisfying as saying that
runoff is a function of rainfall. Experience has taught us
that neither process is quite that simplistic, nor does
either follow a direct linear relationship of cause and
effect. The causal mechanisms that generate a certain
mass load of pollutant in a drainage basin certainly
result from how much mass of that pollutant is applied
to the landscape within the drainage, which in turn is
scoured from the landscape during periods of surface
saturation, transported in, and diluted by runoff. The end
result is a concentration of pollutant in the stormwater
that might be several orders of magnitude greater than
during dry weather flow, the hydrologic period tradition-
ally used to measure and define water quality.
Developing NPS analysis or algorithms for stormwater
quality modeling requires replicating the specific hydro-
graph and its associated chemograph, as well as defin-
ing the mechanisms by which pollutants are scoured
from the land surface, transported in runoff, and pass
through the river system. Total phosphorus (TP), for
example, is transported with the colloidal soil particles
(see Figure 7), so sediment transport and deposition
constitute a key mechanism.
Figure 5. Reduction in annual phosphorus load required to
achieve improved trophic level.
3/5/930:00 3/5/9312:00 3/6/930:00 3/6/9312:00
Figure 6. Storm hydrograph in the Upper Perkiomen watershed
illustrating the dramatic increase in total phosphorus
and suspended sediment during runoff.
Adding to these complications is the question of whether
to model single or multiple events. Is the chemodynamic
process one in which the transport takes place over a
series of storm events, so that each storm moves the
pollutant mass a given distance in the drainage and then
allows it to settle in the channel only to resuspend it with
the next peak of flow? Or does the total mass transport
occur in one single dynamic, from corn field or suburban
lawn to lake, estuary, or other sink, that is hours or days
downstream in the drainage? The issue of how storm-
water transport of pollutants takes place is of paramount
importance in current planning and regulatory imple-
mentation (11) because many of our current BMPs are
relatively ineffective in removing NPS pollutants. This
understanding is critical even as we attempt to intervene
in the pollutant generation process by changing the way
we cultivate the land, fertilize our landscapes, or for that
matter, how we alter the land surface during growth.
113
-------
3,500
3,000
2,500
Q_
Q_
Q_
Q.
2,000
1,000
500
Main Branch rA2 =
0.96 65 Samples
Regression
500 1,000 1,500
Suspended Solids (PPM)
2,000
2,500
Figure 7. The relationship between total phosphorus and suspended sediment concentrations during runoff is strong but varies
with different watersheds.
GIS Evaluation
The GIS data files on land use/land cover that were
created for the UPW show that the bulk of the area is
still quite undeveloped and rural (see Figure 8), with the
steeply sloped and igneous rock areas in the headwa-
ters in forest cover (38 percent) and the valleys in mixed
agriculture (44 percent). The urban/suburban land com-
poses the remaining 18 percent and largely consists of
several older, historic boroughs linked together in a
lineal pattern with widely scattered, low-density residen-
tial areas. Much of the existing housing is turn-of-the-
century at quite high densities, mixed with a variety of
commercial and other uses. This pattern contrasts
sharply with typical large-lot suburban subdivisions. In
fact, these watershed boroughs resemble the "village"
concepts that innovative planning theorists advocate in
a variety of important ways.
The watershed (see Figure 9) is blessed, or cursed,
depending upon one's perspective, with a multiplicity of
local governments including four different counties and
18 different municipalities. This arrangement poses spe-
cial challenges for management program implementa-
tion. Population projections indicate that additional
development will occur at moderate rates throughout the
watershed, reflecting recent trends.
Farming, both crop cultivation and dairying, is a major
existing land use in the watershed, although agriculture
is not especially robust and appears to be declining. This
lack of agricultural vibrancy becomes a major factor in
determining how to impose additional management
measures on agricultural pollution sources. GIS tabula-
tion of agricultural land totals some 19,000 acres above
the reservoir, which can be compared with the estimated
TP and suspended solids (SS) mass transport reaching
the lake. Considering only the agricultural land to be
the source of this NPS input (not quite true) suggests
an average annual yield of 180#/acre/year-SS and
0.22#/acre/year-TP.
This sediment/phosphorus yield is more than sufficient
to maintain a eutrophic condition in the reservoir system.
The problem with this yield, however, is that it is two
orders of magnitude less than commonly accepted
methodologies of soil erosion, such as the universal soil
loss equation (14), would suggest might come from such
a watershed. Analysis of the cultivation practices taking
place on farmland in the watershed estimates soil ero-
sion to be approximately 5 to 10 tons per acre or more
per year, far more than is observed passing out of the
basin into the reservoir. The phosphorus applications on
both cultivated and maintained residential landscapes
also appear much greater than the mass transport actu-
ally measured in the flowing streams, which represent
perhaps 7 percent or less of the annual land application.
The implication for NPS analysis is that the standard
shopping list of either agricultural or urban BMPs might
only reduce the mass transport by a relatively small
fraction, even if successfully applied throughout the
drainage. As Figure 7 illustrates, most of the phosphorus
transport occurs on the colloidal fraction of sediment
particles, which tend to remain in suspension as storm-
waters pass through conventional detention structures,
terraces, or grassed swales.
To consider more radical measures, GIS was used to
determine possible alternatives, such as creating a
stream buffer system (see Figure 10) with various set-
back distances from the perennial stream network, and
to evaluate how great an impact this might have on
agricultural land use and urban development. Land
114
-------
---"" ~ = "' -3%i JHr =" ~>i-^T. -,-, , , " ' - -- .
-'""" *""' "" ~~ - % SJ".P«':!*I*~_,,., .-., «««,-Bu "" ~-
~ -zf^^^r'T^ * -,m *1; -' ^^~,-: --fgr' -T::-
s*f-r ^" .,- -sjsf-c " ' T". _"- --». '*S*^!"Sil*7-5, -;-^'j^^ r55*"
.iL^~= -nJ8" ^lii*HBii'£r... " == . iTTl.:.. HillliMlllllLLlI' ^hmun iLTnnn-- '"" """'"' '"'" = - ?Sfc4llhfc«fca ,"
Agricultural Land
Forested Land
Topography
Figure 8. GIS data files showing land use/land cover characteristics for the Upper Perkomien watershed.
115
-------
Fallow\Pasture
r j
[_ I Cropstrip
Orchard
Urban\Suburban
Trailerpk
Resrural
Resfarm
Golfcourse
Quarry
Figure 9. Existing land use/land cover CIS file for the Upper Perkiomen watershed. The 95-square-mile basin includes portions of
four counties and 18 municipalities.
116
-------
Legend
Forest
Meadow
Water
Developed \
\
Trailer Park
Resrural/Farm \
I
.y
i
--i-----\^^^f^p^-'V-^ - "iX "^----v-T-^v.^!--
:--l - ---%-^;^>:\-l|.-..-::>:..«::;:-':- .?- .-= , ^, .:, -j*- .- -,T- . -.-;.r. ~~Jr "*
₯---*-7n-r&- f
\ :-^ ;^r ^ii ^^,
T:- - -
/ ' y^,/^^M*SaS^i*l^tft^^
jr /...- :"u«P^"" =r =;;-- rj&#c?r.-.v"".."".: '; "= "-:- -:'^
Stream
Structures
Area
of
Detail
Figure 10. GIS analysis of stream corridors allows evaluation of riparian buffer systems, potential agricultural land loss, potential
septic system discharges, and related NFS reduction with selected best management practices.
117
-------
use at varying distances (100 feet, 200 feet, and 1,000
feet) from streams was tabulated, including all land area
in the "active" agriculture categories. This CIS documen-
tation allowed estimation of the significant NPS reduc-
tion in loadings that a riparian corridor management
program could achieve.
In the same way, CIS analysis helped estimate pollutant
loadings from malfunctioning onsite septic systems.
Counts of structures in nonpublicly sewered areas
within varying distances from the stream system were
developed using CIS data files. The nearly 300 potential
systems within a 200-foot radius of those streams drain-
ing into the Green Lane Reservoir identified in this man-
ner, with pollutant generation factors applied, became
the basis of a dry weather pollutant estimation. Although
this approach was dependent on a variety of assump-
tions, alternative approaches of evaluating the problem,
such as field visits to actual onsite systems throughout
the watershed, would not have been feasible.
For urban and suburban development, the management
focus was to estimate NPS loadings from future land
development. CIS was used to demonstrate NPS pollut-
ant load implications of future growth envisioned in the
watershed's keystone municipality, Upper Hanover
Township. Here, an increase of 15,000 residents would
convert 1,772 acres into residential, commercial, and
industrial uses. Nonpoint pollutant loadings generated
by this new land development constituted significant
increases in phosphorus, suspended solids, metals,
oil/grease, and other pollutants, and would reverse any
improvements in Green Lane Reservoir water quality
that recent wastewater treatment plant upgrades
achieved.
From a water quality perspective, future alternative land
use configurations that concentrate development and
minimize ultimate disturbance of the land surface
yielded would substantially reduce NPS pollutant load-
ings into the reservoirs. This entire process of testing
land use implications of different management ap-
proaches for their water quality impacts indicated that
pollutant loads could be minimized far more cost effec-
tively through management actions, both structural and
nonstructural, which varied from the areawide to the
site-specific.
Neshaminy Creek Watershed Stormwater
Management Study
Background
The Neshaminy Creek watershed, including 237 square
miles of mixed urban and rural land uses, lies primarily
in Bucks County, Pennsylvania, and flows directly into
the Delaware River (see Figure 1). The 1978 Pennsyl-
vania 167 Stormwater Management Act, which required
that counties prepare Stormwater management plans for
all 353 designated watersheds in the state, mandated
the Neshaminy study. This act further stipulated that
municipalities then needed to implement the watershed
plans through adopting the necessary municipal ordi-
nances and regulations. In fact, the Neshaminy study
had three water resource management objectives:
Prevent worsened flooding downstream caused by
increased volumes of runoff from land development.
Increase ground-water recharge.
Reduce NPS pollutant loadings from new development.
In the initial study design, water quality and NPS issues
were secondary to flooding concerns. When Pennsylva-
nia's Stormwater management program was conceived,
the state focused on preventing watershed-wide flood-
ing. Clearly, detention basins have become the primary
mode of managing peak rates of Stormwater discharge
site-by-site in most communities. Because detention ba-
sins only control peak rates of runoff and allow signifi-
cantly increased total volumes of water discharged from
sites, however, the increased Stormwater volumes can
theoretically combine and create worsened flooding
downstream. Consequently, most Act 167 planning has
focused on elaborate hydrologic modeling designed to
assess the seriousness of potential cumulative flooding
in watersheds under study.
In the case of the Neshaminy, however, the record sug-
gested that although localized flooding could be an is-
sue, an existing network of eight multipurpose flood
control structures constructed during the 1960s served
to prevent significant flooding. Water quality certainly
was a serious Stormwater concern, however, especially
in the areas flowing into the reservoirs where recrea-
tional use had become intense. Several of the existing
impoundments were multipurpose, their permanent
pools providing critical recreational functions for a bur-
geoning Bucks County population. At the same time, the
proliferation of development in the watershed, with its
increased point and nonpoint sources, had degraded
streams and seriously affected the reservoirs. While the
total stream system in the watershed was of concern,
the future of the reservoirs came to be particularly im-
portant in developing the total Stormwater management
program for the Neshaminy watershed.
The Neshaminy lies at the heart of Bucks County, Penn-
sylvania's primary population and employment growth
county (see Figure 11). Although the Neshaminy water-
shed has already experienced heavy development, es-
pecially in the lower or southern portions, farmsteads
and large areas of undeveloped land still exist, espe-
cially in headwater areas. Agriculture has been a major
land use in the past, but farms rapidly are converting to
urban uses as the wave of urbanization moves outward
from Philadelphia and from the Princeton/Trenton met-
ropolitan areas. Growth projections indicate continu-
118
-------
HESHAMJNY CRKKK WATERSHED
STORM WATER MANAGEMENT PLAN
i^w^
?fejM&
EXISTING LAND COVER
Figure 11. Land use/land cover in the Neshaminy basin of Bucks County, Pennsylvania. The watershed covers 237 square miles
in southeast Pennsylvania.
119
-------
ation of this rapid growth and a continuing change in
existing land use/land cover, together with projected
development with in the required 10-year planning horizon.
Physiographically, the watershed spans both the Pied-
mont and Atlantic coastal plain provinces, with rolling
topography and relatively steep slopes underlain by Tri-
assic formation rock, including the Lockatong, Bruns-
wick, and Stockton formations. This bedrock ranges
from being a poor aquifer (Lockatong) to an excellent
aquifer (Stockton) where the many rock fractures allow
for considerable ground-water yields. Soils are quite
variable, ranging from good loam (hydrologic soil group B)
to clays and other types with poor drainage charac-
teristics (e.g., high water table, shallow depth to bed-
rock). A large proportion of the soils in the watershed are
categorized as hydrologic soil group C, which is mar-
ginal for many stormwater management infiltration tech-
niques (see Figure 12) and produces a relatively large
proportion of direct runoff. With an annual rainfall of 45
inches, base flow accounts for about 12 inches and
direct runoff accounts for 10 inches.
The system of eight stormwater control structures,
which were built over the past three decades under the
federal PL 566 program, have altered the hydrology of
the watershed (15). In addition, in heavily developed
portions of the watershed, impervious surfaces com-
bined with numerous detention basins prevent the bulk
of the precipitation from being recharged, and the vol-
ume of total runoff proportionally increases. An elabo-
rate system of municipal and nonmunicipal wastewater
treatment plants also adds to this alteration of the hy-
drologic cycle. These plants discharge wastewater efflu-
ent that, in some cases, constitutes the bulk of the
stream flow during dry periods. While the impact of NPS
was evident throughout the drainage, it was of special
interest in the impoundment network, especially those
impoundments that were conceived as multipurpose in
function and constitute major recreational resources in
the watershed.
CIS Design
Act 167 requirements and the needs of the hydrologic
and other modeling used in planning both heavily influ-
enced the CIS developed for the Neshaminy. Spatial
data files, including existing land use, future land use,
and soil series aggregated by hydrologic soil groups,
were created by digitizing at a 1-hectare (2.5-acre) cell
resolution. The encoding process that helped design the
CIS used a stratified random point sampling technique
that similar studies had developed and applied (1, 3).
The encoding process used a metric grid of 5-kilometer
sections, subdivided into 2,500 1-hectare cells (100 me-
ters on a side), aligned with the Universal Transverse
Mercator (UTM) Grid System. This grid appears in blue
on U.S. Geological Survey (USGS) topographic maps.
These maps served as the framework of reference for
all data compilation. Within each 100-meter cell, a ran-
domly located point was chosen (see Figure 13) at
which the specific factor was encoded as representative
for the cell, using a digitizer tablet. This approach al-
lowed extraction of the data from the respective source
documents with some rectification necessary for many
types of source maps and photographs.
The combination of soil series and cover in each cell
helped to calculate the curve number and unit runoff per
cell. The 45,000-cell data file was then used to calculate
total runoff for a range of events in each of 100 subbas-
ins that averaged 1.95 square miles each. The resultant
hydrographs, used in combination with a separate linear
data file in CIS describing the hydrographic network of
stream geometry, routed and calibrated the hydrologic
model (TR-20). NPS mass transport loadings were es-
timated on an annual basis by cell, again using the land
use/land cover data file, and total loads summed by
groups of subbasins above critical locations. This issue
was particularly important with respect to the drainage
areas above the impoundments, where NPS pollutants
were of greatest concern.
The soil properties data file was especially useful in
evaluating certain management objectives, such as the
opportunity for recharging ground-water aquifers. The
spatial variation in relative effectiveness of infiltration
BMPs was considered for both quantity and quality miti-
gation because the best methods for NPS reduction
usually include recharge where possible. The soil series
corresponding with new growth areas were classified
regarding their suitability for these BMPs, which are
most efficient on well-drained or moderately well-drained
soil. Thus, the alternative impacts of future growth could
be considered in terms of potential generation (or man-
agement) of NPS loads. A BMP selection methodology
(see Figure 14), which was developed forthe 30 munici-
palities within the watershed, focused on new land de-
velopment applications and considered both water
quantity and quality management objectives. BMP se-
lection is a function of several factors, including:
The need for further peak rate reduction.
The recharge sensitivity of the project site (defined as
a function of headwaters stream location, areawide re-
liance on ground water for water supply, or presence
of effluent limited streams).
The need for priority NPS pollution controls (location
within reservoir drainage).
Development of two "performance" levels of BMP selection
techniques gave municipalities some degree of flexibility
in developing their new stormwater management pro-
grams. This system required only the minimally acceptable
techniques but recommended the more fully effective
ones, hoping that municipalities would strive to incorporate
120
-------
HVDKOLiu.U SUM <;ko|ips
Figure 12. GIS file of hydrologic soil groups in the Neshaminy basin. The 31 soil series are digitized in 45,000 pixels of 1-hectare size.
121
-------
Region With Labeled Zones
29
(4428000mN)
28
A
5
46 47 48 49
(446000mE)
290
288
286
284
282
Zone With Labeled Cells
280
460 462 464 466
468
470
Figure 13. Raster/pixel design of GIS for Neshaminy modeling
study. Each pixel is 1 hectare (2.47 acres).
recommended management measures wherever possi-
ble. The BMP selection methodology also was sensitive
to type of land use or proposed development, assigning
typical single-family residential subdivisions different
BMPs than, for example, multifamily and other nonresi-
dential proposals (including commercial and industrial
proposals). The selection process also determined size
of site to be a factor, differentiating between sites of 5
acres or more because of the varying degrees of cost
and effectiveness of different BMP approaches. The
methodology, if properly and fully implemented, should
achieve the necessary stormwater-related objectives
both quantity and qualitythat the analysis had deemed
necessary (16).
GIS was especially important in its ability to test how
reasonable the BMP selection methodology was. Such
tests included the ability to evaluate, for each municipal-
ity, the following factors:
The nature and extent of the projected development.
The size of development/size of site assumptions.
Other vital BMP feasibility factors such as soils and
their appropriateness for different BMP techniques.
GIS also enabled analysis of the water quantity and
quality impacts of projected growth on a baseline basis,
assuming continuation of existing stormwater manage-
ment practices. Water quality loadings to individual res-
ervoirs and to the stream system could be readily
demonstrated. Because overenrichment of the reser-
voirs was so crucial, researchers could estimate phos-
phorus and nitrogen loadings from projected
development assuming existing stormwater practices,
even on a municipality by municipality basis.
New Jersey Atlantic Coastal
Drainage Study
Background
The third study considered a much larger coastal water-
shed in New Jersey (see Figure 15). The New Jersey
Development
Application
Required
I See Map A I
No Peak
Rate
Factor
Hmpoundment
Drainage
1 Recharge 1
| Sensitive \
4 Non Impoundmen
Drainage
.
1 Impoundment
r~| Drainage
U Non Rechargel
| Sensitive |~
1 Non Impoundm
Drainage
| See Map G| | See Map D
_l Impoundment
| | Drainage
1 Recharge 1 1
I Sensitive | 1
LJ Non Impoundmer
Drainage
1 Impoundment
|~~| Drainage
_| Non Recharge) |
\ -
?nt ^
]
jNon Impoundment!
"I Drain age T
Not Applicable
Required: Multi-Resi and Non-Resi Overs Acres, Porous Pave. With
Underground Recharge Beds for Paved Areas and Infiltration Devices for
Non-Paved Areas, Sized for Peak; Other Uses, Infiltration Devices for
Paved and Nonpaved Areas, Sized for Peak
Not Applicable
Required: Multi-Resi and Non-Resi Overs Acres, Dual Purpose Detention
Basins for Paved/Nonpaved Areas, Sized for Peak; Other Uses, Detention
Basins Sized for Peak
Recommended: All Uses/Sizes, Porous Pave. With Underground Recharge
Beds for Paved Areas; Minimum Disturbance or Wet Ponds/Artificial
Wetlands for Nonpaved Areas, All Sized for Peak
Required: All Uses and Sizes, Porous Pave. With Underground Recharge
for Paved Areas; Minimum Disturbance for Nonpaved Areas
Required: Multi-Resi and Non-Resi Overs Acres, Porous Pave. With
Underground Recharge Beds for Paved Areas and Infiltration Devices for
Nonpaved Areas
Recommended: All Uses/Sizes, Porous Pave. With Underground Recharge
for Paved Areas; Minimum Disturbance and/or Infiltration Devices After
Site Stabilization
Required: All Uses/Sizes, First-Flush Settling Basins for Paved Areas; for
Nonpaved Areas, Minimum Disturbance/Wet Ponds/Artificial Wetlands
Recommended: for All Uses/Sizes, Porous Pave. With Underground
Recharge Beds for Paved Areas; Minimum Disturbance for Nonpaved
Areas
Required: Multi-Resi and Non-Resi Overs Acres, First-Flush Settling
Basin; Other Uses, Detention Basins (No Change)
Recommended: Porous Pave. With Underground Recharge Beds for Paved
Areas; for Nonpaved areas, Minimum Disturbance/Wet Ponds/Artificial
Wetlands
Figure 14. BMP selection methodology used with the GIS database in the Neshaminy basin modeling study.
122
-------
Atlantic Coastal Drainage (ACD) includes an area of
2,086 square miles, with barrier islands (50 square
miles), wetlands/bays/estuaries (285 square miles), and
a unique scrubby pitch pine-cedar forest, known as the
Pine Barrens, largely covering the 1,750 square miles
of mainland interior (see Figure 16). This flat coastal
plain comprises a series of unconsolidated sedimentary
deposits of sand, marl, and clay, which increase in thick-
ness toward the coastline. Over the past 16,000 years,
as the ocean level has risen, the water's edge has
progressed inland to its present position. Ocean cur-
rents and upland erosion and deposition have created a
long, narrow series of barrier islands that absorb the
energy of ocean storms and buffer the estuary habitats
from the scour of waves and currents. Between the
mainland and barrier islands are embayments and es-
tuaries of different sizes and configurations. Inland erosion
and marine sediments have gradually filled many of
these areas, creating extensive wetlands (17).
In this ACD region, new land development and popula-
tion growth have caused significant degradation of water
quality from an increase in both point source and NPS
pollution. Although the array of pollutants is ominously
broad, increased nitrogen and phosphorus loadings
have resulted in enrichment of back bays, estuaries, and
nearshore waters, contributing to algal blooms, declining
finfish and shellfish populations, diminished recreational
Atlantic
Ocean
Legend
Atlantic Coastal Drainage
Cafra Management Area
Pine Barren Region Including
'-1 ~ Cedar-Pine Fringe
Figure 15. The ACD of New Jersey includes approximately 2,000 square miles of land area from the Manasquan River to Cape May.
123
-------
Figure 16. Aerial photograph of New Jersey illustrating the Barrier Islands and estuary system situated along the Atlantic coast.
124
-------
opportunities, and a variety of other problems (18). A
major source of these nutrients is point source sewage
treatment plants (STPs), but the effluent outfalls of almost
all these STPs discharge into nearshore ocean waters
beyond the barrier islands. Thus, NPS pollutants almost
totally dominate the water quality in the estuaries and
back bays (19, 20).
These NPS pollutants, which rain scours from the land
surface and flushes into coastal waters with each rain-
fall, comprise a largely unmeasured and unmanaged
flux of contaminants. Prior research on coastal water
quality has given considerable attention to NPS pollution
generated from paved or impervious surfaces, particu-
larly roadways and parking lots where hydrocarbons,
metals, suspended solids, biologic oxygen demand
(BOD), and other pollutants have been measured.
Although these NPS pollutants are certainly of concern
in New Jersey's coastal waters, the enrichment issue
has led to a focus on NPS pollution produced when
creating large areas of pervious and heavily maintained
landscape, such as lawns and other landscaped areas,
in the sandy soil context of the coastal area. Typically,
significant quantities of fertilizer and other chemicals,
which are applied on these new pervious surfaces, are
naturally low in nutrients. Although a modest portion of
the applied fertilizer runs off directly into surface waters,
larger quantities of soluble pollutants, such as nitrates
and herbicides, quickly percolate down through the
sandy soil, then move rapidly as interflow to the estuary
system.
In this coastal drainage of unconsolidated sediments,
the hydrologic cycle differs from inland watersheds. Of
the 45-inch average annual rainfall, only a small fraction
(2.5 inches per year) becomes direct runoff, with the
balance rapidly infiltrating into the sand strata (21).
Most of the infiltration that reaches the ground water
(20 inches per year) discharges to surface streams
(17 inches per year) within a few hours following rainfall,
producing a lagging and attenuated hydrograph. This
rapid infiltration, combined with the sand texture of the
soil, has a major bearing on the water quality implica-
tions of new land development. Thus, urbanization of
coastal regions has dramatically altered hydrologic
response, with every square foot of new impervious
surface converting what had been approximately 41.5
inches of infiltration into direct runoff to bays and estu-
aries, with a turbid soup of NPS pollutants.
Even in areas that have maintained infiltration, the
coastal soils do not remove NPS pollutants as efficiently
as other areas of New Jersey that overlie consolidated
formations with heavier clay soils. These soils provide a
much more thorough removal of NPS pollutants through
physical, chemical, and biologic processes, as rainfall
percolates through the soil mantle.
With development of coastal areas, increased impervi-
ous areas and changing flow pathways (inlets and storm
sewers) convey nonpoint pollutants introduced by devel-
opment (from both pervious and impervious surfaces)
directly to the coastal waters. In addition, freshwater
recharge to the underlying aquifer decreases with the
increase in impervious surfaces, with resulting in-
creases in saltwater intrusion into the sand aquifers and
contamination of ground-water supply wells along the
coast. Further compounding the loss of the stormwater
for ground-water recharge are increased ground-water
withdrawals necessary for new watersupply. In sum, urban
growth within the ACD, with its 1.13 million permanent
residents (and still growing) and an additional 1.5 million
summertourists, has dramatically altered the natural drain-
age system (and landscape) in a way that significantly
increases the discharge of NPS pollutants (22).
GIS Approach
New Jersey's Department of Environmental Protection
already had developed a computerized GIS system
(ARC/INFO) for environmental analysis and resource
planning, so this study aimed to use existing GIS work
and to refine this GIS system. Although data files for
municipal boundaries, watershed areas, and a variety of
other factors already existed, land use/land cover data
had not been developed and constituted a major work
task. The subsequent land use/land cover file included
the entire 2,000 square miles of the ACD, but this fo-
cused on the urbanized area (212 square miles) that
occupied about 11 percent of the coastal fringe. The end
product was a polygon file that described about 2,500
polygons of urban/suburban land, each averaging about
0.1 square miles (see Figure 17).
Using aerial photographs combined with USGS base
maps and extensive field reconnaissance, each polygon
was classified by:
Land use type.
Percentage of impervious cover and maintained areas.
Degree of maintenance (fertilization) being provided
to these maintained areas.
Although classifying land use type and extent of imper-
vious cover/maintained areas was a relatively straight-
forward evaluation process (rated within one of 11
categories by percentage, 0 to 5 percent, and so forth),
the third variable, degree of maintenance, required spe-
cial treatment and data development procedures. De-
gree of maintenance was translated into high, medium,
and low categories, with high maintenance exemplified
by golf courses or other intensively maintained areas.
Medium maintenance assumed chemical application
rates comparable with those recommended by Rutgers
University state agronomists. Finally, low maintenance
was typified by a wooded or otherwise naturally vegetated
125
-------
yw.'V MLAMTK COOT*. ORMHWE
couuir BOUHOWIES
OCEANIC I15Y '.)
Figure 17. Urban land use polygons digitized for the New Jersey coastal drainage. The 2,500 polygons shown cover approximately
212 square miles (11 percent) of the ACD area of 2,000 square miles.
126
-------
lot and assumed little or no regular chemical application.
Research staff executed considerable field reconnais-
sance to objectify this judgment-based rating technique
(see Figure 18).
Nonpoint Source Analysis
Because the drainage is almost entirely estuarine, the
hydrologic aspects in this study were almost irrelevant
except as a tool to describe the pollutant transport proc-
ess. Such coastal drainage systems do not allow the
measurement of hydrographs and chemographs (see
Figure 6), except on inland riverine segments or se-
lected infrastructure points of discharge (storm sewer
outfalls). Thus, the NPS analysis focused on the pollut-
ant production and transport process, especially the
nutrients applied to the maintained landscapes, which
are a major part of coastal urbanization.
This study required a great deal of effort to produce an
indextable relating urban cover characteristics (percent-
age impervious, amount of chemical application) to NPS
production potential. For each of the 2,500 urban land
polygons CIS described, estimates of the NPS loading
for a number of pollutants were generated. Potential
loadings were then aggregated by subwatershed. Total
NPS loadings could then be compared with point source
loadings for the entire coastal drainage. Interestingly,
the NPS loading dominated water quality in the estu-
arine drainage while the point sources, discharged by
ocean outfalls to nearshore waters beyond the barrier
islands, were the major source of nutrient pollution in
this portion of the coastal environment (see Figure 19).
Given the estimates of NPS pollution, the major ques-
tion involves how to control or reduce these loads. The
suitability of selected BMPs for the reduction/prevention
of pollutant generation was then evaluated and spatially
identified within the drainage (see Figure 20). This figure
evaluated the use of constructed wetland systems as a
structural measure. That is, CIS allowed state regulators
to identify not only what works best in terms of water
quality protection measures, but also where these meth-
ods could work successfully. This analysis was driven
by a detailed evaluation of the combinations of natural
conditions CIS identified within the study area. For ex-
ample, certain BMPs can be applied on soils that have
a certain set of characteristics (permeability, depth to
seasonal high water table) and that are presently in a
given land cover and planned for urbanization.
CIS also aided in evaluating alternative BMP tech-
niques, including reduction in nutrient applications and
land management BMPs such as elimination of artificial
landscapes, again using a combination of natural fea-
tures and land use patterns (see Figure 21). The result
of this analysis considered the relative proximity of ur-
ban land uses to the coastal waters as significantly
increasing the potential for NPS transport. State regula-
tory programs establish minimum setback criteria for
development in sensitive areas, and these criteria may
be modified to consider pollutant production potential
based on CIS delineation of pollutant production.
New Jersey has been striving to develop NPS manage-
ment programs for coastal areas to reduce existing
sources of pollution as well as prevent the creation of
Legend
R = Residential Use
C = Commercial Use
# = Percent Impervious
H = High Maintenance
M = Medium Maintenance
L = Low Maintenance
Figure 18. Classification of urban polygons by land use, percentage impervious cover, and degree of land fertilization.
127
-------
\
Maryland \
Wildwood
N = 224 Tons
P = 52 Tons r!eVe"'V'^e Ocean City
N = 142 Tons N = 151Ton;
P = 33 Tons
Atlantic County
N = 1,183 Tons
P = 276 Tons
Ocean County Monmouth, Bayshore
Ocean County North Monmouth, NE
Ocean County Centra| N = 859 Tons Long Branch, Deal
N = 871 Tons P = 201 Tons Ocean Asbury Park
P = 203 Tons Soutn Monmouth
South
N = 267 Tons
P = 62 Tons
Atlantic Ocean
N = 1,966 Tons
P = 458 Tons
Figure 19. Point and NFS discharges to the ACD. Data shown are in tons of TP and NOs-N per year.
Legend
BMP Suitability
g- (1) Suitable
H- (2) Generally Suitable
M~ (3) Limited Suitability
Unsuitable Areas
D- (4) Unsuitable Soils
^~ Open Water
E3~ Urbanized Areas
Figure 20. BMP analysis using GIS. Files consider soil suitability, current vegetative cover, and BMP criteria on vacant and
developed land parcels.
128
-------
pollution. As most regulatory agencies have discovered,
NPS management programs can be difficult to imple-
ment, especially when confronting issues of land use
management. To substantiate the need for new manage-
ment programs amidst these controversies, the ability to
document causal linkages (i.e., to generate data and
statistics that make the case for NPS pollutant gener-
ators and resultant water quality degradation) is very
important. The need for documentation of various types
is especially great given the less than perfect data re-
cord of water quality in coastal and other waters. All of
these factors come together to make the value of a CIS
system for water quality management very real.
Conclusion
This CIS-driven analysis indicates that NPS pollutants,
especially the nutrients phosphorus and nitrogen, gen-
erated from fertilized fields or maintained landscapes
surrounding new residential, commercial, and other
types of development in drainage systems, contribute
significantly to water quality degradation. In effect, the
particulate-associated phosphorus and the soluble ni-
trates serve as surrogates for the full spectrum of NPS
pollutants that each rainfall washes from the land. A
comprehensive water quality management program
must include structural measures to remove pollutants
this runoff conveys, as well as management of the con-
tributing landscape to reduce (and perhaps eliminate)
the application of these chemicals within the drainage.
In planning new development, management actions
should occur on a variety of levels or tiers. On an
areawide basis, growth should proceed (with guidance
and management) in a manner that would reduce total
pollutant discharges; therefore, the total amount of
maintained area being created should be as concen-
trated as possible. On the more site-specific level,
measures and construction techniques that reduce the
quantity of pollutants generated are essential. Required
development guidelines must include, but not be limited to:
Prevention of excessive site disturbance and ongoing
site maintenance (described as a policy of minimum
disturbance and minimum maintenance).
Use of special materials for reduction of storm-
water runoff (porous pavement and ground-water
recharge).
Use of stormwater treatment systems (water quality
detention basins, artificial wetlands).
In sum, the regulatory framework must contain both
"how to build" guidelines, as well as "where not to build"
guidelines. CIS can be a powerful tool in both of these
processes.
While inland lakes serve as nutrient traps for these NPS
pollutants, perhaps the greatest potential impact is the
gradual process of excessively enriching our coastal
waters. As population continues to migrate to coastal
areas, the importance of protecting this fragile ecosystem
Figure 21. For certain regulatory criteria, the proximity of land uses to the water's edge was a consideration in BMP selection.
129
-------
increases. The pollution that new land development
generates, including the discharge of point source
wastes, should not be allowed to enter coastal waters;
it should not be allowed to destroy the natural balance
that exists between land and water. The concept of
stormwater management takes on an entirely different
meaning when viewed as one of the basic mechanisms
of this NPS pollution transport. For centuries, engineer-
ing of the shoreline has intensively focused on protect-
ing human developments from the ravages of ocean
storms. Now, however, the converse seems to be
emerging: ocean waters need protection from the im-
pacts of human development.
References
1. Bliss, N., T.H. Cahill, E.B. MacDougall, and C.A. Staub. 1975.
Land resource measurement for water quality analysis. Chadds
Ford, PA: Tri-County Conservancy of the Brandywine.
2. Cahill, T.H., R.W. Pierson, Jr., and B.R. Cohen. 1978. The evalu-
ation of best management practices for the reduction of diffuse
pollutants in an agricultural watershed. In: Lohr, R.C., ed. Best
management practices for agriculture and silviculture. Ann Arbor,
Ml: Ann Arbor Science.
3. Cahill, T.H., R.W. Pierson, Jr., and B.R. Cohen. 1979. Nonpoint
source model calibration in the Honey Creek watershed. #R-
805421-01. Athens, GA: U.S. EPA Environmental Research
Laboratory.
4. Cahill, T.H. 1980. The analysis of relationships between land use
and water quality in the Lake Erie basin. Burlington, Ontario:
International Association of Great Lakes Research.
5. Cahill, T.H., J. McGuire, and C. Smith. 1993. Hydrologic and
water quality modeling with geographic information systems. Pro-
ceedings of the Symposium on Geographic Information Systems
and Water Resources, AWRA, Mobile, AL (March).
6. U.S. EPA. 1991. Remote sensing and GIS applications to nonpoint
source planning. Proceedings of the U.S. EPA Workshop for Re-
gion 5 and Northeast Illinois Planning Commission, Chicago, IL
(October 1990).
7. Baker, D.B., and J.W Kramer. 1973. Phosphorus sources and
transport in an agricultural basin of Lake Erie. Proceedings of the
16th Conference, Great Lakes Research, Ann Arbor, Ml (Septem-
ber).
8. Cahill, T.H., and T.R. Hammer. 1976. Phosphate transport in river
basins. Proceedings of the International Joint Committee on Flu-
vial Transport Workshop, Kitchener, Ontario (October).
9. Cahill Associates. 1989. Stormwater management in the New
Jersey coastal zone. Trenton, NJ: State of New Jersey, Depart-
ment of Environmental Protection, Division of Coastal Resources.
10. Delaware Riverkeeper. 1993. Upper Perkiomen Creek watershed
water quality management plan. Lambertville, NJ: Delaware
Riverkeeper/Watershed Association of the Delaware River.
11. U.S. EPA. 1993. Guidance specifying management measures for
sources of nonpoint pollution in coastal waters. EPA/840/B-
92/002. Washington, DC.
12. Soil Conservation Commission. 1982. TR-20, project formula-
tionHydrology. Technical Release No. 20. PB83-223768. Land-
ham, MD: Soil Conservation Service.
13. Cahill, T.H., M.C. Adams, and W.R. Horner. 1990. The use of
porous paving for groundwater recharge in stormwater manage-
ment systems. Presented at the 1988 Floodplain/Stormwater
Management Symposium, State College, PA (October).
14. Soil Conservation Commission. 1974. Universal soil loss equa-
tion. Technical notes, Conservation Agronomy No. 32. Portland,
OR: West Technical Service Center.
15. Cahill, T.H., M. Adams, S. Remalie, and C. Smith. 1988. The
hydrology of flood flow in the Neshaminy Creek basin, Pennsyl-
vania. Jamison, PA: The Neshaminy Water Resources Authority
(May).
16. BCPC. 1992. Neshaminy Creek watershed stormwater manage-
ment plan, Vol. 1: Policy document, and Vol. II: Plan implemen-
tation. Doylestown, PA: Bucks County Planning Commission
(January).
17. Clark, J. 1977. Coastal ecosystems: Ecological considerations for
management of the coastal zone. Washington, DC: The Conser-
vation Foundation.
18. New Jersey Department of Environmental Protection. 1988. The
state of the ocean: A report by the blue ribbon panel in ocean
incidents. Trenton, NJ.
19. Cahill, T.H., M. Adams, C.L. Smith, and J.S. McGuire. 1991. GIS
analysis of nonpoint source pollution in the New Jersey coastal
zone. With Whitney, S., and S. Halsey, New Jersey Department
of Environmental Protection, Division of Coastal Resources. Pre-
sented at the National Conference on Integrated Water Informa-
tion Management, Atlantic City, NJ (August).
20. Cahill, T.H., M. Adams, C.L. Smith, and J.S. McGuire. 1991.
Living on the edge: Environmental quality in the coastal zone.
With Whitney, S., and S. Halsey, New Jersey Department of
Environmental Protection, Division of Coastal Resources. Pre-
sented at the International Conference on Integrated Stormwater
Management, Singapore (July).
21. Martin, M. 1989. Ground-water flow in the New Jersey coastal
plain. Open File Report 87-528. West Trenton, NJ: U.S. Geologi-
cal Survey.
22. National Oceanic and Atmospheric Administration. 1989. Se-
lected characteristics in coastal states, 1980-2000. Rockville, MD:
U.S. Department of Commerce, National Oceanic and Atmos-
pheric Administration, Strategic Assessment Branch, Ocean As-
sessment Division (October).
130
-------
Using GIS To Identify Linkages Between Landscapes and Stream Ecosystems
Carl Richards, Lucinda Johnson, and George Host
University of Minnesota, Duluth, Minnesota
Introduction
Factors that operate on a variety of temporal and spatial
scales influence the structural and functional compo-
nents of stream ecosystems (1). Quantifying the effects
of factors that operate across multiple scales has chal-
lenged aquatic scientists over the last several decades.
Recently, scientists have recognized that they cannot
successfully protect or restore ecosystem integrity with-
out taking into account all appropriate scales; therefore,
they are focusing on understanding interactions between
terrestrial and aquatic components of entire watersheds
(2). Although awareness of the importance of watershed
and landscape-scale influences on streams is growing,
the tools to examine these influences are still in their
infancy.
Most watershed and landscape studies to date have
focused on the role watershed-scale parameters play on
water chemistry (3-5). These studies usually examined
nutrient and sediment inputs from various watershed
land covers. Methods for evaluating the patterns in the
terrestrial segment of the watershed were awkward and
laborious, involving use of planimeters or cutting and
weighing maps. More recent watershed studies have
attempted to integrate both longitudinal and lateral influ-
ences of the terrestrial ecosystems on water quality in
streams and wetlands (6-8). This approach takes advan-
tage of newly available tools (geographic information
systems and multivariate statistics) for quantifying land-
scape structure.
Relatively few studies have examined how watershed
features influence biological communities. Most studies
examining stream biota have concentrated on single
land use types (9, 10) or on the relationship between
watershed land use and stream physical habitat (11,12).
Typically, study designs have not addressed questions
concerning variability of stream communities over rela-
tively large geographic scales.
Our own work centers on identifying linkages between
landscape features (watershed scale) and stream reach
environments (physical habitat, chemistry), and relating
these parameters to major patterns of community vari-
ation. In this manner, landscape and reach environment
interactions probably control the influence that specific
landscape components have on biological communities
(13). This is the general premise when using biological
communities to assess watershed status. To represent
stream biota, we examine benthic macroinvertebrates,
which have been used extensively for biomonitoring
numerous environmental stresses (14). Macroinverte-
brates are sensitive to watershed conditions and exhibit
sufficient stability in assemblage structure over time to
make them useful as long-term monitors of stream
health (15).
This paper presents an overview of our attempts to
identify the relative strengths of landscape variables on
macroinvertebrate communities. We classify landscape
variables into two general categories. The first category,
geology and landscape structure (GEOS), considers
variables that are fixed on the landscape and are largely
uncontrollable by management activities. The second
category, land use (LU), includes variables that have
anthropogenic origins and may be influenced by land
management activities. By understanding the relative
strengths these two sets of variables possess in deter-
mining community structure, we hope to identify specific
species groups that can act as land use and land form
indicators. We also hope to identify ways to predict the
outcome of specific large-scale land management activi-
ties (e.g., silviculture, agriculture) or other large-scale
environmental changes (e.g., global warming) on stream
ecosystems.
Study Area
This study was conducted in the Saginaw River basin,
a 22,562-square-kilometer watershed in east-central
Michigan (see Figure 1) that flows into Lake Huron.
The Saginaw River watershed was chosen for this
study because its component drainages range from
heavily affected agricultural to relatively pristine areas.
131
-------
Figure 1. The Saginaw basin study area.
Dominating the soils in the lake plain are medium- and
fine-textured loams to clays, with sand found in the
outwash plains and channels. Artificial drainage and tile
systems extensively drain the clay regions. Glacial fea-
tures such as ground moraines and outwash plains are
common. The western sector is characterized by rolling
plains with coarse-textured ground moraines. This re-
gion contains a high percentage of the forested land,
while agricultural land use dominates the eastern sector.
The Saginaw basin covers 16,317 square kilometers,
including four major subbasins: the Tittabawassee
(6,734 square kilometers), Shiawassee (3,626 square
kilometers), Flint (3,108 square kilometers), and Cass
(2,331 square kilometers) Rivers. The Tittabawassee
subbasin further divides into three principal water-
coursesthe Chippewa, Pine, and Tittabawassee Riv-
ers. Watersheds adjacent to Lake Huron (Kawkawlin
and East basins) are characterized by low topographic
relief and elevations averaging 203 and 206 meters,
respectively. The Flint and Chippewa/Pine basins aver-
age about 278 meters in elevation. These drainages
also exhibit the greatest variation in topography.
Study Design
The analysis covers 45 stream sites within the study
area. These sites reflect a gradient of land use and
physiographic conditions in the Saginaw River drainage.
Researchers obtained biological, chemical, and physi-
cal samples at one 200-meter stream segment at each
site. In addition, a geographic information system (CIS)
database was compiled reflecting a number of land-
scape parameters for the watershed of each stream
segment.
Sampling Methods
Macroinvertebrates
At each sampling site, we deployed Hester-Dendy arti-
ficial substrate samplers (16) for macroinvertebrate
community characterizations twice, in early summer and
during base flow conditions in late summer and fall of
1991 or 1992. We allowed samplers to colonize for 6 to
8 weeks. In the laboratory, macroinvertebrates were
counted and identified to genus whenever possible. A
series of derived variables from the original species
abundance tables was used to describe community
characteristics. We chose metrics based on their rela-
tive utility for examining macroinvertebrate communi-
ties, as suggested by Barbour et al. (17) and Karr and
Kearns (18). Because macroinvertebrate assemblages
are relatively stable through time (15) and preliminary
analysis indicated no significant differences between
sampling years at stations for which we had 2 years of
data (unpublished data), we combined macroinverte-
brate data into one database.
Chemistry
We assessed nutrients and other chemical properties
related to water quality at each stream site during sev-
eral periods in the summer and fall of 1991 or 1992.
Stream flow during fall sampling was typically less than
median flow rates and was considered to represent base
flow levels. We used the maximum values of samples
taken in June and July to represent summer conditions
and the maximum values from September and October
to represent fall base flow conditions. The nutrients
measured were ammonium (NHs), nitrate-nitrogen, total
nitrogen (TN), orthophosphate (PO4), and total phos-
phorus (TP). In addition, we assessed alkalinity (ALK),
conductivity, total dissolved solids, and total suspended
solids (TSS). Standard methods were used for all
measurements (19).
Physical Habitat
We assessed physical habitat during base flow condi-
tions at each stream site in a stream reach that is at least
8 to 12 times the width of the stream segment. A suite
of quantitative habitat structure measurements and ob-
servations was made at each site. We derived values
for six general habitat attributes:
Substrate characteristics
Instream cover
Channel morphology
Riparian and bank conditions
132
-------
Riffle/Run quality
Pool quality
Landscape Descriptors
Land use patterns, surficial geology, hydrography, and
elevation databases helped to quantify landscape char-
acteristics in the study area (see Table 1). Land use
patterns were derived from existing digital data at the
Michigan Department of Natural Resources (Michigan
Resource Information System [MIRIS] database) (see
Table 2). We based classification of land use/cover cate-
gories on a modified version of the Anderson (20)
scheme, which was constructed specifically for natural
resource applications. The result was the following nine
land use/cover categories:
Urban
Row crop/agriculture
Other agriculture
Herbaceous range land
Shrubby range land
Nonforested wetlands
Forested wetlands
Mixed hardwood forests
Deciduous forests
In this region, nonrow-crop agriculture is largely repre-
sented by pasture, and range lands are predominantly
abandoned fields (old fields).
The U.S. Department of Agriculture (USDA) STATSGO
soils database enabled the compilation of soil data. The
database consists of U.S. Soil Conservation Service
(SCS) soil surveys and includes information on domi-
nant texture and drainage in large landscape units. We
Table 1. Spatial Data Used for Landscape Characterization
Data Layer Source
aggregated soils into simplified categories based on
glacial landform.
We delineated watershed boundaries for each sampling
station on United States Geological Survey (USGS)
topographic maps and digitized them using ARC/INFO
(Environmental Systems Research Institute [ESRI],
Redlands, California). We identified stream order for
each stream segment and coded it as an attribute of the
stream reach file. All databases were transformed into a
common digital format as necessary and projected into
a common coordinate system. We stored data in vector
format and analyzed them in ARC/INFO.
Table 2 lists the landscape variables we derived for each
watershed. Land use/cover values were reported and
analyzed as a percentage of the total watershed area.
Patch heterogeneity measured landscape fragmenta-
tion and was reported as the number of patches per
hectare. We derived slope from elevation data using
ARC/INFO. The standard deviation of elevation was
used as a surrogate measure of topographic variability.
Statistical Analysis
Using redundancy analysis (RDA), a canonical exten-
sion of principal component analysis (PCA), we de-
tected relationships among the individual multivariate
data sets. RDA is a form of direct gradient analysis that
describes variation in a multivariate data set (e.g., habi-
tat variables or macroinvertebrate metrics) based upon
environmental data (21). In RDA, the station scores from
a PCA are regressed on a specified set of environmental
variables with each iteration, and the fitted values of the
regression become new station scores (22). Thus, envi-
ronmental or predictor variables constrain PCA. Two
important outputs from this method are the interset cor-
relations of environmental variables with the RDA axes,
which indicate the environmental variables that have the
strongest influence in the ordination, and the fraction of
Scale
Format Received
Hydrology
Elevation
Land use/cover
Land use/cover
Watershed boundary
Station locality
Soils
Major basin
Quartenary geology
U.S. EPA stream reach
USGS-DEM
USGS
MIDNR
USGS topographic maps
USGS topographic map
USDA SCS STATSGO
USGS topographic
University of Michigan
1:100,000
1 :250,000
1:100,000
1 :24,000
1 :24,000
1 :24,000
1 :250,000
1 :24,000
1 :250,000
ARC/INFO
Digital elevation model
Digital line graph
Intergraph
Manual delineation
Manual digitizing
ARC/INFO
Manual digitizing
Manual digitizing
Key: USGS = United States Geological Survey
DEM = Digital elevation model
MIDNR = Michigan Department of Natural Resources
USDA SCS = United States Department of Agriculture Soil Conservation Service
133
-------
Table 2. Landscape Variables Measured or Derived for Each
Watershed
Results
Variable
Units
Land use/cover
Slope
Standard deviation elevation
Patch heterogeneity
Soils
Total area
Proportional area
Degrees
Meters
Patches/hectare
Proportional area
Hectares
total variance of each predicted variable that is ex-
plained by the RDA axes (22).
Performing Monte Carlo permutation tests determined
the statistical validity of the association between predic-
tor and predicted variables. Tests were conducted by
random permutation of the site numbers in the predictor
variables. We randomly linked the predictor data to the
predicted data. We conducted 99 simulations to approxi-
mate a normal distribution with which to compare our
data with random combinations.
We first determined which of the reach variables had
strong influences on macroinvertebrate distributions by
conducting separate RDAs with physical habitat and
chemistry variables as environmental descriptors. We
then examined the ability of the landscape data to pre-
dict the variation in the important reach variables.
To determine the relative influences of LU and GEOS
landscape variables on stream chemistry and physical
habitat, we used partial RDA, where one landscape
variable type was held constant and variation due to the
other landscape set was examined independently. Us-
ing this approach, total variation in a multivariate data
set can be decomposed in a manner analogous to
analysis of variance (23, 24). For this analysis, we at-
tributed variation in the reach variables (habitat and
chemistry) to four separate components:
The variation in reach variables that LU variables
explained independently of GEOS variables.
The variation in reach variables that GEOS variables
explained independently of LU variables.
The variation in reach variables that both GEOS and
LU variables shared. This shared variation could
have been due to both the dependence of one type
of variable on the other as well as noncausal rela-
tionships (e.g., the types of soil found in a watershed
determine in large part the types of agriculture that
can be practiced).
The variation in reach variables that were unexplain-
able. This may have been attributable to sampling
error, stochastic variation, or other variables not
sampled.
Regional Characteristics
Land Use
Land use within the study region was dominated by
row-crop agriculture (see Table 3). Individual water-
sheds ranged from 14 to 99 percent in agricultural land
uses, with the East basin watersheds exhibiting the
greatest proportion of agricultural land use and the Flint
having the lowest proportion of agricultural land use.
The Chippewa/Pine and Kawkawlin watersheds exhib-
ited the greatest diversity of land use and cover types
within the study region.
Wetlands represented a minor land use component with
most watersheds having between 0 and 15 percent land
area. The Cass and Kawkawlin basins had the greatest
proportion of wetlands, with a median of 6.8 percent for
individual watersheds.
Macroinvertebrates
Considerable variation existed among the major basins
with respect to the 15 macroinvertebrate community
metrics during summer (see Table 4). Metric values for
the Flint, Shiawassee, and Chippewa watersheds were
similar. Sites within the Kawkawlin and East basins
differed considerably from the Flint, Shiawassee, and
East basins in several of the metrics. The Kawkawlin
watershed was notable for high shredder and filterer
proportions and a low proportion of detritivores. The
East basin also had a high proportion of shredders. Both
the East and Kawkawlin basins had lower proportions of
strictly erosional taxa and higher proportions of deposi-
tional taxa than the other major basins.
Taxa at the East and Kawkawlin basins also exhibited
lower oxygen tolerance than at other major basins. In
addition, their Hilsenhoff Biotic Index (HBI) scores
(which are sensitive to oxygen availability) were higher
than other basins, and they had the lowest EPT (Ephe-
meroptera, Plecoptera, Trichoptera) richness. Total rich-
ness at Kawkawlin was relatively high, however.
Richness was highest in the Chippewa/Pine watershed
and lowest in the East basin.
In general, fall patterns of macroinvertebrate metrics
resembled those of summer. The Kawkawlin and East
basins had high HBI scores, low EPT scores, low pro-
portions of erosional taxa, and high proportions of depo-
sitional taxa. The proportion of predators was
exceptionally high in the Kawkawlin basin due to the
abundance and trophic classification of one chironomid
genus.
134
-------
Table 3. Summary of Landscape Metrics in Six Major Basins of the Saginaw River Drainage
Landscape Variables East Basin Cass Flint Shiawassee Chippewa/Pine Kawkawlin
n
Row crops
Other agricultural land
Urban
Deciduous forest
Mixed hardwood
Range: Herb
Range: Shrub
Forested wetlands
Non-forested wetlands
Slope (degrees)
Elevation (meters)
Patch heterogeneity
Watershed area
(hectares)
8
86.4
79.7-98.0
0.2
0.0-1.2
0.8
0.1-2.0
6.4
0.5-12.1
0.03
0.0-0.3
1.0
0.1-3.6
1.3
0.0-2.3
0.1
0.0-0.5
1.7
0.0-5.3
0.15
0.07
206.5
19.8
241.5
102.0
14,968.9
11,132.0
7
58.4
45.1-73.3
0.6
0.4-1.4
1.9
0.4-3.7
16.9
5.8-33.4
0.3
0.01-1.7
6.7
2.1-12.7
4.7
3.2-7.7
0.1
0.0-0.3
6.7
1.9-14.5
0.29
0.12
239.8
6.7
711.4
250.6
38,117.7
58,321.9
8
38.0
25.6-65.5
3.1
0.6-4.1
8.56
1 .4-23.2
17.5
10.2-20.5
2.6
0.0-3.8
13.0
6.6-14.5
6.5
2.6-10.5
2.3
0.3-3.2
3.9
0.4-5.2
0.41
0.11
277.2
22.5
950.4
228.6
28,926.1
19,236.7
5
43.5
26.7-71 .8
0.4
0.1-2.0
10.3
2.4-15.0
19.2
16.4-30.4
0.7
0.4-0.9
12.6
0.9-17.3
7.3
2.2-10.1
0.7
0.0-3.7
3.6
0.7-5.0
0.27
0.10
252.6
47.1
762.7
223.6
46,530.0
50,798.0
15
48.2
13.9-91.3
4.0
0.3-6.6
2.1
0.8-4.0
23.2
3.2-42.1
5.4
0.1-8.5
6.9
1 .0-9.0
5.6
0.7-9.0
1.0
0.0-2.5
4.3
0.1-7.9
0.35
0.10
278.7
34.9
703.6
185.9
53,704.2
44,872.8
3
26.4
18.5-74.9
1.9
0.2-2.3
1.2
1.0-6.8
56.8
13.2-64.5
0.3
0.1-0.3
2.5
0.7-3.1
3.0
2.7-3.2
1.8
0.1-2.2
5.0
1.3-5.9
0.14
.03
203.1
9.8
519.6
156.7
22,240.2
3,848.0
Land use/cover variables
(slope through watershed
(agricultural land through nonforested wetlands) are reported as median and
area) are reported as mean and standard deviation. Land use/cover represents
range; landscape structure variables
proportional areas of each watershed.
Identification of Important
Reach-Scale Variables
Chemistry
RDA showed that chemical variables explained 26 per-
cent of the variation in macroinvertebrate data in sum-
mer and 33 percent in fall. The most important variables
in summer were TN and TSS (see Table 5). Fall
macroinvertebrate communities were influenced by a
greater number of variables, including NHs, TP, ALK,
and TSS.
Physical Habitat
The 13 physical habitat variables explained 37 percent
of the macroinvertebrate data in summer and 46 percent
of the macroinvertebrate data in fall. In summer, the
percentage of deep pools and canopy extent along with
channel dimensions, such as bank full width (BFW) and
bank full depth (BFD), were the most important vari-
ables. In fall, the percentage of fines and deep pools as
well as canopy extent were among the most important
variables (see Table 6).
Landscape Influences on Surface
Water Chemistry
In summer, the landscape data explained 55 percent of
the variation in chemical variables. The proportion attrib-
utable to LU was larger than that attributable to GEOS
(see Figure 2). The two data types shared 12 percent of
100
Physical
Habitat
' Land Use
Geology/
Structure
Summer Fall
Chemistry Chemistry
\T] Shared
| | Unexplained
Figure 2. Results of variance decomposition from partial RDA.
135
-------
Table 4. Mean and Standard Deviation of Macroinvertebrate Metrics Calculated for Summer Collection Periods for Six Major
Basins of the Saginaw River Drainage
n
Chironomidae
Omnivores
Detritivores
Shredders
Gatherers
Filterers
Grazers
Predators
2 Dominants
Total abundance
HBI
Erosional taxa
Depositional
taxa
Species
richness
EPT taxa
richness
East Basin
8
59.1
35.3
19.4
13.7
57.1
34.1
30.3
33.9
59.8
32.9
27.4
38.1
32.2
34.4
1.5
2.2
64.5
25.5
2077
4951
7.1
1.4
25.9
12.4
35.5
13.2
17.2
4.5
5.0
2.7
Cass
7
57.9
20.3
19.1
7.1
69.7
9.7
18.7
5.1
18.7
5.1
23.4
18.5
13.6
16.2
1.2
1.2
54.3
6.1
650
739
5.6
?
36.1
5.5
23.7
9.6
18.3
9.6
5.7
2.8
Flint
8
45.9
27.9
18.1
13.9
75.3
16.0
10.6
6.0
64.3
12.8
22.4
14.1
26.2
21.0
1.5
2.0
50.3
15.2
574
622
5.6
0.8
35.5
9.5
27.5
11.5
22.1
8.2
7.3
3.0
Shiawassee
5
32.2
28.5
14.4
3.8
79.9
6.4
7.7
6.8
65.0
14.1
18.9
12.8
40.1
22.7
1.0
1.0
54.2
9.5
325
91
6.0
0.5
38.9
14.7
27.0
6.6
20.6
4.7
8.0
2.3
Chippewa/Pine
15
45.5
29.6
21.5
9.8
70.4
9.8
14.4
15.6
65.8
17.2
17.6
15.1
25.5
21.1
1.4
0.8
51.9
11.6
497
230
5.1
1.1
36.1
11.0
25.4
10.7
26.6
3.0
10.0
3.7
Kawkawlin
3
67.1
18.6
22.0
16.7
29.0
26.1
51.0
26.9
39.8
30.1
39.9
35.7
25.4
22.9
1.9
0.8
60.0
21.7
433
297
8.1
0.8
14.9
5.3
52.4
6.5
23.3
4.9
3.3
0.5
the variation. The relationship between LU variables and
chemistry was significant (p < 0.05), and the relationship
between GEOS variables and chemistry was not
significant (p > 0.05).
In fall, variation explained by LU was proportionally less
than during summer (see Figure 2). GEOS landscape
variables explained 25 percent of the total variation
while LU variables accounted for less than 10 percent
of the total variation. LU and GEOS variables shared
approximately 8 percent of the variation. In contrast with
summer, GEOS variables were significant and LU vari-
ables were not significant when examined with the
Monte Carlo test (p < 0.05).
The importance of GEOS and LU variables in explaining
variation in the chemistry variables, as well as the total
amount of variation explained, differed considerably
among the chemistry variables. Figure 3 shows only
summer data. For example, the landscape variables
explained almost 80 percent of the total variation in TN
(see Figure 3). The largest proportion of variance was
explained by the shared influences of GEOS and LU.
Alkalinity, which was also well predicted, however, was
much more influenced by variation attributable to LU
variables. In comparison, LU variables explained less
than 45 percent of the total variance in TP, and the
majority of this variance was attributable to GEOS.
To further examine the influence of specific landscape
variables, we compared the various axes in the signifi-
cant partial ordinations (see Figure 4). In summer, when
we observed a significant effect of LU variables on water
chemistry, forested land covers and nonrow-crop agri-
culture had their greatest influence on TSS and ALK. LU
heterogeneity and shrub vegetative cover most strongly
influenced both TN and NHs. In fall, when GEOS vari-
ables had a significant relationship to water chemistry,
ALK and NHs were more influenced by peat land soils
and watershed size. The proportion of sand and gravel
soils, as well as clays, explained much of the variation
in TN and TSS.
136
-------
Table 5. Chemistry Variables That Had a Correlation (r) of at Least 0.30 With One of the RDA Axes With the Summer or Fall
Ordinations; a Monte Carlo Analysis Indicated That Both Summer and Fall Ordinations Were Significant (p < 0.05)
RDA 1
RDA 2
RDA 3
Variable
Summer
Fall
Summer
Fall
Summer
Fall
TN
NH3
TP
P04
TSS
ALK
0.44
0.05
0.12
-0.04
0.01
-0.08
0.03
0.04
0.40
0.19
-0.31
-0.36
-0.04
0.09
0.07
0.01
0.43
-0.16
-0.07
0.52
0.26
-0.08
0.36
0.09
0.05
0.07
0.25
0.26
0.22
0.15
-0.14
-0.09
0.14
0.34
-0.13
0.2
Table 6. Physical Habitat Variables That Had Correlations (r) Over 0.3 With the Ordination Axes; Results of the Monte Carlo
Simulation Indicated That the Fall but Not the Summer Ordinations Were Significant (p < 0.05)
RDA1
RDA 2
RDA 3
Variable
Percentage of
fines
Percentage of
shallows
Wood
Percentage of
deep pools
Erosion
Maximum
depth
Canopy extent
BFW
BFD
Flood ratio
Summer
0.27
-0.06
0.09
0.50
0.16
0.08
0.75
-0.10
-0.09
0.03
Fall
0.47
0.17
0.34
0.54
0.30
-0.14
0.21
-0.23
-0.13
-0.30
Summer
0.04
-0.28
-0.1
0.16
-0.3
0.08
-0.47
0.52
0.32
0.07
Fall
0.28
-0.11
0.02
0.09
-0.31
0.16
-0.42
0.36
0.02
0.23
Summer
0.18
0.32
0.10
-0.17
0.01
-0.39
0.21
0.07
-0.08
-0.41
Fall
-0.09
-0.38
-0.17
0.26
0.27
0.13
-0.25
0.03
0.22
0.02
Landscape Influences on Physical Habitat
GEOS landscape variables attributed for the largest
portion (22 percent) of the explained variation in physical
habitat variables (see Figure 2). LU variables accounted
for 16 percent of the explained variance. The partial
ordination for GEOS but not LU was significant as the
Monte Carlo procedure determined.
As noted with chemistry variables, there were consider-
able differences in the ability of the landscape variables
to predict individual habitat characteristics. Landscape
variables were best at predicting BFW and least power-
ful for predicting the percentage of deep pools (see
Figure 5). BFWwas influenced predominantly by GEOS
and only minimally by LU. The most influential GEOS
variables for BFW related to watershed area (see Figure 6).
Woody debris was predominantly influenced by LU vari-
ables. The most influential LU variables for woody debris
related to forested wetlands. Flood ratio was intermedi-
ate to these examples. Both sets of landscape variables
shared the largest proportion of explained variance for
this parameter.
Discussion
Our studies demonstrate the distinct influences land-
scape features have on stream macroinvertebrate com-
munities through modifying surface water chemistry and
stream habitat. Land use most strongly influences
stream chemistry during summer months when surface
runoff and soil leaching are greatest. In addition, fertil-
izer application in row-crop agriculture is highest in the
first part of the growing season. The strong relationship
between some aspects of land use and stream water
chemistry were similar to those observed in other stud-
ies (4, 6, 7). The specific mechanism by which stream
chemistry influences macroinvertebrates is not clear.
The addition of nutrients can significantly affect stream
productivity (25-27); however, light often limits primary
production in agricultural areas (28-30). Nutrients may
137
-------
Summer Chemistry
Fall Chemistry
120
100
I 80
60
1 20
o
-------
100
60
.i 40
20
-------
Land Use
Geology/Structure
50,
45
-a 40
\ \
B RDA 1 Watershed Area, SD Elev.
i i RDA 2 Outwash Sand and Gravel
81 RDA 3 Lacustrine Clays
-
\\ Jf « K . n J
i
J
Fine Shal Wood Deep Erod MaxD Cnpy BFW BFD Flrt
Figure 6. Factors influencing physical habitat parameters.
10. Jones, R.C., and C.C. Clark. 1987. Impact of watershed urbani-
zation on stream insect communities. Water Resources Bull.
23:1,047-1,055.
11. Bisson, P.A., R.E. Bilby, M.D. Bryant, C.A. Dolloff, G.B. Grette,
R.A. House, M.L. Murphy, K.V. Koski, and J.R. Sedell. 1987.
Large woody debris in forested streams in the Pacific Northwest:
Past, present, and future. In: Salo, E.O., and T.W. Cundy, eds.
Streamside management: Forestry and fishery interactions. Con-
tribution 57. Institute of Forest Resources, University of Washing-
ton, Seattle, WA. pp. 143-190.
12. Everest, F.H., R.L. Beschta, J.C. Scrivener, K.V. Koski, J.R.
Sedell, and C.J. Cedeholm. 1987. Fine sediment and salmonid
production: A paradox. In: Salo, E.O., and T.W. Cundy, eds.
Streamside management: Forestry and fishery interactions. Con-
tribution No. 57. Institute of Forest Resources, University of
Washington, Seattle, WA. pp. 98-142.
13. Richards, C., and G.E. Host. 1994. Using CIS to examine the
influence of land use on stream habitats and biota. Water Re-
sources Bull. 30:729-738.
14. Rosenberg, D.M., and V.R. Resh. 1993. Freshwater biomoni-
toring and benthic macroinvertebrates. New York, NY: Chapman
and Hall.
15. Richards, C., and G.W. Minshall. 1992. Spatial and temporal
trends in stream macroinvertebrate species assemblages: The
influence of watershed disturbance. Hydrobiologia 241:173-184.
16. Ohio EPA. 1987. Biological criteria for the protection of aquatic
life, Vol. II. User's manual for biological assessment of Ohio sur-
face waters. Columbus, OH.
17. Barbour, M.T., J.L. Plafkin, B.P. Bradley, C.G. Graves, and R.W
Wisseman. 1992. Evaluation of EPA's rapid bioassessment ben-
thic metrics: Metric redundancy and variability among reference
stream sites. Environ. Toxicol. Chem. 111:437-449.
18. Karr, J.R., and B.L. Kearns. 1992. Components of biological in-
tegrity: Their definition and use in development of an invertebrate
IBI. In: Simon, T.P., and WS. Davis, eds. Proceedings of the 1991
Midwest Pollution Control Biologists Meeting: Environmental in-
dicators: Measurement and assessment endpoints. EPA/905/R-
92/003. U.S. EPA Region 5, Environmental Sciences Division,
Chicago, IL. pp. 1-16.
19. APHA. 1989. Standard methods for the examination of water and
waste water. American Public Health Association.
20. Anderson, J.R., E.E. Hardy, J.T Roach, and R.E. Witmer. 1976.
A land use and land cover classification system for use with
remote sensor data. Geological Survey Professional Paper 964.
U.S. Government Printing Office, Washington, DC.
21. ter Braak, C.J., and I.e. Prentice. 1988. A theory of gradient
analysis. Advances in Ecol. Res. 8:271-317.
22. Jongmann, R.H., C.J. ter Braak, and O.F. van Tongeren. 1987.
Data analysis in community and landscape ecology. Wageningen,
The Netherlands: Pudoc.
23. Borchard, D., PL. Legendre, and P. Drapeau. 1992. Partialling
out the spatial component of ecological variation. Ecology
73:1,045-1,055.
24. ter Braak, C.J. 1990. Update notes: CANOCO Version 3.0. Agri-
cultural Mathematics Group, Wageningen, The Netherlands.
25. Petersen B.J., J.E. Hobble, T.J. Corliss, and K. Kriet. 1983. A
continuous-flow periphyton bioassay: Tests of nutrient limitation
in a tundra stream. Limnology and Oceanography 28:583-591.
26. Grimm, N.B., and S.G. Fisher. 1989. Stability of periphyton and
macroinvertebrates to disturbance by flash flooding. Ecol.
Monogr. 52:93-110.
27. Pringle, C.M. 1987. Effects of water and substratum nutrient sup-
plies on lotic periphyton growth: An integrated bioassay. Can. J.
Fish. Aquat. 44:619-629.
28. Munn, M.D., L.L. Osborne, and M.J. Wiley. 1989. Factors influ-
encing periphyton growth in agricultural streams of central Illinois.
Hydrobiologia 174:89-97.
140
-------
29. Bushong, S.J., and R.W. Bachmann. 1989. In situ enrichment
experiments with periphyton in agricultural streams. Hydrobiologia
178:1-10.
30. Wiley, M.J., L.L. Osborne, and R.W. Larimore. 1990. Longitudinal
structure of an agricultural prairie river system and its relation-
ship to current stream ecosystem theory. Can. J. Fish. Aquat.
47:373-384.
31. Vannote, R.L., G.W Minshall, K.W. Cummins, J.R. Sedell, and
C.E. Gushing. 1980. The river continuum concept. Can. J. Fish.
Aquat. 37:130-137.
32. Statzner, B., J.A. Gore, and V.H. Resh. 1988. Hydraulic stream
ecology: Observed patterns and potential applications. J. North
Am. Benthological Soc. 7:307-360.
33. Osborne, L.L., and D.A. Kovacic. 1993. Riparian vegetated buffer
strips in water-quality restoration and stream management.
Freshwater Biol. In press.
34. Sedell, J.R., G.H. Reeves, F.R. Hauer, J.A. Standford, and C.P.
Hawkins. 1990. Role of refugia in recovery from disturbances:
Modern fragmented and disconnected river systems. Environ.
Mgmt. 14:711-724.
141
-------
Nonpoint Source Water Quality Impacts in an Urbanizing Watershed
Peter Coffin, Andrea Dorlester, and Julius Fabos
University of Massachusetts at Amherst, Amherst, Massachusetts
Abstract
As part of the larger Narragansett Bay Estuary Project,
the University of Massachusetts Cooperative Extension
Service contracted with the university's METLAND re-
search team to develop a geographic information sys-
tem (CIS) database, generate watershed-wide maps,
perform analyses, and develop a modeling procedure.
The objective was to educate local officials about the
impacts of development on water quality and to help
local boards minimize the effect of nonpoint sources of
pollution.
Because the receiving waters of the Narragansett Bay
are located far downstream in Rhode Island, the up-
stream communities in Massachusetts are reluctant to
enact measures to improve water resources outside of
their jurisdictions. A CIS was used to create awareness
of existing downstream problems and to show the up-
stream communities how development will ultimately
affect water resources in their own backyards.
To nurture this awareness, a "buildout" analysis was
conducted for an entire upstream subwatershed, the
Mumford River watershed, containing parts of four
towns, and roughly 50 square miles. This buildout was
coupled with a loading model using Schueler's Simple
Method to illustrate the potential impacts of future devel-
opment, and encourage local boards to minimize future
nonpoint sources of pollution.
CIS proved its usefulness by developing customized
maps for each town, by generating several "what if
scenarios showing the impacts of different zoning
changes, by facilitating long-range planning for small
towns without professional staff, and by encouraging a
regional perspective on development issues. The entire
planning process was most successful in creating a
series of partnerships that will continue after the grant
expires. The university shared coverages with the state
CIS agency, creating new coverages not previously
available, specifically soils, ownership, and zoning.
Small towns learned about the potential of the new
technology. Students gained from hands-on experience
with real-world problems. State agencies saw their ef-
forts understood at the local level, especially as they
reorganize on a basin approach and begin to implement
a total mass daily loading (TMDL) procedure to coordi-
nate permitted discharges and withdrawals.
As greater emphasis is placed on controlling nonpoint
sources of pollution, more attention needs to be focused
on local boards, who control land use decisions in New
England.
Introduction
Project Description
Narragansett Bay is a vital resource for southern New
England. The health of its waters is critical to the re-
gional economy, supporting fisheries, tourism, and qual-
ity of life. Increased development along the bay's
shorelines and throughout its drainage basin threatens
the quality of these waters, however. The U.S. Environ-
mental Protection Agency (EPA) recognized the threats
to this important water body and designated the Narra-
gansett Bay under its National Estuary Program in 1985.
Completing a Comprehensive Conservation and Man-
agement Plan (CCMP) for Narragansett Bay took 7
years. The CCMP identified seven priority areas for
source reduction or control, including the reduction of
agricultural and other nonpoint sources of pollution. The
nonpoint source strategy identified United States De-
partment of Agriculture (USDA) agencies, conservation
districts, and other public and private organizations as
having principal roles in nonpoint source management.
Whereas the vast majority of Narragansett Bay lies
within the boundaries of Rhode Island, a significant
portion of its pollution load originates in Massachusetts.
Recognizing that the watershed extends beyond state
boundaries, the USDA provided 3 years of funding to
Cooperative Extension and the Soil Conservation Serv-
ice (SCS) in both Massachusetts and Rhode Island to
142
-------
coordinate their efforts in an innovative attempt to re-
duce the impact of nonpoint sources of pollution on
Narragansett Bay. While water quality is a relatively new
focus for Cooperative Extension, it fits well with the
historic mission of extending the knowledge base of the
land-grant colleges out into the community, and provid-
ing training and capacity building for local officials and
community organizations.
With such a large area of concern, the management
team decided to focus on a smaller subwatershed area
in each state for the first 2 years. The strategy was first
to develop a program for the mitigation of nonpoint
source pollution on the smaller scale of a watershed of
roughly 50 square miles, then take the lessons learned
and apply the most appropriate efforts throughout the
larger watershed. By using similar strategies in Rhode
Island and Massachusetts, but choosing subwatersheds
that differ in terms of location relative to the receiving
water, size, staffing, and sophistication, the two states
gained from each other's experience, sharing the suc-
cessful techniques and avoiding each other's mistakes.
For its pilot study, Rhode Island chose Aquidneck Island,
home of Newport, Portsmouth, and Middletown, with a
special focus on protecting surface water supply reser-
voirs. Massachusetts chose an upstream watershed in
the Blackstone Valley, somewhat rural in character, but
rapidly undergoing a transformation to suburbia.
Watershed Description
The Blackstone River drops 451 feet in its 48-mile jour-
ney from Worcester, Massachusetts, to Pawtucket,
Rhode Island. In the 19th century, this drop of roughly
10 feet per mile was ideally suited to providing power to
mills during the early years of the industrial revolution.
By the Civil War, every available mill site was developed,
earning the Blackstone River the name "The Hardest
Working River."
The Blackstone has a long history of pollution. First, the
textile industry, then steel, wire, and metal finishing in-
dustries used the river for power, in their manufacturing
process, and for waste disposal.
In Massachusetts, the Blackstone River is the major
source of many pollutants to Narragansett Bay. Based
on total precipitation event loading calculations, the
Blackstone River is the principal source of solids, cad-
mium, copper, lead, nitrate, orthophosphate, and PCBs
to the bay (1). The Blackstone River has an average flow
of 577 million gallons per day or 23.2 percent of the
freshwater input to the bay.
The watershed area in Massachusetts equals 335
square miles; with a population of 255,682, this results
in a density of 763 people per square mile. The Black-
stone Valley has 9,000 acres in agricultural use, with
more land in hay (4,500 acres) than crops (3,700 acres)
to support its 4,400 animals.
Based on aerial mapping flown in 1987, the Blackstone
Valley has lost 5 percent of its cropland, 9 percent of its
pasture, and 21 percent of its orchards since 1971. The
valley remains more than 60 percent forested, but that
represents a decrease of 5 percent. The forest and
farmlands were lost to development as low density
housing grew by 45 percent, commercial use grew by
15 percent, and transportation grew by 54 percent.
Waste disposal grew 52 percent to 582 acres, and min-
ing, which in this region represents gravel pits, grew 22
percent to 1,100 acres.
Watershed soils consist mainly of compact glacial till on
rolling topography, with 3 to 15 percent slopes. The river
and stream valleys are underlain by glacial-derived sand
and gravel outwash, which provide drinking water to all
towns in the area except Worcester and support the
large gravel pits. The high clay content in the till soils of
the uplands makes for a high water table, which is
beneficial for growing corn but causes problems for
septic systems.
Following a preliminary study of the subwatersheds, the
Mumford River in the Blackstone Valley was selected as
the focus watershed based on its size, location, land
use, and existing water quality (see Figure 1). The Mum-
ford River watershed has an area of 57 square miles,
with a length of 13 miles, and lies within the towns of
Douglas, Northbridge, Button, and Uxbridge. These
towns share the attributes of small, rural communities
undergoing rapid development, with no professional
planning staff (see Figure 2). According to the 1990
Census, Douglas grew 46 percent in 10 years to 5,438;
Uxbridge experienced 24 percent growth to 10,415; Sut-
ton increased 17 percent to 6,824; and Northbridge grew
9 percent to 13,371.
Project Strategy
Because the generation of nonpoint sources of pollution
is so closely tied to land use, and because local boards
composed of citizen volunteers have principal control
over land use in New England, the key focus of this
program is to train local boards to recognize and begin
managing the threat that nonpoint sources of pollution
pose to water quality. Local planning boards, conserva-
tion commissions, and boards of health address land
use issues and can regulate and shape existing and
proposed development. By developing a program to
train local officials, Cooperative Extension can focus its
outreach where it will have the greatest impact in both
the short and long term. Local boards have the strongest
opportunity to comment on how land is to be used as it
undergoes development. Therefore, this project focused
on preventing future deterioration as opposed to fixing
143
-------
Figure 1. Map of Mumford River watershed study area.
existing problems. This is especially appropriate in a
rapidly urbanizing setting.
Both Massachusetts and Rhode Island chose to utilize
CIS technology because of its ability to store, analyze,
transform, and display geographic, or spatial informa-
tion. Its database management and analytical capabili-
ties make it a useful tool for pollution load modeling and
buildout scenario development, while its mapping capa-
bilities make it an excellent tool for sharing information
with local officials. This paper documents a case study
on how CIS technology was used to apply a watershed-
wide pollution loading model and to develop buildout
scenarios for demonstrating to local officials the poten-
tial impacts of future development on water quality.
This project used CIS in four different applications:
Printing customized, large-scale maps: This most ba-
sic application of a CIS proved the most useful for
local officials. It was a revelation for some officials to
see how their current zoning related to actual land
use. In one town, these maps inspired a change in
zoning to protect the area of a future water supply
reservoir. These maps helped officials see how their
towns fit into the regional picture and how their zoning
and land use affected the adjoining towns, and vice-
versa.
Performing "buildout" analysis: A "buildout" analysis
demonstrates the consequences of existing zoning.
It assumes that all land that can be developed will be
developed at some future date. In essence, it is a
spreadsheet that divides the land available for develop-
ment in each zone by the required lot size, subtracting
Northbridge
Legend
I I Agriculture/Open Land
HU Developed Land
V/\ Forest
gg Unforested Wetlands
fH Lakes and Ponds
Scale = 1:90,000
Douglas
Figure 2. Land use/land cover map of Mumford River watershed.
144
-------
a certain percentage for the road network and steep
slopes. It is best used to evaluate different develop-
ment scenarios, substituting different zoning require-
ments.
Applying a watershed-wide pollutant loading model:
CIS provided the input needed to apply the "Simple
Method" for estimating existing and potential pollutant
loads. Future pollution loading was estimated using
a buildout with existing zoning and again assuming
the implementation of cluster zoning. The Simple
Method was compared in one subwatershed with the
Galveston Bay Method, which accounts for the hy-
drologic class of the soils.
Promoting planning for a greenway: Land use maps
were overlaid with parcel ownership to show the
existing network of preserved open space and to
identify those parcels of land having significant wild-
life habitat and recreational value. In one town, these
maps were used to gain funding for planning a river
walk.
Database Development
The most daunting aspect of using a CIS is the prospect
of spending a great deal of time and money creating a
useful database. Fortunately for Massachusetts, many
of the basic coverages needed for regional planning are
housed in a state agency, MASS CIS, and are available
for a small processing fee. These coverages include
most of what appears on the standard United States
Geological Survey (USGS) map: roads, streams, town
boundaries, as well as watershed boundaries and land
use data generated from the interpretation of aerial pho-
tographs. The university entered into an agreement
whereby we gained access to this data at no charge, in
return for sharing the new coverages that the project
would generate.
New coverages needed for the study included: zoning,
soils, sewer and water lines, and land ownership, or
parcels taken from the assessor's maps. The soils maps
were obtained from the SCS, digitized by hand, then the
scale was converted with a computer program, "rubber-
sheeting," to achieve a uniform scale of 1:25,000. All
other new coverages were transferred onto a USGS
topographical map at a scale of 1:25,000, then digitized
directly into the computer. We obtained elevation data,
but the triangulation process used to convert elevation
data to slopes would require so much time and memory
that, for our purpose, deriving a slope map from the four
classes identified on the soils map was sufficient.
While CIS computer programs are powerful enough to
perform most overlay and analysis functions necessary
in nonpoint source pollution load modeling, database
development and accuracy issues can limit the effec-
tiveness of such modeling. The choice of which model
to use is a function of which data are available for input.
Physics-based distributed models are more precise but
require detailed input parameters, beyond the scope of
this project. The extent of our database limited us to
lumped-parameter empirical models. We chose two
such models, the Simple Method and the Galveston Bay
Method.
GIS Applications
The Simple Method
Schueler (2) developed the Simple Method, one of the
simplest lumped-parameter empirical models. The input
data necessary to compute pollutant loading with the
Simple Method are land use, land area, and mean an-
nual rainfall. Land use determines which event mean
concentration (EMC) values and percentage of impervi-
ousness to use in the computation. The amount of rain-
fall runoff is assumed to be a function of the
imperviousness of various land uses. More densely de-
veloped areas have more impervious surfaces, such as
rooftops and paving, which cause stormwaterto run off
the land rather than be absorbed into the soil. The
Simple Method can generate rough figures for annual
pollutant loading within a watershed and can effectively
show relative increases in pollutant levels as land is
developed.
The formula used in the Simple Method is as follows:
(C) * (A) * (2.72)
(load) = (runoff) * (EMC) * (area)
where:
L = pounds of pollutant load per year
P = rainfall depth (inches) over the desired time
interval (1 year)
Pj = percentage of storms that are large enough to
produce runoff (90 percent)
Rv = fraction of rainfall that is converted into runoff
(Rv = 0.05 + 0.009 (I), where I represents the
percentage of site imperviousness)
C = flow-weighted mean concentration (EMC) of
the targeted pollutant in runoff (milligrams per
liter)
A = area (in acres) of the study region
The Simple Method can be applied using a hand-held
calculator or a computer spreadsheet program. For this
project, the calculations were performed entirely within
the ARC/INFO GIS environment, where the input data
were stored. Results were exported to the Excel spread-
sheet program for presentation purposes.
The application of the Simple Method consists of three
major steps.
145
-------
Step 1: Aggregate Land Uses and Obtain Area Fig-
ures for Land Use Categories Within Each Subbasin
The land use coverage in our database has 21 catego-
ries. For the purpose of applying the Simple Method,
these were aggregated into the following six major cate-
gories, based on development density: undeveloped
forest and other open land, large-lot single-family resi-
dential, medium-density residential, high-density resi-
dential, commercial, and industrial. The aggregated land
use categories were matched with study basins from the
Nationwide Urban Runoff Program (NURP) for the pur-
pose of assigning EMC values.
Step 2: Enter Percentage Imperviousness and Event
Mean Concentrations for Each Land Use Type
The TABLES module of ARC/INFO was used to assign
percentage of imperviousness and EMC values to indi-
vidual land use polygons within the watershed's subbas-
ins. The estimated percentage of imperviousness was
obtained from Schueler's guide to using the Simple
Method (2). EMC values for three pollutantsphospho-
rous, nitrogen, and leadwere taken from selected
NURP study basins and were assigned to the aggre-
gated land uses within the watershed.
Step 3: Input the Simple Method's Mathematical
Loading Formula, Calculate Loading Results for
Each Distinct Land Use Area, and Sum Results by
Watershed Subbasin
Finally, the pollutant load was calculated for each dis-
tinct land use area within the Mumford River watershed
by inputting the loading formula through the TABLES
module of ARC/INFO. The mean annual rainfall figure
was assumed to be that of Worcester, Massachusetts,
or 47.6 inches. After calculating loading figures for phos-
phorous, nitrogen, and lead for each distinct land use
area, these numbers were summed for each watershed
subbasin, using the ARC/INFO frequency table report-
ing capability.
The Galveston Bay Method
As an experiment, we applied the Galveston Bay
Method to one of the subbasins to compare results with
the Simple Method. The slightly more sophisticated
Galveston Bay model considers soil drainage charac-
teristics in addition to land use/imperviousness to deter-
mine rainfall runoff. This method is similar to the Simple
Method, in that amount of rainfall runoff and EMCs for
particular land uses are multiplied by land area to deter-
mine total pollutant load (3). Runoff in this method,
however, is calculated using the USDA SCS's TR 55
runoff curve model. The SCS model calculates runoff as
a function of both land use and soil type. Runoff equals
total rainfall minus interception by vegetation, depres-
sion storage, infiltration before runoff begins, and con-
tinued infiltration after runoff begins (4).
The formula used with the Galveston Bay Method has a
structure similar to that of the Simple Method and is as
follows:
P-0.2[(1000/CN)-10]2
P + 0.8[(100(yCN)-10]
(load) =
where:
(runoff
(EMC)
(area)
L = milligrams of pollutant load per year
P = mean annual rainfall amount
CN = runoff curve number, which is a function of
soil type and land use
EMC = event mean concentration
A = area (in acres) of the study region
The application of the Galveston Bay Method consists
of four major steps.
Step 1: Aggregate Land Uses and Obtain Area Fig-
ures for Land Uses Categories. Aggregate Soils Ac-
cording to Drainage Classes
Land use types were aggregated into the same six major
categories as the Simple Method in orderto match EMC
values and to allow for later comparison of the two
pollution loading methods. Soils were aggregated ac-
cording to drainage classes for use with the USDA SCS
TR 55 runoff formula. The SCS identifies four classes of
soils according to their drainage capacity:
Class A = excessively to well-drained sands or
gravelly sands.
Class B = well to moderately drained, moderately
coarse soils.
Class C = moderately to poorly drained fine soils.
Class D = very poorly drained clays or soils with a
high water table.
Step 2: Overlay Soils Data With Land Use Data and
Clip This New Coverage Within the Subbasin
The ARC/INFO CIS overlay capability was used to over-
lay land use and soils maps for the Mumford River
watershed on top of each other. This created new, dis-
tinct areas of different land use and soils combinations.
Because we were only applying this model in one sub-
basin, the subbasin boundary was used in conjunction
with the ARC/INFO "clip" command to cut out (like a
cookie cutter) that portion of the watershed within the
subbasin.
Step 3: Assign Runoff Curve Numbers and EMC
Values to Each New Land Use/Soils Polygon Within
the Subbasin
EMC values were assigned to each distinct land
use/soils area in the same manner as they were as-
signed to land use areas using the Simple Method.
146
-------
Runoff curve numbers were assigned to each distinct
land use/soils area within the subbasin according to
values established by the USDA SCS.
Step 4: Calculate Loading Results for Each Distinct
Land Use/Soils Area and Sum Results for the Entire
Subbasin
Finally, the pollutant load was calculated for each dis-
tinct land use/soils area within the subbasin by inputting
the loading formula through the TABLES module of
ARC/INFO. After calculating loading figures for phos-
phorous, nitrogen, and lead for each distinct land
use/soils area, these figures were then summed for the
subbasin using the ARC/INFO frequency table reporting
capability. Results of this modeling were converted from
milligrams per acre peryearto pounds per acre peryear
to facilitate later comparisons.
Buildout Scenarios
For planning purposes, CIS is most useful in its ability
to quickly generate alternative scenarios. When these
development scenarios are coupled with a pollutant load
model as described above, alternative scenarios can be
evaluated according to their impact on water quality.
This project generated two different scenarios for each
of the four towns in the watershed: a maximum buildout
with existing zoning and a maximum buildout with clus-
tered development.
Maximum Buildout
A maximum buildout scenario was used to show the
worst case for development according to current zoning
regulations (see Figure 3). The result of this buildout is
expressed both in the number of new residential units to
be built and in the area of land to be converted from
undeveloped to residential and other urban uses.
Step 1: Eliminate From Consideration All Land That
Is Already Developed
Step 2: Eliminate From Consideration All Land That
Is Under Water
Step 3: Eliminate From Consideration All Land That
Is Protected From Development
These protected lands included cemeteries, parks, and
all land permanently restricted from development.
Step 4: Reduce the Remaining Amount of Land by
20 Percent To Account for New Roadways and Ex-
tremely Steep Slopes
The remaining land was considered to have "developa-
ble" status. Wetlands were included in this category
because while a house probably would not be built on a
wetland, wetlands can and often do constitute portions
of the required lot size of large residential lots.
Step 5: Overlay the Land Use Coverage With Zoning
and Minimum Lot Size Information
This created new land use areas as a function of zoning.
All forests and fields were converted to a developed
status.
Step 6: Divide Net Developable Land Area Within
Each Zone by Minimum Lot Size Allowed To Obtain
the Number of New Units
Results from the buildout are expressed in the number
of new units. Results can also be shown spatially by
shading in areas on the map according to future density
Legend
Already Developed
<1/4-Acre Lot Size
1/4-to 1/2-Acre Lot Size
1/2- to 1-Acre Lot Size
1-to2-AcreLotSize
>2-Acre Lot Size
! I Protected/Public Land
Scale = 1:90,000
Figure 3. Maximum buildout scenario within the Mumford River watershed.
147
-------
of development (darker shades for higher density, lighter
shades for lower density).
opment can reduce future levels of water pollution, es-
pecially from nutrients (see Figures 4 and 5).
Clustered Buildout
Another alternative development scenario was gener-
ated assuming the implementation of clustered develop-
ment. All areas zoned for lots larger than 1 acre were
changed to cluster zones, where three-fourths of the
land area remains undeveloped, and the remaining one-
fourth of the land area is developed at a density of
1/2-acre lot size. With clustering, an area zoned for
2-acre house lots still supports the same number of new
units, but three-quarters of the land area remains open
space for passive recreation, protected wildlife habitat,
and as a buffer zone to filter runoff.
Step 1: Select All Land Available for Development
Zoned for 1-Acre Lots or Larger
Step 2: Multiply Selected Land by 0.75 and Add to
the Category of Protected Land
Step 3: Multiply Selected Land by 0.25 and Change
the Minimum Lot Size to One-Half an Acre
Step 4: Divide Step 3 by 20 Percent To Allow for New
Roads and Steep Slopes
Step 5: Divide Step 4 by 21,780 (One-Half an Acre)
To Determine Number of New Housing Units
Results
Lumped-parameter empirical models were chosen for
this project and were applied to watershed subbasins
ranging in size from 1 to 20 square miles and having an
average of 4 square miles. The application of the Simple
Method to existing land use conditions allowed for a
comparison of the Mumford River watershed's subbas-
ins for the purpose of identifying the subbasins that
contribute the highest levels of pollutants per acre per
year. The development of a maximum buildout scenario
identified those areas within the watershed that will
sustain the greatest amount of new growth. The appli-
cation of the Simple Method to this maximum buildout
scenario revealed that pollutant levels in surface water
runoff would increase substantially for all subbasins in
the watershed. This finding supports the theory of a
positive relationship between development and in-
creased pollutant levels from surface water runoff.
The development of a customized buildout scenario for
future development identified those areas that are cur-
rently zoned for large-lot residential "sprawl" and that
can support higher development density under cluster
zoning, while protecting a significant amount of open
space that can support a variety of beneficial uses. The
application of the Simple Method to the customized
buildout scenario revealed that the use of cluster devel-
Results determined by applying the Galveston Bay
Method to one subbasin were compared with those
obtained using the Simple Method. The predicted pollut-
ant loading from current conditions differed significantly
Nitrogen Loading Estimates
Total
Whitin Reservoir
West Sutton
Wallis Pond
Tuckers Pond
Stevens Pond
Rivulet Pond
Mouth
Morse Pond
Manchaug Pond
Lackey Pond
East Douglas Jl
Cross Street
Badluck Pond
50 100 150
Percent Change
200
n 1985-Cluster
Buildout
1985 - Maximum
Buildout
Figure 4. Chart showing difference in simple method results
for nitrogen loading between maximum and custom-
ized buildout scenarios.
Phosphorus Loading Estimates
Total
Whitin ._-
Reservoir
West Sutton
Wallis Pond
Tuckers Pond
Stevens Pond
Rivulet Pond
Mouth
Morse Pond
Manchaug Pond
Lackey Pond
East Douglas
Cross Street
Badluck Pond
I
-10 0
10
20 30 40
Percent Change
50
60
70
1985-Cluster
Buildout
1985 - Maximum
Buildout
Figure 5. Chart showing difference in simple method results
for phosphorus loading between maximum and cus-
tomized buildout scenarios.
148
-------
between the two methods; the Simple Method consis-
tently predicted five times the amounts generated by the
Galveston Bay Method.
When the two methods were applied to both the maxi-
mum and customized buildout scenarios, however, the
percentage growth of predicted pollutant loadings was
remarkably similar for both methods; the Simple Method
consistently predicted loadings 10 to 15 percent greater
than the Galveston Bay Method. This indicates that
while the Galveston Bay Method may provide more
accurate results in predicting actual pollutant loading,
the Simple Method is adequate enough for evaluating
and comparing different development scenarios (see
Figures 6 and 7).
Discussion
As states begin to implement a TMDL approach to regu-
lating water quality, they face the quandary of how to
determine the extent of nonpoint source pollution in the
rivers. The crudest method is to subtract from the total
load those quantities generated by point sources and
call all the rest nonpoint source. While this is appropriate
in some settings, it is unacceptable in a watershed with
a long history of pollution because a significant source
of pollution is the resuspension of historical sediments
stirred up by storms. The situation demands the devel-
opment of a model to predict the loading from nonpoint
sources. Only a computer can handle the multiple fac-
tors that interact to generate nonpoint sources of pollution.
As greater emphasis is placed on watershed planning,
the abilities of a CIS to input, store, manipulate, analyze,
and display geographic information become indispensa-
ble. As the scientific community improves its knowledge
base for determining the critical factors influencing non-
point source pollution, CIS technology is improving in
its ability to store and handle large amounts of data.
While a detailed, physics-based distributed model would
be more accurate than the lumped-parameter models
used for this project, they are difficult to apply at the
watershed scale. The real limiting factor is the provision
of all the data coverages needed to apply complex mod-
els. Lumped-parameter models, such as the Simple
Method and the Galveston Bay Method, are ineffective
for accurately predicting pollutant loads, but they are
suitable for comparing and evaluating alternative devel-
opment scenarios.
Time, and the development community, will not wait until
all the answers are known. Local officials continue to
approve development with no thought to the impacts on
water quality. These officials need to be informed about
the implications of haphazard growth. A CIS, with its
ability to generate customized maps and quickly evalu-
ate alternative development scenarios, is a powerful tool
to help local officials visualize how the decisions they
Nitrogen Loading
Total
Uxbridge
Sutton
Northbridge
50 100 150
Percent Change
200
D Galveston Method Percent
Change 1985 to Cluster
ED Simple Method Percent
Change 1985 to Cluster
Galveston Method Percent
Change 1985 to Maximum Buildout
Simple Method Percent
Change 1985 to Maximum Buildout
Figure 6. Chart showing difference between Simple Method re-
sults and Galveston Bay Method results for nitrogen
loading.
Lead Loading
Total
Uxbridge
Sutton
Northbridge
50 100 150
Percent Change
200
D Galveston Method Percent
Change 1985 to Cluster
E3 Simple Method Percent
Change 1985 to Cluster
Galveston Method Percent
Change 1985 to Maximum Buildout
Simple Method Percent
Change 1985 to Maximum Buildout
Figure 7. Chart showing difference between Simple Method re-
sults and Galveston Bay Method results for lead
loading.
make on paper today will have an impact on the land
tomorrow.
References
1. Metcalf & Eddy, Inc. 1991. Assessment of pollutions in Narragan-
sett Bay. Draft report to the Narragansett Bay Project.
2. Schueler, T. 1987. Controlling urban runoff: A practical manual for
planning and designing urban BMPs. Washington, DC: Metropoli-
tan Washington Council of Governments.
149
-------
3. Newell, C., et al. 1992. Characterization of nonpoint sources and
loadings to Galveston Bay, Vol. I, Technical Report: Galveston Bay
National Estuary Program (March).
4. U.S. Department of Agriculture (USDA) Soil Conservation Service.
1975. Urban hydrology for small watersheds. Technical Release
55. Springfield, VA: U.S. Department of Agriculture.
Additional Reading
Arnold, C.L., et al. 1993. The use of Geographic Information System
images as a tool to educate local officials about the land use/ water
quality connection. Proceedings of Watershed Conference, Alexan-
dria, VA.
Chesebrough, E. 1993. Massachusetts nonpoint source management
plan. Massachusetts Department of Environmental Protection, Office
of Watershed Management (October).
Joubert, L, et al. 1993. Municipal training for water quality protection.
Contribution #2845. College of Resource Development, University of
Rhode Island/Rhode Island Agricultural Experiment Station.
Massachusetts Geographic Information System. 1993. MassGIS
Datalayer descriptions and guide to user services. Boston, MA: Ex-
ecutive Office of Environmental Affairs.
McCann, A., et al. 1994. Training municipal decision-makers in the
use of Geographic Information Systems for water resource protection.
Contribution #2927. College of Resource Development, University of
Rhode Island/Rhode Island Agricultural Experiment Station.
Narragansett Bay Project. 1992. Comprehensive conservation and
management plan: A summary (January).
U.S. EPA. 1983. Results of the Nationwide Urban Runoff Program
(NURP), Vols. I and II. Final report prepared by Water Planning Divi-
sion. NTIS PB84185537. Springfield, VA: National Technical Informa-
tion Service.
U.S. EPA. 1992. Compendium of watershed-scale models for TMDL
development. E PA/841 /R-92/002. Washington, DC (June).
150
-------
A GIS for the Ohio River Basin
Walter M. Grayman
W.M. Grayman Consulting Engineers, Cincinnati, Ohio
Sudhir R. Kshirsagar
Global Quality Corporation, Cincinnati, Ohio
Richard M. Males
RMM Technical Services, Inc., Cincinnati, Ohio
James A. Goodrich
Risk Reduction Engineering Laboratory, Office of Research and Development,
U.S. Environmental Protection Agency, Cincinnati, Ohio
Jason P. Heath
Ohio River Valley Water Sanitation Commission, Cincinnati, Ohio
Abstract
Much of the information used in the management of
water quality in a river basin has a geographic or spatial
component associated with it. As a result, spatially
based computer models and database systems can be
part of an effective water quality management and
evaluation process. The Ohio River Valley Water Sani-
tation Commission (ORSANCO) is an interstate water
pollution control agency serving the Ohio River and its
eight member states. The U.S. Environmental Protec-
tion Agency (EPA) entered into a cooperative agreement
with ORSANCO to develop and apply spatially based
computer models and database systems in the Ohio
River basin.
Three computer-based technologies have been ap-
plied and integrated: geographic information systems
(GIS), water quality/hydraulic modeling, and database
management.
GIS serves as a mechanism for storing, using, and
displaying spatial data. The ARC/INFO GIS, EPAs
agencywide standard, was used in the study, which
assembled databases of land and stream information for
the Ohio River basin. GIS represented streams in hydro-
logic catalog units along the Ohio River mainstem using
EPAs new, detailed RF3-level Reach File System. The
full Ohio River basin was represented using the less
detailed RF1-level reach file. Modeling provides a way
to examine the impacts of human-induced and natural
events within the basin and to explore alternative strate-
gies for mitigating these events.
Hydraulic information from the U.S. Army Corps of En-
gineers' FLOWSED model enabled EPAs WASP4 water
quality model to be embedded in a menu-driven spill
management system to facilitate modeling of the Ohio
River mainstem under emergency spill conditions. A
steady-state water quality modeling component was
also developed under the ARC/INFO GIS to trace the
movement and degradation of pollutants through any
reaches in the RF1 representation of the full Ohio River
basin.
Database management technology relates to the stor-
age, analysis, and display of data. A detailed database
of information on dischargers to the Ohio River mainstem
was assembled under the PARADOX database manage-
ment system using EPAs permit compliance system as
the primary data source. Though these three technolo-
gies have been widely used in the field of water quality
management, integration of these tools into a holistic
mechanism provided the primary challenge of this study.
EPAs Risk Reduction Engineering Laboratory in Cincin-
nati, Ohio, developed this project summary to announce
key findings of the research project, which is fully docu-
mented in a separate report of the same title.
151
-------
Introduction
During the past 25 years, computers have been actively
used in water quality management, demonstrating their
potential to assist in a wide range of analysis and display
tasks. Technologies such as geographic information
systems (CIS), database management systems (DBMS),
and mathematical modeling have been applied in the
water quality management field and have proven to be
effective tools. For computers to achieve their full poten-
tial, however, they must become integrated into the
normal programmatic efforts of agencies and organiza-
tions in the planning, regulation, and operational areas
of water quality management.
Recognizing this need for routine use of computer-
based tools, the Ohio River Valley Water Sanitation
Commission (ORSANCO) and the Risk Reduction En-
gineering Laboratory (RREL) of the U.S. Environmental
Protection Agency (EPA) commenced a study in 1990.
The goals of the study included the adaptation, devel-
opment, and application of modeling and spatial data-
base management (DBM) tools that could assist
ORSANCO in its prescribed water quality management
objectives. These goals were consistent with EPAs on-
going programs involving the use of CIS and modeling
technology. The study's goals also coincided with EPAs
Drinking Water Research Division's work over the past
decade, which applied similar technology to study the
vulnerability of water supplies on the Ohio and Missis-
sippi Rivers to upstream discharges.
Methodology Overview
To address the goals of this project, three basic tech-
nologies have been applied and integrated: CIS, water
quality/hydraulic modeling, and DBM. CIS serves as a
mechanism for storing, using, and displaying spatial
data. Modeling provides a way to examine the impacts
of human-induced and natural events within the basin
and to explore alternative strategies for mitigating these
events. DBM technology relates to the storage, analysis,
and display of data. Though these three technologies
have been widely used in the field of water quality
management, integration of these tools into a holistic
mechanism provided the primary challenge of this study.
CIS Technology
The guiding principle in developing the CIS capability
was to maximize the use of existing CIS technology and
spatial databases. The study used ARC/INFO CIS,
EPAs agencywide standard. Remote access of
ARC/INFO on a VAX minicomputer facilitated the initial
work. Subsequently, both PC ARC/INFO and a worksta-
tion-based system were obtained.
EPA has developed an extensive spatial database re-
lated to water quality and demographic parameters. This
served as the primary source of spatial data for the
study. Following is a summary of spatial data used in
this study:
State and county boundaries.
City locations and characteristics.
Water supply locations and characteristics.
Locations and characteristics of dischargers to
water bodies.
Toxic loadings to air, water, and land.
Dam locations and characteristics.
Stream reaches and characteristics.
The primary organizing concept for the water-related
information was EPAs Reach File System (1). This sys-
tem provides a common mechanism within EPA and
other agencies for identifying surface water segments,
relating water resources data, and traversing the nation's
surface water in hydrologic order within a computer envi-
ronment. A hierarchical hydrologic code uniquely identi-
fies each reach. Information available on each reach
includes topological identification of adjacent reaches,
characteristic information such as length and stream
name, and stream flow and velocity estimates. The origi-
nal reach file (designated as RF1) was developed in the
early 1980s and included approximately 70,000 reaches
nationwide. The most recent version (RF3) includes
over 3,000,000 reaches nationwide.
As part of this project, an RF1-level database was es-
tablished for the entire Ohio River basin. The RF3 reach
file was implemented for the Ohio River mainstem and
lower portions of tributaries. River miles along the Ohio
River were digitized and established as an ARC/INFO
coverage to provide a linkage between the reach file and
river mile indexing used by ORSANCO and other agen-
cies along the river. Figure 1 shows the RF1 reach file
representation of the Ohio River basin along with state
boundaries.
The study incorporated several EPA sources of informa-
tion on dischargers to water bodies. The industrial facil-
ity discharger (IFD) file contains locational and
characteristic data for National Pollutant Discharge
Elimination System (NPDES) permitted discharges. De-
tailed permit limits and monitoring information was ac-
cessed from the permit compliance system (PCS). The
toxic release inventory (TRI) system includes annual
loading of selected chemicals to water, land, air, and
sewer for selected industries based on quantity dis-
charged. All water data are referenced to the NPDES
permit number, which is spatially located by reach and
river mile, and by latitude and longitude.
152
-------
0 meter,
147884
Figure 1. RF1 reaches in the Ohio River basin.
Spill Modeling
An important role that ORSANCO fills on the Ohio River
relates to the monitoring and prediction of the fate of
pollutant spills. Typically, ORSANCO serves as the over-
all communications link between states during such
emergency conditions. ORSANCO coordinates and par-
ticipates in monitoring and serves as the information
center in gathering data and issuing predictions about
the movement of spills in the river. In the past, a series
of time-of-travel nomographs, based on National
Weather Service flow forecasts, Corps of Engineers
flow-velocity relationships, and previous experience,
were used to predict the movement of spills. This project
combined a hydraulic model with a water quality model
to serve as a more robust method for making such
predictions.
The U.S. Army Corps of Engineers' FLOWSED model
was selected as the means of predicting daily flow quan-
tities and water levels along the mainstem and portions
of major tributaries near their confluence with the Ohio
River (2). The Ohio River Division of the Corps of Engi-
neers applies FLOWSED daily as part of its reservoir
operations program. The Corps can generate 5-day
forecasts of stage and flow for 400 mainstem and tribu-
tary segments, and ORSANCO can access the results
via phone lines.
EPAs WASP4 water quality model was selected for use
in the project (3). WASP4 is a dynamic compartment
model that can be used to analyze a variety of water
quality problems in a diverse set of water bodies. Because
the primary use of the model in this project is quick
response under emergency situations, only the toxic
chemical portion of the model with first order decay is
being used. The FLOWSED and WASP4 models have
been combined into a user-friendly spatial decision sup-
port system framework described later in this project
summary.
Discharger Database Management System
EPA's PCS and historical records maintained by
ORSANCO furnish a rich source of data on dis-
charge information for the Ohio River. To organize
these data and make them available for analysis, a
database was developed using the PARADOX DBM
system.
The database was established using a relational struc-
ture with a series of related tables (two-dimensional flat
files). Individual tables contain information on facilities,
outfalls, permit limits, monitoring data, and codes used
in the other tables. The NPDES permit number is used
as the primary key in each data table. A mechanism for
downloading and reformatting data from the national
PCS database has been developed along with custom
forms for viewing and editing data, and custom reports
for preparing hard copy summaries. Latitude and longi-
tude values for each facility can provide the locational
mechanism for use of this data in conjunction with CIS.
Integration of GIS/Modeling/Database
Technologies
A major objective of this study was the integration of
CIS, modeling, and DBMS technologies into a holistic
tool for use by ORSANCO. Several integration mecha-
nisms were implemented as summarized below.
Steady-State Spill Tracing
The NETWORK component of the ARC/INFO CIS pro-
vides a steady-state, transportation-oriented routing ca-
pability. This capability was used in an arc macro
language (AML) program to construct a routing proce-
dure for determining downstream concentrations and
travel times. The pollutant may be treated as a conser-
vative element or represented by a first order expo-
nential decay function. This capability has been
implemented for use with the RF1 reach file repre-
sentation of the full Ohio River basin. The user may
select from six flow regimens: average flow, low flow,
and four multiples of average flow ranging from one-
tenth to 10 times average flow. This system gives
ORSANCO the ability to estimate the arrival time of a
spill from any RF1 tributary to the Ohio River mainstem.
153
-------
Sp/7/ Management System
A PC-based spatial decision support system (SDSS)
was built as a spill management system to be a quick
response tool for analyzing and displaying the results of
pollutant spills into the Ohio River. The schematic in
Figure 2 illustrates the components in this computerized
spill management system. The system is implemented
in the C language using a commercial menuing system
and a series of graphic display routines developed at
EPA. Custom, written routines have been used to read
the output from the U.S. Army Corps of Engineers'
FLOWSED model, to generate input files for EPAs
WASP4 model, to create output reports and output plots,
and to provide an animated representation of the con-
centration profiles moving down the river. Figure 3 pre-
sents an example of a graphic output the system
generated. Additionally, the system generates a file in
5-DAY RIVER FLOW
AND STAGE FORECAST
PREPARED
DAILY BY CORPS OF
ENGINEERS USING
FLOWSED MODEL
hooocH
via phone line
ARC/INFO CIS
DATABASE
SPATIAL DECISION SUPPORT SYSTEM
FOR TOXIC SPILL MODELING
Menu driven user interface linked to
EPA WASP 4 water quality model.
Tabular /graphical screen and hard
copy outputs, spill animation and
CIS DBF format output file
Arc View Spatial data
base display system
GRAPHS
CONCENTRATION
REPORT
by
Segment & Time
REPORTS
Figure 2. Schematic representation of spill modeling system process.
Figure 3. Graphic output from the basinwide network spill model.
154
-------
DBF format that may be read by ARC/VIEW (the com-
panion software to ARC/INFO for user-friendly viewing
of spatial data).
Hardware Platform
Within the study, the initial hardware platform was a
combination of local PCs (in Cincinnati) and a remote
access terminal to a VAX computer located at EPA's
National Computer Center in Research Triangle Park,
North Carolina. The final platform, and the one on which
the completed system was installed, comprised a UNIX-
based Data General workstation and a PC workstation.
The full hardware configuration is shown schematically
in Figure 4.
Conclusions
The application of computer-based display, analysis,
and modeling tools in conjunction with CIS technology
proved to be an effective strategy for water quality man-
agement. This study used an existing CIS package and
DBMS in conjunction with existing water quality and
hydraulic models. The study focused primarily on as-
sembling available spatial and relational databases and
integrating the systems to provide a usable, effective tool.
References
1. Horn, R.C., and W.M. Grayman. 1993. Water-quality modeling with
EPA reach file system. J. Water Res. Planning and Mgmt.
119(2):262-274.
2. Johnson, B.H. 1982. Development of a numerical modeling capa-
bility for the computation of unsteady flow on the Ohio River and
its major tributaries. Vicksburg, MS: U.S. Army Engineer WES.
3. Ambrose, R.B., T.A. Wool, J.P. Connolly, and R.W Schanz. 1988.
WASP4, a hydrodynamic and water quality model: Model theory,
users' manual, and programmers' guide. Athens, GA: EPA Envi-
ronmental Research Laboratory. NTIS PB88185095.
UNIX WORKSTATION
1.44MB
FLOPPY
DRIVE
DUAL
CPU
/
150MB
TAPE
DRIVE
8 MM
DAT
2GB
HARD
DRIVE
1/4" TAPE
DRIVE
650MB
HARD
DRIVE
Figure 4. Hardware configuration.
155
-------
Nonpoint Source Pesticide Pollution of the Pequa Creek Watershed,
Lancaster County, Pennsylvania: An Approach Linking
Probabilistic Transport Modeling and GIS
Robert T. Paulsen
The Paulsen Group, Bowie, Maryland
Allan Moose
Southampton College, Southampton, New York
Abstract
The U.S. Environmental Protection Agency (EPA) has
mandated that each state prepare a state management
plan (SMP) to manage pesticide residues in the state's
environment. One aspect of an SMP involves identifying
specific soils and sites that may be vulnerable to the
transport of pesticides into water resources. A recently
developed system identifies vulnerable areas by cou-
pling probabilistic modeling that uses the Pesticide Root
Zone Model (PRZM) with a desktop geographic informa-
tion system (GIS-MAPINFO). A limited test of this sys-
tem succeeded in identifying and mapping individual soil
series in a watershed that were shown to have trans-
ported atrazine to surface and ground water.
During this project, various digital data sources were
evaluated for availability and ease of use, including:
STATSGO.
U.S. Geological Survey (USGS) digital line graphs
(DLGs).
National Oceanic and Atmospheric Administration
(NOAA) climate data.
This study documents hands-on hints and tricks for
importing and using these data.
From 1977 to 1979, the USGS measured the movement
of atrazine off fields of application into water resources
in the Pequa Creek basin in Lancaster County, Pennsyl-
vania (1). Atrazine in surface water appeared at levels
exceeding 20 parts per billion in storm flow and above
the 3 parts per billion maximum contaminant level (MCL)
during base flow from Big Beaver Creek, a tributary to
Pequa Creek. Each soil series in the subbasin was
digitized into a GIS. PRZM allowed simulation of runoff,
erosion, and leaching of atrazine (applied at 2.24 kilo-
grams per hectare in conventionally tilled corn) for each
soil. This process included simulating each soil under
different slopes for an 11 -year period from 1970 to 1980.
Interpreting the results for each soil series determined
the probability distribution of atrazine in kilograms per
hectare for each mode of transport. GIS used these data
to thematically map each soil series for atrazine loss.
The results of this demonstration project suggest that
the Manor silt loam, with slopes varying from 6 percent
to 20 percent, had a high potential to transport atrazine
residues to surface water. This type of analysis could
suggest that this soil series be:
Farmed using conservation tillage.
Managed to install grass waterways or buffer strips
to stop runoff.
Set aside from production to protect water resources.
Digital databases were available for the study area, but
many technical problems were encountered in using the
data. Researchers embarking on these types of model-
ing and GIS projects should prepare themselves for
significant expenditures of time and finances.
Introduction
A significant volume of published literature documents
pesticide residues in ground water, and the volume of
investigations of residues in surface water is expanding.
The growing acceptance of immunoassay techniques
for the determination of pesticide residues in water has
given the field of pesticide monitoring an accurate and
economical analytic methodology. This will result in an
increase in monitoring capability at the federal, state,
156
-------
local, and university levels. These increases in monitor-
ing capability have documented and will continue to
document the occurrence of pesticides in water re-
sources as the result of past transport through the soil
profile. The U.S. Environmental Protection Agency
(EPA) has mandated that each state prepare a state
management plan (SMP) to manage pesticide residues
in the state's environment. Lacking, however, is a reli-
able pesticide screening technique to indicate which
soils, on a countywide scale, may be sensitive to the
transport of a specific pesticide to deep within the soil
profile or to the surface water resources. These assess-
ments would greatly supplement the usability and valid-
ity of SMPs.
Electronic databases such as State Soils Geographic
(STATSGO), Data Base Analyzer and Parameter Esti-
mator (DBAPE), or the SOILS subsets found in Nitrate
Leaching and Economic Analysis Package (NLEAP)
provide easy access to detailed soil data and model
input estimator subroutines, thereby simplifying data en-
try to numerical models. Two groupings define soils: soil
series and soil associations. Soil series are the individ-
ual soil taxa found in a field. Soil associations represent
groups of soil series, usually three or four soil series
occurring together in an area, and are mapped as a
single unit on a county scale. Mapping of most soil
associations across the United States is complete, with
open access to the county scale maps. A digital soils
mapping data set called SSURGO contains many of the
soil series maps for the United States. Climatologic
databases also provide easy access to long-term data
from the National Oceanic and Atmospheric Administra-
tion (NOAA) weather stations, allowing a user the op-
portunity to input realistic climate data to pesticide
transport models.
Many numerical pesticide transport models, such as the
Pesticide Root Zone Model (PRZM), Ground-Water
Leaching Effects of Agricultural Management Systems
(GLEAMS), and Leaching Estimation and Chemistry
Model (LEACHM), can produce transport estimates for
specific pesticides in specific soils. Each model has its
own strengths and weaknesses, and detailing these
characteristics is beyond the scope of this paper. Sev-
eral authors, however, have described comparisons be-
tween models (e.g., Smith et al. [2], Mueller et al. [3],
and Pennell et al. [4]). These numerical models all gen-
erally require extensive site-specific soil, agronomic,
and climatologic databases. The results from these
models are extremely detailed. Their pesticide transport
estimates, however, are only valid for those locations for
which site-specific data are sufficient to allow calibration
of the model. Applying such site-calibrated model results
to larger scales (county scales) is inappropriate.
In one procedure, a user could identify soils and use
transport models that may have a limited ability to retard
pesticides from reaching water resources. This type of
modeling has recently been called probabilistic model-
ing (5). The concept behind this procedure is to use an
existing transport model, such as PRZM, and vary cer-
tain input parameters (e.g., slope, organic content, pes-
ticide Koc) to produce a probability of a given output
being equaled or exceeded. For example, a PRZM
model could be created for a soil series with an average
organic content of 1 percent and a slope of 8 percent in
the eastern corn belt. The model would use the 30 years
of historical climate data for a nearby station. The model
would vary the organic content and surface slope within
given ranges for the soil series and run 1,000 simula-
tions. The analysis could then entail plotting the results
(i.e., monthly runoff loads, erosion loads, and leaching
through the root zone) in a frequency diagram and gen-
erating probability curves. This analysis would allow the
user to estimate the anticipated pesticide losses, runoff,
erosion, and leaching for any given soil in the county.
The soils with greater probabilities for pesticide loss
could be identified and mapped using CIS.
Recent advances in computing speed and efficiency
have reduced the amount of time and expense needed
to run numerical pesticide transport models. This makes
it possible, in a relatively short amount of time, to quan-
titatively model not just one soil series, but the hundreds
of major soil series that occur in an entire state (e.g.,
357 major soil series combined into 464 different soil
associations in Wisconsin). This type of model can be
very useful to the development of SMPs as well as to a
variety of users, including pesticide registrants, bulk
pesticide handlers, custom appliers, county agricultural
extension agents, and individual growers.
Objective
The objective of this study was to use probabilistic mod-
eling analyses and a geographic information system
(CIS) to determine which soil series in a watershed may
contribute to nonpoint source pollution through runoff of
agricultural chemicals. Specifically, the study aimed to
locate a watershed with historical atrazine runoff, map
the soils, and perform transport modeling using histori-
cal precipitation. The results of this procedure would
determine which soil series had a high potential to con-
tribute to the nonpoint source pollution of the watershed.
Once the transport modeling was completed, a CIS
would help map the distribution of the sensitive soil
series. The mapping would act as a base for implement-
ing best management practices (BMPs) to reduce non-
point source pollution.
Background
Ward (1) described the water quality in the Pequa Creek
basin in Lancaster County, Pennsylvania, for the years
1977 through 1979. Flow from Pequa Creek (154-square-
157
-------
Figure 1. Location of Pequa Creek basin, Lancaster County,
Pennsylvania.
mile drainage area) eventually discharges into the
Chesapeake Bay (see Figure 1). The data collection
efforts (6) documented the occurrence of atrazine
(2-chloro-4-ethylamino-6-isopropylamino-s-triazine),
a commonly used herbicide for weed control in corn-
growing regions, and other agrichemicals in both base-
flow and storm-flow conditions of the Pequa Creek. A
subbasin of Pequa Creek, Big Beaver Creek, had the
greatest reported atrazine concentrations during the
sampling period. The maximum reported atrazine con-
centrations at the Big Beaver Creek sampling station,
near Refton, Pennsylvania, were 0.30 parts per billion
during base-flow conditions and 24.0 parts per billion
during storm-flow conditions. The Big Beaver Creek
basin is 20.4 square miles in area, and agriculture con-
stituted about 66 percent of the land use in 1979. Corn
was grown on 26.6 percent of the agricultural lands in
this subbasin. The average rainfall forthis basin is about
37 inches annually (1).
As noted, agriculture represented the major land use in
the area. The primary agricultural soils in Lancaster
County are silt loams (Typic Hapludults and Hapludalfs)
in texture with slopes that range from 0 percent to 8
percent (7). Upon inspection of the air-photo soil se-
ries maps found in the county soil survey, however,
agricultural crops grew on lands with slopes of up to
and exceeding 15 percent, with soils such as the
Manorsilt loam, Pequa silt loam, and Chestersilt loam
(7).
Natural soil organic contents in the agricultural soils
range from 0.1 percent to 2.0 percent. Water contents
of the agricultural soils range from 10 percent to nearly
30 percent. The soil Erosion Factor (K) for the surface
layer ranges from 0.17 for the relatively stable Lingers
Series to 0.43 for the Pequa Series. The greater the
value, the greater the susceptibility to sheet and rill
erosion. The soil Erosion Factor (T in tons/acre/year) for
the entire soil profile ranges from 2 for the relatively
stable Clarksburg Series to 5 for the Elk Series.
Methods
Determining which soil series in the Big Beaver water-
shed may contribute to nonpoint source pollution
through runoff of agricultural chemicals entailed per-
forming a combination of probabilistic modeling analy-
ses and a CIS data manipulation.
Physiographic and Soil Series Boundaries
The orientation of Pequa Creek, Big Beaver Creek, and
other surface water bodies was digitized directly from
the 1:50,000-scale county topographic map for Lancas-
ter County (8). This map also provided the basis for
digitizing the Pequa Creek drainage divide, location of
urban areas, and roadways. The MAPINFO CIS allows
for the creation of boundary files by tracing the boundary
off the topographic map with a digitizing tablet config-
ured to the latitude and longitude coordinates of three
points on the map. The latitude and longitude are dis-
played while the boundary is being traced, allowing the
user to verify the accuracy of the boundary against
known coordinates on the map. CIS contains a self-
checking boundary closure program to ensure that the
polygons are closed and that the boundary contains no
extraneous line segments. These boundary data are
already available from the USGS in a digitized format,
digital line graphs (DLGs). Because digitizing is an easy
task, however, and to minimize costs, the project used
manual digitizing rather than purchase the data.
The roadways and urban areas were digitized to allow
use of standard control points, such as road intersec-
tions and benchmarks, to configure the U.S. Depart-
ment of Agriculture (USDA) Soil Conservation Service
air-photo-based, 1:15,840-scale soil series maps for the
Big Beaver Creek watershed. Known land grid coordi-
nates were noted on the air-photo maps (7). We con-
cluded, however, that using known reference points,
such as roadways and towns, allowed for a better con-
figuration of the digitizing tablet to the air photos and
158
-------
eliminated concerns over scale distortions sometimes
common in air photos.
After configuring the air photos to the digitizing tablet, a
2-square-mile area around the surface water sampling
points was digitized. The next step entailed digitizing all
the mapped soil series units within this area. The loca-
tions of crop areas, as plowed fields, were noted. In
noting forested areas, it became apparent that only
minor acreages were not in agricultural production.
Those mapped units were generally the Manor and
Pequa Series soils with slopes exceeding 25 percent.
Pesticide Transport Modeling
The PRZM pesticide transport model helped to quantify
the ability of several soil series to retard the transport of
atrazine through the root zone as leachate, dissolved in
surface runoff and adsorbed on sediment that moved
during erosion. The PRZM model performed in an un-
calibrated or screening model mode. The input values
for soil properties came from both the EPA DBAPE
database and the Lancaster County Soil Survey (7). The
modeled soil profile was 150 centimeters thick and di-
vided into 5-centimeter compartments. The soil half-life
of atrazine was set at 57 days in accordance with values
that the PRZM manual listed (9). The primary soil prop-
erty that varied in this demonstration project was surface
slope. All other parameters, such as soil organic content,
moisture content, and bulk density, appeared as mid-
point values for the ranges listed in DBAPE.
The agronomic scenario that the model simulated was
for corn grown continuously for 10 years using conven-
tional tillage practices and planted on May 7 of each
year. Atrazine was surface applied at a rate of 2 pounds
per acre (2.24 kilograms per hectare) on May 1 of each
year. For climatic input, the model used the historical
precipitation regimen from 1970 through 1980, as meas-
ured at the Harrisburg, Pennsylvania, station.
PRZM simulations were made for each of the following
soil series:
Chester
Conestoga
Elk
Glenelg
Glenville
Hollinger
Letort
Manor
Pequa
Monthly values were calculated for leachate, runoff, and
erosion per hectare. Unfortunately, no data for the
Pequa Series were available in the DBAPE database;
therefore, this portion of the analyses omitted it. In
addition, analyses of the Manor Soil Series included
more detailed probabilistic modeling where the surface
slope held constant (6-percent slope) and the surface
soil organic content varied to include the high, average,
and low organic contents as listed in DBAPE. PRZM
also calculated the volume of water as evapotranspira-
tion, runoff, and recharge through the root zone.
Results
The results of this study should demonstrate the appli-
cation of transport modeling to the possible protection
of water resources. Regulatory decision-makers should
not consider these results in their current form because
such decisions would require a much more rigorous
simulation strategy to increase the level of confidence in
the data. As a demonstration study, however, the results
do show the usefulness of this approach. Table 1 con-
tains the cumulative frequency data for the simulated
atrazine residues in runoff, erosion, and leaching that
occurred under 30 years of historical precipitation. The
data cover the 12 soils mentioned, with the surface
slope held constant at 6 percent.
Atrazine in Runoff and Erosion
The results of this analysis suggest that the Hagger-
stown Series had the greatest potential for yielding
atrazine in runoff; approximately 50 percent of the simu-
lated monthly atrazine in runoff values equaled or ex-
ceeded 0.0001 kilogram per hectare. Conversely, the
Elk Series yielded the least atrazine to runoff; 50 percent
of the runoff data were at residue levels of 1 x 10~6
kilograms per hectare. Within the Big Beaver Creek
subbasin, the Manor Series had the greatest potential
to yield atrazine in runoff.
As with the runoff data, the Haggerstown Series had the
greatest potential to yield atrazine in eroded sediments,
and the Elk Series yielded the least atrazine in erosion.
Within the Big Beaver Creek subbasin, the Manor Series
had the greatest erosion potential regarding atrazine.
GIS Analyses
After entering the results from the transport modeling
into a database, GIS could produce maps showing the
location of soils with high runoff potentials. Figure 2a
shows the orientation of soil series around the surface
water sampling points in Big Beaver Creek. Figure 2b
represents the same scene but fills in the soils with high
runoff potential. Using this type of analysis can help
areas that may be sources of nonpoint source runoff
contamination. Once identified, these soils can be tar-
geted for alternative management practices that may
reduce the amount of runoff and the degree of nonpoint
source contamination.
159
-------
Table 1. Cumulative Frequency of Simulated Atrazine Residues in Runoff for 12 Major Soils in Lancaster County, Pennsylvania
(values are percentage of data)
Load
(kilograms
per
hectare)
IE-10
IE-9
IE-8
IE-7
IE-6
IE-5
IE-4
IE-3
0.00
0.01
0.10
1.00
10.00
25.00
Bucks
83.33
78.79
74.24
66.67
56.06
47.78
36.36
24.24
15.91
6.82
0.00
0.00
0.00
0.00
Chester
84.85
80.30
76.52
68.94
46.97
35.61
21.97
15.15
6.82
0.00
0.00
0.00
0.00
0.00
Clymer
84.09
80.30
75.00
66.67
56.06
46.21
32.58
21.21
14.39
6.06
0.00
0.00
0.00
0.00
Soil
Connestoga
84.08
81.82
78.03
72.73
61.36
50.00
35.61
21.21
15.12
6.06
0.00
0.00
0.00
0.00
Series
Elk
78.79
75.76
69.70
59.09
50.00
43.18
31.82
21.21
15.15
6.82
0.00
0.00
0.00
0.00
in Lancaster County, Pennsylvania
Glenelg
84.85
80.30
76.52
68.18
56.06
47.73
34.85
21.21
13.64
3.79
0.00
0.00
0.00
0.00
Haggerstown
93.94
93.94
93.94
91.67
84.85
74.24
53.79
37.12
19.70
7.58
0.00
0.00
0.00
0.00
Hollinger
83.33
81.06
77.27
68.94
59.09
47.73
34.85
21.21
15.15
6.06
0.00
0.00
0.00
0.00
Lansdale
88.64
85.61
80.30
75.00
65.15
51.52
40.15
24.24
15.91
6.82
0.00
0.00
0.00
0.00
Letort
85.61
84.85
84.09
83.33
76.52
67.42
46.97
31.82
17.24
4.55
0.00
0.00
0.00
0.00
Manor
86.36
84.61
84.85
75.52
66.67
46.97
31.82
16.67
4.55
0.00
0.00
0.00
0.00
0.00
Lingers
84.09
84.09
81.82
75.76
67.42
53.03
39.39
24.24
15.15
4.55
0.00
0.00
0.00
0.00
This table reads as follows: Given the Elk Series, the first value
IE-10 kilograms per hectare. Similarly, within the same column, 6
than 0.1 kilograms per hectare.
reads that 78.79 percent of the simulated data were greater than or equal to
.82 percent of the data were greater than 0.01 kilograms per hectare but less
Pequa Creek
Pequa Creek
0 0.5 1
Figure 2a. Soil series in the Big Beaver Creek basin.
Detailed Modeling of the Manor Series
Performing an introductory probabilistic modeling exer-
cise allowed further investigation of the potential of the
Manor Series to release atrazine into runoff. The exist-
ing 30-year climate data and the stated agronomic data
were retained from the previous modeling. The organic
carbon content of the surface soil layer, however, was
allowed to vary between the published low, average, and
maximum values found in the DBAPE database. This
exercise followed the principles set forth by Laskowski
et al. (5) and others who describe probabilistic modeling
0.5
Figure 2b. Soils sensitive to atrazine runoff in the Big Beaver
Creek basin.
approaches. In essence, by varying input parameters
within known endpoints, the probabilistic approach can
generate a distribution of pesticide residue values that
statistically reflects the anticipated residues. Parame-
ters to vary may include:
Organic carbon content
Surface slope
Kd (distribution coefficient)
160
-------
Moisture content
By allowing input variables to vary according to a normal
distribution, this approach thereby eliminates some of
the uncertainty associated with pesticide transport mod-
eling. The probabilistic modeling approach requires the
creation of a significant database by performing many
runs (e.g., 1,000 model runs that generate 12,000
monthly values for each soil).
This study included a limited probabilistic modeling ex-
ercise. Table 2 lists the results for the mean atrazine
residues in runoff, erosion, and leaching for the Manor
Soil Series during the:
Entire year
Growing season
Winter months
The surface slope was held constant at 6 percent, but
the soil organic carbon content varied within the publish-
ed range. The means for all months show limited vari-
ation in mean residues. Runoff was by far the major
Table 2. Summary Statistics for Detailed Modeling of the
Manor Soil Series in the Pequa Creek Watershed
(statistics based on 1,080 values)
Mean Atrazine Residue
(kilograms per hectare)
Percent
Organic
Carbon3
All Months
Low
Average
High
Growing Season
Low
Average
High
Winter Months
Low
Average
High
Runoff
0.01381
0.01229
0.01105
0.03290
0.02881
0.02540
0.00014
0.00045
0.00075
Erosion
0.00024
0.00042
0.00057
0.00058
0.00100
0.00130
< 0.000002
< 0.000001
< 0.000003
Leaching
0.00028
0.00012
< 0.000006
0.00017
< 0.000008
< 0.000004
0.00036
0.00014
< 0.000007
Data taken from DBAPE soils database as low, midpoint, and maxi-
mum reported organic contents.
source of atrazine. Erosion and leaching values were on
similar scales (trace amounts).
The greatest runoff and erosion values occurred during
the growing season. The greatest leaching, however,
occurred during the winter months. These results sup-
port the general observations that surface residues run
off during spring and summer but that as the crops grow
and evapotranspiration increases, recharge to ground
water decreases, subsequently limiting pesticide trans-
port to ground water. Conversely, during the winter
months, the surface soil pesticide residues generally
decrease because of exposure to months of photolysis,
hydrolysis, and biodegradation. Subsurface residues
have been protected from degradation, however, and
increased ground-water recharge, due to great reduc-
tions in evapotranspiration, transports the residues
through the soil column.
This limited exercise provided a valuable learning expe-
rience regarding probabilistic modeling. As computing
techniques and hardware advance, the cost in time and
money for each simulation should decrease dramati-
cally. Although researchers tend not to have great faith
in pesticide transport modeling, the advances in this
field will reduce uncertainty and instill greater confi-
dence in the modeling process.
GIS Pitfalls
CIS is a powerful tool and has great promise for use
in environmental problem-solving. Several points or
pitfalls, however, hinder broad acceptance of GIS. As
with most new technologies, cost is the overriding con-
cern in using GIS. Although technical staff and project
scientists understand the power of GIS and the effort
that data preparation requires, management and corpo-
rate staff often do not see the benefits for the costs.
Many managers assume that current GIS systems re-
semble those seen on "Star Trek," and when reality
becomes apparent, managers tend to discard GIS as
too costly and complex. Several points need considera-
tion when contemplating the use of GIS. Although vari-
ous products exist, this discussion focuses on
ARC/INFO and MAPINFO products.
Hardware
Computer hardware is plentiful if the available budget
can support a purchase. Many high- powered GIS pack-
ages (e.g., GRASS, ARC/INFO, INTERGRAPH, IDRISI)
run best on mainframes or minicomputers. Most techni-
cal staff, however, only have access to PC machines.
Corporate purchasing departments more readily expend
funds for PC technology because they will eventually
find use forthese machines even if they are not used for
GIS. A recent ARC/INFO advertisement (August 1994)
lists costs for SUN SPARC minicomputer systems with
ARC/INFO software at $12,000 to $15,000 depending
on configuration.
Minicomputers and mainframes require specialized staff
to configure and maintain the hardware. Today, many
staff level personnel can open and augment their PC
machines with a minimum of external support. GIS per-
formance reflects the tradeoff in hardware, particularly
161
-------
when a considerable amount of data manipulation is
required. For example, if linking discrete depth soil se-
ries data to STATSGO soil associations is necessary,
then a minicomputer system may be best. The postproc-
essed data could, however, be exported to a format that
will run on PC-based systems. If the user wants to
import and manipulate remote sensing imaging (e.g.,
SPOT or Landsat data), then minicomputers are recom-
mended. If the user wants to display already edited
images and preprocessed CIS data, then PC-based
computing may be sufficient. The ultimate use of CIS
drives the hardware selection.
Software
A great number of CIS software packages are available
to meet almost any level of use and expertise. Software
runs under both UNIX and DOS/Windows (denoted as
DOS forthe remainder of this paper) operating systems.
The UNIX-based software tends to be more powerful
and flexible than the DOS-based software. UNIX-based
packages require more specialized staff to optimize
CIS, however.
UNIX-based software packages include GRASS,
ARC/INFO, INTERGRAPH, and IDRISI. Costs vary
from public domain charges for GRASS and IDRISI to
vendor supplied ARC/INFO and INTERGRAPH, which
can cost several thousand dollars each.
DOS versions of ARC/INFO (e.g., PC ARC/INFO,
ARCAD, ARC/VIEW) are also available and provide the
userwith various levels of data editing and manipulation
abilities. Generally, PC ARC/INFO is the same as the
UNIX version, varying in speed of processing. ARCAD
is a CIS engine that uses AutoCAD for drawing and
displaying, giving the user most of the abilities of the
UNIX-based version. ARC/VIEW I was primarily a dis-
play and simple analysis tool. It allowed the user to view,
display, and manipulate existing arc data but did not
support image editing. Currently, ARC/VIEW II provides
more support for image editing and data manipulation.
Costs range from about $3,000 for PC ARC/INFO and
ARCAD (AutoCAD also costs about $2,000) to around
$500 for the ARC/VIEW products.
MAPINFO is a DOS-based CIS that was designed for
marketing and demographic applications. Several re-
searchers, however, have used MAPINFO for environ-
mental applications. The most outstanding feature of
MAPINFO is that it easily imports data layers as it reads
dBASE type files directly. MAPINFO V3 also reads da-
tabase files and recreates them as *.TAB files. In con-
trast to the "coverage and entity" concepts of the
ARC/INFO line of programs, MAPINFO reads latitude
and longitude coordinates and displays the results. This
simplifies data management because many researchers
who have already created custom databases can easily
import those data as long as latitude and longitudes
coordinates are present.
As with the ARC/INFO line of programs, many common
data layers can be purchased for use in MAPINFO.
These layers can be expensively priced, costing ap-
proximately $1,000 per county for roadway, census, and
demographic data. One major lapse is the poor library
of environmental layers, USGS topography, hydrogra-
phy, soil boundaries, or climate stations. MAPINFO does
sell a module that allows users to convert to and from
ARC/INFO coverages so that common data layers can
be established. Experience shows, however, that con-
version programs do not always work as advertised. For
example, large boundary files (STATSGO data for Indi-
ana) do not readily convert from ARC/INFO to MAP-
INFO. Third-party vendors may be needed to convert
data for use in MAPINFO.
One very important factor supporting the use of MAP-
INFO is that it has a business application slant; there-
fore, it is slightly easier to convince corporate
management to invest in CIS because marketing and
sales data (territories) can be relatively easily overlain
onto environmental data.
Finally, some packages that are add-ons to spreadsheet
programs tend not to be powerful or versatile enough for
use in environmental CIS work. These software pack-
ages may be valuable as an introduction to CIS tech-
niques, however.
Data A vailability and Format
After compiling the hardware and software into CIS, the
next step entails accessing data layers such as:
State and county boundaries
Land use covers
Water boundaries
Currently, USGS DLGs for hydrography, land use, trans-
portation, and cultural features are available for minimal
costs. Shareware programs can convert the USGS
DLGS formats into DXF (data transfer files) for import to
CIS packages. These data require conversion to DXF
or ARC coverage type formats for use in either
ARC/INFO or MAPINFO.
The USDA Soil Conservation Service produces digital
data for soil types (STATSGO and SSURGO) that users
can import to ARC/INFO relatively easily. The STATSGO
data cost approximately $1,000 per state and are avail-
able for most states. The detailed soil series maps,
SSURGO, cost approximately $500 per county and are
not available for every county in the United States. Many
data layers are available for direct use by GRASS. As
of yet, however, no convenient conversion utilities exist
to move GRASS data to ARC/INFO or MAPINFO. The
162
-------
U.S. Fish and Wildlife Service now distributes data
layers from the National Wetlands Inventory on the
Internet (enterprise.nwi.fws.gov).
Other data sources available through private vendors
are listed in the MAPINFO and ARC/INFO user guides
and in any issue of CIS World. The user should be
prepared to absorb significant costs if purchasing all the
required data layers.
Conclusion
This study shows that the technology and software exist
for a water resource manager to couple pesticide trans-
port modeling with CIS to identify areas or individual
soils that may contribute to nonpoint source pollution of
water resources. The study used PC-based computing
system and software. Soils maps and hydrographic
maps can be easily digitized for limited cost. A skilled
scientist or technician, without being a CIS expert, can
run CIS to answer specific questions. The technologies
this study demonstrated may be extremely valuable to
managers responsible for producing SMPs.
The pesticide transport modeling performed during
this study was intended for illustrative purposes. More
detailed analyses, and additional simulations, would be
necessary to use these data for regulatory actions or
land use management. The study did, however, succeed
in identifying the Manor Soil Series, with slopes exceed-
ing 6 percent, within the Big Beaver Creek subbasin of
the Pequa Creek basin, as a potential source for
atrazine in runoff.
References
1. Ward, J.R. 1987. Surface-water quality in Pequa Creek Basin,
Pennsylvania, 1977-1979. U.S. Geological Survey. Water Re-
sources Investigation Report 85-4250.
2. Smith, M.C., A.B. Bottcher, K.L. Campbell, and D.L. Thomas. 1991.
Field testing and comparison of the PRZM and GLEAMS models.
Trans. Amer. Soc. of Agric. Eng. 34(3):838-847.
3. Mueller, T.C., R.E. Jones, P.B. Bush, and PA. Banks. 1992. Com-
parison of PRZM and GLEAMS computer model predictions with
field data for alachlor, metribuzin, and norflurazon leaching. Envi-
ron. Toxicol. and Chem. 11:427-436.
4. Pennell, K.D., A.G. Hornsby, R.E. Jessup, and P.S.C. Rao. 1990.
Evaluation of five simulation models for predicting aldicarb and
bromide behavior under field conditions. Water Resour. Res.
26(11):2,679-2,693.
5. Laskowski, D.A., P.M. Tillotson, D.D. Fontaine, and E.J. Martin.
1990. Probability modeling. Phil. Trans. Royal Soc. of London
329:383-389.
6. Ward, J.R., and D.A. Eckhardt. 1979. Nonpoint-source discharges
in Pequa Creek Basin, Pennsylvania, 1977. U.S. Geological Sur-
vey. Water Resources Investigation Report 79-88.
7. Custer, B.H. 1985. Soil survey of Lancaster County, Pennsylvania.
Washington, DC: U.S. Department of Agriculture, Soil Conserva-
tion Service.
8. U.S. Geological Survey. 1977. Lancaster County, Pennsylvania:
County map series (topographic) 1:50,000 scale.
9. Carsel, R.F., C.N. Smith, L.A. Mulkey, D.J. Dean, and P. Jowise.
1984. Users manual for Pesticide Root Zone Model (PRZM), Re-
lease I. EPA/600/3-84/109.
163
-------
Integration of GIS With the Agricultural Nonpoint Source Pollution Model:
The Effect of Resolution and Soils Data Sources on Model Input and Output
Suzanne R. Perlitsh
State University of New York, Syracuse, New York
Abstract
The assessment of agricultural nonpoint source pollu-
tion has been facilitated by linking data contained in a
geographic information system (GIS) with hydrologic
models. One such model is the Agricultural Nonpoint
Source (AGNPS) Pollution Model, which simulates run-
off, nutrients, and sediment from agricultural water-
sheds. Vector-based (ARC/INFO) and raster-based
(IDRISI) GIS systems were used to generate AGNPS
input parameters.
The objectives of this project were to generate AGNPS
input parameters in GIS format from GIS data at differ-
ent resolutions and different levels of detail (soil survey
soils data versus soils data currently available in digital
format from the United States Department of Agricul-
ture). Differences in the AGNPS model sediment out-
put based on the variations in CIS-generated AGNPS
model input were evaluated.
The study also evaluated the influence of cell size reso-
lution and soils data on sediment generated within each
cell in the watershed (SGW), sediment yield from each
cell in the watershed (SY), sediment yield at the water-
shed outlet, and peak flow. Model output was validated
by comparison with measured values at the watershed
outlet for a monitored storm event. Results of this study
indicate that the use of different resolution GIS data and
different soils data sources to assemble AGNPS input
parameters affects AGNPS model output. Higher reso-
lution data do not necessarily provide better results.
Such comparisons could affect decision-making regard-
ing the level and type of data analysis necessary to
generate sufficient information.
Introduction
Agricultural runoff is a major contributor to nonpoint
source pollution. Fifty-seven percent of the pollution in
impaired lakes and 64 percent of the pollution in im-
paired rivers of the United States can be attributed to
agricultural nonpoint source pollution (1). Sediment is
one of the most common agricultural nonpoint source
pollutants and is the largest pollutant by volume in the
United States (2). More than 3 billion tons of sediment
enter surface waters of the United States each year as
a result of agricultural practices (1).
Accurate assessment of the effects of agricultural activi-
ties on water quality within a watershed is vital for
responsible watershed management and depends on
our ability to quantify the spatial variability of the water-
shed and the complex interactions of hydrologic proc-
esses (3). Computer models have been developed to
simulate these hydrologic processes to provide esti-
mates of nonpoint source pollutant loads. Adequate
simulation of a watershed's spatial variability helps pro-
vide the best representation of hydrologic processes
within the watershed.
Preservation of spatial variability within hydrologic mod-
els can be accomplished using a distributed parameter
model. The distributed parameter model is more advan-
tageous than lumped parameter models, which gener-
alize watershed characteristics, because distributed
parameter models provide more accurate simulations of
the systems they model (4). One of these models is the
Agricultural Nonpoint Source (AGNPS) Pollution Model.
AGNPS is a distributed process model because it pro-
duces information regarding hydrologic processes at
grid cells within the watershed, thus enabling preserva-
tion of the spatial variation within the watershed. Distrib-
uted parameter models integrate well with GIS because
GIS can replicate the grid used in a distributed parame-
ter model. Manual compilation of AGNPS input parame-
ters required to evaluate small areas at low resolution
(large grid cells) is relatively easy. Manually assembling
data to evaluate larger areas at finer resolutions be-
comes tedious, however. The integration of GIS data
with the AGNPS model facilitates data assembly and
manipulation (5).
164
-------
Several researchers have integrated AGNPS with CIS
(4-11). Smaller cell sizes within distributed parameter
models are thought to best represent spatial variability
within a watershed (5, 10). Certain AGNPS input
parameters show sensitivity to changes in grid cell size,
affecting sediment yield output (11). The use of CIS to
generate input parameters for the AGNPS model en-
ables analysis of watersheds at higher resolutions than
would be practical using manual methods (5).
Research Hypotheses
The project investigated the following research hypotheses:
AGNPS output at the highest resolution will better
approximate sediment yield at the watershed outlet.
AGNPS output for sediment generated within each
cell in the watershed at highest resolution will best
reflect the watershed processes.
AGNPS output generated from the more detailed soil
survey data will better estimate watershed processes.
The project also investigated other questions: will cer-
tain AGNPS input parameters (cell land slope, soil erodi-
bility [K], the cropping factor [C], and the U.S.
Department of Agriculture [USDA] Soil Conservation
Service [SCS] curve number [CN] show sensitivity to
changes in grid cell size? How does slope affect model
output? Does a qualitatively significant difference exist
between model input parameters and output calculated
from data sets generated at different resolutions with
different levels of detail in soils output?
Objectives/Tasks
The research in this project included analyses of:
Certain AGNPS input parameters generated at differ-
ent resolutions (10- x 10-, 30- x 30-, 60- x 60-, and
90- x 90-meter resolutions).
AGNPS output for sediment yield (SY) and sediment
generated within each cell (SGW) at different resolu-
tions (center cells of 10- x 10-, 30- x 30-, 60- x 60-,
and 90- x 90-meter resolutions).
AGNPS output generated from different levels of de-
tail in the soils input data (soil survey versus STAT-
SGO data sources).
Significance
Version 4.03 of AGNPS was released in June 1994. Ver-
sion 4.03 allows for evaluation of 32,767 cells. This version
allows for cell sizes from 0.01 to 1,000 acres (approxi-
mately 6.36-x 6.36-meter resolution to 2,012-x 2,012-me-
ter resolution). Previous versions of AGNPS limited the
number of cells to 3,200 and the cell size resolution to 0.4
hectares (or 63.25 x 63.25 meters) (12). Reviewed litera-
ture provides no evidence that AGNPS has been used
to evaluate a watershed at 10- x 10-meter resolution.
The soils data in this study were compiled at two differ-
ent levels of detail. Soils data at the 1:20,000 soil survey
level were generated in digital format. This level of detail
was compared with soils data at the State Soil Geo-
graphic (STATSGO) database level with a scale of
1:250,000. Reviewed literature mentions no previous
studies comparing AGNPS output with input generated
from these two different levels of detail in soils input.
Technology for collecting and processing geographic
data is continuously improving. Currently, the United
States Geological Survey (USGS) 1:24,000 digital ele-
vation models (OEMs) are available at 30- x 30-meter
resolution. New satellite technology will enable DEM
data to be available at 10- x 10-meter resolution, or
higher. Certain satellites currently provide land cover
data at 10- to 30-meter resolution (13).
An important objective of this project was to determine
whether higher resolution data provide different results
when routed through AGNPS. Does spatial data at
higher resolutions provide better information? This pa-
per describes the results of an analysis of AGNPS out-
put based on different levels of both resolution and soils
detail in CIS data input sources.
Materials
The Study Area
An ongoing effort is underway to clean up Onondaga
Lake, Onondaga County, New York. To accomplish this
effort, areas contributing agricultural nonpoint source
pollution to the lake are being evaluated.
The Onondaga Lake watershed is approximately 287.5
square miles, with 40 subwatersheds. The subwater-
sheds in the agricultural portion of the Onondaga Lake
watershed (south of Syracuse, New York) have been
isolated for study of their potential nonpoint source con-
tributions to Onondaga Lake (see Figure 1). The study
area watershed (1.84 square miles, 1,177.5 acres) is
one of these agricultural subwatersheds. CIS data were
collected within the Otisco Valley quadrangle (USGS
1:2,400), which includes the southern portion of the
Onondaga Lake watershed. Elevations in the study wa-
tershed range from 1,820 feet to 1,203 feet, with an
average elevation of 1,510 feet. The watershed perime-
ter is approximately 6.5 miles (34,505 feet). The streams
in the watershed flow from south to north to Rattlesnake
Gulf, with a stream length of approximately 3.08 miles
(16,265.4 feet). The stem fall of the main stream stem
is quite steep at 283 feet per mile. The drainage density
of the watershed is 1.67 miles of stream persquare mile.
Land use in the watershed is predominantly agricultural
(82.8 percent).
165
-------
Study Area Watershed
Canty Hill Rd. ! Bar*ell Rd
Figure 1. Onondaga Lake watershed and study area (not to scale).
The AGNPS Model
AGNPS was developed to analyze and provide esti-
mates of runoff water quality, specifically to evaluate
sediments and nutrients in runoff from agricultural
watersheds for a specific storm event (11). To use
AGNPS, a watershed is divided into cells of equal area.
Calculations for each of the model output values are
made within each cell based on the watershed data
contained in each cell. Approximately 1,000 people in 46
different countries use the AGNPS model. Users include
students, university professors, government agencies,
lake associations, and environmental engineers.1
AGNPS was developed in 1987 by the Agricultural Re-
search Service (ARS) in cooperation with the Minnesota
Pollution Control Agency (MPC) and the SCS. The
model runs on an IBM-compatible personal computer.
Data Sources
The CIS packages of ARC/INFO Version 3.4D (14) and
IDRISI Version 4.1 (15) were used to prepare input
parameters for the AGNPS model. AGNPS input pa-
rameters were derived from three base mapsland
use, a DEM, and soils. Table 1 shows the 22 input
parameters that AGNPS required, and the base source
for the data.
The land use map was obtained from a classified
ERDAS image (resolution of 28 x28 meters). The image
was converted to IDRISI, brought into ARC/INFO, and
regridded based on the resolution requirements of each
data set. USGS could not provide a DEM for the study
Personal communication from AGNPS Technical Support, September
1994.
area, so the OEMs were interpolated from points digit-
ized in ARC/INFO based on the Clarke method (16). The
OEMs were interpolated in IDRISI on 10- x 10-, 30- x30-,
60- x 60-, and 90- x 90-meter resolution surfaces.
Soil survey data were obtained from Onondaga County
Soil Survey air photographs. An orthophoto of the
7.5 minute quadrangle was obtained from the USGS
and was used with a zoom transfer scope to ortho-cor-
rect the soil survey data. The corrected soil polygons
were then digitized in ARC/INFO. The Otisco Valley
quadrangle comprises 79 soils mapping units. Thirty-
eight different mapping units occur in the study area
watershed.
The USDA SCS (now the Natural Resource Conser-
vation Service [NARCS]) provides digital soils data
from its STATSGO database. The mapping scale of
STATSGO data is 1:250,000, thus it is best suited for
broad planning and management uses. The number of
soil polygons per quadrangle is between 100 and 400,
and the minimum area mapped is 1,544 acres. The
STATSGO soil data used in this project were obtained
from the Onondaga County Soil Conservation Serv-
ice. Approximately seven STATSGO soil groups were
identified for the Otisco Valley quadrangle. Only one
STATSGO soil type occurs in the study area watershed
(Honeyoe silt loam).
Methods
AGNPS input parameters that showed sensitivity to
changes in grid cell size in previous studies were com-
pared between the resolutions. The AGNPS model was
run eight times using precipitation values from the actual
166
-------
Table 1. AGNPS Input Parameters
# AGNPS Parameter
Root Data Source
General Derivation of Input
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Cell number
Cell division
Receiving cell number
Receiving cell division
Flow direction
SCS curve number
Land slope percentage
Slope shape factor
Average slope length
Manning's n coefficient
USLE K factor
USLE C factor
USLE P factor
Surface condition constant
Chemical oxygen demand
Soil texture
Fertilizer indicator
Pesticide indicator
Point source indicator
Additional erosion
Impoundment indicator
Channel indicator
Watershed map
Not applicable
Aspect map from DEM
Not applicable
Aspect map from DEM
Land use and soils coverage
Slope map from DEM
Algorithm
Table of values
Literature values
SCS and soil survey
Literature values
Literature values
Land use coverage
Land use coverage
Soil survey
Land use coverage
Land use coverage
USGS 1:24,000 map
Field survey, known gullies
1:24,000 map, field survey
Streams coverage
Program written to determine #
No cell division, assumed 0
Program written to determine #
No cell division, assumed 0
Reclassed 1-8 from azimuth map
SML written to determine CN
Provided in slope percentage from IDRISI
Assume uniform slope
Obtained from SCS
Attached to land use database
Attached to soils database
Attached to land use database
Attached to land use database
Attached to land use database
Attached to land use database
Attached to soils database
Assumed for agricultural land class
Assumed for agricultural land class
Points in ARC/INFO and IDRISI
Assume no additional erosion
Assume no impoundments
Assume no significant channel
storm that was monitored. Each time, the model was run
using an input file created with the different input data
sources as follows:
1. 30- x 30-meter resolution-soil survey data.
2. 30- x 30-meter resolution-STATSGO data.
3. 60- x 60-meter resolution-soil survey data.
4. 60- x 60-meter resolution-STATSGO data.
5. 90- x 90-meter resolution-soil survey data.
6. 90- x 90-meter resolution-STATSGO data.
7. Center cells of the 10- x 10-meter resolutionsoil survey
data.
8. Center cells of the 10- x 10-meter resolution
STASGO data.
As grid cell size increases, the time required to assemble
data as well as the space required to store the data files
increase. If a cell size resolution is cut in half, the
number of cells in that coverage quadruples. In the study
area watershed, increasing grid cell size from 90- x 90-
meter resolution to 60- x 60-meter resolution created
778 more cells within the watershed. Moving from
60- x 60-meter resolution to 30- x 30-meter resolution
added 3,646 cells to the watershed, and moving from
30- x 30-meter resolution to 10- x 10-meter resolution
added 40,115 cells to the watershed data set (see
Figure 2). Due to the 32,767-cell limitation of AGNPS
Version 4.03, AGNPS output for SY and SGW at the
10- x 10-meter resolution (which contains 45,104 cells)
could not be obtained. Input parameters at 10- x 10-meter
resolution, however, could be compared with input
parameters at 30- x 30-meter resolution.
Methodology of Data Analysis
The input parameter "maps" were converted to IDRISI
files and combined in a format that could be routed
through the AGNPS model. AGNPS model output for
soil generated within each cell and for sediment yield
was assembled. The 30- x 30-meter resolution maps
were compared with the 60- x 60-meter resolution maps;
the 60- x 60-meter resolution maps were compared with
the 90- x 90-meter resolution maps; and the 30- x 30-
meter resolution maps were compared with the 10- x 10-
meter resolution maps.
A method for comparing maps with different grid sizes
was developed so that maps of different resolutions
167
-------
45,104
90x90
Figure 2. Effect of resolution on study area data set.
60x60
30x30
10x 10
Resolution (meters)
could be compared using the root mean square (RMS)
statistic. This method was selected to provide a means
for statistical analysis that could accommodate the spa-
tial variability of this data set. The 60- x60-meter resolution
maps were "expanded" in IDRISI. This duplicated each
grid cell so that the original 60- x 60-meter resolution
map was equivalent to the 30- x 30-meter resolution
map. The IDRISI command "RESAMPLE" was used to
bring the 60- x 60-meter resolution map data onto the
same grid size as the 90- x 90-meter resolution map.
The center cells of the 10- x 10-meter resolution map
were selected for comparison with the 30- x 30-meter
grid cells. This comparison is based on the fact that the
center cell of the 10-meter resolution is the cell that
best corresponds to the entire cell of 30-meter reso-
lution (see Figure 3).
Once maps were registered on comparable grids, the
RMS difference between the maps of differing resolution
was used to compare the difference between the maps
and the effect of cell size resolution. The RMS statistic
is a measure of the variability of measurements about
their true values. The RMS is estimated by comparing
values in one grid system with the values in the com-
parison grid system. The difference between corre-
sponding values in each grid system is squared and
summed. The sum is then divided by the number of
measurements in the sample to obtain a mean square
deviation. Finally, the square root of the mean square
deviation is calculated. The RMS difference quantifies
the discrepancy between two data sets.
RMS:
(grich -grid2)2
Expand
Resample
Center Cell
30x30
laoVsoV
\
«' 60x60',
\
\
10x 10
Figure 3. Method for comparing cells of differing resolution.
Results
The Storm Event
The storm that was monitored for the purposes of this
field validation occurred on August 28,1994, at approxi-
mately 8:45 p.m.; the storm duration was approximately
1.5 hours. It was a high intensity, short duration thunder-
storm (see Figure 4). A global flow probe (Model FP101)
was used to obtain discharge velocity measurements in
feet per second. These values were then multiplied to
obtain cubic feet per second. The peak discharge occurred
at 11:30 p.m. on August 28, 1994, with a flow discharge
of 11.15 cubic feet per second. The average runoff for
the period was 4.458 cubic feet per second, or 2.42 cubic
168
-------
feet per second per square mile. The runoff volume per
day was 0.09 inches.
Total sediment yield was derived from the analysis of total
solids measured in field samples throughout the 24-hour
storm period. The samples were processed to evaluate
total suspended solids (TSS) using the vacuum filtration
procedure (17). A total of 1.204 tons of suspended sedi-
ment was predicted at the watershed outlet from field
data samples. A LaMotte field nutrient test kit was used
to measure nitrate and phosphate concentrations in the
stream. Nutrient values in this watershed for this storm
event were so small (phosphorous below 0.1 parts per
million and nitrogen 0.3 parts per million), they were not
selected as parameters to be used in evaluating and
validating the AGNPS model. The AGNPS predicted
nutrient output for the storm was not measurable (0.00
parts per million). The low levels of nitrogen and phos-
phorous in the stream channel during the storm event
can be attributed to the time of year in which the stream
was monitored. At the time of field validation, agricultural
activities were not operating.
Results at the Watershed Outlet: Peak Flow
The peak flow values that AGNPS calculated are largest
at the highest resolution and decrease as cell size in-
creases. The peak discharge from the watershed during
the monitored storm event was 11.15 cubic feet per
second. Comparisons of the AGNPS predicted peak
flow and the actual field-validated peak flow showed that
the 30- x 30-meter resolution cells best approximate the
peak flow of the watershed for the sampled storm event.
As grid cells increase from 30 x 30 to 60 x 60 and from
60 x 60 to 90 x 90, the peak flow is underestimated. As
grid cells decrease from 30 x 30 to 10x10, the peak
flow is grossly overestimated (see Table 2).
Results at the Watershed Outlet:
Sediment Yield
Sediment yield at the watershed outlet was determined
to be 1.204 tons. In all of the resolutions, the amount of
sediment deposited at the watershed outlet cell in-
creased as cell resolution increased (see Table 2). For
this particular watershed in this particular storm, the
AGNPS model overestimated the sediment yield pre-
dicted at the watershed outlet at the 10- x 10-, 30- x 30-,
and 60- x 60-meter resolutions and underestimated
sediment yield at the 90- x 90-meter resolution. Table 2
includes the information that AGNPS predicted for the
cell designated as the watershed outlet within each
resolution. (The results reported include output from the
center cells of the 10- x 10-meter resolution data set,
routed through the AGNPS model. Although these val-
ues are reported, the results from this data set cannot
be assumed to approximate the sediment output that
would result had the entire 10- x 10-meter resolution
data set been simulated.)
So/7 Survey Versus STATSGO Data
The Kappa statistic (14,18) was used as an indicator of
similarity to describe the differences between the
AGNPS output for SY and SGW generated from STATSGO
and soil survey data. Results (see Table 3) indicate that no
significant difference exists between the output derived
from the STATSGO and soil survey data inputs within
Relationship Between Streamflow Discharge
and Suspended Sediment
O)
"c
CD
o
c
o
o
"c
CD
E
T3
CD
600 T
500 --
400 --
300 --
200
100 -
M-l t"T I I M 1 1 |ifT'M'"M"l"i'l"1"t'l"i"t'"t"i"t'r">ll'l"l1l"l'"l
CD
O)
ooooooooooooooooooooooooooooo
ococococococococococococococococococococococococococococo
(Nco^mcDr^coajo-^-cNco^-^-cNco^incDr^coajo-^-cNco^incD
CN CN CN CN CN
Time (8/29/94 to 8/29/94)
Figure 4. Storm event hydrograph and pollutograph.
169
-------
Table 2. AGNPS Results at the Watershed Outlet Versus
Actual Field Values
Predicted
Soil Difference Predicted
Resolution Survey From STATSGO
(meters) Data Actual Data
10 x 10
(center cells)
Peak runoff 29.88 +18.73 29.88
rate (cfs)a
Total sediment 4.48 +3.49 4.69
yield (tons)
30x30
Peak runoff 11.27 +0.12 11.27
rate (cfs)
Total sediment 2.84 +1.64 3.11
yield (tons)
60 x 60
Peak runoff 9.96 -1.19 9.96
rate (cfs)
Total sediment 2.11 +0.91 2.12
yield (tons)
90x90
Peak runoff 8.87 -2.28 8.87
rate (cfs)
Total sediment 0.86 -0.34 0.87
yield (tons)
acfs: cubic feet per second.
Table 3. Kappa Coefficient of Similarity
Resolution (meters)
10 x 10 (center)
Soil survey versus STATSGO SY
Soil survey versus STATSGO SGW
30 x 30
Soil survey versus STATSGO SY
Soil survey versus STATSGO SGW
60 x 60
Soil survey versus STATSGO SY
Soil survey versus STATSGO SGW
90x90
Soil survey versus STATSGO SY
Soil survey versus STATSGO SGW
Difference
From
Actual
+18.73
+3.28
+0.12
+ 1.91
-1 .19
+0.92
-2.28
-0.33
Kappa
0.9866
0.9703
0.9859
0.9785
0.8960
0.7542
0.8743
0.6574
Table 4. RMS for AGNPS Sediment Loss Output
Description SGW Pounds SY Pounds
Soil Survey Data Constant
10 centers to 30-meter resolution 168.38 344.22
30- to 60-meter resolution 125.43 739.75
60- to 90-meter resolution 93.50 498.01
30- to 90-meter resolution 29.91 711 .49
STATSGO Data Constant
10 centers to 30-meter resolution 164.45 312.48
30- to 60-meter resolution 125.37 782.07
60- to 90-meter resolution 97.07 501 .81
30- to 90-meter resolution 122.04 747.51
Moving to higher cell resolutions increasingly affects
sediment generated within each cell; the largest differ-
ence in sediment generated within each cell occurs as
cell resolution increases from 30 x 30 to 1 0 x 1 0 meters.
Sediment generated within each cell is least affected by
moving from 60 x 60 to 90 x 90 meters. Sediment yield
per cell is most affected when cell resolution increases
from 60 x 60 to 30 x 30 meters and least affected by
increasing resolution from 30 x 30 to 10x10 meters.
These results prompted an assessment of the methods
used to compare resolutions, to determine whether the
effect on sediment yield between the 30- x 30- and
60- x 60-meter resolutions could result from the method
used in comparing the resolutions (expansion of the
60- x 60-meter resolution). The procedure of comparison
between the 30- x 30- and the 60- x 60-meter resolu-
tions was repeated; however, rather than expanding the
60 x 60 data file, the 60 x 60 data file was "resampled"
onto the 30- x 30-meter resolution grid, then the files
were compared. The RMS results (see Table 5) show
that both methods for comparing data between the reso-
lutions provide essentially the same results. The effect
of resolution on sediment yield per cell is, in fact, greatest
as resolution increases from 60 x 60 to 30 x 30 meters.
Results: AGNPS Input Parameters of Concern
Previous AGNPS analyses have shown sediment yield
(and sediment-associated nutrient yields) to be most
the same resolutions. This may be due to the homoge-
neity of the soil textures in both soils data sets (both
dominated by silty soils).
Effects of Resolution on SGW and SY
AGNPS output for SGW and SYwas evaluated for every
cell within the watershed. The RMS difference was ap-
plied to determine the relative effect of input data reso-
lution on SY and SGW output (see Table 4 and Figure 5).
affected by AGNPS inputs for cell land slope, the soil
erodibility factor (K), the Universal Soil Loss Equation's
Table 5. Comparison of RMS for 30 x 30 to 60 x 60
Method RMS SGW Pounds RMS SY Pounds
Expansion of 60 x 60
to 30 x 30
Resampling 60 x 60
to 30 x 30
Difference
125.43
125.37
0.06
739.75
739.30
0.45
170
-------
RMS Difference
Sediment Generated in Each CellSoil Survey Data
10 to 30
20 40 60 80 100 120140160180
RMS Difference
Sediment Yield per CellSoil Survey Data
60 to 90
30 to 60
10 to 30
0 100 200 300 400 500 600700 800
RMS Difference
Sediment Generated in Each CellSTATSGO Data
60 to 90
30 to 60
10 to 30
0 20 40 60 80 100 120140160180
RMS Difference
Sediment Yield per CellSTATSGO Data
60 to 90
0 100 200 300400 500 600700 800
Figure 5. RMS for AGNPS sediment output.
(USLE) cropping management factor (C) and the SCS
curve number (CN). To address the concerns regarding
the influence of these parameters on sediment yield, the
RMS differences (see Table 6 and Figure 6) and general
statistics (see Table 7) for these parameters were computed.
Discussion
When evaluating the RMS as an indicator of the effect
of resolution on input parameters and output sediment
values, looking at the overall trend between resolutions,
rather than focusing on specific values, is important. The
RMS statistics for the soil erodibility factor (K), the crop-
ping management factor (C), and the SCS curve number
(CN) are least affected by a decrease in cell size reso-
lution from 10x10 meters to 30 x 30 meters. These
parameters are most affected when cell size resolution
decreases from 30- x 30-meter to 60- x 60-meter reso-
lution. As resolution decreases further from 60 x 60
meters to 90 x 90 meters, the effect on RMS decreases.
The small-large-smaller trend in the RMS for these pa-
rameters is the same trend seen in the RMS for sedi-
ment yield throughout the watershed. The sediment
Table 6. RMS Difference: AGNPS Input Parameters of
Concern
Parameters of Concern
K value
Cropping factor
SCS curve number
Slope
10 to 30
0.0058
0.03
2.11
8.17
30 to 60
0.054
0.173
11.40
5.59
60 to 90
0.051
0.015
8.40
3.68
yield within each cell therefore seems to be most af-
fected by these input parameters. The general statistics
for each of these parameters of concern show that very
little difference exists in the values within each resolu-
tion, with the exception of slope. Slope values are higher
at the higher resolutions and decrease as resolution
increases. This is related to the method in which the CIS
calculates slope.
The RMS statistics comparing resolutions for sediment
generated within each cell follow the same trend as the
RMS statistics for slope percentage. As resolution in-
creases, so do the discrepancies between the compared
171
-------
Soil Erodibility (K)
60 to 90
30 to 60
10 to 30
0.01 0.02 0.03 0.04 0.05 0.06
USLE Cropping Management (C)
60 to 90
30 to 60
10 to 30
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
SCS Curve Number (CN)
Slope %
60 to 90
30 to 60
10 to 30
60 to 90
30 to 60
10 to 30
10 12
0123456789
Figure 6. RMS difference for AGNPS input parameter of concern.
data sets. This trend between resolutions indicates that
slope values influence sediment generated within each
cell in the watershed.
The results from the 10- x 10-meter resolution data set
were obtained by selecting the center cells of the
10- x 10-meter resolution data set and routing the data
from this set through AGNPS using the flow pathways
developed for the 30- x 30-meter resolution. The results
do not provide the same information as would be pro-
vided had the entire 10- x 10-meter data set been routed
through AGNPS. The RMS values obtained from com-
parisons of the 10- x 10-meter resolution input parame-
ters with the 30- x 30-meter resolution input data reveal
that little difference exists between the data in these
resolutions. Comparison of the center cell 10- x 10-meter
data set output with field monitored data shows that the
10x10 center cell data overestimates both peak flow
and sediment yielded at the watershed outlet. This can
be attributed to the larger slope values in this resolution.
Conclusion
This study used CIS to generate data files for application
to the AGNPS model. The objectives of this project were
to evaluate the effect of different levels of detail used in
generating the input files on selected input and output
parameters. The results show that, for a watershed with
characteristics equivalent to those of the study area
watershed, differences exist in model output based on
the cell size resolution of the watershed.
The selected cell size resolution directly affects slope
values. The influence of the slope parameter dominates
AGNPS predictions for sediment generated within each
cell and sediment yield at the watershed outlet in the
study area watershed. The indicated parameters of con-
cern have the most influence on sediment yield for each
cell in the watershed. The greatest variation in the indi-
cated parameters of concern and thus the sediment
yield output occurs between the 30- x 30-meter and
60- x 60-meter resolutions. AGNPS estimates for sedi-
ment yield in files generated from STATSGO data were
larger than sediment yields from files generated with soil
survey soils data in the 30- x 30-, 60- x 60-, and
90- x 90-meter resolutions. For this watershed, how-
ever, no significant difference existed between data gen-
erated from soil survey and STATSGO data sources as
indicated by the kappa coefficient of similarity.
Results predicted by the AGNPS model at the water-
shed outlet were compared with results from an actual
storm monitored at the watershed outlet. The 30- x 30-
meter resolution data set provided the most accurate
172
-------
Table 7. Statistics for AGNPS
Value 10 x 10
Input Parameters of Concern
30 x 30 60 x 60 90 x 90
K Value (units of K)
Average
Maximum
Minimum
Standard deviation
C Factor (units of C)
Average
Maximum
Minimum
Standard deviation
CN (units of CN)
Average
Maximum
Minimum
Standard deviation
Slope (%)
Average
Maximum
Minimum
Standard deviation
0.2989
0.49
0.17
0.0453
0.0306
0.076
0
0.0212
71 .004
100
55
9.13
34.13
567
0
33.75
0.2989
0.49
0.17
0.0453
0.0306
0.076
0
0.0211
71.003
100
55
9.08
30.153
224
0
23.45
0.2882
0.37
0.17
0.0513
0.0295
0.076
0
0.0213
70.75
100
55
8.97
27.67
152
0
18.97
0.2889
0.49
0.17
0.0456
0.0295
0.076
0
0.0208
70.68
100
55
8.84
26.38
99
0
16.44
prediction for peak flow at the watershed outlet. AGNPS
output in the 10- x 10-meter center, 30- x 30-meter, and
60- x 60-meter resolutions overestimated the actual
sediment yield recorded at the watershed outlet for the
validated storm event.
For the study area watershed, cell size resolution of
30 x 30 meters seems appropriate based on the accurate
AGNPS model prediction for peak flow when validated
with the field-monitored storm. The steep slopes created
in the 10- x 10-meter resolution data set may lead to an
overestimation of sediment output, rendering data at this
resolution unreliable. At this time, the 10- x 10-meter
resolution is both impractical and infeasible for use with
the AGNPS model.
AGNPS output at the highest resolution does not pro-
vide a better approximation of sediment yield at the
watershed outlet. AGNPS output for sediment gener-
ated within each cell in the watershed at the highest
resolution does not accurately simulate the watershed
processes. AGNPS output generated from the more
detailed soil survey data is not significantly different from
data generated by the STATSGO digital soils database.
This study raises the following questions:
What level of detail is both practical and acceptable
for policy-making and decision-making?
What constitutes a cost-effective analysis?
Ultimately, these questions are best answered on a
case-by-case basis and should be determined based on
the size of the study area and on how the results of the
analysis will be used (i.e., to make a direct land use
decision or for broader planning). For broad planning
analyses on large watersheds, the benefit of digitizing
the soil survey data is outweighed by the cost in time
and effort to generate this detailed database. STATSGO
data may be sufficient. If a direct land management
decision is being made for a small area such as a farm
within a watershed, however, the analysis should use
the most detailed soils data.
Recommendations for Future Work
The original intent of this study was to use the capabili-
ties of AGNPS Version 4.03 to evaluate a watershed
using data generated at a high cell size resolution
10x10 meters. AGNPS h
ad never been used to evaluate data at such a high
resolution. As discovered during this project, the newest
version of AGNPS is not, at this time, capable of han-
dling a data set that has more than 32,767 cells (19).
Once this limitation with the AGNPS model is remedied,
the entire 10- x 10-meter resolution data set should be
routed through the model so that definite conclusions
regarding the applicability of such a detailed data set
can be made.
References
1. U.S. Department of Agriculture (USDA). 1991. Riparian forest
buffers: Function and design for protection and enhancement of
water resources. N1-PR-07-91.
2. Ashraf, M.S., and O.K. Borah. 1992. Modeling pollutant transport
in runoff and sediment. Trans. Amer. Soc. Agric. Eng. 35:1,789-
1,797.
3. Barten, P., and K. Stave. 1993. Characterization of streamflow
and sediment source areas for the Little Beaver Kill Watershed,
Ulster County, NY (July).
4. Vieux, B., and S. Needham. 1993. Nonpoint-pollution model sen-
sitivity to grid cell size. J. Water Resour. Planning and Mgmt.
119(2).
5. Kumar, V. 1993. Geographic information system application for
nonpoint source pollution management. Logan, UT: Utah State
University.
6. Yoon, J., L. Padmanabhan, and L. Woodbury. 1993. Linking Ag-
ricultural Nonpoint Source Pollution Model (AGNPS) to a geo-
graphic information system. Proceedings of the AWRA
Conference on Geographic Information Systems and Water Re-
sources (March).
7. Haddock, G., and P. Jankowski. 1993. Integrating nonpoint
source pollution modeling with a geographic information system.
Department of Geography, University of Idaho.
173
-------
8. Mitchell, J., B. Engel, R. Srinivasan, and S. Wang. 1993. Valida-
tion of AGNPS for small watersheds using an integrated
AGNPS/GIS system. Proceedings of the AWRA Conference on
Geographic Information Systems and Water Resources (March).
9. He, C., J. Riggs, and Y. Kang. 1993. Integration of geographic
information systems and a computer model to evaluate impacts
of agricultural runoff on water quality. Proceedings of the AWRA
Conference on Geographic Information Systems and Water Re-
sources (March).
10. Freezor, D., M. Hirschi, and B. Lesikar. 1989. Effect of cell size
on AGNPS prediction. Paper No. #89-2184. ASAE/CSAE Meet-
ing, St. Joseph, Ml.
11. Young, R., C. Onstad, D. Bosch, and W Anderson. 1989.
AGNPS: A nonpoint-source pollution model for evaluating agri-
cultural watersheds. J. Soil and Water Conserv. 44(2).
12. U.S. Department of Agriculture (USDA). 1987. Agricultural Re-
search Service, AGNPS: Agricultural nonpoint-source pollution
model, a watershed analysis tool. Conservation Research Report
No. 35.
13. Star, J., and J. Estes. 1990. Geographic information systems: An
introduction. Englewood Cliffs, NJ: Prentice Hall.
14. ESRI. 1987. PC ARC/INFO user's guides. Redlands, CA: Envi-
ronmental Systems Research Institute, Inc.
15. Eastman, R. 1993. IDRISI technical reference. Worcester, MA:
Clark University.
16. Clarke, K.C. 1990. Analytical and computer cartography. Engle-
wood Cliffs, NJ: Prentice Hall.
17. Clesceri, L, A. Greenberg, and R. Trussell. 1989. Standard meth-
ods for examination of water and wastewater, 17th ed. pp. 2-75
to 2-76.
18. Rosenfield, G.H., and K. Fitzpatrick-Lins. 1986. A coefficient of
agreement as a measure of thematic classification accuracy.
Photogrammetric Engin. and Remote Sensing 52(2).
19. U.S. Department of Agriculture, Soil Conservation Service. 1995.
AGNPS Newsletter.
174
-------
XGRCWP, a Knowledge- and GIS-Based System for Selection, Evaluation,
and Design of Water Quality Control Practices in Agricultural Watersheds
Runxuan Zhao, Michael A. Foster, Paul D. Robillard, and David W. Lehning
Penn State University, University Park, Pennsylvania
Abstract
The Expert CIS Rural Clean Water Program (XGRCWP)
integrates a geographic information system (CIS), a
relational database, simulation models, and hypertext
mark language documents to form an advisory system
that selects, evaluates, sites, and designs nonpoint
source pollution control systems in agricultural water-
sheds. Its major features include:
Customized CIS functions to obtain spatial and attrib-
ute data and feed them to a rule-based expert system
for selecting feasible control practices.
A user interface for examining the field-specific con-
ditions and recommended control practices on the
screen by clicking on the displayed field boundary
map.
A direct linkage between the CIS spatial data and the
relational attribute data, which allows users to exam-
ine data on the screen interactively.
A graphic user interface to CIS functions, which en-
ables users to perform routine watershed analyses.
Linkage to hypertext reference modules viewable by
Mosaic Internet document browser.
Dynamic access to other models such as the Agricul-
tural Nonpoint Source Simulation Model.
The software environment of XGRCWP is GRASS 4.1
and X-Windows on SUN OS 4.3.1. Its major functions
have been tested for the Sycamore Creek watershed in
Ingham County, Michigan.
Introduction
In 1981, the U.S. Environmental Protection Agency
(EPA) and the U.S. Department of Agriculture (USDA)
initiated the Rural Clean Water Program (RCWP) in 21
agricultural watersheds. This program represents the
most intensive water quality monitoring and implemen-
tation and evaluation of nutrient, sediment, and pesti-
cide reduction practices ever undertaken in the United
States (1). More than a decade of research efforts has
resulted in a wealth of experiences and lessons on
selection, siting, and evaluation of nonpoint source con-
trol practices.
The storehouse of knowledge gained from RCWP is of
little use, however, unless it is properly integrated and
packaged in an easily accessible form. Technology
transfer of this knowledge is therefore critically impor-
tant. To integrate and synthesize the lessons learned
from RCWP, Penn State University initiated an RCWP
expert project. The hypertext-based version of the
RCWP expert system, completed in 1993, can select
and evaluate nonpoint source control systems at a sin-
gle site. Although the hypertext-based version is still
suitable for users who do not have access to geographic
information systems (CIS) data, it is inadequate for the
comprehensive selection and evaluation of control sys-
tems on a watershed basis. It does not provide the user
the spatial reference of a site and requires the user's
subjective judgment for the model input.
The Expert CIS Rural Clean Water Program (XGRCWP)
is the UNIX and X-Window version of the RCWP expert
system, which integrates CIS and the RCWP expert
system to provide decision support at multiple spatial
scales from single fields to subwatersheds to the water-
shed scale. This paper presents the major features of
XGRCWP, including design of the expert system, inter-
face to CIS functions, and linkages to a relational data-
base and simulation models.
Overview
XGRCWP comprises five major components (see
Figure 1):
An expert system for recommending control practices
based on site-specific information.
175
-------
HTML Reference
Modules:
Contaminants
Monitoring
Transport
Case Studies
Expert System:
Recommending
Control Practices
X/Motif
Graphic User
Interface
GIS Functions:
Customized and
Existing Ones for
Watershed Analysis
Relational Database:
Relate Spatial and
Attribute Data
Simulation Models:
AGNPS
GLEAMS (To Be
Added)
Figure 1. Major components of XGRCWP and their relationships.
Custom and existing GIS functions for watershed
analysis and estimation of contaminant loading
potential.
Linkage to fields, soils, and land use databases.
Linkage to the Agricultural Nonpoint Source Simula-
tion Model (AGNPS) (2).
Hypertext mark language (HTML) reference modules.
The X/Motif graphic user interface (GUI) integrates the
five components and allows the user to navigate flexibly
among them. The components are also internally con-
nected in different ways. For example, the expert system
can use the customized GIS functions to retrieve site-
specific information from Geographical Resource Analy-
sis Support Systems (GRASS) (3) data layers and
INFORMIX relational database tables. In addition, the
expert recommendations of control practices can be
displayed and examined using GRASS functions. Fi-
nally, the GIS functions can help generate input to the
AGNPS model, and its output can be converted to GIS
format for additional analyses.
Design of the Expert System
The objective of the expert system is to recommend
feasible control systems (i.e., complementary sets of
control practices to reduce nonpoint source pollution
based on site-specific conditions). One distinct feature
of this system is the combination of two modes of data
acquisition: direct user input and GIS functions.
XGRCWP also has two modes for deriving the expert
recommendations: batch or interactive. This section dis-
cusses these aspects of the expert system as well as its
knowledge base.
Rules for Control Practice Selection
The knowledge base of the expert system includes the
following six site-specific characteristics:
Contaminant of interest and its adsorption characteristic.
Potential level of contaminant loading (low, medium,
or high).
Potential level of contaminant leaching (low, medium,
or high).
Soil hydrologic group (A, B, C, or D).
176
-------
Time of year (during or outside the growing season).
Type of land use (cropland, animal waste, or critical
area).
The user first chooses a contaminant of interest from a
list consisting of four kinds of pesticides (strongly, mod-
erately, or weakly adsorbed, and nonadsorbed) and
eight other contaminants (ammonia, bacteria, sediment,
total nitrogen, total phosphorus, nitrate, orthophospho-
rus, and viruses). The values of other characteristics,
some of which vary with the contaminant specified, can
then be input either directly by the user or by custom
GRASS functions as discussed in the section of this
paper on data acquisition.
The RCWP used 14 general categories of control prac-
tices (see Table 1). Many suitable conditions were es-
tablished for each general category. For example,
conservation tillage is recommended to reduce runoff for
cropland under conditions otherwise favoring loss
through sediment transport, such as a contaminant
strongly adsorbed to the soil (e.g., total phosphorus), the
nongrowing season, and soils with a relatively high run-
off potential (soil group C or D) (see Figure 2). Each
general category includes several specific control prac-
tices. When a general practice category is recom-
mended, the user must decide which specific practice
within that general category to evaluate further by con-
sulting the nonpoint source database (NPSDB) for the
reported research data about this practice or by running
the AGNPS simulation model.
Data Acquisition
The expert system recommends one or more control
systems based on site-specific conditions that are either
directly input by the user or calculated by the customized
GRASS functions. The user always specifies the con-
taminant of interest and the season, while a GRASS
function (R.HYDRO-GRP) always determines the soil
hydrologic group of each field. For the other factors
(loading potential, leaching potential, and application
Table 1. The Best Management Practices Used in the Rural
Clean Water Program
Source control
practices
Nutrient Management (NUTR)
Pesticide Management (PEST)
Structural control Animal waste systems (AWS)
practices Diversion systems (DIV)
Sediment retention and water control (SED)
Terrace systems (TERR)
Waterway systems (WATW)
Vegetative control
practices
Conservation tillage (CT)
Critical area treatment (CAT)
Cropland protection systems (CPS)
Grazing land protection (GLP)
Permanent vegetative cover (PVC)
Stream protection (SP)
Stripcropping (SCR)
class), however, the user has two alternative ways to
decide input values. For example, after the user selects
a contaminant of interest, the program displays the con-
taminant loading potential window (see Figure 3). The
potential level of the selected contaminant can be indi-
cated if the user knows it. Otherwise, the user can let
the GRASS functions derive loading potential from ex-
isting field data.
The direct input option can also be used to help the user
address "what-if questions. When the user selects the
CIS functions to determine the loading potential,
XGRCWP makes a series of calls to appropriate cus-
tomized GRASS functions according to the current con-
taminant of interest. For example, if the contaminant is
total nitrogen, the functions R.MANURE, R.FERT, and
R.B.CONCENTRATION are called to estimate total ni-
trogen from manure, fertilizer, and soil base concentration,
respectively. Another GRASS function, R.NP.LOADING,
is then called to translate the quantitative measure of
loading potential into the qualitative classification (low,
medium, or high) as input to the expert systems. These
GRASS functions generate the inputs by searching and
converting the data from INFORMIX relational data ta-
bles that are associated with the GRASS spatially refer-
enced data, such as field boundary and soil map. Table
2 lists the customized GRASS functions developed for
data acquisition.
Control System Recommendation
XGRCWP derives the expert recommendations for con-
trol systems in two ways: in a batch mode for every field
in a watershed and in an interactive mode for a user-
specified field.
In batch mode, an existing GRASS function, R.INFER,
is used to create a raster data layer for each general
practice category of control practice according to a rule-
set prepared for that general category. For example, the
contents and formats for the conservation tillage prac-
tice are documented in Table 3. The raster data layer for
representing the conservation tillage recommendations
(CTree) is generated by running R.INFER with the ap-
propriate rule. The category value of CTree is 1 at each
point in the data layer where the conservation tillage is
recommended, or 0 otherwise. The R.INFER function is
similarly called for other general practice categories.
Additional GRASS functions can then display or further
analyze the resulting map layers. The batch mode pro-
vides the user the overall picture with a watershed-wide
view of feasible control systems.
In the interactive mode, the field boundary map is dis-
played and the user can specify any field of interest by
clicking the mouse on it. The recommendations and the
site-specific conditions of the field are displayed on the
right half of the screen. The recommended control prac-
tices are also displayed in a popup window for further
177
-------
Adsorb
Nonadsorb
Contaminant
A
B
C | D
Soil Group
Low
Medium
High
Loading
Low
Medium
High
Leaching
Growing] Nongrowing
Season
CL
AW
HSA
Application
Figure 2. Dependency network (AND-OR diagram) for site-specific recommendation of conservation tillage.
examination, such as the specific practices within each
general category, the feasible control systems for non-
point source pollution control, and research data on the
practices. The interactive mode is implemented through
the integration of a Bourne shell script, structured query
language (SQL) commands, a customized GRASS
function (R.RCWP.EXPERT), and GRASS display func-
tions with the Motif GUI. Interactive mode is intended for
detailed consideration of a specific farm.
Interface to GIS Functions
XGRCWP provides a GUI to most of the customized
GRASS functions and some of GRASS'S existing func-
tions (see Figure 4). This interface shields the user from
complex syntax so the user can focus on the subject
(LOW
High
% loitf determined by GIS fanctivn s
\ i,,,,..,,.,*,,.. *,,.,,,. *>
-Tip
Figure 3. The popup window for the potential level of contami-
nant loading.
matter. The GUI makes it easier for the user to perform
routine operations such as estimation of contaminant
loading, identification of critical areas, erosion and runoff
calculation, and other watershed analysis tasks. It also
helps the user make full, effective use of all custom and
some existing GRASS functions.
Linkages to Database and Other Models
Data Structure
The GRASS functions used to generate inputs for the
expert system use the same soils and fields relational
databases as the Water Quality Model/GRASS Interface
under development by the Soil Conservation Service
(SCS) (4). XGRCWP and our custom GRASS functions
were tested forthe Sycamore Creek watershed, Ingham
County, Michigan. In this data structure, spatial data
(e.g., field boundaries, watershed boundaries, soils map
unit boundaries, and elevation data) are saved as
GRASS raster data layers while attribute data (e.g., crop
information, fertilizing schedule, soil information) are
stored in INFORMIX relational database tables. Each
field or soil map unit is assigned a unique identification
(ID) number. The field attribute (INFORMIX) data also
contain this ID number. The linkage between the
GRASS raster map and the INFORMIX data is accom-
plished with a GRASS category label (see Figure 5).
Linkage to Database
To allow the interactive examination of field data from
GRASS raster layers and the associated relational da-
178
-------
Table 2. Summary of the Customized GRASS Functions Developed by Nonpoint Source Agricultural Engineering Research
Group at Penn State University To Generate Inputs for the RCWP Expert System
Name
Descriptions
R.FERT
R.MANURE
R.B.CONCENTRATION
R.NP.LOADING
R.EROSION
R.LEACHING.P
R.HYDRO-GRP
Produces raster maps of total nitrogen or total phosphorous from the scheduled fertilizer applications for
different crops by dynamically retrieving information from a GRASS data layer and INFORMIX data tables
Calculates the total manure on each farm according to animal numbers and types (e.g., dairy cow, beef
cow, horses, swine), allocates manure to the fields on a farm by a user specified strategy (uniformly
spreading or inverse distance weighted distributing method), and finally estimates nitrogen and
phosphorous loading from manure application rate, conversion factor, percentages of transportation
losses, and volatile losses
Estimates nitrogen and phosphorous concentration in parts per million within different types of soils
according to the organic matter contents
Classifies the loading potential of nitrogen or phosphorous into three categories (low, medium, and high)
based on the actual loading from fertilizer and manure and the N or P concentration in soils
Obtains a relative measure of soil erosion severity by dividing the amount of erosion by the tolerance
values of the soils and then reclassifying them into three categories (low, medium, and high)
Estimates leaching index from soil hydrologic group and annual and seasonal precipitation and classifies
it into three categories (low, medium, and high)
Retrieves soil hydrologic group from the INFORMIX database and reclassifies the soil map into soil
hydrologic groups
tabase tables, XGRCWP calls our custom function,
D.WHAT.FIELD.SH, a Bourne shell script that dynami-
cally links GRASS raster layers and the INFORMIX
database tables. When the user clicks on a field, for
example, this function extracts field-specific information
from INFORMIX tables such as field information, fertili-
zation schedule, crop operation schedule, and soil infor-
mation. The D.WHAT.FIELD.SH function then displays
all related soils and fields information for the given field.
It also marks the field boundary map to remind the user
which fields have already been examined.
Linkage to Reference Modules
At any stage of the selection, evaluation, siting, and
design procedure for control practices, the user can
consult reference modules that provide information,
guidance, and data about contaminant properties, trans-
port variables, and examples of applications from
RCWP projects. Four reference modules are available
in the Macintosh version of RCWP expert system: con-
taminants, monitoring, transport, and case studies. We
are currently converting these reference modules into
Mosaic-viewable HTML documents so that they can be
accessed from XGRCWP. Mosaic is a public domain,
Internet-aware document browser that is available for
X-Windows, Macintosh, and Microsoft Windows.
All four modules use graphics to demonstrate design
procedures and contaminant control processes. The
contaminant module provides information about 11 cate-
gories of contaminants cited in RCWP projects and their
impacts on surface and ground-water resources. The
monitoring module describes different aspects of water
quality sampling and analysis systems. The transport
module describes contaminant pathways in surface
and ground water. The case studies module presents
detailed examples from key RCWP projects. These
examples cover both practice selection and implemen-
tation aspects of control systems. The reference modules
serve as a complementary component of XGRCWP.
Linkage to AGNPS
AGNPS is a distributed-parameter, storm event-based
model that estimates runoff, sedimentation, and nutrient
loss in surface runoff within agricultural watersheds (2).
The prototype version of the Water Quality Model/GRASS
Interface developed by SCS conveniently generates an
AGNPS input file for all cells in a watershed from the
spatial and relational soils and fields databases. The
UNIX version of AGNPS can then use this input file.
XGRCWP can call AGNPS directly from its X-Window
interface and convert standard AGNPS model outputs
for all cells in the watershed into GRASS raster format
for display and analysis.
Discussion
The literature on software systems for managing non-
point source pollution in agricultural watersheds is di-
verse and rapidly growing. With few exceptions (5-7),
these decision support systems are purely model-
based, CIS-based (8), or hybrid systems with models
running within a CIS framework (9-14). The addition of
expert system components can overcome some of the
difficulties in primarily model-based systems:
Overly intensive input data requirements.
Inability to handle missing or incomplete data.
Requirements that all inputs be numerically expressed.
179
-------
Table 3. The Rule File for Recommending Conservation Tillage
IFNOTMAP app.class 3
ANDIFMAP contam.feature 2
ANDIFMAP leaching.p 1
ANDIFMAP soil.g 1 2
ANDIFMAP contam.load 1
ANDIFMAP season 2
THENMAPHYP 1 yes, CT
is recommended
i
IFNOTMAP app.class 3
ANDIFMAP contam.feature 2
ANDNOTMAP leaching.p 3
ANDNOTMAP soil.g 1
ANDNOTMAP contam.load 3
ANDIFMAP season 1
THENMAPHYP 1 yes, CT
is recommended
i
IFNOTMAP app.class 3
ANDIFMAP contam.feature 1
ANDNOTMAP contam.load 3
ANDIFMAP season 1
THENMAPHYP 1 yes, CT
is recommended
i
IFNOTMAP app.class 3
ANDIFMAP contam.feature 1
ANDNOTMAP leaching.p 3
ANDIFMAP soil.g 3 4
ANDNOTMAP contam.load 3
ANDIFMAP season 2
THENMAPHYP 1 yes, CT
is recommended
!application class is not high-source area
!contaminant is nonadsorbance
teaching potential is low
!soil groups are A or D
!contaminant loading is low
!nongrowing season
!application class is not high-source area
!contaminant is nonadsorbance
teaching potential is not high
!soil groups are not A
!contaminant loading is not high
!growing season
!application class is not high-source area
!contaminant is strong adsorbance
!contaminant loading is not high
!growing season
!application class is not high-source area
!contaminant is strong adsorbance
teaching potential is not high
!soil groups are C or D
!contaminant loading is not high
!nongrowing season
Manure M/ffffmf I ttati
Alamtre J
Options:
Loatl
_.r. i'' f>i--'-f:P I i~ fj i -:.'.
>!. f - "ill r i
Figure 4. The GUI to the R.MANURE function.
High degree of expertise needed to structure model
input and explain model output relative to the user's
problem context.
The expert system component of XGRCWP also re-
duces the number of model runs needed for decision
support through preliminary, rule-based screening of
control systems at each site of interest in the watershed.
Conclusions
XGRCWP incorporates several kinds of expertise for the
user's benefit:
Subject matter expertise in siting and selecting non-
point source control systems in agricultural water-
sheds.
Expertise in configuring AGNPS model input from the
soils and fields databases.
Expertise in interpreting, explaining, and visualizing
expert system and model input.
The integration of the expert system and the GRASS
CIS makes input to the expert system easier and more
objective. It enhances the expert system's capability
for recommending effective control practices at the
field level to achieve watershed contaminant loading
objectives. XGRCWP is designed as an open structured
180
-------
Category Value
1173
Category Label
FDID1173
Fnin
fert_sch_id
FTID01
fert_name
13-13-13
appl_rate
200.0
fert_name
13-13-13
total_n
13.000
total_p
5.700
Figure 5. Data structure of Sycamore Creek watershed, Ingham County, Michigan.
program and has great potential to be improved easily
and continually according to users' feedback. Ongoing
efforts to enhance the program include:
Developing more rules that incorporate topographical
factors such as slope and slope length for the expert
system so that more site-specific control practices
can be recommended.
Adding dynamic hypertext-based help and reference
to the program.
Establishing intelligent linkages among the expert
system, the CIS functions, and other simulation or
design models for nonpoint source control practices.
Acknowledgments
We thank Vicki Anderson, Ruth Shaffer, and Brent Stin-
son of the Michigan USDA SCS and John Suppnick of
the Michigan Department of Natural Resources for their
assistance in working with the underlying databases in
the Sycamore Creek watershed. This project is sup-
ported by U.S. EPA Grant No. X-818240.
References
1. U.S. EPA. 1992. Extending the RCWP knowledge base to future
nonpoint source control projects. In: Proceedings of the National
RCWP Symposium. EPA/625/R-92/006. Washington, DC.
2. Young, R.A., C.A. Onstad, D.D. Bosch, and WP. Anderson. 1989.
AGNPS: A nonpoint-source pollution model for evaluating agri-
cultural watersheds. J. Soil and Water Conserv. 44(2):168-173.
3. USACERL. 1993. GRASS 4.1 users reference manual. Champaign,
IL: U.S. Army Corps of Engineers Construction Engineering Re-
search Laboratory.
4. USDA. 1993. Users' guide to water quality model/GRASS inter-
face. Fort Collins, CO: U.S. Department of Agriculture, Soil
Conservation Service, Technology Information Systems Division.
5. Barnwell, T.O., L.C. Brown, and W. Marek. 1989. Application of
expert systems technology in water quality modeling. Wat. Sci.
Tech. 21:1,045-1,056.
6. Ford, D.A., A.P. Kruzic, and R.L. Doneker. 1993. Using GLEAMS
to Evaluate the Agricultural Waste Application Rule-Based
Decision Support (AWARDS) computer program. Wat. Sci. Tech.
28(3-5):625-634.
7. Yakowitz, D.S., J.J. Stone, L.J. Lane, P. Heilman, J. Masterson,
J. Abolt, and B. Imam. 1993. A decision support system for evalu-
ating the effects of alternative farm management systems on
water and economics. Wat. Sci. Tech. 28(3-5):47-54.
181
-------
8. Hamlett, J.M., D.A. Miller, R.L. Day, G.A. Peterson, G.M. Baumer,
and J. Russo. 1992. Statewide CIS-based ranking of watersheds
for agricultural pollution prevention. J. Soil Water Cons.
47(3):399-404.
9. Srinivasan, R., and B.A. Engel. 1994. A spatial decision support
system for assessing agricultural nonpoint source pollution.
Water Resources Bull. 30(3):441-452.
10. Srinivasan, R., and J.G. Arnold. 1994. Integration of a basin-
scale water quality model with GIS. Water Resources Bull.
30(3):453-462.
11. Engel, B.A., R. Srinivasan, J. Arnold, and S.J. Brown. 1993.
Nonpoint source (NPS) pollution modeling using models inte-
grated with geographic information systems (GIS). Wat. Sci. Tech.
28(3-5):625-690.
12. He, C., J.F. Riggs, and Y. Kang. 1993. Integration of geographic
information systems and a computer model to evaluate impacts
of agricultural runoff on water quality. Water Resources Bull.
29(6):891-900.
13. Kiker, G.A., G.M. Cambell, and J. Zhang. 1992. CREAMS-WT
linked with GIS to simulate phosphorus loading. ASAE Paper
No. 92-9016. St. Joseph, Ml: American Society of Agricultural
Engineers.
14. Tim, U.S., S. Mostaghimi, and V.O. Shanholtz. 1992. Identification
of critical nonpoint pollution source areas using geographic infor-
mation systems and water quality modeling. Water Resources
Bull. 28:877-887.
182
-------
Integration of EPA Mainframe Graphics and GiS in a UNIX Workstation Environment
To Solve Environmental Problems
William B. Samuels
Science Applications International Corporation, McLean, Virginia
Phillip Taylor
Tetra Tech, Inc., Fairfax, Virginia
Paul Evenhouse
Martin Marietta, Inc., Research Triangle Park, North Carolina
Robert King
Assessment and Watershed Protection Division, U.S. Environmental Protection Agency,
Washington, DC
Abstract
The Assessment and Watershed Protection Division of
the Office of Wetlands, Oceans, and Watersheds has
developed water quality analysis software on the U.S.
Environmental Protection Agency (EPA) mainframe
computer. This software integrates national on-line en-
vironmental databases and produces maps, tables,
graphics, and reports that display information such as
water quality trends, discharge monitoring reports, per-
mit limits, and design flow analyses.
In the past, this graphic software was available only to
users connected to the mainframe with IBM graphics
terminals or PCs with graphics emulation software. Re-
cently, software has been developed that can be used
to: 1) access the EPA mainframe from a UNIX worksta-
tion via the Internet, 2) execute the Water Quality Analy-
sis System (WQAS) procedures, 3) display WQAS
graphics in an X-Window on the workstation, and 4)
download data in a geographic information system (GIS)
format from the mainframe. At the same time, this work-
station can execute ARC/INFO and ARC/VIEW applica-
tions in other X-Windows. This capability allows analysts
to have the power of GIS, the mainframe databases
(e.g., Permits Compliance System [PCS], STORET,
Reach File, Industrial Facilities Discharge File, Daily
Flow File, Toxic Chemical Release Inventory), and the
retrieval/analysis/display software (Environmental Data
Display Manager, Mapping and Data Display Manager,
Reach Pollutant Assessment [RPA], PCS-STORET In-
terface, UNIRAS) available to them on one desktop.
This capability extends the tool set available to GIS
analysts for environmental problem-solving.
This paper discusses application of these tools and
databases to several problems, including EPAs water-
shed-based approach to permitting, and the RPA, an
automated method to identify priority pollutants in water-
sheds.
Introduction
The purpose of this paper is to explore new geographic
information systems (GIS) data integration tools that are
applicable to a wide range of environmental problems,
including the U.S. Environmental Protection Agency's
(EPAs) watershed-based approach to permitting and
the Reach Pollutant Assessment (RPA), an automated
method to identify priority pollutants in watersheds. The
ultimate goal is to make these tools and databases
accessible to a wide range of users.
Understanding aquatic resource-based water quality
management depends on access to and integration of
diverse information from many sources. To date, the
techniques to perform this integration, and thus yield
meaningful analyses supporting environmental deci-
sion-making, are neither fully developed nor docu-
mented. New tools and information resources are now
available, but not used to their full potential, for more
valuable water quality and watershed analyses. EPA
183
-------
headquarters is responsible for ensuring that integrated
data management tools are available for water quality
analyses and data reporting as well as making national
data systems more useful. EPA will accomplish this by
upgrading and crosslinking systems, developing interac-
tive data retrieval and analysis mechanisms, and provid-
ing easy downloading of data to client workstations.
The Assessment and Watershed Protection Division
(AWPD) of the Office of Wetlands, Oceans, and Water-
sheds (OWOW) has developed water quality analysis
software on the EPA mainframe computer (1). This soft-
ware integrates national on-line environmental data-
bases and produces maps, tables, graphics, and reports
that display information such as water quality trends,
discharge monitoring reports, permit limits, and design
flow analyses. In the past, this graphic software was
available only to users connected to the mainframe with
IBM graphics terminals or PCs with graphics emulation
software. Recently, software has been developed that
can be used to:
Access the EPA mainframe from a UNIX workstation
via the Internet.
Execute the Water Quality Analysis System (WQAS)
procedures.
Display WQAS graphics in an X-Window on the
workstation.
Download data in a CIS format from the mainframe.
At the same time, this workstation can execute
ARC/INFO and ARC/VIEW applications in other X-Win-
dows. This capability allows analysts to have the power
of CIS, the mainframe databases (e.g., Permits Compli-
ance System [PCS], STORET, Reach File, Industrial
Facilities Discharge File, Daily Flow File, Toxic Chemical
Release Inventory), and the retrieval/analysis/display
software (Environmental Data Display Manager, Map-
ping and Data Display Manager, RPA, PCS-STORET
Interface, UNIRAS) available to them on one desktop.
This extends the tool set available to CIS analysts for
environmental problem-solving. This paper discusses
how these tools and databases have been applied to
two examples: 1) a watershed-based approach to per-
mitting and 2) the RPA, an automated procedure for
identifying watersheds with priority pollutants.
Mainframe Databases and Tools
The EPA IBM ES9000 mainframe computer, located in
Research Triangle Park, North Carolina, contains a
large volume of digital water quality and environmental
data available on-line through a number of data retrieval
and display tools (see Figure 1). Other documents de-
scribe these databases and tools in detail (2, 3).
This effort focused on showing how these databases
and tools can complement CIS activities. In some cases,
data can be directly downloaded to a workstation in CIS
format. An example of this is accessing EPAs Reach
File (Version 1) from GRIDS (4). In other cases, data-
bases are accessed by mainframe tools, the data are
processed, and a CIS data set is produced that can be
downloaded to a workstation. An example of this case
is the RPA (RPA3) tool that integrates data from the
Reach File, STORET, and PCS to identify priority pollut-
ants in watersheds (5).
The mainframe can be accessed through several paths:
Internet, PC dialup, or dedicated line into a terminal
controller (see Figure 2). In the applications presented
here, the Internet connectivity is emphasized because
this is the mechanism that makes these databases and
tools available to CIS analysts at their workstations.
Figure 3 shows the hardware and software require-
ments for Internet access to the water quality data inte-
gration tools. The basic components are a UNIX
workstation with an X-Window manager, the X3270 soft-
ware, and Internet connectivity. The X3270 software is
required to emulate an IBM 3270 full-screen terminal.
This software is publicly available through EPAs Na-
tional Computer Center User Support. In addition, an
account on the mainframe computer is required. Once
this account is established, an additional software mod-
ule, GDDMXD, is required to map IBM host-based
graphics to the workstation's X-Window. The GDDMXD
software resides on the mainframe and is loaded when
the user logs in. Once the hardware and software are
set up, a single UNIX workstation can provide access to
mainframe and workstation tools and databases on one
desktop (see Figure 4).
Applications
To illustrate how these mainframe and workstation
tools/databases can work together to solve environ-
mental problems, we present two applications. The first
shows a watershed-based approach to permitting; the
second describes the RPA.
Watershed-Based Approach to Permitting
The watershed approach is a process to synchronize
water quality monitoring, inspections, and permitting to
support water quality protection activities on a geo-
graphic basis. It is a coordinated and integrated method
to link science, permits, and other pollution control and
prevention activities to meet state water quality stand-
ards. Numerous local, state, and federal agencies have
recognized watershed approaches as the best way to
manage natural resources effectively and efficiently. Es-
tablishing a schedule for data collection, permit issu-
ance, and other elements of this approach affords the
opportunity to coordinate and integrate other natural
resource management efforts to make better use of
184
-------
NATIONAL DATABASES
1
Reach File FRDS
STORET USGS DLG
PCS LU/LC
TRI Soils
IFD Population
Gage Dun &
Drinks Brad street
Daily Flow ODES
BIOS
V J
^
' \
MAINFRAME TOOLS
EDDM GRIDS
MDDM DFLOW
RPA3 DESCON
IPS5 RF3MSTR
UNIMAP Bookmanager
Sitehelp Pathscan
v }
GIS
Figure 1. EPA mainframe databases/tools and linkage to GIS.
INTERNET
PC or UNIX workstation
with GIS software
DIALUP
DEDICATED LINE
EPA MAINFRAME
Databases
Tools
Figure 2. Access to the EPA mainframe databases and tools.
185
-------
LOCAL WORKSTATION
HARDWARE
UNIX workstation (with X-Windows) on Internet (i.e., DG Aviion)
SOFTWARE
X3270 Software - creates an X-Window, which emulates an IBM 3270
full-screen session (provided by EPA)
EPA MAINFRAME ACCOUNT
GDDMXD software provided on the mainframe to map GDDM
graphics into X-Windows
Figure 3. Hardware and software requirements for accessing water quality data integration tools via INTERNET.
EPA MAINFRAME
WINDOW- X3270
ARC/INFO
WINDOW
ARC/VIEW
WINDOW
DATA DOWN-
LOADING (FTP)
Figure 4. A UNIX workstation environment provides access to mainframe and workstation tools and databases on one desktop.
186
-------
limited local, state, and federal financial and human
resources (6).
This application illustrates how the watershed approach
used CIS and EPA mainframe databases and tools. As
an example, a four-step approach (see Figure 5) has
been developed and applied to an impaired watershed
(Saluda River basin) in South Carolina. Steps one and
two identified watersheds of concern through their
nonattainment of designated uses (see Figure 6) and
highlighted the cause of nonattainment, in this case
pathogens (see Figure 7). The data sets used were U.S.
Geological Survey (USGS) hydrologic unit boundaries,
Soil Conservation Service (SCS) watershed bounda-
ries, and data from the EPA waterbody system, which
were indexed to the SCS watersheds.1 These data sets
were integrated into an ARC/INFO arc macro language
(AMI) to allow users to pose queries and prioritize
watersheds for further investigation.
Once priorities were set, the third step was to evaluate,
in detail, the sources and causes of nonattainment. The
Saluda River basin, which had pathogens as its cause
of nonattainment, was selected for further analyses. In
this step, the mainframe tools supplement the worksta-
1 Clifford, J. 1994. Personal communication with Jack Clifford, U.S.
EPA, Washington, DC.
tion CIS capabilities illustrated so far. A STORET re-
trieval was performed for ambient water quality stations
monitoring for fecal conforms. The STORET stations
were partitioned into three categories (low, medium, and
high) according to the state fecal coliform standard (7),
which reads:
not to exceed a geometric mean of 200/100 ml,
based on five consecutive samples during any 30
day period; nor shall more than 10% of the total
samples during any 30 day period exceed 400/100 ml
The categories in Figure 8 correspond to the standard
as follows:
low: < 200/100 milliliters
200/100 milliliters < medium < 400/100 milliliters
high: > 400/100 milliliters
Figure 8 illustrates the use of ARC/VIEWto visualize the
location of fecal coliform "hot spots" in the Saluda River
basin. Figure 9 focuses on one SCS watershed
(03050109-040) where pathogens cause nonattain-
ment. The locations of industrial and municipal dis-
chargers are plotted, and facilities with fecal coliform
limits and their respective permit expiration dates (cap-
tured from the PCS) are shown in a table included with
the figure. The CIS capabilities used to generate Figure 9
1. IDENTIFY WATERSHEDS OF CONCERN
- Nonattained waters
2. PRIORITIZE WATERSHEDS FOR PERMITTING
Toxic vs. Nonconventiona! vs. Conventional
- Point source vs. Nonpoint source
- Existing/Designated use(s)
- Environmental equity; populations; endangered/sensitive species; etc.
3. EVALUATE SOURCES AND CAUSES
- Ambient water quality
- Location of point and nonpoint source discharges
- Existing controls
4. DEVELOP CONTROLS
- TMDL/WLAs
- NPDES permit limits/controls
Synchronize issuance
Figure 5. Four steps illustrating an example approach to permitting on a watershed basis.
187
-------
Step 1: Identify Watersheds of Concern
n 305b Clean Water Act Support
SsJect Use Category
Overall Use flgua^j-Ht) I Swimming j^CTndafyjCgntactgec
Select SiiBpert Category
(fptscfting I Partially Sue&ortlng Threatened I NotSupssrtlni
^JiBiBct Support Heasiire
FFerceni 5uuport Number of Miles [
aier Quality stations; o
tosic Releass IfiMSMory: 0
Figure 6. Identification of watersheds of concernwatersheds where at least 10 percent of the reaches are not fully supporting
overall designated use (data and ARC/INFO AMLs provided by Jack Clifford).
Step 2: Evaluate Priorities
Figure 7. Watersheds where the cause of nonattainment is pathogens (data and ARC/INFO AMLs provided by Jack Clifford).3
2 See note 1.
3 See note 1.
188
-------
Step 3: Evaluate Ambient Water QualityLow, Medium, and High Levels of Fecal Coliforms
Figure 8. Map of the Saluda River basin showing the location of STORET monitoring stations and fecal coliform levels.
Step 3: Existing ControlsDischargers That Have Limits for Fecal Coliforms
ij Dl rhargm In SCS bs in 41
S^M ,BD OlBljt, OtralltV)
VI Dischargers InSCS t>
1 *_ ^_
y tsea* coiiKusa levels
: I R^th File t -loath ^
_j DnrtUng Waiar Syppt* ^
*
3 Water Quaffly S
! Toxic Chewe.il selaasa iR
'/* . " *
Vf fiaOSOl 03 - SSLODftl
JAv^H'
Figure 9. Focus on SCS watershed 03050109-040 (shaded in gray). Locations of STORET stations, municipal and industrial dis-
chargers are also shown. The table in the upper right highlights dischargers with fecal coliform limits and their respective
permit expiration dates.
189
-------
show that permit issuance is not synchronous, which is
a key element in the watershed approach.
Within this local workstation CIS environment, the attrib-
utes associated with the STORE! monitoring stations
and the PCS dischargers are limited with respect to the
large amount of time series sampling data that exists in
these databases. Figures 10 through 12 illustrate how
an EPA mainframe procedure, the Environmental Data
Display Manager (EDDM), can be accessed from an
X-Window on the CIS workstation to query the entire
STORET and PCS databases and thus provide additional
data analysis and display capabilities to the CIS work-
station. In Figures 10 and 11, a water quality inventory was
performed for a STORET station, and a time series plot
of fecal coliform levels is displayed. In Figure 12, the
limits and discharge monitoring report (DMR) data were
accessed from PCS for a sewage treatment plant. Ex-
cursions beyond the PCS limits for fecal conforms are
easily visualized in the plot.
The fourth and final step in the watershed approach was
to develop controls for achieving water quality stand-
ards. This might include the development of total maxi-
mum daily loads (TMDL), waste load allocations (WLA),
and the synchronization of permit issuance. Another
mainframe tool, the PCS-STORET INTERFACE (re-
ferred to as IPS5 on the mainframe), can be used to
access and compute design flows for TMDL develop-
ment (see Figure 13) and to find all facilities discharging
to a particular reach (see Figure 14), an initial step in the
synchronization of permit issuance.
RPA
The RPA is a procedure on the EPA mainframe that
automates identification of reaches where priority pollut-
ants have been detected. It can be run for a user-
selected state or USGS hydrologic unit.
Section 304(1) of the Clean Water Act (CWA) identifies
water bodies impaired by the presence of toxic substances,
Step 3: Using EDDM To Evaluate Ambient Upstream Fecal Coliform LevelsRegulation
of Upstream Dischargers Is Necessary
ENVIRONMENTAL RISK ASSESSMENT-BRflNDYWlNE RIVER
ENVIRONMENTAL DATA DISPLAY MANAGER (EDDM)
Figure 10. Using EDDM to perform a water quality inventory for STORET station 21SC06WQ S-084 in the Saluda River basin, South
Carolina.
190
-------
Step 3: Using EDDM To Visualize Trends in Ambient Coliform Levels Upstream
of a Discharger
WATERSHED BASED APPROACH TO PERMITTING
SALUDA RIVER WATERSHED -ARCVIEW
HAPPING AND DATA DISPLAY MANAGER (MDDM)
storet_ulm-id
statioa
frequency
mean-long
mean-lat
mean-yalue
ENVIRONMENTAL DATA DISPLAY MANAGER (EDDM)
PCS -STORET INTERFACE
REACH POLLUTANT ASSESSMENT
TMDL DEVELOPMENT - CLARK FORK RIVER
ENVIRONMENTAL RISK ASSESSMENT-BRANDVW1NE SHIS
ENVIRONMENTAL DATA DISPLAY MANAGER (EODM)
STORET RETRIEVAL S-OB4
TIME SERIES - FECAL COLIFORM
PCS FACIL TV LEVELDATA SUMMARY
PCS PIPE LEVELDATA SUMMARY
PLOT OF LIMITS/DMR
Figure 11. From the EDDM water quality inventory table, the fecal coliform parameter (31616) was selected for a time series plot.
identifying point source dischargers of these substances
and developing individual control strategies for these
dischargers. To meet these requirements, the EPA Of-
fice of Wetlands, Oceans, and Watersheds prepared
guidance identifying criteria to be used in reviewing
state reports.
The RPA was designed to address the requirements
under criterion 7 of Section 304(1): identification of state
waters with likely presence of priority toxic pollutants.
This assessment was accomplished by identifying and
summarizing reaches with point source dischargers of
priority pollutants and water quality stations with priority
pollutant data.
Information on the state's waters is summarized using
the USGS hydrologic unit naming convention and the
Reach Structure File (Version 1). Numerous databases
were accessed and analyzed, including the Reach
Structure and Reach Trace File (Version 1), industrial
facilities discharge (IFD) file, STORET parameter file,
PCS, and the STORET water quality file. The IFD file
and PCS provided the facility information. Comparing
information from both data sources identified active fa-
cilities and generated a complete list of facilities by their
assigned reach numbers. Water quality data from
STORET were summarized on reaches with priority pol-
lutant monitoring data. Stations were retrieved with the
following restrictions:
Stations located within the state or hydrologic unit of
interest.
Ambient monitoring stations located on streams,
lakes, or estuaries.
Stations sampled for at least one priority pollutant in
either water, sediment, or fish tissue on or after January
1, 1982.
191
-------
Step 3: Using EDDM To Evaluate Existing ControlsPCS Limits and DMR
PFJ-ENO
PF4-DI1R
PF5-STDRET
PFb-flPERTURE
PF7-UMI TS
PF8-FLUU
ENVIRONMENTAL DATA DISPLAY MANAGER (EDDM
PCS FACILITY LEVEL DATA SUMMSRY
PC5_PIPEJlEWEL_DA_SUMMfl_RV _
PLOT OF LIMITS/DMR
Figure 12.
Using EDDM, PCS data is accessed for discharger SC0023906, Piedmont sewage treatment plant. The windows show
plant location (upper left), facility and pipe summary data (upper right) and time series plot (lower left).
Recently, the RPA program was modified to output files
compatible with CIS. An example of this is shown in
Figure 15. The data in the table portion of this figure are
written to two files as follows:
For each reach in the hydrologic unit, the geographic
coordinates are written to a file in ARC/INFO GEN-
ERATE format.
The attributes associated with each reach (e.g.,
name, length, number of water quality stations) are
written to a delimited ASCII file.
A third file is also automatically generated. This file is an
AMI that GENERATES the line coverage of reaches,
defines and populates the INFO table of attributes, then
joins the attributes in the INFO table to the line coverage
of reaches. Once these three files are created, they are
downloaded to the CIS workstation (via ftp) and proc-
essed by ARC/INFO. In ARC/VIEW, the user can identify
a reach and determine:
The number of water quality stations with priority pol-
lutant monitoring data.
The number and type of industrial facilities with pri-
ority pollutant discharge.
The number of publicly owned treatment works
(POTW) with and without indirect dischargers (see
Figure 15).
In addition to this reach summary data, other tables
(cross-linked to reaches, water quality stations, and dis-
chargers) are produced that summarize the data by pollut-
ant (see Figure 16). In this figure, each pollutant is cross-
linked to the reach where it was detected, and the source
of detection is also identified (i.e., water column, sediment,
fish tissue, NPDES permit limit, Form 2(c) submittal) or
192
-------
Step 4: Using the PCS/STORET INTERFACE for NPDES Permit DevelopmentAnalysis of
Receiving Stream Flow Data
Figure 13. Using the PCS-STORET INTERFACE to access and compute design flows for a specific reach.
Step 4: Using the PCS/STORET INTERFACE To Determine Other Facilities on a Reach for
Wasteload Allocation Purposes
Figure 14. Using the PCS-STORET INTERFACE to list all facilities that discharge to a specific reach.
193
-------
predicted to be in the discharger's effluent based on the
standard industrial classification (SIC) code. More de-
tailed information is also generated. For example, Fig-
ure 17 shows a detailed report for priority pollutants
detected in the water column (similar reports are gener-
ated for sediment and fish tissue). Each pollutant is
cross-linked to a reach and the specific monitoring sta-
tion where it was detected. Basic summary statistics are
also presented.
Figure 18 shows a detailed report for pollutants detected
in the NPDES permit limit. In this figure, each pollutant
is cross-linked to a reach and the specific NPDES dis-
charger containing a permit limit. In addition, each dis-
charger is identified as a major, minor, or POTW.
Figures 15 through 17 show the RPA output in the
foreground and coverages displayed by ARC/VIEW in
the background. Inspection of these figures shows that
the Bush River is a priority pollutant reach containing
seven industrial facilities, four of which discharge priority
pollutants (one pulp and paper mill, three textile facto-
ries). Further examination of the data shows that cad-
mium has been detected in the water column and is also
contained in the NPDES permit limit. It is also predicted
to be in the discharge effluent based on SIC code. In the
water column, cadmium was measured at 10 u,g/L at two
stations sampled in 1988.
Finally, there is a limit for cadmium in NPDES facility
SC0024490 (Newberry plant), a POTW on the Bush
River. The RPA output, linked to CIS, can be used as a
screening and targeting tool for identifying specific
reaches within watersheds where toxic priority pollut-
ants cause water quality degradation.
Summary and Conclusions
The proliferation of CIS workstations, the expansion of
the Internet, and the development of X-Window-based
graphics emulation software (X3270 and GDDMXD) has
afforded analysts the opportunity to use the powerful
analytical capabilities of CIS and the EPA mainframe
databases and tools together on one desktop. Thus, a
user performing a CIS watershed analysis can also
have immediate and complete access to national on-line
databases such as STORET and PCS by opening up a
"window" to the EPA mainframe. This allows detailed
queries to be performed that supplement the data al-
ready being analyzed at the local workstation. This ca-
pability allows users to easily visualize additional data
without having to spend effort in retrieval, downloading,
transforming, and reformatting to make it useful. By
enhancing existing mainframe programs to create out-
put in CIS format, the time spent importing data to the
CIS is reduced and more time can be spent on analysis.
An example of this capability is the RPA program.
Reach Pollutant Assessment: Priority Pollutant Detection in the Bush River
: ] Dlgch&rgan! :n |p| basin dQ '*';':,: tr ^j
"J:* : : , «-if?::*;: '-' :.
,
\ TOMC Ctercicai F
ii i;,*li»
D BASED APPROACH TO PERMITTING
£ 2. SsJfWSV OF SUSERiCAllif SC=RHiS RTACKCS WTh PRICS11V FCUUTSfjT S*3JfKES
ftlilf WflltH rjUflUT" S'lfillOHS IN CATALOGim; UM1 JUMICa
S^q E -'t-gtil
Cat. Unit No. Miles PEKCH Han
CO C2) (3) C-0
001 7.' EflLADfl R
W2 "1,1 SflLABA h
6fN &8.S L «'8KAV
nns 51.s L MURR^v
013 -4.-S UWIM OH
Ot^ B.S WlftOA R
015 1Q.B irtLWm R
SIS t?.5 tf* n
7 1 PLASTIC MOLBIHC &
» V?£A« LLECTSIC
1 (IHtWlC, (.HtFKftLi
Figure 15. Using the RPA procedure to identify specific reaches with priority pollutants.
194
-------
Reach Pollutant Assessment: Cadmium in the Bush River
Figure 16. RPA summary report by pollutant.
Reach Pollutant Assessment: Bush River: Cadmium in the Water Column
emtftool -/liin/csh
POLLUTANT PARSM RFWH RFftCH HAMF
MflME CODE NUMBES
RPA5 DETAIL REPORT CO S
POLLUWTS DETtCTED II4
STOrtFT S'ATTOH TD WO. OF
103? 3Q5S50M53 ftit&'i S ZfSCfiflWj S-fiS8
1G2? 3050109030 LITTLE H 21SCSQWQ S-C34
1027 3D501D90S2 COROHACft
102? SQSQIMC1^ 5HAPA fi
102? ^osoiiBcie sysiit B
ibroii 321DS 3D50109059 5kl.MA P
ifcros 32-irtS SegO'lGSPS,! S'SDV R
ft tot 1Q34 3D501D9C18 3JSH R
H tDt 1034 3050109053 riCOOV B
lO.UUU
1G-OM
10. DUD
10, Mfl
in.nno
10.COD
1G.C03
Figure 17. RPA detail report: pollutants detected in the water column.
195
-------
Reach Pollutant Assessment: Permitted Industries for CadmiumBush River
cmdtool -/iln/csh
R0AS DtTAIl REPORT [4) 8V POLLLTAHT
POLLUTANTS INCLUDED T> NPOE3 PERMIT LIMIT
POLLUTANT MME
PflRAH REACH
COPE &UM8£ft
Acenaphthylene
3 FIBERS/COLUMBIA PtftXT HAIOB 5D1
&COUOJSi? ftLiltl) UttRS/L'OUJMlil^ HflHT NftJQR 001
btD303191 HiUIKt!^ S CQ/SmE'J MJLI. NflJOP U01
SCOOfi3^7 £UI£D fie£RS/COiU»9STft PiftUT i^J*< Qyi
SCQOC3557 PLLIED FIEER5/COLUMBIA PLANT MftJtiR DQ1
SCOOD3557 fttLIEtt rZBTRS/COlUieXA PIWT WJ08 KM
SC0003557 fttllEB FIBEK/COlUnBIft PLANT MAJOR 301
SCOOD355? ALLIED FIBERS/COLUMBIA PLANT KUOR D01
5COD^43S NS'iWE&m'/PBOPGSED PifeNT HT^OR POTW 001
sroc2oro2 LAU^EHS MATOR POTW im
REfiCh HftP - PRIOBiTV POUUTffNTS J
Figure 18. RPA detail report: pollutants included in the NPDES permit limit.
Two examples were presented as illustrations of how
CIS and the mainframe databases and tools can work
together.
In the first example, EPA's watershed-based approach
to permitting, a four-step approach, was outlined, show-
ing how a combination of local CIS functions and remote
mainframe databases and tools were used in each step
of the process. The end result was the targeting and
prioritizing of watersheds of concern, and a detailed look
at where and why water quality standards were not
being met.
In the second example, the RPA program along with CIS
was used to identify and map toxic priority pollutants and
cross-link them to reaches, media (water column, sedi-
ment, fish tissue), NPDES dischargers, and monitoring
stations. This analysis focused on what the toxic pollut-
ant problems are and where they occur.
Acknowledgments
The authors wish to thank Jack Clifford, AWPD, OWOW,
EPA, for providing data and ARC/INFO AMLs that were
used in the South Carolina watershed example. The
authors also extend their appreciation to Duane More
and Tom Lewis (both with Martin Marietta), who provided
support in setting up the X3270 and GDDMXD software.
References
1. Taylor, P.L., P. Evenhouse, L. Hoelman, T. DeWald, W.B. Samuels,
and O. Hansen. 1988. STORETWater quality analysis system.
Presented at the State/EPA Data Management Conference,
Raleigh, NC (June).
2. U.S. EPA. 1992. Office of Water environmental and program
information systems compendium, information resources manage-
ment: Tools for making water program decisions. EPA/800/B-
92/001. April.
3. Samuels, W.B., PL. Taylor, P.B. Evenhouse, T.R. Bondelid, PC.
Eggers, and S.A. Hanson. 1991. The environmental display man-
ager: A tool for water quality data integration. Water Resources
Bull. 27(6):939-956.
4. U.S. EPA. 1993. Geographic Resources Information and Data Sys-
tem (GRIDS): User guide. Office of Information Resources Man-
agement, National CIS Program.
5. Samuels, W.B. 1990. Reach pollutant assessment user's guide.
McLean, VA: Science Applications International Corporation.
6. Washington State Department of Ecology (WDEC). 1993. Water-
shed approach to water quality management. F-WQ-93-029.
October.
7. State of South Carolina. 1992. South Carolina code of regulations,
Chapter 61, Regulation 68 and 69. Amended April 24, 1992.
196
-------
Wetlands Applications
-------
Wetlands Mapping and Assessment in Coastal North Carolina:
A CIS-Based Approach
Lori Sutter and James Wuenscher
Division of Coastal Management, State of North Carolina Department of Environment, Health,
and Natural Resources, Raleigh, North Carolina
Introduction
The coastal area of North Carolina covers 20 counties
and over 9,000 square miles of land area, about 20
percent of the state (see Figure 1). It also includes over
87 percent of the state's surface water. The North Caro-
lina Coastal Management Program (NC CMP) is re-
sponsible for managing this area to meet the goals set
forth in the Coastal Area Management Act (CAMA)
(North Carolina General Statute [NCGS] 113A, Article
7). These goals provide a broad mandate to protect the
overall environmental quality of the coastal area and to
guide growth and development in a manner "consistent
with the capability of the land and water for develop-
ment, use, or preservation based on ecological consid-
erations" (NCGS 113A-102(b)(2)).
Figure 1. HU and county boundaries in the North Carolina
coastal area.
Much of the North Carolina coastal area consists of
wetlands, which, in many areas, constitute nearly 50
percent of the landscape. These wetlands are of great
ecological importance, in part because they occupy so
much of the area and are significant components of
virtually all coastal ecosystems, and in part because of
their relationships to coastal water quality, estuarine
productivity, wildlife habitat, and the overall character of
the coastal area.
Historically, close to 50 percent of the original wetlands
of the coastal area have been drained and converted to
other land uses (1-3). Although agricultural conversion,
the largest historical contributor to wetlands loss, has
largely stopped, wetlands continue to disappear as they
are drained or filled for development. Conflicts between
economic development and wetlands protection con-
tinue to be a major concern, with many coastal commu-
nities considering wetlands protection to be a major
barrier to needed economic development.
Because wetlands are such a dominant part of the
coastal landscape and are vitally important to many
aspects of the area's ecology, their management and
protection is a major concern of the NC CMP. The State
Dredge and Fill Act (NCGS 113-229) and the CAMA
regulatory program stringently protect tidal wetlands, or
"coastal wetlands" as law and administrative rules call
them. Coastal wetlands are designated areas of environ-
mental concern (AECs), with the management objective
"to give highest priority to the protection and manage-
ment of coastal wetlands so as to safeguard and per-
petuate their biological, social, economic and aesthetic
values; and to coordinate and establish a management
system capable of conserving and utilizing coastal wet-
lands as a natural resource essential to the functioning
of the entire estuarine system" (15A NCAC 7H .0205).
North Carolina law does not, however, specifically pro-
tect nontidal freshwater wetlands. State protection of
199
-------
freshwater wetlands is limited to the regulatory authority
provided under federal laws for state agency review of
federal permits; in this case, §404 permits granted by
the U.S. Army Corps of Engineers. Under §401 of the
Federal Water Pollution Control Act (33 USC 1341), a
Water Quality Certification from the North Carolina Divi-
sion of Environmental Management (DEM) is required
for a 404 permit to discharge fill material into wetlands.
Section 307 of the federal Coastal Zone Management
Act (CZMA - 16 USC 1451 et seg.) also requires that
404 permits be consistent with the enforceable rules and
policies of the NC CMP. The standards for consistency
are the use standards for AECs and wetlands policies
stated in the applicable local land use plan. Other than
AECs, the NC CMP has no consistent policies regarding
wetlands. A few local land use plans include policies to
protect freshwater wetlands, but most do not.
Wetlands Conservation Plan
In 1991, the CZMA §309 Assessment of the NC CMP
revealed NC CMP's weakness in protecting nontidal
wetlands (4). The assessment demonstrated that both
opponents and proponents of wetlands protection con-
sidered the current system inadequate. Economic de-
velopment interests found the 404 regulatory program
to be unpredictable and inconsistent, often resulting in
the loss of needed economic growth in coastal counties.
Environmental interests felt that the program allowed
the continued loss of ecologically important wetlands. As
a result, the assessment identified wetlands manage-
ment and protection as one of the primary program
areas in need of enhancement.
The North Carolina Division of Coastal Management
(DCM) developed a 5-year strategy (5) for improving
wetlands protection and management in the coastal
area using funds provided under the Coastal Zone En-
hancement Grants Program established by 1990
amendments to §309 of the federal CZMA. The Office
of Ocean and Coastal Resources Management (OCRM)
in the National Oceanographic and Atmospheric Admini-
stration (NOAA), U.S. Department of Commerce admin-
isters the §309 program. Funds provided under this
program, particularly Project of Special Merit awards for
fiscal years 1992 and 1993, supported the work reported
in this paper. A grant from the U.S. Environmental Pro-
tection Agency (EPA) for a Wetlands Advance Identifica-
tion (ADID) project in Carteret County, North Carolina,
also funded this work.
The key element of DCM's strategy for improving wet-
lands protection is the development of a wetlands con-
servation plan for the North Carolina coastal area. The
plan has several components:
Wetlands mapping inventory
Functional assessment of wetlands
Wetlands restoration
Coordination with wetlands regulatory agencies
Coastal area wetlands policies
Local land use planning
The obvious first step in developing a wetlands conser-
vation plan is to describe the wetlands resource. An
extensive geographic information system (CIS) wet-
lands mapping program is helping to accomplish this
first step by producing a CIS coverage of wetlands by
wetland type for the entire coastal area. The CIS cover-
age allows generation of paper maps for areas within
any boundaries available in CIS format. This is the
subject of the first part of this report.
One weakness of the 404 program is that, for individual
permits, it attempts to apply the same rules and proce-
dures equally to all wetlands, regardless of the wetland
type and location in the landscape. This approach can
result in permits being granted for fill of wetlands of high
ecological significance or permits being denied to pro-
tect wetlands of little significance. Neither outcome is
desirable because the result may be the loss of either
vital wetland functions or beneficial economic activity.
This is an unsatisfactory way to manage wetland re-
sources in an area such as the North Carolina coast,
where:
A high proportion of the land is wetlands.
Many of the wetlands are vital to the area's environ-
mental quality.
Economic stimulation is sorely needed.
To help overcome this weakness in the current wetland
regulatory framework, the Wetlands Conservation Plan
includes an assessment of the ecological significance of
all wetlands to determine which are the most important
in maintaining the environmental integrity of the area.
This will result in a designation of each wetland polygon
in the CIS coverage as being of high, medium, or low
functional significance in the watershed in which it ex-
ists. The procedure by which this occurs is the subject
of the second part of this report.
The remaining components of the Wetlands Conserva-
tion Plan comprise the means by which the results of the
wetland mapping and functional assessment steps will
be used to improve wetland protection and management.
Close coordination with other state and federal agencies
involved in wetlands protection and management has
been an important component of the entire effort.
Agency representatives have been involved in develop-
ing the methods used, and the agencies will receive
copies of the resulting maps for use in their own plan-
ning and decision-making. Policies for protection of wet-
lands of varying functional significance will be proposed
to the Coastal Resources Commission to serve as the
200
-------
basis for consistency review of 404 permit applications.
Wetland maps and functional assessment results will
also be provided to local governments for use in local
land use planning, and DCM will work with local govern-
ments to increase local involvement in the wetlands
regulatory structure.
While the wetland maps themselves are useful for land
use planning and helping to find suitable development
sites, simply knowing where the wetlands are located is
insufficient information for many purposes. Any area for
which a 404 permit application is in process has been
officially delineated as a wetland by the Corps of Engi-
neers. The value of wetland maps to the regulatory
review agencies at this stage is limited to determining
the relationship of the site to other wetlands in the area.
While, ideally, all wetlands should be avoided in plan-
ning development, avoiding wetlands completely in the
coastal area is difficult, and avoiding all wetlands in any
extensive development is virtually impossible.
The results of the functional assessment will provide
additional information about the ecological significance
of wetlands. This information will be valuable to wetland
regulatory review agencies in determining the impor-
tance to an area's environmental integrity of protecting
a particular site for which a permit to fill has been
requested. It will also enable development projects to be
planned so as to avoid, at all reasonable costs, the most
ecologically important wetlands. An accurate functional
assessment of wetland significance, then, is the most
valuable component of the Wetlands Conservation Plan.
Wetlands Mapping Inventory
An important, initial step in developing a comprehensive
plan for wetlands protection is to understand the extent
and location of wetlands in the coastal area. When
developing mapping methods, DCM quickly realized
that the more than 9,000-square-mile coastal area was
too large for any mapping effort in the field (see Figure
1). To complete this task in an accelerated timeframe,
DCM needed to use existing data compatible with CIS.
Reviewing the existing data revealed that most are not
applicable for one of two reasons: (1) available wetlands
data are based on older photography, and (2) more
recent data are not classified with the intent of wetlands
mapping. These data types, used independently, are
inappropriate for use in a coastal area wetlands conser-
vation plan. In addition, the classification schemes used
in the existing methods are too complex or not focused
on wetlands.
While several data sets were believed to be inappropri-
ate if used exclusively for wetlands mapping in coastal
North Carolina, each contained useful components.
DCM elected to combine three primary layers of data
and extract the most pertinent information from each
layer. DCM selected the National Wetlands Inventory
(NWI) because its primary purpose is to map wetlands.
Unfortunately, these maps were based on photography
from the early 1980s in coastal North Carolina, and
many changes have occurred in the landscape since
that time. NWI also omitted some managed wet pine
areas from its maps; DCM wished to include these areas
because they are important to the ecology of the North
Carolina coastal area. DCM also selected detailed soils
lines for use in its mapping efforts. While soils alone
should not be used to identify wetlands, soils can be
very useful in identifying marginal areas. Finally, DCM
also employed thematic mapper (TM) satellite imagery
in its methods. This data layer was not developed as a
wetlands inventory; however, the imagery is more recent
than the soils and NWIs. DCM desired to incorporate the
benefits of each of these data sources into its mapping
techniques.
The information provided by this mapping exercise will
be useful to county and municipal planners in helping
guide growth away from environmentally sensitive ar-
eas. For this reason, DCM elected to pursue mapping
on a county by county basis. In addition, a single county
allowed DCM to focus methodology development to a
limited geographic area to refine its methods. Carteret
County was selected as a methods development labo-
ratory because data were available for the area and
because Carteret has a large number of representative
wetlands.
Data Descriptions
The U.S. Fish & Wildlife Service produces the NWI for
all wetlands in the country. For the coastal North Caro-
lina area, these vector data were developed from
1:58,000-scale color infrared photography taken during
the winters of 1981, 1982, and 1983. Photointerpreters
delineated wetland polygons on clear stabilene mylar
taped over the photographs. After an initial scan of the
photographs to identify questions or problem signatures,
the photointerpreters reviewed areas in the field. They
performed approximately one-half to one full day of field
verification per quadrangle (quad) (6). Features were
compared with U.S. Geological Survey (USGS) topo-
graphic maps for consistency. Following completion of
the 'draft' paper maps, the Regional Coordinator re-
viewed the data. After approval as a final map, each
quad was digitized. Initially, the North Carolina Center
for Geographic Information and Analysis (CGIA) digit-
ized the coastal North Carolina NWI maps, and later, the
NWI Headquarters in St. Petersburg, Florida, who sub-
contracted the task, digitized them. Digital maps were
obtained initially from 1/4-inch tape transfer and later
from direct access to NWI via the Internet.
CGIA provided digital, detailed soil lines, which also are
vector data based on 1:24,000 quads. County soil sci-
entists delineated soil boundaries on aerial photographs
201
-------
based on slope, topography, vegetative cover, and other
characteristics. This process occurs in any soil survey.
After appropriate personnel approved the lines, a quali-
fied soil scientist recompiled them onto orthophoto
quads. CGIA scanned or manually digitized these lines.
The coverage incorporated databases describing soil
characteristics, which were then released for use.
The Landsat Thematic Mapper (TM) imagery was clas-
sified as part of the Albemarle-Pamlico Estuarine Study
(APES). To provide complete coverage for the southern-
most region of DCM's jurisdiction (Onslow, Render,
Brunswick, and New Hanover Counties), DCM con-
tracted with CGIA and the North Carolina State Univer-
sity (NCSU) Computer Graphics Center to have that
area processed identically to the APES region. These
data provide a raster-based coverage of approximately
30-meter pixel resolution. Some of the imagery was
taken at high tide, which precludes some near-water
wetlands from appearing in certain areas. Using
ERDAS, imagery specialists grouped similar spec-
tral signatures into one of 20 classes. DCM used these
data in two formats: filtered and unfiltered. The unfiltered
information was vectorized with the ARC/INFO GRID-
POLY command. To remove some of the background
noise in the coverage, it was filtered using ERDAS 'scan'
with a Majority filter of 5 by 5 pixels, then vectorized with
the ARC/INFO GRIDPOLY command.
Methods
Within each county, mapping is based on 1:24,000
USGS quads. After completion, each quad is assembled
into a countywide coverage, which eventually is assem-
bled into a coastal area coverage. The initial step in the
mapping process is to ensure completion of the base
layers described previously. Reviewing for errors at
early stages prevents confusion in correction later in the
process; therefore, the importance of the preliminary
techniques cannot be overemphasized. The NWI data
are first inspected to ensure complete coverage. If parts
of the quad are missing, the error is investigated and
corrected. Omissions may be areas of severe cloud
cover on the photography or areas neglected during the
digitization process. Next, the coverage is reviewed for
missing label points. Any omissions are corrected based
on the finalized version of the published NWI paper map.
Appropriate NWI staff are contacted for the necessary
information. At this time, labels are verified for typo-
graphical misentry. If not corrected, these errors could
lead to confusion later in the mapping process.
Once the label errors are detected and corrected, the
polygons are reviewed for completion. Verifying every
line in the areas of coastal North Carolina densely
populated with wetlands is impossible, but the lines are
reviewed for completeness. NWI staff again must pro-
vide necessary information for any omissions. When the
map is approved, technicians ensure projection of the
quad to the State Plane Coordinate System. If this has
not been completed, the ARC/INFO PROJECT com-
mand is employed.
The soils information is prepared in a similar manner to
the NWIs, with questions being directed to the county
soil scientist. Prior to the steps described previously,
soils must be verified for completeness. Because soils
are mapped by county boundaries and DCM maps by
quad, some files must be joined in quads that intersect
county boundaries. At this time, the quad must be
checked for differing abbreviations between counties.
Discrepancies are handled on a case-by-case basis.
When an abbreviation describes different soils in differ-
ent counties, a temporary abbreviation is created for one
of the counties. If a single soil is described by two
abbreviations across counties, both abbreviations are
incorporated into the classification scheme.
The Landsat data do not require additional verification.
Review of this layer is often helpful, however, to ensure
that the geographic boundaries match. Cases where
landforms do not appear to match require investigation
of the discrepancies. If the area is mis registered, this
layer might be omitted from the analyses. To date, no
area has been mapped without this imagery.
The hydrogeomorphology of a wetland is unique in de-
fining the wetland's function (7). Because these maps
serve as the base for additional wetland projects (as
described later in this report), an accurate determination
of this characteristic is essential. Prior to the overlay
procedure, technicians add a new item, hydrogeomor-
phic (HGM), to the NWI coverage. Because DCM con-
siders both vegetation and landscape position in its
classification (discussed later), riverine, headwater, and
depressional wetland polygons are assigned an HGM of
'r,' 'h,' or 'd,' respectively. The digital line graphs (DLGs)
of hydrography are essential in this step of the procedure.
All wetlands that are adjacent to streams or rivers are
considered in the riverine HGM class and are desig-
nated as riverine polygons. This class should include all
bottomland hardwood swamps and some swamp for-
ests. It rarely includes any of the interfluvial wetland
types. If it does, it is a small section of a large interfluvial
flatwood from which a small stream emerges. Only the
polygons adjacent to the stream are considered riverine.
Headwaters are defined as linear areas adjacent to
riverine areas that do not have a stream designated on
the hydrography data layer. Because these unique sys-
tems form the transition between flatwoods and riverine
wetlands, they are treated specially. Finally, polygons
that exist on interfluvial divides are designated as flats
or depressional wetlands. This class should not include
any wetlands along streams.
202
-------
The complete data coverages are overlaid to create a
new, integrated coverage that often approaches 100,000
polygons. Each polygon has many characteristics assigned
to it, including the Cowardin classification assigned by
the NWI, the soil series provided by the detailed soil
lines, the unfiltered land use/land cover code, the filtered
land use/land cover code from the Landsat TM imagery,
and the HGM classification assigned in the previous
step.
Based on these characteristics, each polygon is as-
signed to one of DCM's classes through an automated
ARC/INFO model using an arc macro language (AMI).
Personnel from the NWI and the North Carolina Depart-
ment of Environment, Health, and Natural Resources
Division of Soil and Water Resources have reviewed the
classification of the Cowardin types into DCM wetland
types. The classes that DCM currently recognizes are
upland, salt/brackish marsh, estuarine shrub scrub, es-
tuarine forest, maritime forest, pocosin, bottomland
hardwood, swamp forest, headwater swamp, hardwood
flatwoods, piney flatwoods, and managed pinelands.
DCM also classifies soils as hydric or nonhydric based
on List A of the U.S. Soil Conservation Service (SCS)
List of Hydric Soils.
The base of the map is the NWI polygon coverage.
Some NWI polygons are omitted from the DCM maps
because they are temporarily flooded, but on nonhydric
soils or because recent TM imagery indicates these
areas are currently bare ground. The managed pineland
wetland group on DCM maps includes areas that NWI
considers uplands, identified as pine monocultures on
the imagery, and that occur on hydric soil.
In addition, DCM also provides a modifier to some of
these polygons. DCM notes if NWI has determined that
the area has been drained or ditched. Areas designated
as wetlands at the time of the NWI photography that
currently appear as bare ground on the TM imagery are
designated as 'cleared' on the maps. Many of the
cleared areas would no longer be considered jurisdic-
tional wetlands. These modifiers are useful indicators of
the impacts wetlands sustain from human activities.
Initiation of an interactive session follows completion of
the automated procedure. This session considers land-
scape characteristics that are not easily described to
a computer model in correcting the classification. This
is especially important in distinguishing bottomland
hardwood swamps from hardwood flats. Both contain
deciduous, broad leaf species of trees and can be tem-
porarily flooded. The hydrology of these systems, how-
ever, is completely different. All bottomland hardwood
swamps, for example, must be adjacent to a river where
they receive seasonal floodwaters from the channel.
Conversely, hardwood flatwoods should be located on
interfluvial flats and not adjacent to any streams. Water
is not introduced into hardwood flatwoods via a channel;
rather, precipitation and ground water provide the water
for this system. Polygons that are adjacent to rivers or
estuaries but do not have a distinct channel designated
in the hydrography coverage are considered headwater
swamps.
During the course of methodology development, staff
members visited at least 371 sites in the field. As staff
members encountered new Cowardin classes, they
would verify that the polygons were being placed into
the correct DCM categories. If they determined that a
particular Cowardin class was systematically misidenti-
fied, they updated the algorithm for automation. While
this method does not provide for a usable accuracy
assessment, it allowed development of the most accu-
rate methods.
The accuracy of these data is unknown at this time. An
accuracy assessment of the data is anticipated in the
near future. This assessment will allow map users to
understand the strengths and limitations of the data. It
also will provide an overall summary of data error.
Functional Assessment of Wetlands
Certain initial considerations shaped the approach and
methods used in developing a wetlands functional as-
sessment procedure. The procedure needed to fit within
the context and objectives of the Wetlands Conservation
Plan for the North Carolina coastal area as described
above. This context, and the opportunities and limita-
tions it imposed, had considerable influence on the spe-
cific procedure developed.
Because we are dealing with a large geographic area
with many wetlands, we recognized from the outset that
we needed a method we could apply to large land areas
without site visits to each individual wetland. This ruled
out the many site-specific functional assessment meth-
ods that were applied in other contexts. Almost of ne-
cessity, a CIS-based approach was chosen. That meant
we would have to use information available in CIS for-
mat and make use of CIS analytical techniques. The
wetland mapping on which the functional assessment is
based was performed using CIS, so the basic digital
data were available.
The primary objective was to produce information about
the relative ecological importance of wetlands that would
be useful for planning and overall management of wet-
lands rather than to serve as the basis for regulatory
decisions. While we could not visit every wetland, the
goal was to predict the functional assessment value that
a detailed, site-specific method would determine. We
wanted to be able to predict in advance what the wetland
regulatory agencies would determine as a wetland's
significance so that the resulting maps would identify
those wetlands where a 404 permit would be difficult or
impossible to obtain. The resulting information would
203
-------
then be useful in determining where not to plan devel-
opment. This would benefit potential permit applicants
by preventing ill-advised plans that would be unlikely to
receive permits and simultaneously serve to protect the
most ecologically important wetlands. The result of the
procedure, then, is not a substitute for a site visit in
making regulatory decisions, but a predictor of what a
site visit would determine.
A primary consideration was that the procedure be
ecologically sound and scientifically valid, based on the
best information available about the functions of wet-
lands. It needed to be based on fundamental principles
of wetlands and landscape ecology rather than on arbi-
trary or subjective decisions.
Finally, the procedure was to be watershed-based. This
requirement was primarily because consideration of a
wetland's role in its watershed is the soundest basis for
determining its ecological significance, but also because
the other components of the Wetlands Conservation
Plan, including wetland mapping and restoration plan-
ning, are based on watershed units. The watersheds
being used are 5,000- to 50,000-acre hydrologic units
(HUs) delineated by the SCS as illustrated in Figure 1.
The North Carolina coastal area comprises 348 of these
HUs. Watershed units of any size, however, could be
used without changing the validity of the watershed-
based considerations used in the procedure.
These initial considerations result in a summary defini-
tion of the functional assessment procedure. It is a CIS-
based, landscape scale procedure for predicting the
relative ecological significance of wetlands throughout a
region using fundamental ecological principles to deter-
mine the functions of wetlands within their watersheds.
The functional assessment procedure is meant to be
used with CIS data for regional application. It is not a
field-oriented, site-specific method that involves visiting
individual wetlands and recording information. A CIS-
based procedure is the only practical approach for deal-
ing with a large geographic area with many wetlands in
a limited amount of time.
This CIS-based approach can make information on wet-
land functional significance available for broad regions
in advance of specific development plans. The informa-
tion is then available for planning to help avoid impacts
to the most ecologically important wetlands. In this
sense, the North Carolina procedure is unlike other
functional assessment techniques that are designed for
use in a regulatory context or that require field data for
each wetland.
Data Requirements
Because the procedure uses CIS analysis, it requires
digital information in CIS format. CIS data layers used
in the procedure include:
Wetland boundaries and types (the topic of the first
section of this report).
Soils maps.
Land use/land cover.
Hydrography.
Watershed boundaries.
Threatened and endangered species occurrences.
Estuarine primary nursery areas.
Water quality classifications.
In the North Carolina coastal area, these data layers
either already existed and were available from the CGIA
or were developed as part of the Wetlands Conservation
Plan. Because other projects funded most of the data
acquisition and digitization, developing the necessary
CIS databases was not a major cost.
The soils coverage consists of digitized, detailed county
soils maps produced by SCS and digitized by CGIA. The
soils coverage allows identification of the soil series
underlying a wetland, and the properties of the series
are used to determine soil capacity for facilitating the
wetland's performance of various functions.
The land use/land cover data layer was produced for the
APES from interpretation of satellite TM imagery (8). It
is used to determine land cover and uses surrounding
each wetland and in the watershed.
The basic hydrography coverage consists of 1:24,000-
scale USGS DLGs. Because the functional assessment
procedure uses stream order as an indicator of water-
shed position, stream order according to the Strahler
system was determined manually and added to the DIG
attribute files.
As described previously, the watersheds used in the
procedure are relatively small HUs delineated by SCS.
DCM contracted with CGIA to have these boundaries
digitized for the coastal area. During the digitization
process, the watershed boundaries were rectified to
USGS and DEM boundaries of larger subbasins to en-
sure that the HUs could be combined into larger water-
shed units.
A data layer produced by the North Carolina Natural
Heritage Program is used to identify threatened and
endangered species occurrences. The North Carolina
Division of Marine Fisheries maintains the coverage of
primary nursery areas, and the Division of Environ-
mental Management developed a map of water quality
classifications that was digitized by CGIA.
The ways in which these data layers are used to determine
values for various parameters in the functional assess-
ment procedure are described later in this report. The
CIS procedures have been automated using ARC/ INFO
204
-------
AMI on a Sun workstation. The AMI programs are
available from DCM to anyone planning to use the pro-
cedure elsewhere.
Because the assessment procedure was designed for
CIS analysis, the choice and expression of individual
parameters have been shaped to some extent by the
CIS data available and the capabilities and limitations of
ARC/INFO techniques and AMI automation. DCM was
fortunate to have a relatively large amount of CIS data
readily available. For use in other areas, the procedure
could be modified to use different CIS coverages. At
least the first five databases listed above, however, are
essential to its basic propositions.
Classification Considerations
The HGM classification system for wetlands (7) classi-
fies wetlands into categories based on landscape position
(geomorphic setting), water sources, and hydrodynam-
ics (direction of water flow and strength of water move-
ment). It is being increasingly used as the basis for
wetland classification and functional assessment sys-
tems. HGM classification focuses on the abiotic features
of wetlands rather than on the species composition of
wetland vegetation as do most traditional wetland clas-
sification schemes.
Several features of the HGM classification system make
it a useful starting point for an assessment of wetland
functions. Because the HGM system is based on geo-
morphic, physical, and chemical properties of wetlands,
it aggregates wetlands with similar functions into
classes. The HGM class of a wetland, in itself, indicates
much about the ecosystem functions of the wetland. The
HGM approach also forces consideration of factors ex-
ternal to the wetland site, such as water source. This
helps relate the wetland to the larger landscape of which
it is a part and puts consideration of the wetland's func-
tions in a landscape and watershed context.
Three HGM classes are used as the starting point for
the North Carolina functional assessment procedure. All
wetlands are first classified as one of the following:
Riverine
Headwater
Depressional
Riverine wetlands are those in which hydrology is deter-
mined or heavily influenced by proximity to a perennial
stream of any size or order. Overbank flow from the
stream exerts considerable influence on their hydrology.
Headwater wetlands exist in the uppermost reaches of
local watersheds upstream of perennial streams. Head-
water systems may contain channels with intermittent
flow, but the sources of water entering them are precipi-
tation, overland runoff, and ground-water discharge
rather than overbank flow from a stream. Depressional
wetlands, including wet flats and pocosins, generally are
not in direct proximity to surface water. While they may
be either isolated from or hydrologically connected to
surface water, the hydrology of depressional wetlands is
determined by ground-water discharge, overland runoff,
and precipitation.
The functions of wetlands in these different HGM
classes differ significantly. Riverine wetlands regularly
receive overbank flow from flooding streams and, thus,
perform the functions of removing sediment and pollut-
ants that may be present in the stream water and pro-
viding temporary floodwater storage. Headwater and
depressional wetlands cannot perform these functions
because they do not receive overbank flow. Headwater
wetlands occur at landscape interfaces where ground
water and surface runoff coalesce to form streams.
Headwater wetlands provide a buffer between uplands
and stream flow so they can perform significant water
quality and hydrology functions. While depressional wet-
lands do not perform buffer functions, they often store
large amounts of precipitation or surface runoff waters
that otherwise would more rapidly enter streams. Wet-
lands in all HGM classes can perform important habitat
functions.
Because the wetlands in these different HGM classes
are functionally different, their functional significance is
assessed using different, though similar, procedures. If
the same procedure were used for all HGM classes,
depressional wetlands would always be considered of
lower functional significance simply because they are
not in a landscape position to perform some of the water
quality and hydrologic functions of riverine and headwa-
ter wetlands.
In addition to HGM classes, wetland types identified by
dominant vegetation are used at several points in the
functional assessment. This reflects a recognition that
the biologic properties of a wetland site considered to-
gether with its hydrogeomorphic properties can provide
a more detailed indication of its functions than either
taken alone. The HGM class of a wetland, as a broad
functional indicator, determines which assessment pro-
cedure to use. Within each HGM class and correspond-
ing assessment procedure, wetland type determines the
level or extent of specific parameters.
The wetland types used are those typical of the North
Carolina coastal area. They result from a clumping of
the Cowardin classes used on NWI maps into fewer
types with more intuitively obvious type names (e.g.,
swamp forest, pocosin), as described previously. These
wetland types are used in the wetland maps that form
the starting point for the functional assessment.
Wetland types are used in the procedure as indicators
of functional characteristics. Correlations between wet-
land type and wetland functions were determined from
205
-------
statistical analysis of field data from nearly 400 sites. At
each site, the presence or absence of a list of functional
indicators was recorded. Dr. Mark Brinson of East Caro-
lina University developed the functional indicators lists,
in part. Dr. Brinson served as primary scientific consult-
ant in developing the HGM classification system and the
field sampling methodology.
Wetland types differ in other areas, so their inclusion in
this procedure limits its use in its current form to the
southeastern coastal plain. Adaptation of the procedure
for use in other areas would require either extensive field
sampling as was performed in coastal North Carolina or
a more arbitrary clumping of wetland types based on
best professional judgment. Other methods of wetland
classification could be used, provided wetlands are clas-
sified in such a way that functional characteristics of the
wetland types are constant and can be determined by
field sampling, literature values, and/or professional
judgment. The procedure could be applied directly to
NWI polygons if these are the only wetland map base
available.
In addition to wetland type, several other parameters are
used as indicators of the existence or level of specific
wetland functions. These include both site-specific pa-
rameters, such as wetland size and soil characteristics,
and landscape considerations, such as watershed posi-
tion, water sources, land uses, and landscape patterns.
CIS analysis determines values for these parameters
based on the data layers discussed above. They could
be determined manually, but the process would be very
labor intensive.
Unlike assessment procedures that depend solely on
information that can be collected within a wetland, this
procedure relies heavily on factors external to the wet-
land site itself. Relationships between a wetland and the
landscape within which it exists are integral considera-
tions in determining wetland functional significance.
Characteristics of the landscape surrounding a wetland
are often more important determinants of its functional
significance than are the characteristics of the wetland
itself. Of the 39 parameters evaluated in the procedure,
21 are landscape characteristics, and 18 are internal
characteristics of the wetland itself.
While we believe this emphasis on a wetland's land-
scape context is a more ecologically sound approach to
functional assessment than site-specific methods, it re-
quires a great deal more information than could be
collected within the wetland itself. The procedure is
based on CIS data and analysis, not only to make it
suitable for regional application, but because CIS pro-
vides the most practical way to analyze the spatial rela-
tionships of landscape elements and their properties.
Structure of the Assessment Procedure
The assessment procedure uses a hierarchical structure
that rates individual parameters and successively com-
bines them to determine the wetland's overall functional
significance. The complete hierarchical structure is illus-
trated in Figure 2. It consists of four levels:
Overall functional significance of the wetland.
Specific functions and risk of wetland loss.
Subfunctions.
Parameters evaluated to determine the level and ex-
tent of functions.
The objective of functional assessment is to determine
an individual wetland's ecological significance in its wa-
tershed and the larger landscape. The highest hierarchi-
cal level, or end result of applying the procedure, then,
is the wetland's overall functional significance.
The second hierarchical level includes the four primary
factors that are considered in determining the wetland's
functional significance (see Figure 3). The overall eco-
logical significance of a wetland is determined by the
degree to which it performs, or has the capacity to
perform, specific functions. The broadest grouping of
wetland functions includes water quality functions, hy-
drologic functions, and habitat functions. The nature of
the landscape and the water characteristics of the wa-
tershed in which a wetland functions also determine
ecological significance to some extent. These factors
determine the potential risk to watershed and landscape
integrity if the wetland functions were lost. Including a
"risk factor" as a basic consideration in functional as-
sessment also provides a means of considering cumu-
lative impacts and the practicality of replacing lost
functions through mitigation in determining a wetland's
overall significance.
Each primary function of wetlands is actually a combi-
nation of separate, more specific subfunctions. Water
quality subfunctions include the removal of nonpoint
source pollutants from surface runoff and the removal of
suspended or dissolved pollutants from flooding
streams. Hydrology subfunctions include storage of pre-
cipitation and surface runoff, storage of floodwater from
streams, and shoreline stabilization. Habitat subfunc-
tions include providing habitat for both terrestrial species
and aquatic life. Several considerations that, while not
truly wetland functions, are called subfunctions for par-
allelism also determine risk factor. The subfunction lev-
els of the assessment procedure are illustrated in
Figures 4 through 7.
Properties of the wetland and its surrounding landscape
determine the extent to which a wetland performs these
different subfunctions. The assessment procedure re-
fers to these properties as "parameters." Parameters
206
-------
Overall Significance
Functions/Risk III I
Subfunctions | I | I it
III IIIIIII Illll | I I III I "l Illll
Figure 2. Overall hierarchical structure of the functional assessment procedure.
Wetland's Overall
Ecological
Significance
Hydrology
Functions
Habitat
Functions
Risk
Factor
Figure 3. Assessment level two: Primary wetland functions and risk factor.
Water Quality
Functions
Nonpoint Source
Function
Floodwater
Cleansing Function
Figure 4. Water quality subfunctions.
make up the levels in the hierarchical structure that are
actually evaluated based on fundamental ecological
considerations. Parameter values, in turn, are combined
to produce ratings for the subfunctions. Future reports
will explain in detail all parameters evaluated in the
assessment procedure and document them for scientific
validity. This paper discusses only the parameters under
the nonpoint source removal subfunction of the water
quality function for illustration (see Figure 8).
The first parameter determining a wetland's significance
in removing nonpoint source pollutants from surface
runoff water is whether the water contains sediment,
nutrients, or toxic pollutants in significant quantities. This
is evaluated in the "proximity to sources" parameter
based on the land uses surrounding the wetland. If
agricultural fields or developed areas from which pollut-
ants are likely to enter surface runoff largely surround
the wetland, the wetland's potential for removing non-
point source pollutants is high. If, on the other hand,
natural vegetation from which runoff water is likely to be
largely unpolluted mostly surrounds the wetland, its po-
tential for removing significant pollutants is low.
Proximity to sources is an "opportunity" parameter. That
is, it determines whether a wetland has the opportunity
to remove pollutants from surface runoff by considering
how likely the runoff water is to be polluted. The other
parameters for this subfunction are "capacity" parame-
ters that measure the wetland's ability to perform the
function if the opportunity is present. Opportunity and
capacity parameters are treated differently in determining
a wetland's overall significance to prevent a wetland from
being rated lower simply because present opportunity
does not exist. This is discussed in more detail below.
The second parameter considered in determining a wet-
land's significance in nonpoint source removal is its
proximity to a surface water body. If runoff entering a
wetland would otherwise directly enter surface water,
the wetland's significance as a filter is greaterthan if the
wetland is far removed from surface water. In that case,
pollutants in runoff could either settle out or be removed
by other means before they enter surface water as
pollutants.
207
-------
Hydrology
Functions
Surface Runoff
Storage
Floodwater
Storage
Shoreline
Stabilization
Figure 5. Hydrology subfunctions.
Habitat
Functions
Terrestrial
Wildlife
Aquatic
Life
Figure 6. Habitat subfunctions.
Risk
Factor
Landscape
Character
Water
Characteristics
Replacement
Difficulty
Restoration
Potential
Figure 7. Risk factor subfunctions.
Nonpoint Source
Function
Proximity to
Sources
Proximity to
Water Body
Watershed
Position
Site
Conditions
Wetland
Type
Soil
Characteristics
Figure 8. Parameters evaluated under nonpoint source pollutant removal subfunction.
208
-------
The third parameter is the position of the wetland in its
watershed. Several studies have documented that
headwater wetlands are most effective in removing non-
point source pollutants (9-11). Thus, the higher in its
watershed a wetland is located, the higher is its signifi-
cance in nonpoint source removal.
Two subparameters, wetland type and soil charac-
teristics, determine the value of the fourth parameter,
site conditions. By virtue of their typical microtopogra-
phy, hydrology, and vegetative structure, some wetland
types more effectively retain and filter surface runoff
than do other types. Some soil series are more effective
than others in retaining and chemically transforming
pollutants. Each subparameter is rated, and their com-
bined values produce a rating for the site conditions
parameter.
A similar evaluation of specific parameters is performed
to derive significance ratings for other wetland subfunc-
tions. In all cases, CIS analysis determines parameter
values based on the data layers described above. Some
parameters, such as wetland type in the nonpoint source
illustration, are surrogates or indicators of other wetland
properties that actually determine the wetland's func-
tional capacity. The limitations of CIS data and tech-
niques necessitate the use of indicator parameters.
Evaluation Procedure
The objective of the assessment procedure is to deter-
mine an individual wetland's ecological significance in
the watershed in which it exists. Ecological significance
is divided into three broad classes (high, medium, and
low) rather than attempting to derive a specific numerical
"score." This is partly because of the procedure's initial
application in an EPA ADID project performed by DCM
in Carteret County, North Carolina. Standard ADID pro-
cedure is to classify wetlands into three groups:
Areas generally unsuitable for the discharge of
dredged or fill material.
Areas that require a project-by-project determination.
Possible future disposal sites for dredged or fill material.
These groups correspond to the H, M, and L used in the
assessment procedure.
The approach of classifying wetlands into three broad
functional significance classes is also used, however,
because it is feasible with our current understanding of
wetland function. Attempting to assign a specific value
along a numeric continuum of functional significance
greatly exaggerates the precision with which we can
realistically apply current knowledge. The three signifi-
cance classes used in the assessment procedure pro-
vide the information necessary to meet the procedure's
objectives without going beyond the realm of reasonable
scientific validity.
As explained above, the basic evaluation is performed
at the parameter level. An H, M, or L value is assigned
to each parameter as it relates to the performance of the
wetland subfunction being considered. For example, if
the soils underlying a wetland have properties that are
highly conducive to the function being considered, the
soil characteristics parameter is rated H; if soil proper-
ties are less conducive to performing the function, the
parameter is rated M; and if soil properties are not at all
conducive to the function, the parameter is rated L. All
individual parameters under a given subfunction receive
similar ratings.
The individual parameter ratings are then combined to
give an H, M, or L rating for each subfunction. The
subfunction ratings are combined into a rating of the
wetland's significance in performing each of the primary
wetland functions. Finally, the ratings for primary func-
tions are combined into an overall rating of the wetland's
functional significance.
The process of successively combining ratings up the
structural hierarchy is the most complex aspect of the
assessment procedure. The combining, as well as the
evaluation of individual parameters, is based on funda-
mental ecological principles about how wetlands and
landscapes function. Because the ecological processes
themselves interact in complex ways, combining ratings
is much more complex than a simple summation of
individual ratings. Some parameters are normally more
important than others in determining the level at which
a wetland performs a specific function and, thus, must
be weighed more heavily in determining the combined
value. In some cases, different combinations of individ-
ual parameter ratings result in the same level of func-
tional significance. Each possible combination of
parameters must then be considered.
The automated version of the assessment procedure
maintains all individual parameter ratings and combina-
tions in a database. Because the combining process is
complex, the reason a wetland receives an overall H, M,
or L rating may not be intuitively obvious. The database
makes it possible to trace through the parameter, sub-
function, and primary function ratings that result in a
wetland's overall rating.
This database also allows consideration of specific wet-
land functions individually. For example, in a watershed
targeted for nonpoint source pollution reduction, one
management objective may be to give the highest level
of protection to wetlands most important in performing
this function. The database allows examination of each
wetland for its significance in nonpoint source removal
and production of a map of wetlands rated according to
their significance for this single function.
Individual function ratings in the database can also
be used to improve planning, impact assessment, and
209
-------
mitigation for development projects that affect wetlands.
If alternative sites are available, such as alternative
corridors for a highway, the alternative with the least
impact on the wetland function considered most impor-
tant in the watershed can be identified. Rather than
simply minimizing acres of wetland impact, the objective
would be to minimize impacts to the most important
wetland functions. Environmental assessment of wet-
land impacts can identify specific functions to be lost.
Mitigation can be improved by giving priority to sites with
the highest potential for performing the same functions.
Future reports will explain detailed procedures for evalu-
ating individual parameters and combining them into
functional ratings. This paper illustrates only the water
quality nonpoint source removal subfunction. The rating
system for this subfunction is summarized in Figure 9.
Four parameters are evaluated to determine the signifi-
cance of the nonpoint source removal subfunction.
Because (d), the site conditions parameter, has sub-
parameters below it, it is first evaluated using a relatively
simple procedure. If conditions typical of the wetland
type and characteristics of the underlying soil are both
highly conducive to removal of pollutants in runoff water
entering the wetland, the site conditions parameter is H.
If either the wetland type or the soil is not at all conducive
to pollutant removal and the other subparameter is no
more than somewhat conducive, the site conditions pa-
rameter is L. Any other combination results in an M.
Parameters
(a) Proximity to Sources
(b) Proximity to Surface Water
(c) Watershed Position
(d) Site Conditions
(1) Wetland Type
(2) Soil Characteristics
Evaluation Procedures
Site Conditions
H Both Parameters H
M Other combinations
L One parameter L and neither H
NPS Subfunction
H (a) & (b) H and (d) at least M or
(c) & (d) H and (b) at least M
M Other combinations
L Two of (b), (c), & (d) L
Figure 9. Parameters evaluated under nonpoint source pollut-
ant removal subfunction.
Following evaluation of all parameters, they are com-
bined to evaluate the significance of the wetland in
removing nonpoint source pollutants. Two combinations
result in the wetland being evaluated as highly signifi-
cant in performing this function. First, if the wetland is
adjacent to both a significant source of polluted runoff
(a = H) and a permanent surface water body into
which the runoff would flow if the wetland were not
there (b = H), and has site conditions that are at least
reasonably efficient in catching, holding, and removing
pollutants from the runoff (d at least M), it receives an
H. Alternatively, even if the wetland is not adjacent to a
pollutant source, it receives an H if it is in the headwaters
of the watershed (c = H), site conditions are highly
conducive to pollutant removal (d = H), and it is at least
close to an intermittent stream (b at least M).
On the other hand, if any two of parameters (b), (c), and
(d) are evaluated L, the significance of the wetland for
nonpoint source pollutant removal is Low. That is, the
wetland is evaluated as L for this function if any of the
following conditions exist:
The wetland is not close to surface water (b = L) and
downstream in the watershed (c = L).
The wetland is not close to surface water (b = L), and
its site conditions are poor for pollutant removal (d = L).
The wetland is downstream in the watershed (c = L)
and has poor site conditions (d = L).
Any combination of parameter evaluations other than
those resulting in an H or L results in the wetland being
evaluated as of moderate significance for removing non-
point source pollutants. This example is typical of evalu-
ation procedures used for all subfunctions. More often
than not, the evaluation procedures are complex and
multifarious in their reasoning and application. Hope-
fully, though, they are scientifically valid based on cur-
rent knowledge of wetland ecology.
Opportunity and Capacity
The concepts of opportunity and capacity for a wetland
to perform a given function were briefly discussed
above. For a wetland to actually perform a function, it
must have both the opportunity and the capacity for the
function. In terms of the nonpoint source example, a
source of potentially polluted runoff must enter the wet-
land to provide an opportunity, and the wetland must
have the internal capacity to hold the runoff and remove
the pollutants before releasing the water. Factors exter-
nal to the wetland usually determine the opportunity to
perform a function, while properties of the wetland itself
along with its landscape position determine the capacity
to perform the function.
Because the assessment procedure is a landscape
scale procedure that evaluates the functions a wetland
210
-------
performs in relation to its surroundings, essentially every
subfunction includes opportunity parameters. A func-
tional assessment that is too heavily dependent on op-
portunity parameters, however, is static and rapidly
becomes invalid as land uses change. A wetland that is
bordered by natural forest today can be bordered by a
young pine plantation or a subdivision under construc-
tion by next year. The fact that a wetland does not have
the opportunity to perform certain functions today does
not mean that it will not have the opportunity in the
future. If an assessment of wetland significance is to
remain valid overtime in a landscape subject to change,
opportunity parameters alone cannot be determinative.
The evaluation procedure for the nonpoint source sub-
function explained above is an example of how the
assessment procedure handles this situation. The op-
portunity for a wetland to receive polluted runoff water
from surrounding lands (a = H) can result in an evalu-
ation of H for this subfunction if other properties are also
present, but it does not have to be present for a wetland
to be evaluated H. Other parameters (c and d = H, and
b at least M) that give a wetland a high capacity to
remove nonpoint source pollutants can also result in an
H. Conversely, lack of present opportunity (a = L) does
not result in an evaluation of low significance for this
function. At least two of the other parameters must be L
for the wetland to be evaluated as L.
These conventions hold throughout the procedure. A
present high opportunity to perform a function can result
in an evaluation of high significance for the function, but
high capacity can also result in an H evaluation even if
present opportunity is lacking. Lack of present opportu-
nity alone never results in an evaluation of low signifi-
cance for a function. High opportunity is treated
essentially as a "bonus" consideration that can result in
a higher evaluation for a wetland than its capacity alone
would indicate but that will never result in a lower evalu-
ation because of its absence.
Overriding Considerations
Several considerations are of such importance in the
North Carolina coastal area that their presence alone
will result in a wetland evaluation of high significance.
These parameters are evaluated first as either true or
false, and if one or more of them is true, the rest of the
evaluation procedure is not performed.
The first overriding consideration is whether the wetland
is a salt or brackish marsh meeting the definition of
"coastal wetland" as set forth in North Carolina statutes
(NCGS 113-229(n)(3)) and rule (NCAC 7H .0205(a)).
Coastal wetlands in North Carolina are designated by
law as highly significant. Consequently, the assessment
procedure evaluates them automatically as H and in-
cludes no considerations for differentiating among the
functional significance of these wetland types.
The second overriding consideration is whether the wet-
land is adjacent to an officially designated primary nurs-
ery area (PNA). All designated PNAs are included in
"areas of environmental concern" in the NC CMP and
are protected by a specific set of regulations. They are
areas where initial postlarval development of finfish and
crustaceans takes place and, thus, are critical to estu-
arine fish and shellfish populations. Wetlands adjacent
to PNAs are highly important in maintaining water qual-
ity and appropriate salinity gradients in these critical
areas and are automatically evaluated as of high func-
tional significance.
The third overriding consideration is whether the wet-
land contains threatened or endangered species. If a
known threatened or endangered plant or animal spe-
cies on either federal or state lists is present, the wetland
is evaluated as highly significant. The determination is
based on information obtained from the North Carolina
Natural Heritage Program.
The fourth overriding consideration is whether the wet-
land includes all or part of a critical natural area as
designated by the North Carolina Natural Heritage Pro-
gram. If so, the site is considered of high significance.
CIS data layers maintained by the Natural Heritage
Program also help make this determination.
Verification
Throughout the development and initial application of
the assessment procedure, we have checked and veri-
fied its validity. Parameter evaluations and combination
procedures are based on the best wetland science avail-
able in the scientific literature. The validity and accuracy
of the CIS databases used to apply the procedure
have been verified to the extent possible. Following sec-
tions of this report fully document any assumptions
made about wetland ecology, CIS data, or CIS analytical
techniques.
An advisory panel of wetland scientists familiar with the
wetlands of coastal North Carolina and representatives
of several state and federal wetland-related agencies
reviewed every step of the procedure's development.
While their review does not represent an endorsement
of the procedure or its results by the agencies or indi-
viduals included, it does indicate the level of peer review
the procedure has received.
During development of the procedure, field visits were
made to nearly 400 wetland sites to gather data on func-
tional indicators. On these same site visits, a field-based
functional assessment procedure, the Wetland Rating
System developed by the North Carolina Division
of Environmental Management, was applied. This pro-
vides the basis for a field verification of the assessment
procedure.
211
-------
Discussion
As we continue to understand more about the role of
wetlands in maintaining a healthy environment, the use-
fulness of wetlands locational data continues to grow in
importance. Spatial data can assist county planners in
guiding development away from environmentally sensi-
tive areas. Landowners now have the capability to look
at a map and realize very quickly that wetlands exist on
a given area of land. In addition, economic development
councils can use this information to plan development in
areas attractive to a particular industry. If a new busi-
ness or industry wishes to locate in an area positioned
such that the wetlands permitting process could be
avoided, maps showing lands void of wetlands could be
a significant tool to the economic development council.
The representation of these wetlands' ecological signifi-
cance dramatically increases the utility for these data.
While paper maps can be distributed to all interested
parties, digital data also are available to public agencies
who have CIS capabilities. In Carteret County, for exam-
ple, a publicly installed workstation will be made avail-
able with these data installed. The county government
will be able to view wetlands in the context of cadastral
boundaries that already are on CIS. Information about
sensitive resources made available prior to any devel-
opment will, hopefully, lead development away from
environmentally sensitive areas.
References
1. Hefner, J.M., and J.D. Brown. 1985. Wetland trends in the south-
eastern United States. Wetlands 4:1-12.
2. Dahl, I.E. 1990. Wetlands losses in the United States 1780s to
1980s. U.S. Department of the Interior, Fish and Wildlife Service,
Washington, DC.
3. DEM. 1991. Original extent, status, and trends of wetlands in
North Carolina. Report No. 91-01. North Carolina Department of
Environment, Health, and Natural Resources, Division of Envi-
ronmental Management, Raleigh, NC.
4. DCM. 1992. Final assessment of the North Carolina coastal man-
agement program. Report to the Office of Ocean and Coastal
Resource Management, NOAA, U.S. Department of Commerce,
performed under the Coastal Zone Enhancement Grants Pro-
gram (January 10). Section 309, CZMA, North Carolina Division
of Coastal Management, Raleigh, NC.
5. DCM. 1992. Final strategy for achieving enhancements to the
North Carolina coastal management program. Proposal to the
Office of Ocean and Coastal Resource Management, NOAA, U.S.
Department of Commerce, performed under the Coastal Zone En-
hancement Grants Program (March 25). Section 309, CZMA,
North Carolina Division of Coastal Management, Raleigh, NC.
6. Hefner, J.M., and K.K. Moorhead. 1991. Mapping pocosins and
associated wetlands in North Carolina. Wetlands 2(Special Is-
sue):377-389.
7. Brinson, M.M. 1993. A hydrogeomorphic classification for wet-
lands. U.S. Army Corps of Engineers Waterways Experiment Sta-
tion Report WRP-DE-4. Washington, DC.
8. Khorram, S., H. Cheshire, K. Siderelis, and Z. Nagy. 1992. Map-
ping and CIS development of land use/land cover categories for
the Albemarle-Pamlico drainage basin. Report No. 91-08-NC.
Raleigh, NC: North Carolina Department of Environment, Health,
and Natural Resources.
9. Leopold, L.B. 1974. Water: A primer. San Francisco, CA: WH.
Freeman and Co.
10. Whigham, D.F., C. Chitterling, and B. Palmer. 1988. Impacts of
freshwater wetlands on water quality: A landscape perspective.
Environ. Mgmt. 12(5):663-671.
11. Novitski, R.P. 1979. The hydrologic characteristics of Wisconsin
wetlands and their influence on floods, streamflow, and sediment.
In: Greeson, P.E., J.R. Clark, and J.E. Clark, eds. Wetland func-
tions and values: The state of our understanding. Lake Buena
Vista, FL: American Water Resources Association.
212
-------
Decision Support System for Multiobjective Riparian/Wetland Corridor Planning
Margaret A. Fast and Tina K. Rajala
Kansas Water Office, Topeka, Kansas
Kansas has numerous programs that affect riparian cor-
ridors and associated wetlands. These programs in-
clude planning, monitoring, assistance, research, and
regulatory activities. Although administration of these
programs often overlaps, integration of program objec-
tives into a holistic, multiobjective approach to resource
planning and management has been lacking. A large
amount of resource data was routinely collected and
compiled, but no effective way had been developed to
integrate these data into the decision-making process.
The Kansas Water Office (KWO) was awarded a grant
in September 1992 from the U.S. Environmental Protec-
tion Agency (EPA) to develop a geographic information
system (CIS) decision support system (DSS) that would
enable the state to augment its ability to manage ripar-
ian/wetland corridors. The project used CIS to differen-
tiate between reaches of a stream corridor to evaluate
their environmental sensitivity. The Neosho River basin,
one of 12 major hydrologic basins in Kansas, was used
as a pilot to demonstrate the feasibility of the concept.
The KWO will use the DSS to help target sensitive areas
in the Neosho basin for further planning activities. The
project will also benefit other state agencies in their
riparian/wetland corridor efforts. The implementation of
planning objectives may involve local units of govern-
ment and, ultimately, private landowners.
Major phases of the project included:
A needs assessment study
A feasibility analysis
A system design
Construction of the DSS for the Neosho River basin
A final evaluation of the DSS capabilities
An interagency project advisory group (IPAG), consist-
ing of representatives from eight agencies directly or
indirectly involved in riparian and wetland protection
activities, was formed to assist in project design and
evaluation.
Major steps involved in designing the DSS included:
Selection and CIS development of databases used
for riparian corridor evaluation.
Creation of riparian corridor segments.
Development of an analysis methodology to apply to
corridor segments.
Evaluation of the DSS.
Databases Selected for Decision Support
System Development
Many types of data were reviewed for the DSS. Several
were not used due to the costs associated with geo-
graphically referencing the data, given the current data
format.
The databases listed in Table 1 are available in the DSS.
During the system design phase of the project, the IPAG
identified the need to develop a pilot study area for the
DSS. The IPAG had difficulty understanding how a DSS
would use geographically referenced data sets (cover-
ages). Before committing to a design for the develop-
ment of a basinwide system, the IPAG decided first to
develop a pilot study area, with a specific focus (appli-
cation), that could be on-line and demonstrated early.
This would allow time for further refinement of the scope
of work and identification of coverages to be developed
prior to basinwide development of the DSS. For the pilot
study application, the IPAG chose to assess the value
and vulnerability of the riparian areas in two 11-digit
hydrologic unit code (HUC11) watersheds to allow the
user to evaluate a corridor segment and compare be-
tween segments and to prioritize or target segments for
further planning activities.
As development of data layers progressed for the pi-
lot, the IPAG quickly determined that the DSS project
213
-------
Table 1. DSS Database List
DSS Name Data Description
Source
Boundary
Buffer
Channels
Con_ease
Contam
Corridor
County
Dams
Dwrapp
Gages
Geology
Huc11
HydMOOk
Kats
Landc
Lc_stats
MDS
NPS
Perenial
Pop
PPL
Publand
Roads
Sections
Streamev
Neosho River basin boundary
Riparian corridor
Stream channelization
Conservation easements
Water contamination
Riparian corridor
County boundaries
Dam structures
Water appropriations
United States Geological
Survey (USGS) stream gaging
stations
Surface geology
11 -digit hydrologic unit
boundaries
Hydrology
Kansas water quality action
targeting system
Land cover
Land cover statistics
Minimum desirable stream
flow monitoring gages
Nonpoint source pollution
Perennial hydrology
Population
Populated places
Public lands
Roads
Section corners
1981 stream evaluation
Soil Conservation Service (SCS) HUC11 drainage basins; 1 :100,000-scalea
Original buffer on mainstem Neosho and Cottonwood; 147 corridor segments split
on tributary confluences
Division of Water Resources (DWR) legal description of locations
Locations of important natural resources that could be purchased by the state from
willing landowners for conservation protection
Kansas Department of Health and Environment (KDHE) contamination locations3
Final riparian corridor; 63 corridor segments developed from HUC11 boundaries
Kansas Geological Survey (KGS) cartographic database; 1 :24,000 scale3
DWR legal descriptions of locations
DWR legal descriptions of locations3
USGS latitude-longitude descriptions; CIS cover developed by USGS
KGS 1 :500,000-scale3
SCS HUC11 drainage basins; 1 :100,000-scale3
USGS 1:100,000-scale digital hydrology3
KDHE target valuable and vulnerable scores by HUC11 drainage basin
1 :100,000-scale developed from satellite imagery by the Kansas Applied Remote
Sensing Program, University of Kansas3
Summary statistics on land cover by corridor segment
Subset of USGS gaging stations
Target watersheds identified in the Kansas Water Plan
Reselected perennial streams from 1:100,000 USGS digital hydrology
Urban land cover (from landc) with 1980 and 1990 Census population data
Geographic names information system (GNIS) entries for Kansas; GIS cover
developed by USGS
State and federally owned land digitized from 1:100,000-scale USGS quad maps
USGS 1:100,000-scale digital roads3
KGS cartographic database; 1 :24,000-scale3
U.S. Fish and Wildlife stream evaluation study; Kansas Department of Wildlife and
Parks (KDWP) provided data on paper maps
T_and_e Threatened and endangered
species
Temussel Threatened and endangered
species
Tigrcity City boundaries
Twp Townships
Watrfowl Water fowl locations
Wq_eff Water quality: effluent
Wq_grnd Water quality: ground water
Wqjake Water quality: lake
Wq_strm Water quality: stream
Stream locations of state and federal identified threatened and endangered
species; KDWP provided data on paper maps
Locations of state endangered floater mussels; KDWP provided data on paper maps
U.S. Census 1:100,000-scale TIGER line data; boundaries only, areas not named
(use with PPL)
KGS cartographic database; 1:24,000-scale3
KDWP locations and counts of annual waterfowl migration; data developed from
paper maps (Restrict public distribution of data per KDWP request.)
KDHE sampling sites; GIS cover developed by KDHE
KDHE sampling sites; GIS cover developed by KDHE
KDHE sampling sites; GIS cover developed by KDHE
KDHE sampling sites; GIS cover developed by KDHE
Data available at the Kansas Data Access and Support Center (DASC).
214
-------
parameters would have to be limited to the riparian
corridor along the mainstem of the Neosho and Cotton-
wood Rivers. The costs associated with developing ri-
parian corridor segments for all perennial waters in the
Neosho basin was fargreaterthan the available funding.
Creation of Riparian Corridor Segments
A buffer width of one-half mile (one-quarter mile from
each stream bank) for the mainstem of the Neosho and
Cottonwood Rivers was used to produce the riparian
corridor. If more time and funding had been available,
riparian corridors for all perennial streams in the Neosho
basin could have been developed. The development of
this second view of data, organized by the HUC11 wa-
tershed, would then have been useful for individual
watershed analysis because all perennial streams in the
watershed could be analyzed.
The intersection of the HUC11 basin boundaries seg-
mented the corridor. In several instances, small sliver
polygons were produced where the HUC11 boundary
paralleled the river within the 1/4-mile corridor. The sliver
polygons were dissolved into the majority HUC11. In
other words, this project assumed that the 1/4-mile cor-
ridor buffer was more accurate and useful than the
1:100,000-scale HUC11 boundary.
Many of the HUC11 boundaries that the Soil Conserva-
tion Service (SCS) developed actually follow the course
of the Kansas streams, ratherthan intersect them. When
this occurred along the Neosho and Cottonwood Rivers,
we found that the resulting opposing corridor segments
did not always balance with an equivalent length. Also,
some HUC11 boundaries would first follow the river,
then cross the river. This resulted in corridor segments
that encompass both sides of the river for a portion of
the segment and follow only one side of the river for
another portion of the segment. To address these situ-
ations, the KWO arbitrarily added intersections to create
equivalent left and right bank corridor segments and to
create corridor segments that encompassed either one
side of the river or both sides of the river.
Once the corridor segments were finalized and numbered,
the corridor segment identification number (corrseg-id)
was attached to the other CIS covers. This allows the
reselection of data for a given corridor segment, using
Boolean expressions in the DSS.
Development of an Analysis
Methodology: Land Use
The IPAG determined that one of the most significant
factors associated with the quality of the riparian corridor
is land cover. Land cover was analyzed for the riparian
corridor segments; the CIS cover lc_stats contains sum-
mary statistics for each corridor segment. The calcula-
tions discussed in the following paragraphs identify the
data found in the lc_stats cover. Due to the size of the
land cover data set in the Neosho River basin, the DSS
includes only the land cover within the riparian corridor.
One way of identifying corridor segments in need of
protection or remedial action is to determine the ratio of
the number of acres in the corridor segment that contain
the preferred riparian land cover types (grasses, woods,
and water) to the number of acres that contain the least
preferred types of land cover (crops and urban areas).
The corridor segments can then be ranked according to
that ratio.
Other calculations are useful:
bad_pct: percentage of the corridor segment that con-
tains crop and urban land cover types.
bad_tbad: percentage of all crop and urban land
cover for the entire riparian corridor that resides in
the corridor segment.
'type'_pct: percentage of the corridor segment that is
crop, grass, wood, water, and urban. Type' refers to
each of the five land cover types; lc_statsuses: a
separate value for each (e.g., crop_pct).
'type'_t'type': percentage of each type of land cover
for the entire riparian corridor that resides in the cor-
ridor segment (e.g., crop_tcrop).
'type'_acres: total acreage of each type of land cover
in the corridor segment (e.g., crop_acres).
good_acres: total acreage of grass, wood, and water
in the corridor segment.
bad_acres: total acreage of crop and urban in the
corridor segment.
Another significant benefit of the DSS is the ability to see
where the land cover types are in relation to the river.
As an example, the ability to identify corridor segments
that have crop land extending to the river on both banks
is useful because they are the segments most vulner-
able to bank erosion. Those segments can then be
targeted for further remedial activities planning.
Decision Support System Requirements
The DSS data sets were developed and analyzed using
ARC/INFO on a UNIX-based workstation. The final cov-
ers were then exported and transferred to a microcom-
puter for use in ARC/VIEW. Hardcopy prints are printed
to a Tektronix Phaser III color wax printer with 18 Mb of
RAM, running in Postscript mode.
The DSS data sets total 26 Mb. ARC/VIEW version 1
requires 8 Mb of RAM to load the program. To run the DSS
efficiently, a 486DX-66 with 16 Mb of RAM is preferred.
The DSS is slower on a 486DX-33 with 8 Mb of RAM. It
was not tested on any other PC configuration, so a con-
figuration in between the two may be satisfactory.
215
-------
Processing GIS Data
Reselecting the perennial streams in the Neosho basin
and further identifying the mainstem of the Neosho and
Cottonwood Rivers using United States Geological Sur-
vey (USGS) 1:100,000-scale hydrography can be time
consuming. Perhaps the River Reach III covers should
replace that data in the future.
Attaching census data to the urban land cover polygons,
as was done for the pop cover, is not recommended.
Use of the TIGER line files and cover would give a more
accurate distribution of the population. Because less
than 5 percent of the riparian corridor had urban land
cover, the KWO did not use the pop cover in its evalu-
ation. Several summary covers of TIGER and census
data will soon be available from DASC.
Clipping the other ARC/INFO covers to the Neosho
basin and attaching the corrseg-id, using the identity
command, was unremarkable.
Processing Non-GIS Digital Data
Channels and dams were in digital format but were not
in ARC/INFO format. The files were processed using the
LeoBase conversion software from the KGS, then gen-
erated into ARC/INFO covers. Some records were lost
in the conversion. The LeoBase program fails to convert,
or incorrectly converts, legal descriptions for sections
that do not have four section corners (e.g., northeastern
Kansas). The Division of Water Resources is in the
process of attaching latitude-longitude to the point loca-
tions. Processing these data should take only a few
hours at most.
Processing Nondigital Data
Several covers were developed on contract from paper
maps or legal descriptions. They were: conservation
easements (con_ease), public lands (publand), stream
evaluation (streamev), threatened and endangered spe-
cies (t_and_e and temussel), and waterfowl (watrfowl).
Most of the data for these covers were drafted on a
1:100,000-scale USGS quad map and digitized. The
stream evaluation data were developed using a
scanned paper map of the coded streams as a backdrop
for the 1:100,000-scale hydrography; the digital streams
were reselected and coded.
In summary, KWO's GIS personnel needed approxi-
mately 275 hours to develop the riparian corridor seg-
ments, process the land cover data and summary
statistics, export the covers, transfer and import the
covers for ARC/VIEW, and assist in the development
and presentation of the DSS demo. Contract personnel
spent approximately 183 hours developing GIS covers
for the DSS. This does not include the time spent iden-
tifying the perennial and mainstem hydrology in the
USGS 1:100,000-scale hydrology.
Final Evaluation of the Decision
Support System
In its final evaluation of the system, the IPAG determined
the system to be useful and an excellent start at consoli-
dating a variety of data that have application for riparian
corridor/wetland issues. Many IPAG members found
ways to use the DSS in their own programs. Additional
comments on the system evaluation are as follows:
Concern about the lack of complete wetland data.
The land cover data available could not identify wet-
land areas.
Need for more detailed woodland data. Again, the
resolution of the land cover data precluded detailed
identification of woodland areas. The Kansas Biologi-
cal Survey (KBS), the KWO, and EPA are now pur-
suing options to develop more detailed land cover
data, including wetlands and woodlands.
The lack of information on the tributaries did not allow
full basin analysis, which would be desirable. This
issue is addressed in the "construction" discussion
above.
Desirability of expanding the project with elevation
and temporal data.
Lack of definition of the floodplain. Federal Emer-
gency Management Agency (FEMA) floodplain data
are not easily incorporated into a GIS. Other options,
including satellite imagery of the flood of 1993, will
be evaluated.
Project development requires extensive communica-
tion between program people and GIS technicians.
This can be a daunting task due to the technical
vocabularies involved and the many other ongoing
activities of the participants.
Consideration of the requirements for transferring the
project to other potential users. GIS applications gen-
erally use large databases. User microcomputers
may not have the CPU, RAM, and storage capacity
necessary for the DSS application and often have a
limited number of options for data transfer.
Concern about costs and time associated with the
expansion of the DSS to other basins in the state.
This project was focused on one of the 12 major
hydrologic regions in Kansas. Funding options, pro-
ject scope, and system refinements based on the
physical characteristics of the other basins need to
be pursued.
The KWO learned that clearly defining a single DSS
application at the outset of the project is critical. The
KWO originally believed that the DSS could be devel-
oped with general descriptions of the broad range of
program applications, utilized by multiple agencies, that
could benefit from the DSS. Each participating agency
216
-------
could bring its programs and needs to the IPAG for areas for further planning activities, the IPAG became
discussion; the resulting DSS would then serve those more confident in its advisory role. Upon completion of
multiple programs and needs. Instead, the ambiguity of the project, the IPAG members could readily identify
the objective confused the IPAG. Once the IPAG chose how the DSS could be enhanced, modified, or directly
to focus on a single application, the assessment of used in their own programs.
riparian corridor value and vulnerability to target priority
217
-------
Design of GIS Analysis To Compare Wetland Impacts on Runoff in Upstream
Basins of the Mississippi and Volga Rivers
Tatiana B. Nawrocki
Natural Resources Research Institute, University of Minnesota, Duluth, Minnesota
Introduction
The attention given in hydrologic studies to wetlands
differs significantly between the United States and Rus-
sia at the present time. Fundamental theories and
mathematical models are developed in both countries to
describe hydrologic processes and impacts of water-
shed conditions on surface runoff. In the United States,
however, theoretical investigations are directly pointed
at wetlands and are supported by large-scale field stud-
ies and advanced technological capabilities to manage
spatially distributed information. Unlike in Russia, in the
United States, special scientific symposia are devoted
to wetland hydrology, where major tasks for hydraulic
and hydrologic research needs are formulated. Among
these tasks are the understanding and assessment of
relationships between various hydrologic modifications
and wetland functions, especially wetland flood convey-
ance and water quality protection functions (1). Water-
shed-scale comprehensive field studies of wetland
functions are underway, for example, at constructed
experimental wetlands in the Des Plaines River basin in
Illinois (2). A new long-term goalstrategic restoration
of wetlands and associated natural systemshas been
formulated (3).
The intensive efforts of many U.S. scientists yielded
numerous results and attracted more attention to the
complicated nature of wetlands processes. Wetlands
were evaluated as runoff retention basins, and it was
found that, in northwestern states, up to 12 inches of
water could be accumulated per wetland acre (4, 5).
Overtime, piecemeal loss and degradation of wetlands
in many areas of the United States have seriously de-
pleted wetland resources. Researchers also discovered
that adverse impacts from wetland degradation could
appear indirectly with little obvious spatial or temporal
connections to sources. As described by Johnston (6):
Cumulative impacts, the incremental effect of an
impact added to other past, present, and reasonably
foreseeable future impacts, has been an area of
increasing concern. . . . Impacts can accumulate
over time or over space and be direct or indirect. An
indirect impact occurs at a location remote from the
wetland it affects, such as the discharge of pollut-
ants into a river at a point upstream of a wetland
system.
The process of solving environmental problems related
to wetlands is increasingly complex. Analyzing diversi-
fied data over increasingly broad areas becomes essen-
tial for making competent decisions.
Comparing wetland hydrologic functions in headwaters
of the Mississippi River (United States) and the Volga
River (Russia) could provide additional information
about how alternative management strategies affect
runoff, peak flow, and water quality under changing
climates. A macro-scale "field experiment" in both of
these naturally similar areas is already underway. Wet-
land conservation as opposed to drainage is now the
prevailing policy in the upper Mississippi basin. In Rus-
sia, however, economic problems have prevented this
type of policy from becoming a priority. Instead, peat
mining, reservoir construction on lowlands, and drain-
age for farming and private gardening are common.
This project, which is being implemented at the Natural
Resources Research Institute (NRRI), University of Min-
nesota, Duluth, has the following goals:
Developing a multilayered hierarchical base of geo-
graphic information system (GIS) data for headwater
watersheds of the Mississippi and Volga Rivers.
Developing a comparative analysis of wetland im-
pacts on the hydrology of the rivers.
Studying the relationships between natural and hu-
man-induced factors on wetland functions under cli-
mate change and variable strategies of wetland
conservation.
218
-------
Defining criteria and thresholds for wetland system
stability with regard to flood risk and water quality.
Outlining recommendations for wetland management
in the headwaters.
The methodology for comparative assessments in-
volves statistical analysis, hydrologic models, CIS, and
remote sensing. Representative watersheds will be
studied in more detail, and procedures for scaling infor-
mation from the local to the regional level will be
developed.
Input Data
In recent years, U.S. governmental and state agencies,
as well as a number of private companies, have ex-
pended considerable efforts to compile the available
data in CIS format for multidisciplinary analysis of wa-
tershed problems. Among the major sources of informa-
tion essential for studies of wetland hydrologic functions
are:
The National Wetlands Inventory, conducted by the
U.S. Fish and Wildlife Service.
Digital elevation maps (OEMs) developed by the U.S.
Geological Survey (USGS).
Major and minor watershed boundaries, outlined for
Minnesota by the Minnesota Pollution Control Agency
(MPCA).
The water quality sampling network from the U.S.
Environmental Protection Agency (EPA).
The digital chart of the world (DCW), issued by the
Environmental Science Research Institute (ESRI) in
scale 1:1,000,000.
These and other sources, listed in Table 1, were used to
compile the map illustrations for this paper.
Almost no similar data in CIS form could be found for
the territory of the former USSR, however. Any specific
data (e.g., detailed maps, hydrology records, water
quality sampling data) are generally in paper files dis-
persed among many agencies and are hard to obtain.
The forms of information storage and means of its analy-
sis are out of date, and most maps exist in single or few
copies in paper files. In Russia, the time lag grows
between the dynamic changes in the environment and
the traditional pattern and inertia of management
structures.
The CIS situation in Russia developed some interesting
paradoxes. During the first few decades of space pro-
grams, certain state agencies accumulated an outstand-
ing bank of world image data. When economic
hardships hurt the previously privileged space industry,
numerous joint ventures with foreign companies were
created to distribute images on the world market for hard
currency. These data are hardly available for domestic
uses, however. The current domestic price for image
data is 20,000 to 25,000 rubles for a black-and-white
picture of an area 60 by 60 kilometers, or 60,000 to
70,000 rubles for the same image on a computer disk.
With the present level of funding for scientific research,
the price is too high. Security regulations still restrict
access to later data, showing land use changes.
Another paradox is scientists' attitude toward their data.
Abandoned by the state, agencies and institutes are
reluctant to share their specific data in multidisciplinary
projects. Data files are now a commodity. Accomplishing
an overlay and integration of special data coverages,
which is essential to any watershed CIS study, is almost
impossible.
The third paradox is the attitudes of local, regional, and
central authorities toward CIS. Many authorities are still
ignorant about the potential of this technology. Those
who are knowledgeable prefer not to promote CIS for
watershed-related tasks because it involves land use
analysis. With the onset of land privatization, the best
pieces of property (e.g., the waterfront lots adjacent to
drinking water reservoirs in the Moscow region) are
rapidly allocated to the most powerful landowners. Thus,
limiting access to this kind of information is deemed
safer.
Experts in Russia have not yet applied CIS to wetland
hydrology studies because CIS is still a very new and
mostly unfamiliar technology. This makes the current
study unique both for its results and for its application of
CIS methodology.
Closer review of data sources indicates that most of the
input data for the project is available, though dispersed
among many agencies. Table 1 is a preliminary list of
data and data sources.
Project Design
The project addresses the following questions:
How do the extent and positioning of wetlands in the
headwaters of large rivers affect runoff and peak
flow?
What are the spatial relationships between wetland
and other land uses regarding flood risk and water
quality under variable climate conditions?
What is the role of wetlands for diffuse pollution pre-
vention and sediment deposition control under alter-
native management?
What determines major criteria for wetland conserva-
tion in headwaters, ensuring environmentally sustain-
able development under multiobjective land and
water resource uses?
219
-------
Table 1. Data Sources for a GIS Study of Wetlands in the Basins of the Mississippi (United States) and the Volga (Russia)
Data
Level 1
Base maps
Stream network
Urban and rural areas
Wetlands, unclassified
Forests
Agricultural lands
Level 2
Watershed boundaries
Digital elevation maps
Digital orthophotos
Soils
Hydrologic records
Water quality records
Land uses
Wetlands, classified
United States
DCW3
DCW3
DCW3
NWIa
DCW3
LMDb'c
MPCAa
USGSa
USGS, LMICa
SCSb'c
USGSb
EPA/MPCAa'b
LMIC, LSAT3
NWIa
Russia
DCW3
DCW3
DCW3
DCW3
MGUb
MGUb
RWRCC
NA
CD, RPIb'c
RPIC
CHMb'c
CHM, RCP, RPIC
LSAT, CD, RPI3'C
RPIC
Data available in GIS ARC/INFO format.
bDatabases; needed conversion to ARC/INFO.
cData available in paper files; needed digitizing.
Key:
CD = Commercial distributors
CHM = Russian Committee on Hydrometeorology
DCW = ESRI digital chart of the world
EPA = U.S. Environmental Protection Agency
LSAT = Satellite image data
LMD = Published literature and map data
LMIC = Minnesota Land Management Information Center
MGU = Moscow State University
MPCA = Minnesota Pollution Control Agency
NA = Data not available
NWI = U.S. National Wetland Inventory
RCP = Russian State Committee on Natural Resources and Conservation
RPI = Miscellaneous planning agencies and research institutes
RWRC = Russian Water Resources Committee
SCS = U.S. Soil Conservation Service
USGS = U.S. Geological Survey
Tasks established to address these questions include:
Developing a multilayered hierarchical base of GIS
data for headwater watersheds of the Mississippi and
Volga basins.
Performing a comparative analysis and simulation of
wetland impacts on hydrology and water quality at
representative watersheds.
Deriving the relationships between natural and hu-
man-induced factors and wetland functions under cli-
mate change with regard to variable strategies of
wetland conservation.
Defining the criteria and thresholds for wetland sys-
tem stability with regard to flood risk and water quality
parameters.
Outlining recommendations for land and water re-
sources management and wetland positioning in the
headwaters.
GIS is the essential tool for manipulating and integrating
the many types of spatial data on water resources, soils,
vegetation, land use, economics, and the environment.
GIS compiles many sources (e.g., maps, field notes,
remote sensing, statistical data) into a consistent, inter-
pretable database used for specific scientific goals and
development decisions. The user can run GIS ARC/INFO
software on workstation and PC platforms and apply the
hierarchical approach to GIS data management, devel-
oped earlier (7). At the task level, data resolution and
corresponding modeling tools vary.
Level 1 contains the basic reference information for
large regions (e.g., the Upper Volga and the Minnesota
portion of the Upper Mississippi River basins). It covers
an area of several hundred thousand square kilometers,
with a map scale approaching 1:1,000,000. Landsat
thematic map data and the DCW (8) are used as
sources of data at this level. Vogelmann et al. (9) dem-
onstrated the methodology for detection of freshwater
wetlands using remote sensing data based on maximum
220
-------
likelihood supervised classification. A graphic data file
on CIS focuses on basic physical characteristics such
as stream network, geology, soils, wetland classifica-
tion, and other major land uses. A complementary tabu-
lar database or attribute file contains information on
stream flow, water quality, pollution sources, and wet-
land impacts on material fluxes. At this level, the general
physiographic and statistical information is accumulated
and analyzed, territories are classified, and major prob-
lems and typical case study watersheds are defined.
This information is compiled from literature, cartographic
data in paper and digitized form, statistics, and space
image data.
In Level 2, the more detailed CIS analysis and scenario-
based modeling is implemented at the watershed scale
with a map scale of approximately 1:25,000. The water-
shed demonstration focuses on alternative approaches
to priority-setting in wetland management, climate im-
pact analysis, and resulting interactions with landform,
soils, biosphere, and runoff. The sources of data are
special, topographic maps and air photo interpretation.
Simulation studies of water balance and fluxes among
the various reservoirs are implemented at Level 2. De-
veloping procedures for scaling information from the
local to regional level is the important task at this level.
CIS assists in handling the input parameter library and
analyzing the output. CIS studies, involving area meas-
urements and distribution analysis, evaluate cumulative
impacts on runoff and its quality from the loss of wetland
area, caused by drainage or filling, under stationary and
changing climate.
Wetland functions are considered under two sets of
scenarios. Management scenarios compare different
wetland and farming allocations, conservation practices,
and agricultural chemical use. Climate scenarios as-
sume rainfall and temperature changes under global
warming. Scenario-based simulation is applied in the
analysis of watershed runoff, wetland moisture regime,
soil erosion, and water quality processes.
Methodology
CIS database structure is related to the selected meth-
odology. CIS serves as a linking tool for input-output
data analysis and transfer between models, used at
different levels and stages of studies.
Scientists in both the United States and Russia devel-
oped statistical methods to obtain quantitative relation-
ships between stream flow and wetland area in the river
basins. Johnston (6) summarized the U.S. findings:
Empirical equations for predicting streamflow, de-
veloped by U.S. Geological Survey in Wisconsin
and Minnesota, indicate that flood flow is propor-
tional to the negative exponent of wetlands and
lakes ratio on a watershed (10,11). This means that
relative flood flow is decreased greatly by having
some wetlands in a watershed, but a watershed with
a large proportion of wetlands does not reduce flood
flow much more than a watershed with an interme-
diate proportion of wetlands. For example, predicted
flood flow was 50 percent lower in Wisconsin water-
sheds with 5 percent lakes or wetlands than it was
in watersheds with no lakes or wetlands, but in-
creasing the proportion of lakes and wetlands to 40
percent decreased relative flood flow by only an
additional 30 percent (12).
Other estimates agree that wetland encroachment on a
watershed of less than 25 percent generally has a mini-
mum influence on peak flow (5, 13, 14).
Johnston and colleagues (15) applied these equations
to watersheds in central Minnesota. They found that a
watershed with 1.6 percent lakes and wetlands had a
flow per unit watershed area that was 10 times the flow
predicted for a watershed with 10 percent lakes and
wetlands, while watersheds with 10 to 50 percent lakes
and wetlands had about the same flood flow per unit
area.
Statistical analysis indicates that peak discharge in-
creases with decreasing wetland area within the drain-
age basin. The regression equation defines the
approximation for northwestern Minnesota (16):
QE.R A A 0.677 n N-0.506
AM=58.4AW (Ls)
LS=100(AL + AM)/AW+1
where:
QAM = arithmetic mean of the annual series, cubic
feet per second.
Aw = watershed area, square miles.
AL = lake area within the watershed, square miles.
AM = marsh area within the watershed, square miles.
A similar statistical approach was developed for peak
flow determination in Russia. Maximum flow discharge
from snow melt is calculated for the central European
zone as (17):
Qm=K0*hp*S1 *S2*S3/(A + 1)n
where:
Qm = flow discharge, cubic meters per second.
KQ = coefficient of flood concurrence, KO = 0.006
for plain river basins.
hp = calculated flood runoff for given probability,
millimeters.
A = drainage area, square kilometers.
n = coefficient, n = 0.17.
221
-------
B! = lake storage coefficient, if lake area is less
than 1 percent of A, then S1 = 1.
S2 = pond and reservoir storage coefficient, 82 =
0.9 with ponds and S2 = 1 without ponds.
S3 = combined wetland and forest storage
coefficient.
S3 = 1 - 0.8 Ig (0.05 Sf + 0.1 Sw + 1), where Sf
and Sw are forest and wetland area,
percentage to total drainage area.
Calculated flood runoff for a given probability, hp, is
determined based on average flood runoff h, millimeters,
coefficient of variation Cv, and tabulated parameter F, as
follows:
hp = (1 + F * Cv) h
h = K, hk
hk = 100 millimeters for Moscow region.
K, = land surface coefficient, Kt = 0.9 for plains
and sandy soils, K, = 1.1 for hills and clay
soils.
The studies mentioned above generally agree with an
assumption that the incremental loss of wetland area
would have a small effect on flood flow from watersheds
with 10 percent up to 40 to 50 percent wetlands, but a
large effect on flood flow from watersheds with less than
10 percent wetlands.
The existence of similar thresholds was found in relation
to wetlands abilities to intercept pollutants. As Johnston
stated (6),
The same 10 percent threshold was identified by
Oberts (18) for suspended solids, a measure of
water quality function. Stream-water draining water-
sheds having 10 to 20 percent wetlands had about
the same loading of suspended solids, so the con-
tribution of suspended solids was relatively constant
per unit area of watershed. However, the water-
sheds with less than 10 percent wetlands had load-
ing rates per unit area that were as much as 100
times greater than the loading rates from the water-
sheds with more than 10 percent wetlands.
CIS could be especially helpful in determining the im-
pacts on downstream water quality of the spatial posi-
tioning of wetlands within watersheds. Studies prove
that the location of wetlands can affect their cumulative
function with regard to water quality (6). In an earlier
work (15), Johnston developed an index of wetland
location and applied it to a landscape-level CIS study of
urban and rural stream watersheds in central Minne-
sota. The index is formulated as:
where:
PWP = relative wetland position.
j = stream order (19) of water quality sampling
station.
i = stream order of wetland.
Ai = area of Ah order wetlands.
Calculated values for the index ranged from 0 (i.e., all
wetlands were on streams of the same order as that of
the sampling station) to 2.6 (i.e., average wetland posi-
tion was 2.6 stream orders upstream of the sampling
station). Watersheds with wetlands located close to
sampling stations had significantly better water quality
(i.e., lower concentrations of inorganic suspended sol-
ids, fecal coliform, and nitrate; lower flow weighted con-
centrations of ammonium and total phosphorus) than
watersheds with wetlands located far from sampling
stations (6).
The review of methodological approaches, as shown
above, indicates that parameters describing wetland ex-
tent, positioning, and land surface characteristics are of
universal significance for any comprehensive water-
shed-scale wetland study.
In the current study, CIS is used at Level 1 to evaluate
wetland area per watershed and to develop input pa-
rameters for relative wetland position assessment. The
comparison and selection procedures for case study
watersheds in the Volga and Mississippi basins are
based on these values. The parameters, derived from
CIS, are as follows:
1. Total watershed area.
2. Lake, pond, and reservoir area.
3. Forest area.
4. Wetland area.
5. Ratio of wetland area to total watershed area.
6. Wetland area by subwatersheds of different order.
7. Relative wetland extent by subwatersheds of different
order.
8. Land surface coefficients.
Parameters listed in groups 1 through 4 are obtained
directly from CIS attribute tables as values of "area"
items for the respective land cover polygons. Parame-
ters 5 through 7 require calculations relating values of
area items for different polygon coverages. Land surface
coefficients (group 8) could be determined indirectly
based on basic soil, land cover, and topography data.
Most U.S. methodologies use hydrologic soil groups,
222
-------
based on soil permeability, rates of infiltration, and Soil
Conservation Service (SCS) runoff curve numbers (20).
Some Russian methodologies have adopted similar em-
pirical land surface coefficients. For example, for central
European Russia, this value varies from Kt = 0.9 for
plains and sandy soils, to Kt = 1.1 for hills and clay soils
(17).
Level 2 of analysis applies two hydrologic simulation
models:
The Agricultural Watershed Runoff and Water Quality
Model (Agricultural Nonpoint Source Pollution Model
[AGNPS]), developed by the Agricultural Research
Service of the U.S. Department of Agriculture, con-
tains explicit procedures to evaluate the impacts of
management practices and landscape feature posi-
tioning on watershed runoff. AGNPS is a cell-based
runoff model that estimates water volume, peak flow,
eroded and delivered sediment, chemical oxygen de-
mand, and nutrient export from watersheds (20-22).
The Forest Runoff Watershed Model (FRWM) com-
bines analytical and numerical methods for solving
hydro- and thermodynamics equations (23). This
model considers snow melt constituent in runoff in
more detail than does AGNPS. Hydrologic simulation
is based on physical process descriptions for snow
cover dynamics, freezing and thawing of soil, soil
moisture dynamics in frozen and thawed soils, inter-
ception of liquid and solid precipitations by vegeta-
tion, surface runoff, ground-water aquifers, and
channeled streams.
Both models use a similar set of watershed input data,
derived from CIS (e.g., elevations, slopes, channel
slopes, stream network configuration, soil texture, land
cover). The methodology, linking CIS with hydrologic
models, was already tested in the wetland study project
at the Voyageurs National Park in Minnesota. The
ARC/INFO GRID module was used to derive watershed
variables for input to AGNPS. CIS then presented and
interpreted the scenario-based results of the simulation
(24). The typical stages of such an analysis and inter-
pretation for a watershed-scale area are presented in
Figures 1 through 5.
Case Study Watersheds
The areas where wetland impacts on runoff are evalu-
ated are located in Minnesota (United States) and Mos-
cow and adjacent regions (Russia) (see Figures 6
through 15). They have mixed urban, rural, recreational,
and forest land uses. Both regions have a variety of
development pressures. The relative effects of different
alterations in watershed management are distinguished
and quantified. CIS provides metrics for comparative
assessments and analysis of related variables for both
areas.
Table 2 and Figures 7 through 14 present a general
overview of wetland extent in both areas. The case
study subwatersheds used for more detailed analysis
will include tributaries of the second and third order. At
this stage, several watersheds are considered for more
detailed analysis. The limitations imposed by data avail-
ability as well as by project resources could affect the
final selection. Table 2 serves, therefore, as a prelimi-
nary overview of several areas that could potentially be
adopted for more detailed studies.
CIS analysis shows that in the Upper Volga, wetlands
extent very much depends on allocation of populated
areas. The heavily urbanized Moscow metropolitan area
affects a large territory of many thousands of square
kilometers. The ratio of wetlands as a percentage of total
land area is one-tenth of that in the neighboring Tver
area, which has the same size but a smaller population
(see Figure 11). In areas of intensive agriculture (e.g.,
the Pronya basin located southeast of Moscow), almost
all wetlands were drained and have not existed for
several decades.
In Minnesota, despite the growing urbanization (e.g., the
Twin Cities area [7,330 square kilometers]), about half
of the presettlement wetlands still remain (25); wetlands
occupy 442 square kilometers, or over 6 percent of the
land area; and shallow lakes constitute an additional 114
square kilometers (1.56 percent). Some watersheds
within the Twin Cities metropolitan area have a high
wetland percentage, such as 18.9 percent in the Lam-
berts Creek watershed. Intensive studies with CIS ap-
plication of landscape feature functioning and wetland
impacts on stream flow and water quality demonstrated
an innovative approach and made detailed databases
available for this area (15).
Preliminary comparative analysis indicates that two
pairs of case study watersheds could be initially se-
lected for further studies in the Mississippi and Volga
basins:
Upstream watersheds with wetlands area of 15 to 20
percent (Tver region in Russia and Cass and adja-
cent counties in Minnesota).
Tributary watersheds downstream with wetlands area
of 1 to 2 percent (the Istra basin in Russia and sub-
watersheds of the Minnesota River basin, located in
Sibley, Scott, and adjacent counties in Minnesota).
Case study watersheds in the Mississippi and Volga
basins are situated on gently rolling plains in mixed
forest zones with southern portions extending into the
forest/steppe and prairies. The Quaternary sediments
are of glacial, glaciofluvial, lacustrine, and alluvial origin.
Wetlands have hydric soils with various degrees of gley
process development and/or peat accumulation, varied
by wetland type and soil moisture regimen (28). The
annual precipitation is 500 to 600 millimeters with similar
223
-------
Variable
Semiconstant
Constant
Climate
Resource Use
and
Control Practices
Soil and
Physiographic
Factors
t
Weather
Data
i
Hydrological
Runoff
Land Cover
and
Biota
J -
Water
Quality
Soil
Properties
_ _ t
Soil
Quality
Elevation
t
Biological
Wildlife
Habitat
Landscape Feature
Positioning Alternatives
and Recommendations
Improved Methodology for
Spatial Process Analysis
Input
Parameters
Processes
Analysis Based on
Models and CIS
Output Results
Demonstrated
on CIS
Figure 1. Conceptual framework of linking GIS and models for environmental management.
Hypsometric curve
With Ponds
40 -
20 ..
v
340
360 380
Value
400
Outlet
Elevations (meters)
II <350
era 350 to 360
m 360 to 370
m 370 to 380
&B >380
Beaver Ponds
Beaver Wetlands
Wooded Wetlands
Coniferous Forest
Deciduous Forest
0 500 1,000 1,500 2,000
Figure 2. Stream network configuration derived from GIS ele-
vation map.
Figure 3. Scenarios of land use.
224
-------
With Ponds
With Ponds
500 1,000 1,500 2,000
500 1,000 1,500 2,000
Figure 4. Scenario-related hydrologic curve numbers.
Table 2. Comparative Data on Wetland Extent in Minnesota and in the Upper Volga Basin (8, 25-27)
Figure 5. Patterns of sediment transfer between cells (%),
+ deposition, - erosion.
Total Area
Wetland Area
Region
United States
Minnesota
Beltrami Co.
Cass Co.
Hubbard Co.
Le Sueur Co.
Hennepin Co.
Sibley Co.
Wright Co.
Scott Co.
Lambert Creek
Russia
Tver region
Moscow region
Istra basin
Pronya basin
(Square Kilometers)
205,940.30
7,923.04
6,256.38
2,624.81
1 ,204.82
1 ,588.07
1 ,555.38
1 ,852.97
982.56
9.51
10,000.00
10,000.00
1 ,827.38
10,200.00
(Square Kilometers)
30,500.00
3,909.23
1,505.29
283.22
28.31
36.37
24.26
24.27
8.06
3.69
1,169.08
165.39
24.07
a
Percentage
14.80
49.34
24.06
10.79
2.35
2.29
1.56
1.31
0.82
18.90
16.90
1.70
1.32
a
Wetland area is insignificant and not identified by available maps.
225
-------
1 - Doibiza, 2 - Istra, 3 - Pronya
Figure 6. Location of study areas in the Volga basin.
Volga Basin, Russia
Land Features
Urban Lands
H Peat Bogs
a Wetlands
£±J
M Reservoirs and Lakes
il
/J Rivers
71 Roads
Statistics %
Total Area 100.00
Urban 12.38
Water 1.58
Wetlands 1.66
Kilometers
10
20 30
40
Figure 7. Wetlands and urban lands in the Moscow region.
226
-------
Land Features
Urban Lands
1 Peat Bogs
I Wetlands
&J
H Inundated
^
I] Reservoirs and Lakes
7] Rivers
?] Roads
Statistics %
Total Area 100.00
Urban 1.22
Water 4.64
Wetlands 16.93
Kilometers
10 20 30 40
Figure 8. Wetlands and urban lands in the Tver region.
Wetlands, 1954
1 Wetlands, 1980s
Urban Lands
Reservoirs and Lakes
30
40
Figure 9. Wetland decline since 1954, Moscow region.
227
-------
Volga Basin, Russia
£59
Land Features
Urban Lands
(Population, Thousands)
Peat Bogs
Wetlands
Reservoirs and Lakes
Rivers
Roads
£| Watershed
Elevation (meters)
Square
Kilometers %
Total Area 1,827.38 100.00
Urban 61.31 3.36
Water 40.43 2.21
Wetlands 24.07 1.32
Kilometers
10
20
30
40
Figure 10. Istra watershed in the Moscow region.
Wetlands Moscow Region, Russia
Populated Areas
t ] <10 EH2 20-29 m 40-49 ^ffl 75-99
10-19 [T-T] 30-39 50-74 100
Kilometers
50 100 150 200
Figure 11. Land cover, percentage of total in Moscow and Tver regions.
228
-------
Presettlement
Current
Wetland Drop, Percentage
and Location of Study Areas
0-10
11-20
21-30
31-40
41-50
51-60
61-70
>70
Kilometers
100 200 300 400
Figure 12. Wetlands in Minnesota, percentage of total area.
Mississippi Basin, Minnesota
Land Features
H Urban Lands
j Forests
H Wetlands
3S55S
H Reservoirs and Lakes
2 Rivers |/\/| Roads
2| Watershed |/'v'| Counties
Statistics
Total Area
Urban
Water
Wetlands*
100.00
0.04
14.5
18-24
'Indicated by County, %
Kilometers
10
20
30
40
Figure 13. Wetlands and urban lands in Cass County area.
229
-------
Mississippi Basin, Minnesota
a.
Land Features
1 Urban Lands
] Forests
Wetlands
Reservoirs and Lakes
Rivers I/S/I Roads
Watershed |/ V' | Counties
Statistics %
Total Area 100.00
Urban 9.0
Water 4.5
Wetlands* 1-2
Kilometers
'I -'! I I
10 20 30 40
Figure 14. Wetlands and urban lands in Twin Cities area.
Mississippi Basin, Minnesota
Major Watershed I/ v I Clean Streams
/ \''| Minor Watersheds L-'Vl Impaired Quality
Urban
Lakes
Sampling Sites
Kilometers
i
10
20
30
40
Figure 15. Minnesota River watershed in Twin Cities area.
230
-------
seasonal distribution in both areas. Average runoff
ranges from 150 to 250 millimeters (7, 29).
Conclusion
The current status of the project indicates that most of
the input data is available, though dispersed among
many agencies. In both the United States and Russia,
research methodologies have been developed and ap-
plied to study landscape feature impacts on runoff quan-
tity and quality based on simulation and statistical
analysis. The comparative analysis of hydrologic and
diffuse pollution processes on watersheds in the Upper
Mississippi and Upper Volga basins will allow derivation
of metrics of wetland loss relative to impacts on runoff
and water quality.
The applications of CIS to watershed hydrology are
currently much more advanced in the United States than
in Russia. Initiatives emerging in the United States,
however, could considerably promote CIS use in Rus-
sia. Such promotion is beneficial for several reasons.
First, this kind of cooperative political activity is in full
compliance with the 1992 Freedom Support Act, ap-
proved by the U.S. Congress. Second, support of CIS
as a new information technology will create a favorable
infrastructure in many bilateral economic fields and busi-
nesses. Third, a better meshing of the CIS systems in
the two countries will lead to further international coop-
eration in responding to global changes.
Project implementation also helps meet the goal of pro-
viding a basis for sound environmental, technical, and
economic decision-making on the use of natural re-
sources. This knowledge is essential in developing prac-
tical guidelines for sustainable economic development
through applied research and technologies.
Acknowledgments
The Water Quality Division of MPCA provided valuable
assistance with CIS data for Minnesota. The National
Science Foundation (NSF/EAR-9404701) contributed
research support. Any opinions, findings, and conclu-
sions or recommendations expressed in this material
are those of the author and do not necessarily reflect the
views of the National Science Foundation.
References
1. Wetland management: Hydraulic and hydrologic research needs.
1987. In: Wetland hydrology: Proceedings of the National Wet-
land Symposium, Chicago, IL (September 16-18). Association of
State Wetland Managers, Inc. pp. 331-333.
2. Sather, J.H. 1992. Intensive studies of wetland functions: 1990-
1991 research summary of the Des Plaines River Wetland dem-
onstration project. Technical Paper No. 2. Wetland Research, Inc.
3. Hey, D.L. 1985. Wetlands: A strategic national resource. National
Wetlands Newsletter 7(1): 1-2.
4. Kloet, L. 1971. Effects of drainage on runoff and flooding within
the Pembina River basins, North Dakota and Manitoba. Unpub-
lished report by the Bureau of Sport, Fisheries, and Wildlife (De-
cember).
5. Simon, B.D., L.J. Stoerzer, and R.W. Watson. 1987. Evaluating
wetlands for flood storage. In: Wetland hydrology: Proceedings
of the National Wetland Symposium, Chicago, IL (September
16-18). Association of State Wetland Managers, Inc. pp. 104-112.
6. Johnston, C.A. 1994. Cumulative impacts to wetlands. Wetlands
14(1):49-55.
7. Higgins J.M., T.B. Nawrocki, and N.A. Nazarov. 1993. Hierarchi-
cal approach to integrated watershed management: Joint
TVA/Russian demonstration project. Proceedings of AWWA
CONSERV93 Conference, Sessions 4B-1 through 7C-3, Las
Vegas, NV. pp. 1,177-1,197.
8. Environmental Science Research Institute. 1993. Digital chart of
the world.
9. Vogelmann, J.E., F.R. Rubin, and D.G. Justice. 1991. Use of
Landsat thematic mapper data for fresh water wetlands detection
in the Merrimack River watershed, New Hampshire (unpub-
lished).
10. Conger, D.H. 1971. Estimating magnitude and frequency of
floods in Wisconsin. Open File Report. Madison, Wl: U.S. Geo-
logical Survey.
11. Jacques, J.E., and D.L. Lorenz. 1988. Techniques for estimating
the magnitude and frequency of floods in Minnesota. Water Re-
sources Investigation Report 87-4170. St. Paul, MN: U.S. Geo-
logical Survey.
12. Novitzki, R.P. 1979. Hydrologic characteristics of Wisconsin's
wetlands and their influence on floods, stream flow, and sedi-
ment. In: Greeson, P.E., J.R. Clark, and J.E. Clark, eds. Wetland
functions and values: The state of our understanding. Minneapo-
lis, MN: American Water Resources Association, pp. 377-388.
13. Larson, L.A. 1985. Wetlands and flooding: Assessing hydrologic
functions. Proceedings of the National Wetland Assessment Sym-
posium, Portland, ME (June 17-20). pp. 43-45.
14. Ogawa, H., and J.W Male. 1986. Simulating the flood mitigation
role of wetlands. J. Water Res. Planning and Mgmt. 112(1).
15. Johnston, C.A., N.E. Detenbeck, and G.J. Niemi. 1990. The cu-
mulative effect of wetlands on stream water quality and quantity:
A landscape approach. Biogeochemistry 10:105-141.
16. Moore, I.D., and C.L. Larson. 1979. Effects of drainage projects
on surface runoff from small depressional watersheds in the
North Central Region. WRRC Bulletin 99. Water Resources Re-
search Center, University of Minnesota.
17. Maslov, B.S., I.V. Minaev, and K.V. Guber. 1989. Reclamation
manual. Rosagropromizdat (in Russian).
18. Oberts, G.L. 1981. Impacts of wetlands on watershed water qual-
ity. In: Richardson, B., ed. Selected proceedings of the midwest
conference on wetland values and management. Navarre, MN:
Freshwater Society, pp. 213-226.
19. Horton, R.E. 1945. Erosion development of streams. Geol. Soc.
Amer. Bull. 56:281-283.
20. Chow, V. 1964. Handbook of applied hydrology. New York, NY:
McGraw-Hill.
21. Wishmeier W.H., and D.D. Smith. 1978. Predicting rainfall erosion
losses: A guide to conservation planning. Agriculture Handbook
No. 537. U.S. Department of Agriculture.
231
-------
22. Young, R.A., C.A. Onstad, D.D. Bosch, and W.P. Anderson. 1987.
AGNPS: Agricultural nonpoint-source pollution model. U.S. De-
partment of Agriculture Conservation Research Report 35.
23. Nazarov, N.A. 1988. Model formation of the flood hydrograph of
Northern Plain Rivers. Water Resour. 15(4):305-315.
24. Nawrocki, T., C. Johnston, and J. Sales. 1994. CIS and modeling
in ecological studies: Analysis of Beaver Pond impacts on runoff
and its quality. Voyageurs National Park, Minnesota, case study.
NRRI Technical Report 94/01 (February).
25. Anderson, J.P., and W.J. Craig. 1984. Growing energy crops on
Minnesota wetlands: The land use perspective. Minneapolis, MN:
Center for Urban and Regional Affairs, University of Minnesota.
pp. 1-88.
26. Mitsch, W.J., and J.G. Gosselink. 1986. Wetlands. New York, NY:
Van Nostrand Reinhold.
27. Brown, R.G. 1987. Effects of wetland channelization on storm
water runoff in Lamberts Creek, Ramsey County, Minnesota. In:
Proceedings of the National Wetland Symposium on Wetland
Hydrology, Chicago, IL (September 16-18). pp. 130-136.
28. SCS. 1975. Soil taxonomy: A basic system of soil classification
for making and interpreting soil surveys. Agriculture Handbook
No. 436. Soil Conservation Service, U.S. Department of Agriculture.
29. USGS. 1988. Water resources data: Minnesota, water year 1988.
Vol. 2, Upper Mississippi and Missouri River basins. U.S. Geo-
logical Survey Data MN-88-1.
232
-------
Water Quality Applications
-------
Vulnerability Assessment of Missouri Drinking Water to Chemical Contamination
Christopher J. Barnett, Steven J. Vance, and Christopher L. Fulcher
Center for Agricultural, Resource, and Environmental Systems, University of Missouri,
Columbus, Missouri
Introduction
In 1991, the Missouri Department of Natural Resources
(MDNR) implemented the Vulnerability Assessment of
Missouri Drinking Water to Chemical Contamination pro-
ject. MDNR's Public Drinking Water Program (PDWP)
contracted with the Center for Agricultural, Resource,
and Environmental Systems (CARES) to conduct this
assessment. They designed the project to determine
which, if any, public water supplies are threatened by
chemicals being tested under the Safe Drinking Water Act.
Under Phase II of the Safe Drinking Water Act, the
United States Environmental Protection Agency (EPA)
required that all public drinking water systems be rou-
tinely monitored for 79 contaminants beginning January
1,1993. If a selected chemical parameter is not detected
in an area that would affect a water supply (where
"detected" is defined as used, stored, manufactured,
disposed of, or transported regardless of amount), then
the water supply need not be tested for that chemical.
Instead, that system would be granted a use waiver,
meaning that the state would not test for that chemical.
EPA grants use waivers for 43 of the 79 contaminants.
Use waivers can result in considerable cost savings.
Because use waivers are granted based on the spatial
relationship between drinking water sources and con-
taminant sources, accurate positional data needed to be
collected for those items. A geographic information sys-
tem (CIS) was used to store and analyze this informa-
tion in a spatial context.
Water Sources
Water sources, as defined for this study, are the points
where water is drawn from a river, lake, or aquifer for
use in a public water supply. Our efforts focused primar-
ily on the development of the water source layers for the
CIS. These layers, containing wellheads, impoundment
intakes, and river intakes, were created in house or
obtained from state and federal agencies. MDNR
regional office personnel inspected these water source
layers in the spring of 1993. Since these personnel
routinely inspect Missouri public drinking water supplies,
their knowledge of these locations is exceptional.
The updated water source information was mapped
on 1:24,000-scale USGS topographic quadrangles at
the regional offices, then entered into the CIS. MDNR's
PDWP provided available attribute information, which
was associated with these layers. The layers offer the
most accurate and current information available. Only
the community (e.g., cities, subdivisions, mobile home
parks) and the nontransient, noncommunity (e.g.,
schools, large businesses) water supply systems were
considered for water source mapping. This study did not
consider private wells.
The information is stored in the CIS in the form of
geographic data sets or layers. The wellhead layer con-
tains 2,327 public wells and their attributes (e.g., well
depth, casing type). The majority of the wellheads are
located in the Ozarks and Southeast Lowlands. Natu-
rally poor ground-water quality prohibits a heavy reli-
ance on ground water for drinking water in other areas
of the state. The surface water impoundment layer con-
tains 105 points representing the intake locations for
systems that rely on lake water. Additionally, the drain-
age basin and lake area are mapped for these systems.
The majority of the systems that rely on lake water are
located in northern and western Missouri. The final layer
represents the systems that use river water. The major-
ity of the 50 intakes are located on the Mississippi and
Missouri Rivers and on the major streams in the Grand
and Osage River basins.
Contaminant Sources
Contaminant sources, as defined for this study, are the
points or areas where existing databases indicate the
presence of a chemical contaminant. Incorporation of
contaminant data into the CIS proved to be the most
difficult task. These data usually contained very precise
235
-------
information about what contaminants were found at a
site and who was responsible, but the quality of the
locational information was often poor.
Ninety-three state and federal databases were reviewed
for contaminant information before performing the final
use waiver analysis. The contaminant information was
broken into two separate types, contaminant sites and
pesticide dealerships. The contaminant sites were loca-
tions at which certain chemicals were known to exist.
The pesticide dealerships were dealerships licensed to
distribute restricted use pesticides. Information about
contaminant sites was extracted from the databases
and entered into Microsoft Excel, a spreadsheet pro-
gram. The small amount of data with coordinate (lati-
tude/longitude) or map information was readily
converted to the CIS. The majority of the contaminant
records, however, contained only address information,
often appearing as a rural route address or post office
box number.
While the water source locations were being verified,
personnel at the MDNR regional offices reviewed the
contaminant site records. The regional office personnel
were familiar with their respective territories and could
assist CARES personnel in locating the contaminant
sites. The Missouri Department of Agriculture pesticide
use investigators provided additional information about
the locations of contaminant sites. All contaminant
source information was also mapped on the 1:24,000-
scale USGS topographic quadrangles and transferred
to the CIS.
Of more than 2,800 contaminant sites found in these
databases, 88 percent were geographically located and
used in the study. At this time, the contaminant site layer
contains 2,493 points representing the information col-
lected on the 43 chemical contaminants required by
MDNR. Each point contains a seven-digit chemical code
indicating the chemical it represents and serving as a
link to the chemical contaminant files. The contaminant
sites tend to be concentrated more in urban areas than
rural areas. Even though this layer is being continually
updated, the basic distribution of contaminant sites re-
mains the same.
A second contaminant source layer represents Mis-
souri's licensed pesticide dealers. This information is
included to indicate potential contamination even though
specific chemicals at dealership locations are not
known. At this time, we have been able to locate 1,344
dealerships out of 1,650. Two types of dealerships are
included in the layer, active dealers and inactive dealers.
Of the active dealerships in 1991, 91 percent were found
and entered into the CIS. Of the inactive dealerships, 79
percent were located.
Spatial Analysis
The final parameters for the use waiver analysis were
developed from EPA and MDNR guidelines and account
for the capabilities of the CIS. These parameters were
designed to present a conservative list of the systems
that needed to be tested for the possible presence of
studied chemicals. Parameters forthe wellhead analysis
are as follows:
A 1/4-, 1/2-, and 1-mile radius around each wellhead
was searched for contaminant sites and pesticide
dealerships (see Figure 1). Any contaminant sources
found within those radii were reported to PDWP.
(PDWP requested that the results of the three radius
analyses be reported, but the 1/2-mile radius was
used to determine the issue of the use waiver.)
Any wellheads found within a contaminant area were
denied a use waiver for that contaminant.
Each highway and railroad within 500 feet of a well-
head was recorded. This indicates the threat posed
by the transport of chemicals near wellheads.
Additionally, the percentage of the county planted in
corn, soybeans, wheat, sorghum, tobacco, cotton,
and rice was listed for each well to indicate the threat
posed by agricultural chemical use within that county.
The parameters for the systems relying on lake water
are as follows:
Any contaminant sources found within a surface
water impoundment drainage basin caused the asso-
ciated intake(s) to fail use waiver analysis for those
contaminants.
X = Contaminant
Figure 1. Use waiver search radius distances.
236
-------
Any area of contamination overlapping a drainage
basin caused the associated intake to fail use waiver
analysis for that contaminant.
Transportation corridors passing through a drainage
basin were noted to indicate the threat posed by
transport of chemicals within the basin.
The percentage of the county planted in the seven
crops mentioned above was listed to indicate agricul-
tural chemical use within the drainage basin.
Many of the rivers that supply water to systems in Mis-
souri have their headwaters outside the state. To fully
evaluate the potential for contamination within those
drainage basins, we would have to collect data for large
areas outside of the state. For example, the Mississippi
and Missouri River drainage basins cover large portions
of the United States. Because collecting data for those
areas would be impractical, we have recommended to
MDNR that use waivers not be granted to river supplies.
The following provides details on how the analysis was
performed. The CIS searches around each wellhead for
each radius and notes which contaminant sites affect
which wellheads. If a contaminant falls within that ra-
dius, we recommend that the wellhead be monitored. In
this example, the well is affected by one contaminant
within the 1/4-mile radius, two within the 1/2-mile radius,
and four within the 1-mile radius.
Results
The results of the use waiver analysis indicate which
systems may be affected by the use of a chemical near
a water source. Several results show the substantial
savings realized from our analysis. For example, the
analysis showed that only five wells serving four public
drinking water systems were potentially affected by di-
oxin and should be monitored. By not testing the remain-
ing systems for dioxin, the state can realize a
considerable cost savings, as the test for dioxin is the
most expensive test to perform.
The final wellhead system analysis shows that the
1/2-mile buffer analysis affected a total of 447 well-
heads in 241 systems. That is, a chemical site or pesti-
cide dealership was found within 1/2 mile of 447 public
wellheads. A result form was generated for each of the
1,340 systems in the state listing each well or intake and
the potential threat posed by nearby contaminant
sources.
The cost of testing all wellhead systems for all 43 con-
taminants without issuing use waivers is more than $15
Table 1. Estimated Cost Savings for Public Drinking Water
Systems
Method
Estimated
Total Cost
Estimated
Mean Cost
per System
Estimated
Total Cost
Savings
No use waiver $15,533,100 $12,200 $0
With use waiver $1,813,900 $1,400 $13,719,200
million (see Table 1). According to our analysis, CARES
estimates that only $1.8 million need be spent to monitor
vulnerable wells. Therefore, the state can save more
than $13.5 million in monitoring costs.
Summary and Recommendations
To date, the investment the state made in the vulnerabil-
ity assessment project has provided many benefits. The
state saved several million dollars in testing costs and
developed several spatial and nonspatial databases that
will have many uses. In addition, the project established
a basic framework for future assessments, which EPA
requires on a regular basis.
The basic data required for use waiver analysis are the
locations of water sources and the locations of potential
contamination sources. CARES determined that the
available data did not contain the information necessary
to map these locations or that the data were of question-
able quality. Many layers required update and correc-
tion. Considerable effort was necessary to improve
existing locational information for both water source lay-
ers and chemical contaminant files. Local knowledge of
an area was heavily relied upon to determine accurate
locations, particularly contaminant sites. The vast ma-
jority of these sites contained only the address as the
geographic reference. An address is not a coordinate
system; it does not indicate a fixed location on a map.
Because the location of any chemical detection site is
of vital importance, state and federal agencies that col-
lect these data need to record more complete geo-
graphic information. Ideally, a global positioning system
could be employed to generate coordinates. Realisti-
cally, the recording of legal descriptions or directions
from an easily located point would substantially improve
the quality of the current databases.
In many cases, data resided in digital format; however,
due to regulations or lack of agency cooperation, they
could only be distributed in paper format. Reentering
data from paper format into digital format required con-
siderable time and expense. Interagency cooperation
should be emphasized to reduce unnecessary data entry.
237
-------
Reach File 3 Hydrologic Network and the Development of GIS Water Quality Tools
Stephen Bevington
Water Quality Section, Division of Environmental Management, North Carolina Department of
Environment, Health, and Natural Resources, Raleigh, North Carolina
Introduction
The application of geographic information system (GIS)
tools to water quality management is limited by the lack
of geographically referenced data describing the surface
water environment. Ongoing efforts at the local, state,
and federal level are producing a multitude of GIS data
coverages describing land use/cover and relevant water
quality data files. As these data coverages become
available, water quality managers will need to develop
new analysis techniques to take advantage of the vast
amount of geographically referenced data. A key step in
the development of analytical tools for water quality
management will be the development and maintenance
of a coverage describing the structure and hydrology of
surface waters.
Reach File 3 (RF3) is one potential source of surface
water maps and topology for the development of a
CIS-based water quality analysis tool. This paper de-
scribes a pilot project designed to examine the suitability
of RF3 as a network system for the collection, integra-
tion, and analysis of water quality data.
To be considered an appropriate water quality analysis
tool, RF3 should provide the following functions:
Present a working environment that allows users to
explore geographic relationships between surface water
features, landmark features, and data coverages.
Allow users to select specific stream segments, in-
cluding all points upstream and downstream of a
given point.
Provide tools to assist users in partitioning water quality
databases into hydrologically meaningful subsets.
Reach File 3
RF3 is a hydrographic database of the surface waters
of the United States. The database contains 3 million
river reaches mapped at 1:100,000 scale. The source
for RF3 arcs were digital line graphs (DLGs).
Attribute data for RF3 arcs include the major-minor DIG
pairs, stream name, water-body type, stream order, and
a unique identifying reach number. The unique reach
numbers are structured in such a way as to provide a
logical hydrologic framework. Reach numbers can be
used to sort the database for all reaches in any specified
watershed or locate all upstream or downstream
reaches.
The U.S. Environmental Protection Agency (EPA) origi-
nally designed RF3 as a tabular data set. It evolved into
a GIS data coverage, and EPA and the U.S. Geological
Survey (USGS) will likely maintain it as a surface water
mapping standard. At present, RF3 as a GIS data layer
is not widely used for water quality applications.
RF3 Pilot Study: Upper Yadkin River Basin
The Upper Yadkin River basin (USGS h03040101) was
selected to test RF3 water quality applications (see
Figures 1 and 2). The Upper Yadkin was chosen be-
cause of the availability of water quality and stream flow
data layers in that area. Also, the Upper Yadkin RF3 file
contained arcs depicting lakes and double-line rivers as
well as simple stream networks. These two-dimensional
water features present interesting complications to net-
work routing and path-finding.
Figure 1. The Upper Yadkin River watershed, North Carolina
and Virginia.
238
-------
Figure 2. RF3 hydrography for the Upper Yadkin River basin.
Two forms of point source data were used in the study:
National Pollutant Discharge Elimination System
(NPDES) wastewater discharge points and USGS
gages. The NPDES coverage includes data on the per-
mit limits such as daily flow, dissolved oxygen, bio-
chemical oxygen demand (BOD), and ammonia. The
USGS gage coverage includes data on several flow
statistics for each USGS gage in the basin. Both data
layers contain information about the location of the site
and stream with which it is associated.
Coverages of counties and cities were also made avail-
able for geographic orientation.
Preparing the Network
The original RF3 file received from the USGS had sev-
eral topological issues that needed to be addressed
before RF3 could function as a stream network. First,
not all arcs were connected to each other (see Figure 3).
The ARC/INFO command TRACE was used to select all
connected arcs. This revealed three major blocks of
connected arcs and many isolated arcs. The three major
blocks were easily connected in ARCEDIT by extending
the main tributary links between the blocks. Processing
of the isolated arcs was not pursued for this study.
Complete processing of arcs for this RF3 basin would
not be difficult or time consuming, with the possible
exception of the many arcs surrounding the lake. A
functional network encompassing a high percentage of
the arcs was not difficult to achieve, however.
The second network issue concerned the direction of the
arcs. RF3 has all arcs oriented toward the top of the
watershed, with the exception of one side of double-line
streams. Arcs that make up double-line streams are
oriented up one side of the double-line section and down
the other (see Figure 3). Clearly, this complicates rout-
ing. To allow for accurate downstream routing, arcs on
the downward-facing side of the stream were flipped
using ARCEDIT. With all arcs in the network facing
upstream, most hydrologic routes can be traced. Given
the network system alone, upstream routing from dou-
ble-line streams does not function properly, ignoring all
tributaries on one side of the double-line stream.
Double-Line Stream Routing
Many possible solutions exist for the problems caused
by double-line streams. Some involve improving the
network (e.g., by adding center-line arcs down the mid-
dle of double-line streams). This would involve not only
adding arcs but establishing conductivity with all tribu-
taries. This option will involve significant topological
changes to RF3. To maintain compatibility with other
239
-------
VTH ' rf^^^^VW^
^mt-i^^'
V-Jv- -'>'''(-." ' J
' " --XVS/'
Figure 3. Original conductivity of RF3 hydrography.
RF3 and DIG sources, this option should be considered
only as part of a major RF3 upgrade.
At the other end of the technological spectrum, one
could simply instruct users to watch for double-line
streams and select arcs from both sides of the river.
Users may have trouble with this option, however, if they
are not working at an appropriate scale to easily differ-
entiate between double- and single-line streams.
A third option is to program an arc macro language
(AMI) to check for double-line streams and run up-
stream traces from both sides of the stream. The diffi-
culty in this method is to find the appropriate starting
place on both banks. The algorithm developed to do this
goes as follows:
Select stream segment and trace upstream. (Results
in incomplete trace.)
Find the minimum segment and mile of selected dou-
ble-line streams.
Unselect all double-line streams below minimum seg-
ment and mile.
Add to selection all non-double-line streams.
Trace from original point both upstream and down-
stream. (Results in completed upstream trace.)
Results and Conclusions
AMLs and menus were written that can perform up-
stream and downstream traces on the RF3 stream net-
work and select data points within 500 feet of the
stream. Lists of attributes can be returned to the screen.
This system is easy to use and can be used to quickly
identify general watersheds and water quality data
points. An AMI can be used to trace upstream from a
double-line stream given only one point on the stream
(see Figure 4). The success of these methods suggests
that two-dimensional surface water features can be suc-
cessfully integrated into RF3 water quality analyses.
This system could be further developed to support poly-
gon analysis using the ARC command BUFFER. Other
developments could include the procedures to write se-
lected attributes to files and increased flexibility for the
screen environment and outputs.
This pilot project demonstrates only a few of the poten-
tial applications of RF3 to water quality. Success in this
pilot project suggests that RF3 is a potentially valuable
water quality analysis tool. It may also be a valuable tool
for demonstrating the results of water quality analyses
to managers or the public.
Because RF3 will require some processing before network
algorithms can be run, it is important to plan for the inte-
gration of RF3 into other CIS tools and data coverages.
240
-------
Figure 4. Upstream and downstream traces of RF3 hydrography.
Ongoing efforts to update RF3 may address some of proceed in a way that is compatible with ongoing efforts
these problems. If RF3 is to be developed into a produc- to update RF3 and the development of new data
tive water quality management tool, it is important to sources.
241
-------
EPA's Reach Indexing ProjectUsing GIS To Improve Water Quality Assessment
Jack Clifford
Office of Wetlands, Oceans, and Watersheds, U.S. Environmental Protection Agency,
Washington, DC
William D. Wheaton and Ross J. Curry
Research Triangle Institute, Research Triangle Park, North Carolina
Abstract
The Waterbody System (WBS), which the U.S. Environ-
mental Protection Agency (EPA) originally developed to
support preparation of the report to Congress that Sec-
tion 305(b) of the Clean Water Act requires, is a poten-
tially significant source of information on the use support
status and the causes and sources of impairment of U.S.
waters. Demand is growing for geographically refer-
enced water quality assessment data for use in inter-
agency data integration, joint analysis of environmental
problems, establishing program priorities, and planning
and management of water quality on an ecosystem or
watershed basis.
Because location of the waterbody assessment units is
key to analyzing their spatial relationships, EPA has
particularly emphasized anchoring water bodies to the
River Reach File (RF3). The reach file provides a nation-
wide database of hydrologically linked stream reaches and
unique reach identifiers, based on the 1:100,000 U.S.
Geological Survey (USGS) hydrography layer.
EPA began the reach indexing project to give states an
incentive to link their water bodies to RF3 and to ensure
increased consistency in the approaches to reach index-
ing. After a successful 1992 pilot effort in South Carolina,
an expanded program began this year. Working with
Virginia, a route system data model was developed and
proved successful in conjunction with state use of PC
Reach File (PCRF), a PC program that relates water
bodies to the reach file. ARC/INFO provides an extensive
set of commands and tools for developing and analyzing
route systems and for using dynamic segmentation.
One important advantage of the route system is that it
avoids the necessity of breaking arcs; this is an impor-
tant consideration in using RF3 as the base coverage in
a geographic information system (GIS). Using dynamic
segmentation to organize, display, and analyze water
quality assessment information also simplifies use of the
existing waterbody system data. Because of the variabil-
ity in delineation of water bodies, however, other states
used a number of different approaches. Working with
these states has defined a range of issues that must be
addressed in developing a consistent set of locational
features for geospatial analysis.
Wider use of these data also depends upon increased
consistency in waterbody assessments within and be-
tween states. Several factors complicate the goal of
attaining this consistency in assessment data:
The choice of beneficial use as the base for assess-
ment of water quality condition.
The historical emphasis on providing flexible tools
to states.
The lack of robust standards for assessment of water
quality condition.
This paper explores possible resolutions to the problem
of building a national database from data collected by
independent entities.
Section 305(b) of the Clean Water Act and
the Waterbody System
Background of Section 305(b)
Since 1975, Section 305(b) of the Federal Water Pollution
Act, commonly known as the Clean Water Act (CWA),
has required states to submit a report on the quality of
their waters to the U.S. Environmental Protection Agency
(EPA) administrator every 2 years. The administrator
must transmit these reports, along with an analysis of
them, to Congress.
State assessments are based on the extent to which the
waters meet state water quality standards as measured
242
-------
against the state's designated beneficial uses. For each
use, the state establishes a set of water quality criteria
or requirements that must be met if the use is to be
realized. The CWA provides the primary authority to
states to set their own standards but requires that all
state beneficial uses and their criteria comply with the
'fishable and swimmable' goals of the CWA.
Assessments and the Role of Guidelines
EPA issues guidelines to coordinate state assessments,
standardize assessment methods and terminology, and
encourage states to assess support of specific benefi-
cial uses (e.g., aquatic life support, drinking water sup-
ply, primary contact recreation, fish consumption). For
each use, EPA asks that the state categorize its assess-
ment of use support into five classes:
Fully supporting: meets designated use criteria.
Threatened: may not support uses in the future unless
action is taken.
Partially supporting: fails to meet designated use cri-
teria at times.
Not supporting: frequently fails to meet designated
use criteria.
Not attainable: use support not achievable.
In the preferred assessment method, the state com-
pares monitoring data with numeric criteria for each
designated use. If monitoring data are not available,
however, the state may use qualitative information to
determine use support levels.
In cases of impaired use support (partially or not sup-
porting), the state lists the sources (e.g., municipal point
source, agriculture, combined sewer overflows) and
causes (e.g., nutrients, pesticides, metals) of the use
support problems. Not all impaired waters are charac-
terized. Determining specific sources and causes re-
quires data that frequently are not available.
States generally do not assess all of their waters each
biennium. Most states assess a subset of their total
waters every 2 years. A state's perception of its greatest
water quality problems frequently determines this sub-
set. To this extent, assessments are skewed toward
waters with the most pollution and may, if viewed as
representative of overall water quality, overstate pollu-
tion problems.
Assessment Data Characteristics
Each state determines use support for its own set of
beneficial uses. Despite EPA's encouragement to use
standardized use categories, the wide variation in state-
designated beneficial uses makes comparing state uses
an inherent problem. This affects the validity of aggre-
gation and use of data across state boundaries. Com-
parably categorizing waters into use support categories
also poses a problem; different states apply the qualita-
tive criteria for use support levels in very different ways.
Further limiting the utility of Section 305(b) data is that
data are aggregated at the state level and questions
about the use support status of individual streams can-
not be resolved without additional information. While
some states report on individual waters in their Section
305(b) reports, EPA's Waterbody System (WBS) is
the primary database for assessment information on
specific waters.
State monitoring and assessment activities are also
highly variable. States base assessments on monitoring
data or more subjective evaluation. The evaluation cate-
gory particularly differs among states.
Waterbody System
The WBS is a database and a set of analytical tools for
collecting, querying, and reporting on state 305(b) infor-
mation. It includes information on use support and the
causes and sources of impairment for water bodies,
identification and locational information, and a variety of
other program status information.
As pointed out earlier, although some states discuss the
status of specific waters in their 305(b) reports, many do
not. The WBS is generally much more specific than the
305(b) reports. It provides the basic assessment infor-
mation to track the status of individual waters in time
and, if georeferenced, to locate assessment information
in space. By allowing the integration of water quality
data with other related data, the WBS provides a frame-
work for improving assessments.
WBS has significant potential for management planning
and priority setting and can serve as the foundation for
watershed- and ecosystem-based analysis, planning,
and management. In this respect, it can play a vital role
in setting up watershed-based permitting of point
sources. The primary function of WBS is to define where
our water quality problems do and do not exist. WBS is
increasingly used to meet the identification requirement
for waters requiring a total maximum daily load (TMDL)
allocation. It can serve as the initial step in the detailed
allocation analysis included in the TMDL process. In
addition, WBS is an important component of EPA par-
ticipation in joint studies and analyses. For instance,
EPA is currently participating with the Soil Conservation
Service (SCS) in a joint project to identify waters that
are impaired due to agricultural nonpoint source (NPS)
pollution. WBS can also anchor efforts to provide im-
proved public access at the state and national levels to
information on the status of their waters.
It is important to recognize that use of WBS is voluntary.
Of the 54 states, territories, river basin commissions, and
Indian tribes that submitted 305(b) reports, approximately
243
-------
30 used the WBS in the 1992 cycle. While submissions
for the 1994 cycle are not complete, we anticipate about
the same level of participation. This represents about a
60-percent rate of participation in WBS, which may be
the limit for a voluntary system. This severely limits use
of WBS assessment data for regional and national level
analysis. If data at the national level are required, man-
datory data elements, formats, and standards may be
necessary.
EPA is currently attempting to achieve consistency
through agreement with other state and federal agen-
cies. The recent work of the Interagency Task Force on
Monitoring offers hope for eventual consensus on the
need for nationally consistent assessment data and mu-
tually agreed upon standards for collection, storage, and
transfer. The Spatial Data Transfer Standards already
govern spatial data, allowing movement of data between
dissimilar platforms. The Federal Geographic Data
Committee provides leadership in coalescing data inte-
gration at the federal level; it provides a model for gov-
ernment and private sector efforts. This level of
cooperation, however, has not always been present in
water assessment data management. Assuming that
national and regional assessment data are needed, if
consistency is elusive through cooperative efforts, regu-
lations may be necessary. Developing a national data-
base may not be feasible without a mutual commitment
by EPA and the states to using common assessment
standards.
WBS was originally developed as a dBASE program in
1987. It has undergone several revisions since then, and
the current Version 3.1 is written in Foxpro 2.0. The
WBS software provides standard data entry, edit, query,
and report generation functions. WBS has grown sub-
stantially in the years since its inception, primarily in
response to the expressed needs of WBS users and
EPA program offices. The program's memory require-
ments and the size of the program and data files, how-
ever, are of growing concern to state WBS users and
the WBS program manager. Because of the wide range
of WBS user capabilities and equipment, users must be
equipped to support an array of hardware from high
capacity Pentium computers to rudimentary 286 ma-
chines with 640 Kb of memory and small hard disks. This
range makes memory problems inevitable for some users.
While WBS contains over 208 fields, exclusive of those
in lookup tables, approximately 30 fields in four files
provide the core data needed to comply with 305(b)
requirements. These fields contain:
Identification information for the water body.
The date the assessment was completed.
The status of use support for beneficial uses.
The causes and sources of any use impairment in
the water body.
The uses WBS considers are both state-designated
uses and a set of nationally consistent uses (e.g., overall
use, aquatic life support, recreation) specified in the
305(b) guidelines. The other essential piece of informa-
tion is the geographic location of the water body, which
the remainder of this paper discusses in detail.
Significant differences exist in the analytical base as well
as in assessments. EPA provided little initial guidance
on defining water bodies; therefore, states vary widely
in their configurations of water bodies. Water bodies are
supposed to represent waters of relatively homogene-
ous water quality conditions, but state interpretation of
this guidance has resulted in major differences in water-
body definition.
Initially, many states developed linear water bodies, and
these were often very small. The large number of water
bodies delineated, however, created significant difficul-
ties in managing the assessment workload and were not
ideal in the context of the growing need for watershed
information. Some states, such as Ohio, developed their
own river mile systems.
As discussed below, some states indexed their water
bodies to earlier versions of the reach file, and therefore,
the density of the streams these water bodies include is
fairly sparse. Recently, many states have redefined
their water bodies on the basis of small watersheds
(SCS basins, either 11-digit or 14-digit hydrologic unit
codes [HUCs]).
Locating water bodies geographically is a necessary
prerequisite to assessing water quality on a watershed
or ecosystem basis. The WBS has always included
several locational fields, including county name and
FIPS, river basin, and ecoregion. These fields have not
been uniformly populated, however. One of the WBS
files includes fields for the River Reach File (RF3)
reaches included in the water body. While a few states
had indexed their water bodies to older versions of the
reach file (RF1 and RF2), however, no state had indexed
to RF3 until 1992.
In 1992, EPA initiated a demonstration of geographic
information system (CIS) technology in conjunction with
the South Carolina Department of Health and Environ-
mental Control. This project involved:
Indexing South Carolina's water bodies to RF3.
Developing a set of arc macro languages (AMLs) for
query and analysis.
Producing coverages of water quality monitoring sta-
tions and discharge points.
Using CIS tools in exploring ways to improve water
quality assessments.
244
-------
South Carolina has defined its water bodies as SCS
basins.
The results have been very encouraging. First, South
Carolina took the initial coverages and decided they
needed much more specificity in their use support de-
terminations and their mapping of the causes and
sources of impairment. As a result, they mapped these
features down to the reach level. Next, they decided that
they needed better locational information, so they used
global positioning satellite receivers to identify accurate
locations for discharges and monitoring stations. They
then used CIS query and analysis techniques to relate
their monitoring and discharge data to their water quality
criteria. South Carolina is using CIS to actively identify
water quality problems and improve their assessments.
In 1993, EPA worked cooperatively with several states
to index their water bodies to the reach file. Virginia, the
next state to be indexed, demonstrated the successful
use of PC Reach File (PCRF) software (described later
in this paper) for indexing water bodies to the reach file.
Ohio and Kansas also are essentially complete. Each of
these states required a somewhat different approach
than Virginia. The need for flexibility in dealing with
states on reach indexing issues is essential. Existing
waterbody delineations often represent considerable in-
vestment; therefore, EPA must provide the capability to
link the state's existing assessment data to the reach file
in order to encourage state buy-in.
Figure 1 shows the results of Ohio's indexing of a typical
cataloging unit (CD). Figure 2 reflects part of the output
of the Kansas work. We can link use support, cause, and
source data to each of these water bodies now. In the
future, we hope to map these attributes at a higher level
of resolution, down to the reach segment level. CIS has
proven to be a useful assessment tool. With higher reso-
lution, it should prove to be even more helpful in identify-
ing water quality problems, picking up data anomalies,
and assessing management actions, strategies, and poli-
cies. This entire process has taught us much and has
strengthened enthusiasm for place-based management.
The Reach Indexing Project
Georeferencing the Waterbody System
Purpose and Overview
The reach indexing project is designed to locate water
bodies using RF3 as an electronic base map of hydrog-
raphy and to code RF3 reaches with the specific water-
body identifier (WBID). After linking water bodies to their
spatial representation, they can be queried and dis-
played with assessment data located in WBS files.
Reach indexing includes several steps. First, the state
must supply waterbody locations and WBIDs. The next
step entails developing a set of procedures for indexing.
Finally, the coded RF3 data must be produced.
Input data to the indexing process includes:
A list of valid WBIDs. In most cases, the state has
already input these identification numbers to the WBS.
Information about the location of each water body. Lo-
cational information may be found in marked-up
paper maps showing waterbody locations or electronic
files from WBS containing waterbody indexing
66022
66023
66024
66025
67001
66001
66002
66003
66004
66005
66006
66007
fifinnp
DDUUO
66009
66010
= 66011
66012
«*«** 66013
ooocxj 66014
= 66015
ca=t 66016
fifim 7
DDU 1 /
66018
= 66019
__-. 66020
Figure 1. State of Ohio water bodies in CU 04100008.
245
-------
Figure 2. State of Kansas water bodies in CU 11070202.
expressions (discussed later), orit may be embedded
in the WBID itself.
A complete set of RF3 data for the state being indexed.
Depending on the type of information the state supplies,
procedures used to index water bodies can be almost
fully automated, semiautomated, or completely manual.
The final result of the indexing processes is a set of RF3
coverages that contain a WBID attribute. This product
allows querying and displaying of assessment data,
which is collected and stored by water body, in a CIS
environment.
The Reach File Database
The reach file is a hydrographic database of the surface
waters of the continental United States. Elements within
the database represent stream segments. The elements
were created for several purposes:
To perform hydrologic routing for modeling programs.
To identify upstream and downstream connectivity.
To provide a method to uniquely identify any particu-
lar point associated with surface waters.
The unique reach identifier has succeeded in associat-
ing other EPA national databases, such as STORET, to
surface waters. Any point within these databases can be
associated with and identified by a specific location on
any surface water element, such as a reservoir, lake,
stream, wide river, or coastline.
There are three versions of the reach file. The first was
created in 1982 and contained 68,000 reaches. The
second version, released in 1988, doubled the size of
Version 1. The third version (RF3) includes over
3,000,000 individual reach components.
The base geography of RF3 is derived from U.S. Geo-
logical Survey (USGS) hydrographic data (1:100,000
scale) stored in digital line graph (DIG) format. Unlike
DIG data, which are partitioned by quad sheet bounda-
ries, RF3 data are partitioned by CU. A CU is a geo-
graphic area that represents part or all of a surface
drainage basin, a combination of drainage basins, or
a distinct hydrologic feature. The USGS uses CUs
for cataloging and indexing water-data acquisition
activities.
The continental United States comprises over 2,100
CUs. CUs are fairly small; for example, 45 units fall
partially or completely within the state of Virginia (see
Figure 3).
RF3 is a powerful data source used in hydrologic appli-
cations for many reasons, including the following:
RF3 has spatial network connectivity that topological
upstream/downstream modeling tools use.
246
-------
Figure 3. Cataloging units in Virginia.
RF3 has attributes that describe connectivity, which
offers the ability to accomplish upstream/downstream
navigation analytically (without topological networking).
RF3 has a simple and consistent unique numbering
system for every stream reach in the United States.
RF3 has built-in river mileage attributes that describe
upstream/downstream distances along river reaches.
Use ofRF3 in the Indexing Process
When importing Reach File data from EPA's mainframe
computer, an arc attribute table (AAT) is automatically
built for each RF3 coverage. The AAT contains the
standard AAT fields, plus the items found in Table 1.
The CD item stores the USGS CD number of this piece
of RF3. Every arc in the coverage has the same value
for CU.
The SEG item stores the number of the stream seg-
ment to which the particular arc is assigned. SEG num-
bers start at 1 and increase incrementally by 1 to 'N' for
each CU. A SEG could represent all the arcs of a
mainstream, the arcs of a tributary, or piece of a main-
stream or tributary. SEG numbers were defined in the
production of RF3.
Ml stores the marker index for each particular arc. The
Ml resembles a mile posting along a stream. In reality,
the Ml field does not truly measure mileage along the
RF3 stream network. It does, however, represent a
method of producing a unique identifier (in combination
Table 1. Fields Found in Arc Attribute Table
12070104-ID CU SEG Ml
UP
DOWN
1
2
3
4
5
6
7
1 20701 04
1 20701 04
1 20701 04
1 20701 04
1 20701 04
1 20701 04
1 20701 04
1
1
1
2
3
3
4
0.00
1.30
2.10
0.00
0.00
1.15
0.00
-1
-1
-1
-1
-1
-1
-1
0
0
0
0
0
0
0
with the CU number and the SEG number) for every
reach in the United States (see Figure 4).
Together CU, SEG, and Ml uniquely identify every arc
in RF3 nationwide. These three items are combined in
the redefined item called RF3RCHID. This provides a
powerful scheme for consistently identifying locations
along streams everywhere in the country.
Along with the AAT file, a second attribute file is auto-
matically created for RF3 coverages. This file is always
named COVER.DS3. The DS3 file stores a wealth of
information about arcs in the coverage. Some of the
important fields in the DS3 file contain:
Upstream and downstream connectivity for navigat-
ing along reaches.
Codes to describe the type of reach (e.g., stream,
lake boundary, wide river).
DIG major and minor attributes.
247
-------
4.05
2.4
o
1.30
WBID
KS- KR-04-R001
KS-KR-04-W020
KS-KR-04-W030
1.0
o
WBID WBNAME
KS-KR-04-R0001 Mainstem
KS-KR-04-W020 Tributaries
KS-KR-04-W030 Lakes
Figure 4. RF3, SEG, and Ml data elements.
Waterbody Locations
Because states define water bodies, they provide the
only information on waterbody location. South Carolina
was indexed to RF3 in 1992, followed by Virginia. Vir-
ginia indexed its water bodies using the PCRF program
instead of indexing in a CIS environment with ARC/INFO.
PCRF is a PC-based system that indexes water bodies
and locates other assessment data from WBS. PCRF
stores the definitions of water bodies (including their
location) in a file that is linked to other WBS database
files that contain information about the assessment
status and quality of the waters.
A water body is a set of one or more hydrologic features,
such as streams, lakes, or shorelines, that have similar
hydrologic characteristics. Water bodies are the basic
units that states use to report water quality for CWA
305(b) requirements. Depending on the state's assess-
ment goals and resources, water bodies can be defined
in several ways, including (see Figure 5):
All streams within a watershed
All lakes and ponds within a watershed
Sets of streams with similar water quality conditions
PCRF stores locational data for a water body with a
unique WBID. WBS uses this WBID as a common field
to relate the water body's definition and location to de-
scriptive data about the water body's assessment status
and quality. The two most important files used in PCRF
are the SCRF1 and SCRF2 files.
The SCRF1 file simply lists the unique water bodies by
state. Table 2 offers an example. The most relevant data
for reach indexing in this file are the WBID, WBNAME,
and WBTYPE, as defined by the state. The WBID, as
stated, is a unique identifier for each water body the
state has defined. The WBNAME stores a verbal de-
Figure 5. Potential definitions of water bodies.
scription of the water body. Finally, the WBTYPE de-
fines the type of water body; for example, R is for river,
L is for lake.
The SCRF2 file contains an explicit definition of each
water body. Because of the complexity involved in de-
fining water bodies, this file may include more than one
record for each water body. The SCRF2 file can be
considered a waterbody definition language because it
contains specific codes, attributes, and keys that can be
converted into specific reaches on the RF3 data, if read
properly (see Table 3). The WBBEGIN and WBEND
fields contain explicit CD, SEG, and Ml attributes to
define the location of the starting point and ending point
for the water body. The WBDIR field contains an attrib-
ute that describes whether to go upstream or down-
stream from the WBBEGIN to the WBEND. In addition,
a blank WBEND field denotes that the water body
should include all upstream or downstream reaches
(depending on the WBDIR) of the WBBEGIN reach.
Virginia used PCRF to create an SCRF2 file that con-
tains reach indexing expressions for all of their defined
water bodies. ARC/INFO macros were then written to
process this file and expand the expressions into the set
of specific arcs that compose each water body. The
macros will be described in more detail later.
Table 2. Example of SCRF1 File Data
WBID WBNAME WBTYPE
WBSIZE
KS-KR-04-R001
KS-KR-04-W020
KS-KR-04-W030
KS-KR-04-W040
KANSAS
RIVER
LOWER
WAKARUSA
RIVER
MUD CR
CAPTAIL CR
R
R
R
R
15.20
61.60
39.43
15.63
248
-------
Table 3. Example of SCRF2 File Data
WBID WBDIR WBBEGIN
WBEND
RFORGFLAG
KS-KR-04-R001
KS-KR-OR-W020
KS-KR-04-W030
KS-KR-04-W040
U
D
U
U
10270104001 0.00
10270104005 10.80
10270104059 12.05
10270104038 0.00
1027010400115.20
102701040050.00
102701040078.10
2
2
3
3
States that have not already generated indexing expres-
sions in PCRF must provide locations in some other
way. The most basic method is for the state to supply a
set of 1:100,000 USGS quad sheets that they have
marked up with locations of each water body. The
maps can be used in conjunction with a digitizer to
manually select the appropriate RF3 reaches and
code them with the WBID.
The state of Ohio created a CIS database of its river
reaches several years ago. The CIS coverage is rep-
resentational in nature. The stream reaches are 'stick-
figures' only. Generally, they fall along the paths of the
actual streams, but they are schematic in nature and do
not show the true shape of streams. The CIS layer,
however, contains the attributes of Ohio's stream reach
numbering system, which is used to identify water bod-
ies as well. Ohio's river reach coverage contains infor-
mation on the locations of water bodies and is being
manually conflated to transfer the WBIDs to RF3. The
conflation process will be covered later in this paper.
The state of Kansas had previously defined its water
bodies on RF2, the precursor to RF3. Kansas defined
some indexes by a set of RF3 SEG numbers in a CD
and some by the RF3 reaches in a small watershed
polygon within a CD. The locations were, in effect, de-
fined within the WBID itself.
Indexing Procedures
Procedures developed for performing waterbody index-
ing include automated, semiautomated, and manual
systems.
Automated Indexing Procedures
As stated, Virginia used PCRF to perform the indexing
operation. The state delivered an SCRF2 file containing
indexing expressions for all of its water bodies. AMI
programs were created to read the SCRF2 file and
select the reaches specified by each indexing expres-
sion. The selected sets of reaches were then coded with
the appropriate WBID. The macros were designed to run
on one RF3 CD at a time, so the operator specified runs
of up to 10 CDs at a time. The macros had to accom-
modate indexing expressions that included:
Select reaches upstream of a specified location.
Select reaches on a reach-by-reach basis.
Select reaches within a given polygon area.
Select shorelines of lakes or ponds given latitude and
longitude coordinates.
Select reach downstream from a specified location.
Kansas water bodies were also indexed through an
automated process. Kansas supplied an ARC/INFO
coverage of small watershed polygons (sub-CD poly-
gons) containing a watershed identifier. The state's
WBID contained all other information necessary to de-
termine the RF3 CD and the set of reaches making up
each water body. An example of a Kansas WBID is
KS-KR-02-W030. This is explained by the following:
KS refers to the state. All WBIDs in Kansas begin
with KS.
The second component (in this case KR) is an ab-
breviation of the basin in which the water body falls.
KR indicates that this water body is in the Kansas-
Lower Republican River basin.
The third component contains the last two digits of
the eight-digit CD number. Although basins comprise
several CDs, the last two digits of each CD in a basin
are unique; therefore, between the basin (e.g., KR)
designation and the last two digits of the CD (e.g.,
02), the complete eight-digit CD number in which the
water body falls is defined.
The next letter (in this case W) denotes whether the
water body is defined by a watershed polygon (W),
an RF2 SEG (R), or a lake or pond shoreline (L).
Finally, the WBID ends with the number of the poly-
gon (in this case 030) that contains the reaches for
the water body in the watershed coverage.
The completed macros could index the entire state in a
single run provided that all the WBIDs were contained
in single file.
In all cases, Kansas has indexed to RF2 reaches. Only
RF3 reaches originally created in RF2 production, there-
fore, are coded with a WBID.
249
-------
Manual Indexing Procedures
Because Ohio already has a coverage of river reach
codes, WBIDs from this coverage had to be transferred
to the RF3 reaches they represent. This entailed using a
manual conflation process. The operator displayed a CD
of RF3 along with the Ohio river reach system for the
same area. In a simple process of 'pointing and clicking,'
the operator first selected an Ohio river reach arc, then
the RF3 arcs that seemed to coincide. As each RF3 arc
was selected, it was coded with the WBID of the pre-
viously selected Ohio river reach arc.
Other states that have no means of describing water
bodies in electronic files may have to mark up paper maps
to show waterbody locations. These maps can then be
used in a manual process of selecting RF3 reach and
coding them with WBIDs either in ARC/INFO or in PCRF
Using the Route System Data Model To
Store Water Bodies
Because water bodies can be defined as noncontiguous
sets of arcs and portions of arcs, a robust linear data-
base model is necessary to model these entities.
ARC/INFO's route system data model seems well suited
for this application. The route system data model allows
one to group any set of arcs or portions of arcs into
routes. Each route is managed as a feature in itself.
Attributes of water bodies are stored in a route attribute
table (RAT) and relate to all the arcs defined as the water
body. Figure 6 helps illustrate the route system model.
Each route comprises one or more arcs or sections of
arcs. ARC/INFO manages the relationship between arcs
and routes in the section table (SEC). The structure of
the SEC, which is an INFO table, is defined in Table 4.
Tables reflects how the sections that make up the above
routes would appear.
cover.ratwbs
WBS#WBS-ID
1 1
WBID
KS-KR-04-R0001
KS-KR-04-W020
KS-KR-04-W030
Table 4. Definition of Structure of SEC INFO Table
ROUTELINK* The route upon which the section falls
ARCLINK* The arc upon which the section falls
F-MEAS The measurement value at the beginning of the
section
T-MEAS The measurement value at the end of the section
F-POS The percentage of the distance along the arc at
which the section begins
T-POS The percentage of the distance along the arc at
which the section ends
SEC# The internal identifier of the section
SEC-ID The user identifier of the section
Table 5. How Sections Appear in SEC INFO Table
ROUTE-
LINKS ARCLINK8 F-MEAS T-MEAS F-POS T-POS SEC# SEC-ID
1
1
1
2
3
3
3
1
2
3
4
5
6
7
0
1.30
2.10
0
0
1.15
0
1.30
2.10
4.05
1.20
1.15
2.4
2.5
0
0
0
0
0
0
0
100
100
100
100
100
100
100
1
2
3
4
5
6
7
1
2
3
4
5
6
7
Figure 6. The route attribute table containing waterbody data.
Representing Water Bodies as Routes
ARC/INFO offers several ways of grouping sets of arcs
into discrete routes. One can use ARCEDIT to select a
set of arcs to group them into a route, or ARCSECTION
or MEASUREROUTE in ARC to group arcs into routes.
The method described here uses the MEASUREROUTE
command. This method requires that the AAT or a re-
lated table has an attribute containing the identifier of
the route to which an arc should be assigned. In the
application the authors employed, they converted the
SCRF2 file into an INFO table containing, for each arc
in the coverage, the RF3RCHID of the arc and the WBID
to which the arc should be assigned. The WBID item is
used to group arcs into routes. One route exists for each
unique WBID. Table 6 illustrates the table used in the
MEASUREROUTE command method. This table is re-
lated to the AAT of the RF3 coverage by the RF3RCHID.
An RAT is automatically created for the coverage, which
now can be related to other WBS assessment files for
display and query. Figure 4 illustrates the RAT. The most
important characteristic of the file is that it has only one
record for each water body. This simplifies the display
and query of water bodies based on water quality data.
Using EVENTS for Subwaterbody Attributes
Water bodies, as states define them, often constitute a
gross aggregation of the water in an area. States often
have more specific data about particular stretches of
streams within a water body. A system is needed to
250
-------
Table 6. Table Used in MEASUREROUTE Command Method
To Group Arcs Into Routes
$RECNO
RF3RCHID
WBID
1
2
3
4
5
6
7
8
9
1 02701 04
1 02701 04
1 02701 04
10270104
10270104
10270104
10270104
10270104
10270104
1 0.00
1 1.30
1 2.10
2 0.00
3 0.00
3 1.15
40.00
5 0.00
6 0.00
KS-KR-04-R0001
KS-KR-04-R0001
KS-KR-04-R0001
KS-KR-04-W020
KS-KR-04-W020
KS-KR-04-W020
KS-KR-04-W020
KS-KR-04-W030
KS-KR-04-W030
4.05
2.4
o
1.0
1.30
WBID
KS- KR-04-R001
KS-KR-04-W020
KS-KR-04-W030
query and display data at the subwaterbody level.
ARC/INFO's dynamic segmentation tools and event ta-
bles are useful for this application. Once water bodies
have been defined and reporting methods have been set
up based on those water bodies, the task of redefining
them is cumbersome.
Event tables can help to keep these waterbody defini-
tions yet still offer the ability to store, manage, and track
data at the subwatershed level. Event tables are simple
INFO files that relate to route systems on coverages.
This concept and data structure can work in conjunction
with the predefined waterbody system. We have already
seen how a route system called WBS is created in RF3
to group arcs into waterbody routes. This works quite
well when displaying water bodies and querying their
attributes. A route system based on the WBID cannot,
however, act as an underlying base for subwaterbody
events because the measures used to create the WBS
route system are not unique for a particular route. For
example, in the route depicted in Figure 7, three loca-
tions are defined as being on WBID KS-KR-04-W020 and
having measure 1.0.
The mileage measurements along SEG, however, are
always unique (see Figure 8). To use EVENTS, therefore,
a second route system must be created based on the RF3
SEG attribute, which provides a unique code for each CD.
The ARCSECTION command, instead of the MEASURE-
ROUTE command, is used to create the SEG route
system. This is because the measurement items (Ml on
the AAT and UPMI on the DS3) already store the
summed measures along particular SEGs. Table 7 lists
the contents of the resulting RAT table.
Because the name of the route system is SEG, the
SEG# and SEG-ID are the names of the internal and
user IDs. The SEG item contains the actual SEG num-
ber in the RF3 coverage. Because the SEG numbers for
each RF3 CU coverage start at 1 and increase incre-
Figure 7. Measurements along SEGs.
O
Figure 8. Events located on RF3 data.
Table 7. Route Attribute Table
SEG-ID SEG#
SEG
mentally by 1, the SEG item looks much like the SEG-ID
and SEG#.
Event tables contain a key item, the WBID or SEG, to
relate them to the appropriate route system (see Figure 8).
They also contain locational information on where to lo-
cate the events on the route (either WBID to indicate the
water body on the WBS route or SEG to identify the
route in the SEG route system). Separate event tables
can then relate use support, causes, and sources as
linear events. FROM and TO store the starting and
251
-------
ending measures for each event. Using event tables
allows us to apply many useful cartographic effects
(e.g., hatching, offsets, text, strip maps). Events can be
queried both in INFO and graphically. Event data can
help in producing overlays of two or more event tables.
An event table can display use support information (see
Table 8). WBS users can update their event tables using
RF3 maps supplied by EPA without having proficiency
in ARC/INFO. ARC/VIEW2 is expected to support
events and route systems. This will give users powerful
tools for spatial query of assessment data. Developing
event tables would also display and query data on the
causes and sources of use impairment. These events
can be offset and displayed to show the areas of inter-
action. More permanently, preparing line-on-line over-
lays can show intersections and unions.
An alternative approach is to use an EVENT-ID as a
unique identifier for each event. The SEG field stores the
number of the route (SEG) upon which the event occurs.
FROM and TO store the beginning and end measures
along the route upon which the event occurs. WBID
contains the identifier of the water body upon which the
event occurs (see Table 9). An event can occur within
a single SEG, across two or more SEGs, within a single
water body, or across two or more water bodies.
Additional attribute tables can be created to store de-
scriptive attributes for each event. These tables would
resemble the SCRF5 and SCRF6 files except that in-
stead of using the WBID to relate to a water body, a field
called 'EVENT-ID' would link the use, cause, and source
data to a particular event (see Table 10).
Both approaches offer some advantages. In either case,
they allow us to map our water quality assessment data
and communicate it in a meaninaful and useful wav.
Table 8. Event Table That Reflects Use Support Information
USE
SEG FROM TO WBID USE SUPPORT
1 0.80 1.30 KS-KR-04-R0001
1 1.30 2.10 KS-KR-04-R0001
1 2.10 2.31 KS-KR-04-R0001
1 0.50 1.30 KS-KR-04-R0001
3 0.00 1.15 KS-KR-04-W030
4 0.00 2.5 KS-KR-04-W040
21 Fully
21 Partial
21 Not supported
40 Threatened
21 Fully
40 Not supported
Table 9. Using EVENT-ID as a Unique Event Identifier
EVENT-ID SEG FROM TO WBID
1 1 0.80 1.30
1 1 1.30 2.10
2 4 0.00 2.5
Table 10. Using EVENT-ID To Link Use,
Data to an Event
EVENT-ID ASCAUSE
1 900
1 -9
1 0500
2 1200
2 0900
KS-KR-04-R0001
KS-KR-04-R0001
KS-KR-04-W040
Cause, and Source
ASSOURC
1200
1100
1100
9000
8100
252
-------
Environmental Management Applications
-------
Ecological Land Units, GIS, and Remote Sensing:
Gap Analysis in the Central Appalachians
Ree Brannon, Charles B. Yuill, and Sue A. Perry
West Virginia University, Morgantown, West Virginia
Abstract
The gap analysis team in West Virginia is assessing the
state's natural communities as part of a nationwide,
comprehensive planning effort. Underrepresented or
unrepresented habitats represent gaps in the present
network of conservation lands and conservation activi-
ties. After identifying these gaps, we can assess
whether our current management direction will maintain
natural diversity and will prevent additional species from
being classified as threatened or endangered.
The relationship between vegetation and ecological
variables serves as the basis for classifying ecological
land units. To characterize the ecological land units,
many layers of physical data can be integrated in a
geographic information system (GIS). Satellite imagery
and videography map existing conditions over the state.
The existing vegetation is classified to reflect physiog-
nomic and floristic elements to correlate with vertebrate
and butterfly habitat requirements. This correlation of
vegetation and wildlife habitat creates mappable habitat
types. Analysis of these habitat types with land-
ownership data indicates where the species-rich areas
occur in the landscape and whether the most species-
rich areas are protected.
Introduction
To respond to the urgency of habitat loss and its effect
on species diversity, scientists must implement a meth-
odology for rapid assessment and documentation of
natural communities at a scale pertinent for regional
management activities.
Geographic information systems (GIS) and remote
sensing support the development of ecological land
classifications over large regions. CIS-based mapping
of ecological land classes allows users to combine and
display environmental variables for spatial modeling and
refinement of ecological land units (1).
The Gap Analysis Project is a comprehensive planning
effort for land conservation in the United States. The
objective of the Gap Analysis Project is to identify spe-
cies, species-rich areas, and vegetation types underrep-
resented or unrepresented in existing biodiversity
management areas. Unprotected communities are the
gaps in the conservation strategy. The Gap Analysis
Project is not merely identifying communities with the
largest number of species; its ultimate goal is to identify
clusters of habitats that link the greatest variety of
unique species.
Local areas with considerable diversity of habitat or
topography usually have richer faunas and floras (2).
Nature reserves, which incorporate a variety of habitats,
may be the best guarantee for long-term protection of
biodiversity. By protecting species-rich regions, we can
reduce the enormous financial and scientific resources
needed to recover species on the brink of extinction.
The West Virginia Gap Analysis Project began in 1991.
The objective is to map existing vegetation and to use
that as the foundation to model potential distribution of
vertebrate and butterfly species. High cost precludes
intensive field inventory and monitoring of wildlife.
Therefore, habitat modeling is critical to predict wildlife
species composition and potential ranges over the vari-
ous landscape types of West Virginia. Lastly, the vege-
tation map will provide a record of the existing habitat to
use in monitoring changes due to human activities and
natural disturbances.
Pilot Study Area
Distribution of wildlife and plant communities will be
modeled for the entire state. Initially, we will focus on a
smaller pilot study area. This region includes approxi-
mately 50,000 hectares in the central Appalachian
Mountains and spans several physiographic provinces
and vegetative communities. Generally, soils in the pilot
study area are of two kinds: acidic soils that develop a
255
-------
clay horizon from extensive leaching over time and
younger soils that are found on steep slopes and where
environmental conditions, such as cold climate, limit soil
development (3, 4).
The vegetation types include spruce-fir, oak-pine, high
elevation bogs, northern hardwoods, Appalachian oak,
mixed-mesophytic, open heath barrens, and grass
balds. The mixed-mesophytic, the most diverse in West
Virginia, lies primarily west of this area, but localized
stands do occur in the lower elevations. Cover types
within the Appalachian oak and mixed-mesophytic types
are not discrete and will be difficult to delineate.
The pilot study area includes a variety of land uses, such
as residential, commercial, industrial, mining, and agri-
culture. Portions of the Monongahela National Forest in
this area are the Fernow Experimental Forest, and the
Otter Creek and Laurel Forks Wilderness Areas.
Methods
The following discussion describes the methods formu-
lated and data compiled for the West Virginia Gap Analy-
sis Project.
Describe Ecological Land Units With Existing
Vegetation
Davis and Dozier (1) note that a landscape can be
partitioned by ecological variables, which contributes to
an ecological land classification. This process is applied
frequently to analysis and mapping of natural resources.
Davis and Dozier classified vegetation in California
based on the documented associations of vegetation
with terrain variables. They based this approach on the
assumptions that the arrangement of natural landscape
features is spatially ordered by an ecological interde-
pendence among terrain variables and that actual vege-
tation is a reliable indicator of these ecological
conditions. Similar documentation exists for the distribu-
tion of vegetation types in West Virginia, and the gap
analysis team is proceeding along a similar course.
West Virginia lies in two major provinces, the Eastern
Broadleaf Forest and the Central Appalachian Broadleaf
Forest-Coniferous Forest-Meadow Provinces (5). Within
these provinces are several broad vegetation types. The
gradient diagram in Figure 1 (6) illustrates the range of
these types. The vertical axis represents elevation in
feet. Three vegetation types emerge distinctly along the
elevation gradient. The horizontal axis is not quite as
explicit. This gradient spans moist, protected slopes to
dry, exposed ridgetops, and the vegetation types are
much less distinct. Drier oak and pine types occur al-
most exclusively on exposed ridgetops. The vegetation
types along the horizontal axis are the mixed mesophytic
forest association of the Allegheny and Cumberland
Mountains and can have 20 to 25 overstory and under-
story species per hectare in North America (7).
The distribution of the vegetation along gradients such
as elevation and soil moisture lends itself to a CIS
analysis. Physical data such as elevation and soil mois-
ture regime can be incorporated into a CIS. These lay-
ers of information can be manipulated graphically or
mathematically to model the spatial distribution of vege-
tation types or to provide useful ancillary data for clas-
sification of satellite imagery (see Figure 2). Much of this
data is digital and can be used to substantially reduce
the time required to develop a database. To standardize
output, members of the national gap team have
specified ARC/INFO as the software to generate final
products.
Classify Satellite Imagery To Create a Map of
Current Distribution
Remote sensing provides an effective means to classify
forests, and the Gap Analysis Project has successfully
used it in the western United States (1, 8, 9).
The Gap Analysis Project is using Landsat Thematic
Mapper imagery in all states to standardize the baseline
information. The hypothesis is that the spectral data
from the imagery is related to the distribution of the
ecological land units and land use across the landscape.
The data include all spectral bands, except thermal, for
the entire state. The West Virginia project is using two
seasons of data, spring and fall. Temporal changes,
which record phenologic variation in the deciduous spe-
cies, enhance classification accuracy. The spectral reso-
lution is 30-meter pixels. This is equivalent to
approximately 1/6 hectare (1/2 acre). Our final product
will be a series of 1:100,000 maps. The minimum map-
ping unit is 40 acres.
The mountainous terrain in the Appalachian Mountains
offers disadvantages and advantages for using remotely
sensed data. Irregular topography can cause inconsis-
tencies in the spectral data that diminish the classifica-
tion accuracy. Similar cover types may have different
spectral signatures; for instance, if one stand is in sun-
light and the other is shaded. Also, phenology can vary
due to microclimatic influences. Conversely, topo-
graphic features influence the distribution of vegetation
types, and ancillary data, such as digital elevation mod-
els (OEMs), enhance classification results of the im-
agery. The West Virginia gap analysis team selected the
following strategy for image classification.
1. Stratify the imagery using ecological units based on
a hierarchical scheme. Bauer et al. (10) found that
an initial stratification of physiographic regions was
necessary to reduce the effect of broadscale
environmental factors caused by changes over
latitude. Therefore, stratification enhanced the
256
-------
5,000
4,500
4,000
3,500
3,000
2,500
2,000
1,500
1,000
500
MESIC
Red Spruce
Forest ._. .,
Heath
Barren
Red Spruce -
Yellow Birch
Northern Hardwoods
Forest
Red
--------- , Oak
._--"' " ', ' Forest /
x' Hemlock I . ,'' /
and ,' ,-' /' '
Hemlock- ' -' / ,-'*
Cove / Hardwoods ,' / ,-'
Hardwoods ' ''
White /'
Oak /
1 / ' '
, / ; i
/ Red Oak - / chestnut/
/ / White Oak / oak ''
Forest / ,'
Scarlet/
,' \ ,' I Black Oak
Coves Flats Sheltered Open Slope
Canyons Draws/Ravines Slopes NE, E, S, W, NV
Grass
Bald
/
Pitch/
Shortleaf
Pine
Heath
Scrub
Pine
Post/
Blackjack
Oak
Ridges
V XER\C
Figure 1. Environmental gradients for vegetation.
efficiency of the training data. An interagency
committee, including ecologists from the Monongahela
National Forest, West Virginia Division of Forestry,
and geologists from the State Geologic Survey,
generated a draft map of physiographic regions.
They delineated sections based on geomorphology
and climate. Sections were divided into subsections:
those most typical of the section or those that are
transitional, or irregular, to the section. Figure 3 is a
draft map of these sections and subsections in West
Virginia.
2. Classify stratified imagery using the ancillary data.
High resolution imagery has not been used widely in
the eastern United States, where forests are not
homogeneous stands of relatively few tree species
as they are in the West. Researchers who classified
eastern forests from satellite imagery attempted to
find the most distinctive spectral band combinations
for discriminating cover types (10-13). One recent
technique (14) uses a nonparametric approach that
combines all spectral and informational categories to
classify imagery. We are testing a variety of methods
such as nonparametric processes, traditional
clustering techniques, and use of derived vegetation
indexes to find the most successful method.
3. Assess accuracy with random plots from
videography. Videography will be acquired in the
spring of 1995. Aerial transects, which extend the
length of the state, will be flown with approximately
30-kilometer spacing. By regulating flight altitudes,
the resolution per frame can be captured at 1
kilometer per frame. About 7,000 frames will be
collected, which make up a 3 percent sample of the
state. About 2 to 3 percent of the videography frames
will be field verified. With this strategy, we will test
the effectiveness of using videography, instead of
intensive field plots, to verify classification of satellite
imagery. Areas of special interest, which the
systematic transects may not capture, will require
separate flights. The bulk of the videography will
provide training data for supervised image
classification. The remaining frames will be used to
assess the accuracy of the classification.
;>, Streams
."I;;-- Ecological Subregions
";:>- Digital Elevation Models
i-* Soils
Agricultural, Tilled
High Elevation, Red
Spruce
Low Elevation, Coniferous
Forest (Hemlock)
Open Grassland, Deep
Soil (Grass Bald)
High-Density Urban
Landscape Unit
Classes
Figure 2. GIS and the development of ecological land units.
257
-------
Figure 3.
Legend
g] Pilot Study Area
Physiographic regions of West Virginia and pilot
study area.
In summary, 100 percent of the state will be classified
using the Thematic Mapper imagery. Aerial video-
graphy, covering approximately 3 percent of the
state, will help to verify the image classification, and
2 to 3 percent of the videography will be verified from
transects on the ground.
4. Determine sources of data for image classification.
Due to the increasing interest in CIS, digital data are
more readily available from a variety of sources,
such as the federal government, state agencies, and
private companies. Acquisition of available data sets
can substantially reduce the time and cost of
database development. Users must bear in mind,
however, that databases are developed with differing
objectives and techniques, so one must consider
scale and standards of production when deciding
which data sets are appropriate for project design.
The West Virginia gap team determined that the
following CIS coverages are important for image
classification.
The U.S. Geological Survey's (USGS's) graphic infor-
mation retrieval and analysis system (GIRAS) ear-
marked land use/land cover data. The classification was
done several years ago, and although these data are not
current, they provide excellent information concerning
urban and agricultural land use. Land-cover categories
represent Level II classifications from Anderson's (15)
system. The maps are produced at a 1:250,000 scale,
so they require few CIS operations to piece together a
regional coverage of land use.
The Southern Forest Experiment Station mapped U.S.
forestland using advanced very high resolution radiome-
ter (AVHRR) satellite imagery (16). This imagery is rela-
tively current (1991 to 1992), but the resolution is coarse
at 1 kilometer per pixel (100 hectares or 247 acres). The
classes are based on Forest Inventory and Analysis
plots established by the U.S. Department of Agriculture
(USDA) Forest Service and Kuchler's (17) potential
natural vegetation types. We are using the maps to
depict broad changes in forest type over a region, such
as the state of West Virginia. This coverage does not
show land use.
The eastern region of the Nature Conservancy has com-
pleted a draft of the classification of the terrestrial com-
munity alliances (18). The classification hierarchy is that
prescribed by the national gap team, and as such, re-
flects physiognomic and floristic characteristics neces-
sary for correlating vegetation structure and floristic
composition with vertebrate habitat requirements. The
descriptions include the range of alliances and charac-
teristic species of the overstory, understory, and herba-
ceous layer. This provides information on associated
species not detectable by image classification.
The National Wetlands Inventory data are available digi-
tally. Maps have been digitized at a scale of 1:24,000,
and the classification scheme is from Cowardin et al.
(19). Coverages come with attribute data for each poly-
gon, arc, or point as needed. A labor-intensive effort is
required to join the maps in a CIS for an area the size
of West Virginia, but the information will be invaluable
for masking water and forested wetlands on the satellite
imagery. The U.S. Fish and Wildlife Service includes
detailed instructions for converting the data to coverages.
Field data, much of it already digital, has been acquired
from many sources. Commercial timber companies pro-
vided data for timber stand composition and age groups.
The USDA Forest Service ecologist conducted transects
throughout the Monongahela National Forest to charac-
terize ecological land units. Forest inventory and analy-
sis plots are also available. We acquired these data to
verify videography classification.
The USGS has digital data of terrain elevations. The
West Virginia gap analysis team acquired 3 arc-second
OEMs as an additional band in the satellite imagery. We
will use these data to generate coverages of slope,
aspect, and elevation classes to further stratify the
physiographic regions of the state. This will increase the
accuracy of the classification.
The Soil Conservation Service created a statewide da-
tabase called STATSGO, produced at a scale of
1:250,000. For West Virginia, the map of the soil map-
ping units consists of approximately 450 polygons. Each
mapping unit is an aggregation of soil components that
occupy a certain percentage of the mapping unit area.
The database is extensive and includes information on
soil attributes such as soil taxonomy, soil chemistry, soil
structure, and interpretations for erodability and wildlife
habitat. For an attribute such as soil temperature, each
component has an individual value so that each
mapping unit may have several different values for soil
258
-------
temperature. Attribute information is difficult to query in
ARC/INFO, where there is a one-to-many relationship
between polygons and database entries (for instance,
each mapping unit, or polygon, has several soil compo-
nents). We found that exporting the attribute information
from ARC/INFO to another software package such as
Excel was easier. The values can be aggregated by
attribute and then imported into ARC/INFO to produce
individual coverages such as soil texture, soil depth, or
soil group. STATSGO data can provide useful informa-
tion for the physical variables that influence vegetation,
such as soil moisture and nutrient availability.
To review, the project researchers will first identify the
physical parameters that govern the distribution of
ecological land units and the existing vegetation in the
state. Then, the team will gather applicable ancillary
data of physical data in CIS to support the image clas-
sification. Once the imagery has been classified, the
wildlife models can be incorporated.
Integrate Terrestrial Vertebrate Models
Once classification is completed for the state, the gap
analysis team will integrate terrestrial vertebrate mod-
els. Concurrent with the image classification, the team
will develop a species profile for each vertebrate species
known to occur in the state. These profiles, when com-
pleted, will be condensed into rule-based models for
associated species that can be linked to the ecological
land units (see Figure 4). This step will create habitat
types. After integration, the team will generate maps that
display species richness of vertebrates for each habitat
type (see Figure 5). These maps link spatial data to the
species database. This enables users to identify areas
in the landscape that combine habitats with the greatest
number of unique species. When a coverage of land-
ownership is overlaid on this map, land managers or
conservation groups can take a proactive stance to seek
protection of critical habitats. Additional analyses that
Wildlife
Database
Landownership
Assess
Protection
Status
Figure 4. Informational flow chart for wildlife data.
users can perform are displays of the potential distribu-
tion of vertebrate groups, such as upland salamanders.
Another analysis would be to report the species that
occur in the fewest habitats and that would be most
vulnerable to landscape changes. Clearly, CIS provides
a powerful environment for quick and efficient retrieval
of spatial data for management decisions.
Summary
To summarize, the West Virginia gap analysis team is
assessing the natural communities in the state as a part
of the national comprehensive planning effort. We need
to conduct the assessment rapidly, compiling existing
information and integrating these data with CIS and
remotely sensed data. Ecological land units are classi-
fied according to the relationship between vegetation
and ecological variables. Satellite imagery is used to
map existing conditions over the state. The existing
vegetation is classified to reflect physiognomic and flor-
istic elements to correlate with vertebrate and butterfly
habitat requirements.
The gap analysis team is using many widely available
data sets such as OEMs, land use/land cover, wetlands
inventory, and soils data. While these can reduce the
time and cost of developing an ecological database,
they do present implications for project design and ac-
curacy. When the user combines maps of different
scales, accuracy is constrained by the map with the
smallest scale. Additionally, data sets may be con-
structed with objectives for an intended use that is not
compatible with project needs. The classification of
AVHRR data reflects forest types but not land use, so
another source may be required for these data.
The Gap Analysis Project is not a substitute for intensive
biological studies at a fine scale. It is merely a quick
assessment at a broad scale to provide information on
existing conditions. While accumulating data and mod-
eling potential wildlife distributions, we will identify
where inventory data may be lacking. Additional work
must be done to verify wildlife models and the classifi-
cation of vegetation, but this preliminary analysis will be
a valuable framework that will direct future studies of
biological diversity. Finally, this effort will provide a data
set that can be used to monitor changes to land cover
and land use.
Acknowledgments
This study is being funded by the Cooperative Research
Units Center, National Biological Survey, United States
Department of the Interior. This manuscript is published
as Scientific Article No. 2499 of the West Virginia Agri-
culture and Forestry Experiment Station.
259
-------
Legend
Very High
n High
E3 Moderately High
m Moderate
H Moderately Low
IB Low
H Very Low
ea Comm. Timber Lands
H Public Lands
N
Figure 5. Relative species richness.
References
1. Davis, F.W., and J. Dozier. 1990. Information analysis of a spatial
database for ecological land classification. Photogramm. Eng.
and Remote Sensing 56:605-613.
2. Scott, J.M., B. Csuti, K. Smith, J.E. Estes, and S. Caicco. 1991.
Gap analysis of species richness and vegetation cover: an inte-
grated biodiversity conservation strategy. In: Kohm, K., ed. Bal-
ancing on the brink of extinction: the Endangered Species Act
and lessons for the future. Washington, DC: Island Press.
pp. 282-297.
3. Stephenson, S. 1993. Upland forests of West Virginia. Parsons,
WV: McClain Printing Co.
4. Buol, S.W, F.D. Hole, and R.J. McCracken. 1980. Soil genesis
and classification, 2nd ed. Ames, IA: Iowa State University Press.
5. Bailey, R.G. 1994. Ecoregions and subecoregions of the United
States (map). U.S. Department of Agriculture, Forest Service.
Revised February 23.
6. Fortney, R. 1994. Vegetation pattern of central Appalachian
Mountains. (Unpublished).
7. Barbour, M.G., J.H. Burk, and WD. Pitts. 1987. Terrestrial plant
ecology, 2nd ed. Menlo Park, CA: Benjamin/Cummings Publishing
Company.
8. Davis, F.W, D.M. Stoms, J.E. Estes, J. Scepah, and M. Scott.
1990. An information systems approach to the preservation of
biological diversity. Int. J. Geogr. Info. Syst. 4:55-78.
9. Scott, J.M., F. Davis, B. Csuti, R. Noss, B. Butterfield, C. Groves,
H. Anderson, S. Caicco, F. D'Erchia, T.C. Edwards, Jr., J. Ulliman,
and R.G. Wright. 1993. Gap analysis: A geographic approach to
protection of biological diversity. Wildl. Monogr. 123:1-41.
10. Bauer, M.E., T.E. Burk, A.R. Ek, PR. Coppin, S.D. Lime, T.A.
Walsh, O.K. Walters, W Befort, and D.F. Heinzen. 1994. Satellite
inventory of Minnesota forest resources. Photogramm. Eng. and
Remote Sensing 60:287-298.
0 10
Kilometers
11. Benson, A.S., and S.D. DeGloria. 1985. Interpretation of Land-
sat-4 thematic mapper and multispectral scanner data for forest
surveys. Photogramm. Eng. and Remote Sensing 51:1,281-1,289.
12. Hopkins, PR, A.L. Maclean, and T.M. Lillesand. 1988. Assess-
ment of thematic mapper imagery for forestry application under
lake states conditions. Photogramm. Eng. and Remote Sensing
54:61-68.
13. Schriever, J.R., and R.G. Congalton. 1993. Mapping forest cover-
types in New Hampshire using multitemporal Landsat Thematic
Mapper data. Papers presented at ASCM/ASPRS Annual Con-
vention and Exposition, New Orleans, LA (February), pp. 333-342.
14. Cetin, H., T.A. Warner, and D.W Levandowski. 1993. Data clas-
sification, visualization, and enhancement using n-dimensional
probability density functions (nPDF): AVIRIS, TIMS, TM, and geo-
physical applications. Photogramm. Eng. and Remote Sensing
59:1,755-1,764.
15. Anderson, J.R., E.E. Hardy, J.T Roach, and R.E. Witmer. 1976.
A land use and land cover classification system for use with
remote sensor data. U.S. Geological Survey Professional Paper
965. Washington, DC: U.S. Geological Survey.
16. Zhu, Z., and D.L. Evans. 1992. Mapping midsouth forest distri-
butions: AVHRR satellite data and CIS help meet RPA mandate.
J. Forestry 27-30.
17. Kuchler, A.W. 1964. Potential natural vegetation of the contermi-
nous United States (map and manual). American Geographical
Society Special Publication 36. New York, NY: American
Geographical Society.
18. Sneddon, L., M. Anderson, and K. Metzler. 1994. A classification
and description of terrestrial community alliances in the Nature
Conservancy's eastern region: First approximation. Washington,
DC: U.S. Department of the Interior Fish and Wildlife Service.
19. Cowardin, L.M., V Carter, F.C. Golte, and T.E. LaRoe. 1979.
Classification of wetlands and deepwater habitats of the United
States. Report No. FWS/OBS-79/31. Washington, DC: U.S.
Department of the Interior Fish and Wildlife Service.
260
-------
A GIS Strategy for Lake Management Issues
Michael F. Troge
University of Wisconsin, Stevens Point, Wisconsin
Abstract
Lake management plans are crucial to the sustained life
of a lake as it experiences pressures from human as well
as environmental activities. As proven in the past, geo-
graphic information systems (GIS) can meet the needs
of most if not all environmental entities. Applying GIS to
lakes and lake management, however, is a fairly new
concept because most previous work focused on the
terrestrial realm. Future studies must address problems
relating to dimension, but adopting certain methods (i.e.,
cross-sectional coverages) can help lake managers
plan for critical lake issues. By using sufficiently planned
coverages, lake quality data management coverages
can increase storage and/or analysis efficiency. After
evaluating certain management criteria, a lake manage-
ment plan can be derived and set up as a coverage.
These criteria can then correspond collectively to form
management zones within a lake. Each of these zones
has its own set of management goals to which all lake
users must strictly adhere.
Introduction
The importance of maintaining lake quality has long
concerned recreationalists and ecologists. The multifac-
eted interrelationships of the lake environment, how-
ever, usually make proper assessment and analysis of
lake quality information difficult. Over the last decade,
assessment has become easier due to the increased
use and acceptance of geographic information systems
(GIS). This computer-based tool has allowed successful
integration of water quality variables into a comprehen-
sible format.
One area of the environmental sciences that has neglected
GIS is lake management. This paper presents an alternative
method for using a traditional two-dimensional GIS for
viewing, querying, and displaying three-dimensional in-
formationin this case, lake quality information and
lake management criteria. Lakes, unlike geologic enti-
ties, offer a three-dimensional realm that humans can
fully penetrate without a great amount of effort. Lakes
also contain a complete aquatic environment of physi-
cal, chemical, and biological entities that humans can
effectively observe and analyze. This paper does not dis-
cuss the issue of dimension; however, future studies, pri-
marily those relating to the creation of three-dimensional
GIS, should address this issue.
GIS allows incorporation of a multitude of environ-
mental variables (e.g., water chemistry, geologic
strata) into a synergism of the many coexisting vari-
ables of the lake environment. The ability of a GIS to
"capture, manipulate, process, and display spatial or
georeferenced data" is now well known and accepted
(1). Surprisingly though, GIS is rarely used for lake
management databases and associated water quality
analysis.
The few examples that exist include Schoolmaster's
Texas Water Development Board System (2), which
examined water use on a county basis, and RAISON
GIS (3), which is an expert computer system imple-
menting proper application of hydrologic principles to
a particular lake. Many other systems are simply data-
base collection storehouses of lake information, such as
the Galveston Bay National Estuary Program (4).
One notable example is the LAKEMAP program (5).
This extensive and comprehensive GIS spans the
entire United States covering approximately 800,000
lake sampling sites from the U.S. Environmental Pro-
tection Agency's (EPA's) STORET system. LAKEMAP
uses both a database management and mapping dis-
play system, allowing retrieval of information for spe-
cific sites or aggregation of regional areas. This GIS
is unique because it examined the creation of stand-
ards that could be used across the country in data-
base development and the presentation of that data.
Using GIS in a lake management study inspires many
questions because of the lack of existing research and
the absence of any true standards. For example, how
should one create a lake quality database for general
purpose management? Is visualizing the integration of
several variables within the lake ecology possible? Can
261
-------
one examine temporal changes in pH? Many technical
and logistical CIS questions therefore existed when the
Legend Lake study began.
Background of the Legend Lake Study
Legend Lake (see Figure 1) is located in Menominee
Reservation, which is in Menominee County in north-
eastern Wisconsin. Legend Lake is a 1,230-acre im-
poundment comprising eight natural drainage lakes that
a single stream once connected. In the late 1960s, a
plan was introduced to convert this area into an im-
poundment/recreational area, and construction soon be-
gan. The ecology and hydrology had not been seriously
evaluated since the development was finalized, hence
the Legend Lake project was designed in cooperation
with EPA, Wisconsin Department of Natural Resources
(WDNR), Menominee Reservation and County, Legend
Lake District, and the Legend Lake Property Owners
Association. This intensive study spanned the qualita-
tive and quantitative aspects of surface water, ground
water, sediment, and aquatic plants, as well as human
influences on the Legend Lake watershed and sur-
rounding areas.
Figure 1. Legend Lake with basin identifiers.
Initially, questions needed to be answered regarding
aquatic plants and sediment and their influences on lake
management strategies. Because the scope of the Leg-
end Lake study included subjects such as surface- and
ground-water chemistry, land use and development,
septic system impacts, and recreational stress, all these
factors needed to be considered in determining optimum
management strategies. In addition, the study ad-
dressed the question of GIS's ability to alleviate some
technical aspects of deriving and presenting a lake man-
agement plan. Given these questions, the goal was to
integrate collected data into a CIS database to create a
prototype standard for future lake studies, as well as to
present new techniques for visualization and analysis of
lake quality data.
CIS/Database Creation
Several techniques were used to best manage the data
for a three-dimensional system: cross-sectional cover-
ages (described later), data summary coverages, and
multidate coverages. The latter two coverage types are
traditional coverages that contain general lake informa-
tion excluding water column data. These techniques
work fairly well for data storage and visualization; how-
ever, they were not sufficient for determining the geo-
graphic areas within the lake that required intensive
management decisions as opposed to areas that
needed little attention.
During formulation of a new coverage design, the study
focused on the littoral zone, which is usually defined as
that area of the lake with a depth of 15 feet or less. Most
management concerns deal with the littoral zone be-
cause most recreational and ecological activities occur
in this zone. One of the most pressing issues in lake
management is aquatic plant growth. The fact that most
aquatic plant growth is confined to the littoral zone rein-
forced the decision to use the littoral zone as the primary
sink for potential management decisions.
To curtail the dimensional problem, depth was basically
ignored. This allowed for easier delineation of areas
within the lake. This, in turn, facilitated classifying areas
into management zones to implement varying degrees
of activity, ranging from casual to intensive efforts. Thus,
management recommendations for a particular zone
were the same at a depth of 2 feet as at a depth of 12
feet. This greatly reduced complication of the model and
centered effort on the areal extent of the lake. It also
facilitated visualization of management decisions by
professionals and lay people.
Problems can arise when combining depths for man-
agement considerations, as this technique did. For ex-
ample, a littoral zone that contains a gentle slope usually
does not receive the same attention as a littoral zone
with a very abrupt slope because the littoral zone with
the gentle slope contains more area. Thus, if a lake
manager recommends restricting boat traffic in a man-
agement zone (a section of the littoral zone) that has a
gentle slope to promote wildlife habitat, the amount of
area available for boaters would decrease significantly.
In this situation, primary activity would be most crucial
in shallow areas where wildlife or waterfowl predominate
rather than in more open water areas where boaters
predominate. Situations like these may require the crea-
tion of management subzones when setting up a lake
management CIS and database to increase efficiency
of lake area use and increase support by lake users.
Many criteria can affect the decisions made for a
management zone. For instance, if the lake manager
recognizes excessive, unhealthy weed growth in a par-
ticular zone, the lake manager may recommend exten-
sive weed harvesting to neutralize the situation. An
adjacent zone may have very little weed growth and
may not require weed harvesting. Criteria such as these
must be recognized before constructing the CIS. Table 1
lists some common criteria to consider. Generally, man-
agement criteria include anything that is influent on the
262
-------
Table 1. Criteria To Consider When Creating Lake
Management Plans
Environmental Variables
Artificial Variables
Aquatic plants
Sediment
Ground water
Surface water
Wildlife/waterfowl
Fisheries
Climate
Wind
Geology/geography
Adjacent natural land cover
Natural nutrient loading
Hydrologic characteristics
Adjacent land use
Septic systems
Development
Construction sites
Fuel leaks
Shoreland zoning
Population density
Primary lake use (recreational)
Nutrient loading
Visitor use
shoreline and lake itself and any influence the lake has
on the shoreline or adjacent shorelands.
A primary concern when accessing lake data from a
computer database is being sure to query the correct
lake. For example, searching a database for all data on
"Sand Lake" would be a legitimate action, except that
the database may include close to 200 lakes with the
name of Sand Lake. This is one of the main reasons why
the WDNR developed a system known as the Water
Mileage System (6). Based on logical criteria, each
water body (e.g., lakes, streams, sloughs) receives a
unique six- or seven-digit number called the waterbody
number. Thus, if the number 197900, assigned to Sand
Lake near Legend Lake, is the query subject, then the
output should include all data for this particular Sand
Lake. Because having a unique identifier for each spe-
cific entity in a CIS database is ideal, the waterbody
number was used, and all sampling performed on this
lake will be linked with this number.
Examples of Types of Coverage
Management Zone Coverages
Figure 2 and Table 2 together show how a potential lake
management CIS and plan might work. The lake man-
ager can easily manage and frequently update this sys-
tem if necessary, or the system can serve as a long-term
plan to consult for all decision-making. A plan of this sort
specifically emphasizes areas that need intensive man-
agement over areas that may need frequent monitoring.
It provides specific instructions for plan implementation,
leaving little guess-work forthe manager. This technique
is also visually informative to the lake user because the
user can easily discern areas of concern. This example
is hypothetical, but a plan is being formulated based on
the information collected during the Legend Lake study.
Cross-Sectional Coverages
Another technique currently included in the Legend
Lake study entails the z dimension. Cross-sectional
views of each lake basin, derived from 1992 lake con-
tour maps, provided a more detailed description of the
lake bottom. These cross sections were then digitized
and transformed into CIS coverages. For each lake
basin, 22 tests were conducted on the deepest part of
the lake at several different depth intervals along the
water column. These data provided valuable information
on the way various chemical and biological attributes
react to depth. Using CIS, a point could represent each
depth where data were collected. These points could
actually act as labels for polygons based on depth.
For example, if performing a series of analyses on a lake
(maximum depth of 10 feet) at 3-foot, 5-foot, and 8-foot
depth intervals, the labels on the cross-sectional cover-
age would be placed at these respective depths. Thus,
labels would be positioned at depths of 3, 5, and 8 feet.
Because these labels represent cross-sectional poly-
gons, the 3-foot depth label may represent a polygon
with boundaries at 0 feet and 4 feet. The 5-foot depth
label may represent a polygon with boundaries at 4 feet
and 6 feet, and the 8-foot depth label may represent a
polygon with boundaries at 6 feet and the lake bottom
(10 feet). Those who are familiar with lake ecology
understand that no clear-cut boundaries distinguish
where chemical values jump from one measurement to
another without a gradual transition. All users of a lake
quality CIS must be made aware of these types of
inaccuracies (see Figure 3).
Figure 2. Hypothetical management zones for a section of Legend Lake that correspond to the management plans in Table 2; black
areas indicate depths greater than 15 feet.
263
-------
Table 2. Hypothetical Management Plans for a Section of Legend Lake (see Figure 2)
Management Zone Management Plan (brief explanations)
I
IV
V
VI
VII
VIII
IX
X
XI
XII
XIII
XIV
XV
High recreation area, high plant growth, frequent harvesting; frequently monitor water quality
Moderate recreation; manage for fish habitat
Moderate recreation; manage for fish habitat
Open water, high recreation/possible fish habitat; consider subzoning
High-grade wildlife habitat; restrict human contact
Moderate recreation; manage for fish habitat
Open water, high recreation, increased shoreland development; frequently monitor water quality
Adjacent to high recreation area, possible fish habitat; manage for fish and aquatic habitat
Adjacent to high recreation area, shoreland development; monitor water quality
Increased development; frequently monitor water quality
aPrime wildlife/waterfowl habitat adjacent to high recreation area; restrict human presence (hot spot)
Open water, high recreation area; frequently monitor water quality
Open water, high recreation area, possible fisheries and wildlife habitat; consider subzoning
Excessive aquatic plant growth (species listed) choking out preferred species; potential wildlife/waterfowl
habitat, fisheries potential; continual harvesting; restrict human presence
High-grade slope with little plant growth, potential for increased sedimentation; monitor shoreland development
Critical area between good habitat and high recreation area; monitor extensively.
Figure 3 shows a cross section from one of the larger basins,
Basin F, in the Legend Lake system (see Figure 1). The
shades of gray represent ranges of temperature, with the
lightest being the coldest and the darkest being the
warmest. The thermocline can be located roughly in the
middle. Each colored section represents a depth range
where certain chemical attributes were collected. Using
ARC/VIEW, the user can choose these areas with a
pointer (mouse) and gain access to the database that
contains all the sample results for this depth range.
Conclusion
These ideas are still preliminary as the Legend Lake
study analysis concludes. Clear-cut discussions and
recommendations will become available at a later date,
although some observations can be made at this point.
First, future research should focus on creating and de-
veloping a three-dimensional CIS, not to be confused
with three-dimensional or cartographic models. Ideally,
lake management plans should consider depth. This
paper did not include depth because of the dimensional
factor. Depth was not compatible with ourtwo-dimensional
CIS, and the presentation quality was not sufficient to relay
our results in the form of management zones. We must
develop a three-dimensional CIS to address the problems
of the three-dimensional environment in which we live.
Lastly, maps are the best form for communicating this
information to professionals and the public. Maps can
also easily confuse people, however. Proper carto-
graphic techniques are a necessity (7). Significant effort
must be devoted to map creation to ensure a successful
plan and successful relationships between lake manag-
ers and lake users. CIS and map-making are closely
related. Both the planning stages and the database
development phase of the lake quality CIS should em-
phasize this point. At an early stage of the process,
management criteria should be determined, and all play-
ers or potential players must be included. A poorly
planned project can lead to a failed CIS.
Creativity may offer new ideas in map development. For
instance, animation (8) has some unique traits. Trend
analysis using animation may produce the best visual
results. Techniques such as these augment our methods
of communication, and some are very revolutionary.
Remember, however, that cartographic principles still
must apply.
Acknowledgments
I would like to thank Dr. Keith Rice of the University of
Wisconsin, Stevens Point, for his review efforts as well
as Dr. Byron Shaw and Steven Weber for their technical
support on lake management issues. Also, I would like
Figure 3. Cross section from one of the larger basins, Basin F,
in the Legend Lake system.
264
-------
to thank members of the University of Wisconsin,
Stevens Point, Geography/Geology Department for their
support and for the use of their equipment.
References
1. Fedra, K. 1993. Models, GIS, and expert systems: Integrated
water resources models. Proceedings of the HydroGIS '93: Appli-
cation of Geographic Information Systems in Hydrology and Water
Resources Conference, Vienna. International Association of Hy-
drological Sciences Publication No. 211.
2. Schoolmaster, F.A., and P.G. Marr. 1992. Geographic information
systems as a tool in water use data management. Water Re-
sources Bull. 28(2):331-336.
3. Lam, D.C.L., and D.A. Swayne. 1993. An expert system approach
of integrating hydrological database, models and CIS: Application
of the RAISON system. Proceedings of the HydroGIS '93: Appli-
cation of Geographic Information Systems in Hydrology and Water
Resources Conference, Vienna. International Association of Hy-
drological Sciences Publication No. 211.
4. Rifai, H.S., C.J. Newell, and P.B. Bedient. 1993. CIS enhances
water quality modeling. CIS World 6(8):52-55.
5. Samuels, WB. 1993. LAKEMAP: A 2-D and 3-D mapping system
for visualizing water quality data in lakes. Water Resources Bull.
29(6):917-922.
6. Fago, D. 1988. Retrieval and analysis system used in Wisconsin's
statewide fish distribution survey, 2nd ed. Report 148. Wisconsin
Department of Natural Resources Bureau of Research.
7. Ahner, A.L. 1993. Modern cartography plays essential role in GIS.
CIS World 6(10):48-50.
8. Loucks, D.P. 1993. Interactive multimedia, GIS, and water re-
sources simulation. Proceedings of the HydroGIS '93: Application
of Geographic Information Systems in Hydrology and Water Re-
sources Conference, Vienna. International Association of Hydro-
logical Sciences Publication No. 211.
265
-------
A Watershed-Oriented Database for Regional Cumulative Impact Assessment
and Land Use Planning
Steven J. Stichter
Division of Coastal Management, State of North Carolina Department of Environment, Health,
and Natural Resources, Raleigh, North Carolina
Introduction
In 1974, North Carolina passed the Coastal Area Man-
agement Act (CAMA) to guide growth and development
in the state's coastal zone. Today, the Division of Coastal
Management (DCM), under the direction of the gover-
nor-appointed Coastal Resources Council, implements
CAMA. DCM's jurisdiction covers the 20 counties that
border either the Atlantic Ocean or the Albemarle-Pamlico
estuary.
This coastal region comprises a diverse set of human,
animal, and plant communities. Abroad array of coastal
plain ecosystems occurs in this area, from the barrier
dunes and maritime forests of the outer banks to cedar
swamps and large pocosin complexes of interior areas.
This area includes some of the state's fastest growing
counties and some that are losing population. Urban
centers such as Wilmington do exist, but the region
remains primarily rural.
In recognition of the 20th anniversary of the passage of
CAMA, the governor designated 1994 as the "Year of
the Coast." Associated celebrations, panels, and studies
highlighted the unique features of the North Carolina
coast, successes of coastal management in the state,
and unresolved problems and concerns. Problems re-
main despite protection efforts by various agencies. For
instance:
Fish landings have dropped dramatically of late.
Shellfish Sanitation recently closed a set of shellfish
beds located in outstanding resource waters.
Shellfish statistics show that the quality of the state's
most productive coastal waters continues to decline.
Because coastal North Carolina as a whole is growing
more rapidly than any other section of the state,
pressures on coastal resources can only continue to
increase.
Declining water quality and associated sensitive habi-
tats, resources, and animal populations have prompted
several state agencies to develop new approaches to
environmental protection that incorporate a broader,
natural systems perspective. The North Carolina Divi-
sion of Environmental Management is developing river
basin plans to guide point and nonpoint water pollution
control efforts. The DCM has begun work to assess and
manage the cumulative and secondary impacts of de-
velopment and other land-based activities by using
coastal watersheds as the basis for analysis. The goal
of this work is to expand the regulatory and planning
programs in order to better address cumulative impacts.
This paper describes the approach that DCM has devel-
oped for cumulative impacts management, with special
emphasis on the use of a geographic information system
(CIS). The project described here is scheduled for com-
pletion by the fall of 1996.
Cumulative Impacts Management
The concept of cumulative impacts management (1) is
not new to North Carolina's coastal program. CAMA
requires the consideration of cumulative impacts when
evaluating development permits within defined areas of
environmental concern. A permit must be denied if "the
proposed development would contribute to cumulative
effects that would be inconsistent with the guidelines.. .."
Cumulative effects are defined as "impacts attributable
to the collective effects of a number of projects and
include the effects of additional projects similar to the
requested permit in areas available for development in
the vicinity" (2). Despite this directive, few permitting
actions have been denied because of cumulative ef-
fects; the existence of limited impact data and a dearth
of viable analysis approaches have restricted applica-
tion of this rule.
Since the passage of the National Environmental Policy
Act (NEPA) in 1969, many attempts have been made to
266
-------
define and assess cumulative impacts. The Council on
Environmental Quality developed the most familiar defi-
nition in its guidelines for NEPA implementation. It de-
fines cumulative impact as:
... the impact on the environment which results
from the incremental impact of the action when
added to other past, present, and reasonably fore-
seeable future actions, regardless of what agency
or person undertakes such other actions. Cumula-
tive impacts can result from individually minor but
collectively significant actions taking place over a
period of time (3).
Although this definition focuses the discussion of cumu-
lative impacts, it provides little guidance on how to carry
out such an analysis. Selecting both an appropriate time
frame for the assessment (how far into the past and
future to carry the analysis) and appropriate boundaries
for the study (municipal or county boundaries, water-
sheds, ecoregions) are but two of the questions that
require answers to successfully investigate cumulative
impacts. Such decisions become even more complex
when incorporating the limits imposed by available data
and existing management structures.
Rigorous cumulative impact analysis is a difficult propo-
sition. It requires identification of all sources of degrada-
tion that affect a given resource. The next step involves
assigning relative significance to each of these sources
along with any impacts that result from additive or syn-
ergistic interaction between sources. Assessment of the
impacts of a pier on surrounding sea grasses, for in-
stance, must include not only impacts related to the
structure, such as shading and wave or current changes,
but also such ambient impacts as natural wave and wind
effects, upland runoff, and varying salinity.
All these investigations require the availability or collec-
tion of baseline environmental data at an appropriate
spatial and temporal scale. Quantifying all the sources
and causal pathways that affect a resource is extremely
complicated in all but the simplest of systems. Assigning
proof of significant impact is difficult unless the cause is
clear and direct.
Because of the difficulties associated with assigning
cause in cumulative impact analysis, especially in a
regional review, DCM has chosen a different approach.
It is focusing instead on locating areas at high risk to
cumulative impacts. Impacts management studies and
responses can then target the areas at greatest risk of
degradation. Changing the scale of analysis from the
site to the region requires applying some simplifying
assumptions. The first assumes that any existing re-
source degradation results from the cumulative impact
of all sources within the system boundaries. Locating
such areas is relatively straightforward because most
natural resource fields have developed measurements
and indicators for locating degraded resources. The
second assumption claims that a sufficiently intensive
concentration of activities within a limited area will result
in cumulative impacts on the affected system. Determin-
ing a threshold beyond which impacts cause degrada-
tion is much harder than locating already degraded
resources because the level of such a threshold de-
pends upon both the strength or spatial concentration of
the impacts and the sensitivity of the resource.
Working from these simplifying assumptions, DCM's
first step in assessing regional cumulative impacts is to
identify areas within coastal North Carolina that exhibit
symptoms of resource degradation, contain a concen-
tration of activities that affect resources, or contain a
concentration of sensitive resources. The use of catego-
ries of resources and impacts have helped to focus this
search. These eight cumulative impact, high-risk area
categories are:
Impaired water quality
High potential for water quality impairment
Sensitive ground-water resources
Impaired air quality (present or potential)
Historical rapid growth
Anticipated high growth
High-value resources
Productive and aesthetic resources
This set of categories is presently under public review. The
next step is to develop indicators of the presence of im-
pacts or resources appropriate to each of these categories.
These indicators, when applied to a database of information
about the study area, will help identify those locations at
high risk as defined by the eight categories.
The regional cumulative impacts assessment approach
that DCM developed is a hybrid of various assessment
techniques. The overall approach is grounded in the
theory and methods of site-specific cumulative impact
assessment. Determination of high-risk categories and
appropriate indicators and indexes is closely associated
with both relative risk assessment procedures and geo-
graphic targeting. By focusing on known causes and
effects of cumulative impacts on terrestrial and aquatic
natural resources instead of attempting to quantify all
impact pathways, available data and analysis techniques
can help assess relative risk of cumulative impacts.
A Watershed Database for Cumulative
Impact Assessment
High-risk categories and indicators of degradation or
sensitivity are useless without information on the location
of sensitive resources and impact sites. Consequently,
a comprehensive database of information about coastal
267
-------
North Carolina is central to cumulative impacts manage-
ment in this area. The form of any database determines
what types of questions to ask it; the selection of
boundaries has been central to this study.
County boundaries constitute the most typical reporting
unit in DCM operations. Counties determine the
boundaries of DCM's jurisdiction, and the great majority
of statistics used in planning and assessment are avail-
able primarily or solely by county. County size, hetero-
geneity, and the small number of counties available for
comparison, however, have made county boundaries
inappropriate for this project. Because the study focuses
on impacts on natural resources, clearly the most appro-
priate boundaries would relate more directly to those
resources.
Although using a single set of boundaries may not be
appropriate for assessing impacts on all resources,
management constraints limit the choice to one bound-
ary type. Because the primary resources of concern are
water based, watersheds were considered most appro-
priate. Surface waters receive the integrated effects of
activities within a watershed; such boundaries fit intui-
tively with the concept of cumulative impact assess-
ment. The number of water-related resources of concern
also supported this choice. This analysis used small
watersheds (5,000 to 50,000 acres) delineated in 1993
by the Soil Conservation Service for the entire state of
North Carolina.
The Population, Development, and Resources Informa-
tion System (PDRIS), which was designed for this pro-
ject, is a PC-based, watershed-oriented database that
contains the following information about the coastal
area:
Natural resources
Population and housing
Agricultural activities
Economic activities
Development activities
Table 1 includes a list of database fields. The presence
and extent (or absence) of each of the features that this
database represents will be available for each coastal
watershed. The small watershed orientation of this study
is only possible because of the availability of CIS; the
volume and complexity of the watershed boundaries
preclude any other assessment tool. In fact, 348 of these
watersheds fall wholly or partially in the 20-county re-
gion. Figure 1 shows a map of these small watersheds.
This map indicates county boundaries and shorelines in
solid lines and the watershed boundaries in gray.
Data Needs and GIS Analysis
Over the past 5 years, North Carolina has actively col-
lected a large amount of natural resource and base map
information in GIS form. Research and funding associated
with the Albemarle/Pamlico Estuarine Study (APES), a
national estuary program study, spurred much of this
data development in the coastal area. The state main-
tains a central repository for geographic data at the
North Carolina Center for Geographic Information and
Analysis (CGIA). Table 2 lists the general types of infor-
mation available from the state database. The availabil-
ity of data in GIS form is but one criterion for selecting
a data set for use in this analysis. To be useful, the scale
and accuracy of the data must be appropriate to the
analysis.
Data Scale
The majority of data in the state's GIS database was
collected at a scale of 1:100,000. Broader use and
interest will probably urge the development of data lay-
ers at finer scales. A recently released layer of closed
shellfish waters, for instance, was created at 1:24,000
scale. This prompted an update of the associated shore-
line coverage to the same base scale. A handful of state
departments and divisions, including Coastal Manage-
ment, now use global positioning systems to collect
even more precise locational information. This scale
suits DCM's regional cumulative impacts scan, which is
based on summary values for entire watersheds. More
detailed intrawatershed planning and analysis would
require finer scale data. A scale of 1:24,000 delineated
the watershed boundaries in this project.
Mixing these 1:24,000 boundaries with the 1:100,000
data sets, however, can cause problems. For instance,
a number of watersheds were designated for the large
open water areas in Albemarle and Pamlico sounds.
Although these should comprise exclusively water, over-
lay analysis of these watershed boundaries on the Tl-
GER-derived census boundaries (1:100,000 scale)
resulted in the assignment of small population counts to
some of these watersheds. Individually locating and
correcting such discrepancies is necessary.
Database Accuracy
Data layer accuracy problems are difficult to identify and
assess. Because other agencies developed the majority
of data used in this project, these source agencies must
be relied upon for accuracy assessment of the source
data. CGIA, steward of the state GIS database, adheres
to National Map Accuracy Standards for all GIS data that
it maintains. CGIA delivers metadata reports with any
data; these reports include the source agency, collection
date, and scale for the information used to derive the
GIS layer. Descriptions of data lineage (collection and
processing procedures), completeness, and positional
268
-------
Table 1. Population, Development, and Resource Information System: Database Fields
Agriculture: Livestock and Poultry
Beef feedlots (< 300 head, > 300 head)
Dairy farms (< 70 head, > 70 head)
Hog farms (< 200 head, > 200 head)
Horse stables (< 200 head, > 200 head)
Poultry farms (< 15,000 birds, > 15,000 birds)
Agriculture: Farming
Land in farms (acres, % of HU)
Land with best mgmt. practices (acres, % of HU)
Land w/o best mgmt. practices (acres, % of HU)
Land in conservation tillage (acres, % of HU)
Land w/o conservation tillage (acres, % of HU)
Harvested cropland (acres, % of HU)
Hay crops (acres, % of HU)
Irrigated land (acres, % of HU)
Pasture land (acres, % of HU)
Row crops (acres, % of HU)
Primary
Estuarine waters (acres, % of HU)
Freshwater lakes
HU name
Receiving HU
Receiving water body
Primary water body
Secondary water body
Shoreline
Waterways w/vegetated buffers (miles, % of HU)
Population 1970
Population 1980
Population 1990
Population growth 1970 to 1980
Population growth 1980 to 1990
Counties
Total HU size
Land area (acres, % of HU)
Water area (acres, % of HU)
Stream length (miles)
Stream order (miles, % of stream length)
Development
Building permitsall residential
Building permitsamusement/recreation
Building permitsmultifamily residential
Building permitsone-family residential
Building permitshotels and motels
Building permitsretail
Building permitsindustrial
Highway mileage:
Total (miles)
Primary (miles, % of total)
Secondary (miles, % of total)
Paved (miles, % of total)
Unpaved (miles, % of total)
Rail lines (miles)
Increase of primary & secondary roads (miles, %)
Increase of paved vs. unpaved roads (miles, %)
Economic
Ag-related business (number, employees, income)
Farms (number, employees, income)
Fisheries business (number, employees, income)
Forestry/wood-using business (number, employees, income)
Lodging establishments (number, employees, income)
Manufacturing establishments (number, employees, income)
Marinas (number, employees, income)
Mining establishments (number, employees, income)
Recreation business (number, employees, income)
Restaurants (number, employees, income)
Retail establishments (number, employees, income)
Ground Water
Ground-water contamination incidents
Ground-water class (acres, % of HU)
Ground-water contamination area (acres, % of HU)
Ground-water capacity use areas (acres, % of HU)
Land and Estuarine Resources
Anadromous fish streams (miles, % of streams)
Coastal reserve waters (acres, % of HU)
Coastal reserve lands (acres, % of HU)
Federal ownership:
National parks (acres, % of HU)
National forests (acres, % of HU)
Military reservations (acres, % of HU)
USFWS refuges (acres, % of HU)
Federal ownershipother (acres, % of HU)
State ownership:
Game lands (acres, % of HU)
State parks (acres, % of HU)
State forests (acres, % of HU)
State ownershipother (acres, % of HU)
Natural heritage inventory sites (count)
Primary nursery areas (acres, % of water area)
Private preservation (acres, % of HU)
Secondary nursery areas (acres, % of water area)
Threatened/endangered species habitat
Water supply watersheds (acres, % of HU)
Land Use
Total wetland area (acres, % of HU)
High-value wetlands (acres, % of HU)
Medium-value wetlands (acres, % of HU)
Low-value wetlands (acres, % of HU)
Predominant land cover
Population and Housing
Average seasonal population
Peak seasonal population
Units without indoor plumbing
Units with septic tanks
Units on central water systems
Units on central sewer
Units with wells
Permits
Air emission permitsPSD
Air emission permitstoxic
CAMA minor permits
CAMA general permits
CAMA major permits
CAMA exemptions
CWA Sect. 404/10 permits
Landfill permitsmunicipal
Landfill permitsindustrial
Nondischarge permits
NPDES permitsindustrial
NPDES permitsother
NPDES permitsPOTW
Stormwater discharge permits
Sedimentation control plans
Septic tank permits
Shellfish
Shellfish waters (acres, % of HU)
Shellfish closurespermanent (acres, % of HU)
Shellfish closurestemporary (acres, % of HU)
Water QualityOpen Water
Class B waters (acres, % of water area)
Class C waters (acres, % of water area)
HQW waters (acres, % of water area)
NSW waters (acres, % of water area)
ORW waters (acres, % of water area)
Swamp waters (acres, % of water area)
SA waters (acres, % of water area)
SB waters (acres, % of water area)
SC waters (acres, % of water area)
WS-I waters (acres, % of water area)
WS-II waters (acres, % of water area)
WS-III waters (acres, % of water area)
269
-------
Table 1. Population, Development, and Resource Information System: Database Fields (Continued)
Water QualityStreams
Class B streams (miles, % of streams)
Class C streams (miles, % of streams)
HQW streams (miles, % of streams)
NSW streams (miles, % of streams)
ORW streams (miles, % of streams)
Swamp water streams (miles, % of streams)
SA streams (miles, % of streams)
SB streams (miles, % of streams)
SC streams (miles, % of streams)
WS-I streams (miles, % of streams)
WS-II streams (miles, % of streams)
WS-III streams (miles, % of streams)
Key
HU = hydrologic unit
PSD = point source discharges
POTW = publicly owned treatment work
NPDES = National Pollutant Discharge Elimination System
HQW = high-quality waters
NSW = nutrient-sensitive waters
ORW = outstanding resource waters
SA = saltwater classification A
SB = saltwater classification B
SC = saltwater classification C
WS1 = water supply classification 1
WS2: water supply classification 2
WS3: water supply classification 3
Water QualityUse Support
Algal blooms (count, extent/severity)
Fish kills (count, extent/severity)
Streams fully supporting (miles, % of streams)
Streams support threatened (miles, % of streams)
Streams partially supporting (miles, % of streams)
Streams nonsupporting (miles, % of streams)
Waters fully supporting (acres, % of water area)
Waters support threatened (acres, % of water area)
Water partially supporting (acres, % of water area)
Waters nonsupporting (acres, % of water area)
accuracy are not available from these standard metadata
reports, however.
DCM's cumulative impacts analysis also incorporates
information not available from the state CIS database.
Some of this information, such as business locations, is
available from private data providers. Other information,
especially agricultural statistics, does not presently exist
in CIS form. Non-GIS formats include county statistics,
voluntary compliance databases with self-reported co-
ordinates, and other tabular databases. Typically, little
quality control has been performed on any coordinate
information. When the data originate from other state
agencies, DCM is often the first user of the data outside
of the source agency.
CIS Analysis Procedures
This study involves no sophisticated CIS analysis pro-
cedures. CIS helps to generate summary statistics by
watershed for each of the database features. CIS draw-
ing and query operations allow analysis of database
accuracy. If the data are acceptable, the next step re-
quires overlaying the watershed boundaries on the fea-
ture and assigning the appropriate watershed codes to
all features that fall within the study area. Statistics can
then be generated on the number of points, length of
lines, acreage of polygons, or a total of any other nu-
meric field in the feature coverage. Finally, the resulting
summary file is converted to the format that PDRIS
requires. A macro has been developed to complete this
analysis. This macro generates a page-size reference
map, performs the overlay, generates the watershed
summary statistics, and converts the statistics to the
final PC format.
Extra steps are necessary to analyze any information
that does not already exist as a CIS coverage. Typically,
these are tabular summaries associated with a specific
boundary layer, such as county or U.S. Census statis-
tics. These cases entail overlaying the watershed
boundaries on the reporting unit boundaries; the data
are distributed to the watershed in direct proportion to
the percentage of the unit that falls into the watershed.
For instance, if a census tract falls 30 percent into
watershed A and 70 percent into watershed B, 30 per-
cent of the total tract population will be assigned to
watershed A and the remainder to B. After performing all
assignments, summary statistics are again generated
by watershed. This procedure assumes that the distri-
bution of the feature is even across each reporting unit.
Rarely is this a valid assumption, but when the units are
considerably smaller than the watersheds, as is the
case with census tracts and blocks, this assumption
introduces only limited errors. Watershed estimates
based solely on county statistics, however, can be
grossly inaccurate. When working with county informa-
tion, therefore, using covariate information that ties
more precisely to specific locations is necessary. Crop-
land location derived from the LANDSAT land cover
layer, for instance, can be used to better distribute
county-level agricultural statistics.
Database and Analysis Documentation
Data documentation is essential to this project. Given
the large number of fields in the final database and the
270
-------
Figure 1. Watersheds in the North Carolina coastal area.
correspondingly large number of data types and
sources, such documentation is key to understanding
the quality of the individual database components as
well as easing future database additions and updates.
Because the results of this cumulative impact analysis
exercise will be used to extend DCM's resource man-
agement efforts, documenting data sources and analy-
ses will be critical if any decisions made based on this
information are disputed.
A metadata database has been developed to document
PDRIS data sources and analysis procedures. For each
Table 2. A Sample of North Carolina GIS Database Contents
Type Examples
database entry, fields exist for a description and contact,
collection methodology, and geographic extent of the
source data. Data selection, overlay, and conversion
procedures are also documented, along with any as-
sumptions made in the analysis. In addition, recording
data source, analysis procedure, and the date facilitates
future database updates. Fields also record accuracy
assessments for positional and attribute accuracy, logi-
cal consistency, and completeness.
The restrictions listed above regarding source data ac-
curacy assessment, however, have limited their use.
Once an entry is made to the PDRIS, all project team
members receive metadata reports along with a refer-
ence map for a final review of completeness of the
source data, data selection, and analysis logic. Figure 2
shows an example of a blank metadata worksheet.
Status of the Cumulative Impacts
Assessment
DCM is presently gathering, verifying, and analyzing
information for entry into the PDRIS. Although each
source data layer was checked for accuracy before use,
the logical consistency of each of the database entries
relative to the other components also needs addressing.
One example of such database inconsistencies is the
watersheds that are covered entirely by water but also,
according to the database, support a resident popula-
tion. These inconsistencies could result from problems
related to scale, differing category definitions, data inac-
curacies, or errors in the GIS conversion or analysis at
DCM. Database precision is essential for an accurate
analysis and for general support of DCM's cumulative
Coverage
Natural resource
Base data
Ownership
Permits, waste sites
Cultural, population
Fishery nursery areas
Natural heritage sites
Detailed soils
Closed shellfish areas
Water quality use classes
Detailed wetlands maps
Hydrography (24K, 100K)
Roads/transportation
County and city boundaries
LANDSAT-derived land cover
Federal and state ownership
NPDES permit site
Landfills, hazardous waste, Superfund
TIGER boundaries, census information
Historic register sites, districts
Coastal North Carolina
Statewide
Varied
Coastal North Carolina
Statewide
Varied
Statewide
Statewide
Statewide
Coastal North Carolina
Statewide
Statewide
Statewide
Statewide
Statewide
271
-------
General Information:
Field
Description
Database
Definition
Source Data Description:
Contact
Data
Scale
Sample Method
Geographic Extent
Database Entry:
Procedures
Assumptions
Accuracy Assessment:
Rating
Logic Test
Error Comment
Value Range
Update Procedure:
Next Update
Procedure
Overall
Positional
Units
Attribute
Logical
Consistency
Completeness
Source
Figure 2. Population/Development database: Data dictionary
impacts approach. Because the watershed database pro-
duced for this project will be widely available, errors and
inconsistencies will undermine support for the rest of the
project. Careful documentation of data sources, limita-
tions, and analysis assumptions and procedures will pro-
vide useful support should problems or concerns arise.
Once database development is sufficiently complete
(the database encompasses much dynamic data and
could be constantly updated), indexes describing each
of the cumulative impact, high-risk areas must be final-
ized. Applying these indexes to the database will allow
identification of the watersheds at highest risk to cumu-
lative impacts. Discussions held concurrently with index
development will determine which management re-
sponses are appropriate to each high-risk category.
Possibilities include strengthened land use planning re-
quirements, new permit standards, or the designation of
a new type of environmental critical area.
Although the data-intensive approach that DCM has
chosen relies heavily on a CIS, the greatest challenges
in this project do not lie in the CIS analysis. Applying this
watershed-based analysis to existing political jurisdic-
tions will be a more difficult undertaking. A convincing
demonstration of the importance of including a natural
systems perspective into a development permitting sys-
tem, land use plan, or even economic development
strategy, will ultimately contribute more to environmental
protection in coastal North Carolina than any individual
regulation that emanates from this project.
Summary
Twenty years after the passage of CAMA, DCM has
developed a framework for a consistent approach to the
problem of cumulative impacts of development. The
approach and PDRIS database combine existing natural
resource management techniques to locate areas of the
coast at greatest risk of serious impairment from cumu-
lative impacts. The availability of natural resource data
at an acceptable scale (1:100,000) eases the develop-
ment of the database essential to this analysis. The
simultaneous development of a set of comprehensive
small watershed boundaries for the state, along with the
initial planning of this project, provided the final critical
component to DCM's approach.
Perhaps more importantly, both DCM and individual
local governments will have a large volume of informa-
tion on natural units, which will provide an important,
new perspective on the problems and prospects for local
governmental action.
This project will not solve all problems related to cumu-
lative impacts. The PDRIS will provide little support for
site-specific or within-watershed cumulative impacts
analysis; such an analysis at fine scales requires a much
more precise database. By providing a broader-scale
framework for this discussion, however, DCM's regional
cumulative impacts study will hopefully further discussion,
understanding, and management of cumulative and
secondary impacts on natural systems.
References
1. Wuenscher, J. 1994. Managing cumulative impacts in the North
Carolina coastal area. Report of the Strategic Plan for Improving
Coastal Management in North Carolina. North Carolina Division of
Coastal Management.
2. North Carolina General Statutes (NCGS) 113A-120.
3. 40 CFR §1508.7.
272
-------
A GIS Demonstration for Greenbelt Land Use Analysis
Joanna J. Becker
Environmental Planning Services, Santa Rosa, California
Purpose
The goal of this project was to demonstrate what analy-
ses could be undertaken with a GIS program without
substantial GIS training ortime input. The demonstration
attempted to show how planning staff and decision-makers
could easily and usefully employ GIS. It was not intended
as a complete study of all possible variables. Only avail-
able data were used. Diverse techniques were pre-
sented while keeping the content as simple and relevant
as possible. The project was designed as a demonstra-
tion using regional scale data and was combined with
another parcel-specific demonstration that showed
urban GIS applications.
The demonstration showed the following modeling
techniques:
Buffer zones
Combination of variables (overlays)
Weighting of values
Absolute value variables
Reclassification of final values
Site Location
The San Luis Obispo watershed comprises an area of
approximately 84 square miles. The watershed drains
into the Pacific Ocean at Avila, California. The major
creek in the watershed, San Luis Obispo Creek, is a
perennial creek, but many of its tributaries have only
seasonal flow. Agriculture and grazing are the major
land uses in the watershed, although a significant num-
ber of areas have been developed. Growth of these
areas is moderate to limited but has a pronounced effect
on the watershed. The watershed also supports a large
amount of riparian and other natural vegetation. Figure 1
demonstrates the distribution of land cover/land use
within the watershed.
Watershed
Study Perimeter
City Greenbelt
Pacific Ocean
Figure 1. A generalized map of the San Luis Obispo area show-
ing the location of the two ARC/INFO GIS study areas:
(1) the parcel analysis in the Dalideo area and (2) the
regional analysis in the city greenbelt stippled area
surrounding the city. All boundaries are approximate
and are for schematic purposes only (1, 2).
Background
The City Council approved the open space element of
the San Luis Obispo General Plan in January 1994 and
identified a greenbelt area that extended from the Urban
Reserve Line approximately to the boundaries of the
San Luis Obispo Creek watershed. The intent of this
greenbelt area is to provide a buffer to the city and to
preserve the agricultural and natural resources of the
area.
273
-------
The data forthe watershed were already available through
the work of the Landscape Architecture Department of
California Polytechnic State University at San Luis
Obispo for a study of San Luis Obispo Creek (1), and
the city's greenbelt area lay approximately within those
boundaries (see Figure 1). Variables were selected that
could be extracted from the available data.
The data for the creek study were initially entered into
workstation ARC/INFO in a polygon format. They were
then transferred to MacGIS, a PC raster program, for
simplicity of use. The final product was then transferred
back to ARC/INFO as a grid format and viewed in a PC
version of ARC/VIEW using DOS files.
In interpreting the overlay of values, the assumption was
made that the occurrence of high values for the most
variables would result in the most suitable land for that
land use. This was presented as a range of values
derived from the total values divided into three approxi-
mately equal groups of high, medium, and low. In addi-
tion to providing a composite analysis, however, any one
of the data sets can be queried separately such that, for
example, slopes greater than 20 percent could be iden-
tified or two layers such as storie index and distance
from roads could be compared.
Criteria
Note that the ratings of high, medium, and low are based
on available data, and the rating of low implies no suit-
able use. In addition, these ratings do not imply that
categories rated low could not be used for a particular
land use but rather that other land uses might be more
appropriate. For example, open space use was rated
low for flatter slopes but only because this category
would likely be more suitable for agriculture.
The demonstration used an existing cell size of the data
on the MacGIS program of 75 meters per side, which is
assigned during the initial conversion process. There-
fore, buffer zones are in aggregates of 75 meters. This
size cell does not allow for minute analysis but reduces
the size of the files, which may become extensive in
raster format.
In presenting the final analysis, land contained within the
urban reserve limit line has been excluded.
Procedure
Initially, eight variables from the available data were
deemed suitable for this analysis:
Slope
Storie index (indicating soil fertility)
Distance from major roads
Distance from creeks
Erosion hazard1
Oak woodlands
Land use compatibility
Grasslands2
After selecting the six variables, the categories were
receded to conform to a rating of high, medium, or low.
Each land use was then evaluated separately for every
variable except for combining the variables of slope and
distance from creeks for rangeland analysis. In this case,
composite values were assigned to the two variables, then
receded to produce a high, medium, or low rating.
After obtaining maps for each of the variables according
to land use, the maps were compiled to indicate the
density of overlays for each land use category. In as-
sessing the suitability of land for the three land uses, the
values of all the variables, except land use, were aggre-
gated and a rating system developed. In addition, a
double weight was assigned to the storie index in evalu-
ating agriculture (because this is a primary index for
considering prime agricultural land). If less than 75 me-
ters, the distance from creeks also received added
weight in consideration of open space preservation (be-
cause this is likely to ensure the least erosion and
pollution to the waterways). The weighting then altered
the scores as follows:
Land Use
Agriculture
Rangeland
Open space
Attributes
5
4
5
Number of Values
6
4
6
After the values had been assigned for each ranking,
further receding established three categories of high,
medium, and low for each land use.
The land use buffer was added to this receded aggre-
gate map and resulted in an additional three values due
to the interaction of the buffer with each category. These
additional values were receded according to each land
use to produce a final map with three values.
The Urban Reserve Area was overlaid on the final map
to exclude urban areas.
Assumptions
In determining what properties would be most suitable for
each land use, the following assumptions were made.
1 After reviewing the material, erosion hazard was eliminated because
it was similar to the Soil Conservation Service (SCS) storie index
data while the identification of native grasslands fell only within the
area currently designated as open space land, so it was not included.
2 See above note.
274
-------
Open Space
This land is desirable to preserve as open space be-
cause of the existence of scenic or significant natural
resources. It could also be land that is inappropriate for
other uses due to the presence of such factors as steep
slopes or poor soils. A distinction is made in the final
map between land that is designated open space for
recreational uses, such as parks, and land preserved for
habitat or species protection. Open space adjacent to
an urban area would be rated high if public accessibility
was desirable but low if its purpose was resource pro-
tection or preservation. Separate maps based on two
types of proposed uses present the contrast in analysis.
The analysis of the variables thus was rated as follows:
A. Land with steep slopes and therefore less suited for
other purposes.
B. Land that has oak woodland vegetation resources.
C. Riparian land.
D. Low storie index indicating a lack of suitability for
agriculture or rangeland.
E. At least approximately one-eighth of a mile from a
major road to avoid negative impact on wildlife.
F. Either approximately one-quarter of a mile from ur-
ban areas if designated to protect resources or adja-
cent to urban areas if designated to serve as parks and
recreation.
Agriculture
This land use includes all forms of agricultural activity;
obviously, its suitability for specific crops and practices
would vary. The determination of suitability would need
to be made on a site-specific basis.
For the purpose of general agricultural suitability, the
highest land suitability for agriculture was a rating of the
variables as follows:
A. Land that does not have oak woodland.
B. Land that is not close to perennial creeks (to avoid
fertilizer/pesticide runoff contamination).
C. The flattest slopes.
D. The highest storie index.
E. Proximity to a major road (considered an advantage
for trucking and farm equipment access).
Rangeland
Some types of livestock can graze under most condi-
tions, but for purposes of this analysis, land more suited
for either open space or agricultural designation was
rated above that of rangeland. The major limitation to
suitability of land for rangeland was a combination of
steep slopes and proximity to creeks. A rating of medium
for the other variables was considered the most desir-
able for rangeland purposes.
Details of the Variables
Storie Index
In determining the most suitable uses according to soil
fertility, the SCS storie index rating was used, with a
modification of the categories to three to accommodate
the ratings of high, medium, and low that were used
throughout the analysis. Therefore, the first two SCS
categories of excellent and good were combined into
Category 1. Categories fair and poor were combined
into Category 2, and categories very poor, nonagricul-
tural, urban, and mines were combined to compose
Category 3.
Subsequently, receding was undertaken to prioritize
these categories according to land use:
Agriculture
Rangeland
Open space
Roads
High for Category 1
High for Category 2
High for Category 3
The five principal arterials of the watershed were used
in this analysis:
Highway 1
Highway 227
Los Osos Valley Road
U.S. 101
Avila Valley/San Luis
A buffer on each side of the road was created, which
was then receded into the three categories of more than
75 meters, 75 meters to 150 meters, and greater than
150 meters. Each land use was then evaluated accord-
ing to these criteria, with agriculture deemed the most
suitable closest to the roads and open space the least
suitable.
Slope
The slope categories in the San Luis Obispo watershed
data set were divided into a number of classes, which were
receded into the most appropriate grouping of three
classes for the analysis of the three land uses. The existing
categories were not altered for the purpose of the demon-
stration, so they do not necessarily represent the most
ideal slopes for the particular uses. A separate category of
less than 10 percent slopes was provided for agriculture
because most agricultural practices require flat land.
275
-------
Slopes between 10 and 21.5 percent would be limited
to such activities as orchards or vineyards.
Agriculture:
Open space:
< 10 percent
10 to 21.5 percent
> 21.5 percent
< 10 percent
10 to 46 percent
> 46 percent
High (H)
Medium (M)
Low (L)
L
M
H
No slope analysis was undertaken for rangeland be-
cause this category was combined with that for streams
(see Streams section).
Urban Adjacency
Urban adjacency was treated as an overlay of the ag-
gregate map of the other variables because it is an
absolute value. That is, this variable has no ranking.
Land is either within or outside the buffer. The suitability
of each land use adjacent to urban areas was deter-
mined, then the aggregate map was adjusted according
to a comparison of the receded aggregate values with
the designated ranking of land use suitability.
The first step was to recede the existing data on land
uses (interpreted from 1989 aerial photography ob-
tained from the United States Department of Agriculture
[USDA], Agricultural Stabilization and Conservation
Service, Atascadero, California) into urban/commercial
areas and nonurban/noncommercial areas. A buffer
zone of approximately one-quarter of a mile was then
applied around the urban/commercial areas and an
analysis undertaken of suitability for the three land uses
to be located within this buffer.
In making the analysis, the following assumptions were
made:
Land Use
Ranking
Agriculture L
Rangeland M
Open space H
L
Conflict with dust, noise,
pesticides, and urban use
Fire hazard of open
grassland near buildings
Most suitable if used as
parks/recreation
Least suitable if
designated for
habitat/wildlife protection
Therefore, the analysis provided for two planning alter-
natives for open space, with the scenarios presented as
separate maps overlaid on the aggregate map for the
other rangeland variables.
In interpreting this map, the combined values were rated
according to the above criteria, with any values lower
than the desired ranking receiving a low value no matter
what the aggregate value had been. This produced the
following results:
Agriculture: A rating of low for any land within the
buffer zone.
Rangeland: A rating of medium for any land within
the buffer zone except that rated low for the aggre-
gate map.
Open Space: A rating equivalent to the rating of the
aggregate map if the land was to be used for
parks/recreation, or a rating of low for any land within
the buffer zone if the land was to be used for habitat
protection.
Streams
The original data for streams (from 1:24,000-scale
USGS maps), which included both intermittent and per-
ennial creeks, were used for rangeland analysis. A buffer
of 150 meters was then applied to these streams, and
these data were combined with the slope analysis. This
combination was important in evaluating the erosion
hazard and resulting stream pollution caused by nitro-
gen waste and hoof disturbance. The complete stream
complex for the watershed area was therefore evaluated
using the following matrix:
Stream Buffer
< 75 meters
75 to 150 meters
> 300 meters
<21.5
percent
1/L
4/M
7/H
Slope
21.5 to 46
percent
2/M
5/H
8/H
>46
percent
3/L
6/M
9/M
The stream data were then modified for agricultural and
open space land use analysis to indicate only the per-
ennial creeks as defined by the California Department
of Fish and Game. Buffers were created for this as
follows:
Stream Buffer
< 150 meters
150 to 300 meters
> 300 meters
Oak Woodlands
Open Space Agriculture
H
M
L
L
M
H
The receding of oak woodland data for agriculture was
different than for the other two land uses because
the presence of oak woodlands is not conducive to
agriculture:
No oak woodlands
< 10 percent
> 10 percent
H
M
L
276
-------
The suitability of oak woodlands for open space and References
rangeland was ranked as follows: ^ Ha||ock B G L s Bowker WD Bremer and D N Long 19g4
Open Space Rangeland Nutrient objectives and best management practices for San Luis
* * " Obispo Creek. San Luis Obispo, CA: Coastal Resources Institute,
California Polytechnic State University.
< 33 percent L M
2. Community Development Department of San Luis Obispo. 1994.
33 to 75 percent M H Open space element for the city of San Luis Obispo general plan.
> 75 percent - high H L san Luis Obispo, CA.
277
-------
GIS as a Tool for Predicting Urban Growth Patterns and Risks
From Accidental Release of Industrial Toxins
Samuel V. Noe
University of Cincinnati, Cincinnati, Ohio
Introduction
The catastrophic Bhopal incident demonstrated to the
world what could happen when industry and population
are geographically incompatible. Many believe that the
large urban population "should not have been there." A
recent publication, "New York Under a Cloud," presents
a frightening map of New York State that indicates po-
tential areas of serious population exposure to acciden-
tal releases of chemicals stored by area industries and
municipalities.
Conventional urban planning and administrative prac-
tices at the local level do not adequately provide for the
minimization of these risks. Local jurisdictions on the
fringes of metropolitan areas may be particularly ill-
equipped to respond and plan effectively. Their elected
officials, supported by minimal professional staffs and
unaware of specific potential risks, may be more inter-
ested in soliciting new industrial development along with
the tax base it brings. They therefore create industrial
zones without restricting facilities that may generate
hazardous substances and without recognizing the pos-
sibility that underground aquifers, which are current or
potential sources of drinking water, may underlie these
zones. Jurisdictions often permit facilities that could rou-
tinely or accidentally release toxic substances into the
air without due regard for prevailing wind patterns or
existing or projected urbanized areas that may be
affected.
Although the available data and methodology have
some gaps, much of the knowledge required to provide
adequate protection from these risks exists. We know
how to identify the hazardous substances that these
sites may produce or store and how to calculate the
types and levels of risks associated with them. We can
accurately map the locations of streams, underground
aquifers, and their catchment areas. Although with less
precision, we also can indicate the areas more likely to
receive the outfall of airborne and waterborne pollutants.
At the other end of the equation, we can construct
models for projecting patterns of urban growth. Those
responsible for planning, however, have not made the
connections between these techniques. Hazardous fa-
cilities sites are thus still permitted in areas that place
existing urban residents and their drinking water sup-
plies at risk, and new urban development grows in areas
polluted by existing hazardous substance sites.
Clearly, this situation displays a need forthe coordinated
application of scientific risk assessment techniques and
new approaches to regulating urban development.
Equally critical, however, is the need to give greater
attention to formulating appropriate public policy meas-
ures at the local and state levels for dealing with the
complex disputes that surround these issues.
Project Background
The project this paper describes addressed these
needs. It was undertaken by a team of faculty from the
University of Cincinnati's School of Planning, Depart-
ment of Environmental Health, and the College of Law.
The study team focused on the accidental release of
hazardous materials both into the air and into the
ground-water supply. The team's purpose was to de-
velop an integrated approach to scientific risk assess-
ment, environmental analysis, urban planning, and
policy analysis to address conflicts between:
Expected patterns of suburban residential growth.
The need to safeguard existing and new residential
areas, and their water supplies, from toxic chemical
pollution.
The promotion of industrial development on the pe-
ripheries of urban regions, which often leads to the
proliferation of hazardous substances sites.
The need for effective regulation of these sites in
complicated multijurisdictional environments.
278
-------
This project examined these issues within a 100-square-
mile area on the northern edge of metropolitan Cincin-
nati. The study area is not yet completely urbanized but
lies in the path of urbanization. It contains a significant
number of industrial or storage facilities that house sup-
plies of hazardous materials. A major aquifer serving as
a public water supply source passes under the area.
Approximately 17 local jurisdictions fall within the study
area: two counties, six townships, and nine municipali-
ties. The area encompasses an intricate mix of agricul-
tural, residential, and commercial land uses. In addition,
several major industrial concentrations, as well as a
number of jurisdictions, are aggressively soliciting new
industrial employment. Because of its proximity to most
major employment sites in southwest Ohio and to a
variety of large retail complexes, the area is experiencing
rapid residential development.
Projecting Areas of Future Development
The study team used PC ARC/INFO geographic infor-
mation systems (CIS) to project the locations of future
residential and industrial growth in the study area, to
show the locations of areas at various degrees of risk
from either airborne or waterborne industrial toxins, and
to reveal the potential areas of population exposure
resulting from the overlap of these areas.
Although this paper does not describe the models used
to project residential and industrial growth in the study
area, it does include the criteria used to make projections.
The criteria we used to project residential growth were:
Travel times to major employment concentrations in
the region.
Proximity to interstate highways, interchanges, and
main trunk sewers.
Avoidance of areas composed of steep slopes and
flood plains.
Land currently zoned for agriculture or housing.
The criteria for projecting industrial areas were:
Relatively flat, not in a flood plain, and zoned industrial.
Proximity to existing industrial development, main
trunk sewers, and interstate highways.
Relative aggressiveness of local jurisdictions in at-
tracting industrial development.
Identifying Areas at Risk From
Airborne Releases
The study team determined the model project areas at
risk from airborne releases by using information avail-
able from the Ohio Environmental Protection Agency.
This information, which included the location, identity,
and quantity of hazardous materials recently stored in
the study area, was available as a result of reporting
requirements mandated by several federal statutes. The
study team assumed that a similar ratio reflecting sites
containing hazardous materials to the area of industri-
ally developed land would continue into the future.
Based on this assumption, the study team could ran-
domly project new potential release sources.
The algorithm and associated software used for calcu-
lating the plume size of aerial dispersion of hazardous
chemicals was Aerial Locations of Hazardous Atmos-
pheres (ALOHA). The National Oceanic and Atmos-
pheric Administration (NOAA) developed this system,
which is in wide use by government and industry for the
preparation of emergency contingency plans. This soft-
ware is available for use on any Macintosh computer, or
any IBM-compatible with an Intel 80286 (or better) CPU.
This software employs three classes of variables in
calculating the plume dispersion for a specific chemical:
Chemical variables
Meteorological options
Source strength options
Chemical variables include both physical properties of
the chemical and parameters that define the human
health effects of the chemical. In the latter case, the two
variables are the threshold limit value (TLV) and the
immediately dangerous to life and health (IDLH) value.
The TLV is a measure of chronic toxicity of the chemical
in humans. It represents the maximum concentration of
the chemical in air to which a human can safely be
exposed for 8 hours per day on a daily basis. The IDLH
is a measure of acute toxicity of the chemical in humans.
The IDLH represents the maximum concentration to
which a human can be exposed for a short time and not
experience death or some other severe endpoint.
Meteorological options describe the ambient atmos-
pheric conditions into which the chemical disperses.
ALOHA has the capability of downloading real-time data
from NOAA satellites. This case, however, employed
average meteorological conditions for the study area
over the course of a year. The variables this study used
were atmospheric inversion height (or no inversion),
wind velocity, air temperature, ground roughness (rural
or urban), and stability class (a combined variable de-
scribing cloud cover and incoming solar radiation).
The source options quantify the amount of chemical
being released and describe how the chemical is re-
leased (instantaneous or continuous).
The ALOHA model provides a procedure forshowing the
IDLH and TLV risk zones from a single accidental re-
lease of a single chemical, given specified conditions of
atmospheric stability, wind direction, and air tempera-
ture. Obviously, climatic conditions change daily, so
areas surrounding a single industrial site experience
279
-------
different degrees of risk depending on the variability of
these conditions. Moreover, a single site that can poten-
tially release more than one chemical poses a higher
risk to surrounding areas. Finally, when release plumes
from two or more sites that are located relatively close
to each other overlap, risks also increase. To account
for these factors of climatic variation and overlapping
chemical release plumes, the study team constructed
the model described below.
In discussing these procedures, bear in mind that the
risk factors are relative. Sufficient data are not available
to estimate the absolute probability of an accidental
release of industrial toxins into the air. Consequently, no
absolute risk levels can be estimated.
1. We acquired NOAA climatic statistics for a full year
for the weather station nearest the study. NOAA
tabulates eight wind directions (N, NE, E, SE, S, SW,
W, and NW). We sorted the average daily wind
speeds and average daily temperatures according to
the daily prevailing wind direction. Thus, for each of
the eight wind directions, it was possible to determine
the number of days in the year that the prevailing
wind comes from that direction, as well as the
average daily wind speed and temperature.
2. We then input the temperature and wind speed data,
derived as explained above, into the ALOHA model
to prepare plots of the IDLH and TLV zones, or plumes,
for each of the eight wind directions. Individually, the
plumes emanated downwind from the source of the
release. In our study, IDLH plumes varied from 0.17
miles to greater than 10 miles. TLV plumes varied
from 0.52 miles to greater than 10 miles.
When the plumes for the eight different wind directions
were combined, the results were translated into a
pattern of wedges representing various plume lengths
in each of the eight different wind directions. These
risk levels vary according to the number of days per
year the wind blows in each direction from the source
of the release. Plumes blowing in different directions
vary in length according to average temperature and
wind velocities for the days the wind blew in each
direction. The numbers in the wedges represent risk
factors assigned as indicated in Table 1.
Table 1. Assigned Risk Factors
Frequency of Wind
Risk Factor
0-25 days/year
26-50 days/year
51 -75 days/year
76-100 days/year
1
2
3
4
3. When a single industrial location employed more
than one chemical, the IDLH and TLV risk patterns
for each were overlaid on one other. Where the
overlap occurred, we added the risk factors together.
4. Finally, when the risk patterns from two or more sites
overlapped, we added together the risk factors
assigned to overlapping areas. The overlap capabilities
of CIS allowed us to easily draw and combine the risk
patterns, superimposed on a map of the industrial
sites under consideration.
When we compared the eight existing combined IDLH
and TLV risk patterns with the risk patterns that appear
when six projected new sources of releases and pro-
jected areas of residential growth are added, the
changes are rather dramatic. Of course, we must note
that the projected sources of potential releases are hy-
pothetical and their locations selected at random. In
reality, precise locations of areas at greater risk cannot
be predicted with any degree of certainty. We can rea-
sonably deduce from the maps, however, that any sub-
stantial increase in industries storing or using toxic
chemicals that might be accidentally released into the
air can compound risk levels much more than might be
expected. The maps we produced indicated five risk
levels.
The study results also showed that an industry capable
of generating accidental releases of airborne toxins will
very likely place at risk residents not only of the same
community but those in neighboring jurisdictions as well.
This is particularly true on the fringes of metropolitan
areas where highly fragmented political boundaries ex-
ist. This fact complicates tremendously the ability of
each jurisdiction to protect its citizens and suggests a
need for a more comprehensive approach to regulation
than conventional land use zoning measures that each
locality administers.
Identifying Areas at Risk From Accidental
Releases Into Ground-Water Supplies
As the introduction of this paper indicates, local govern-
ment usually manages conventional land use planning,
while air quality is largely a state or federal responsibility.
Thus, decisions regarding the regulation of industrial
location may not account for the types of industrial
operations proposed, the possible use of hazardous
materials, and possible risks to local residents from
accidental release of toxins into the air. The same prob-
lem applies to the location of industries that have the
potential for accidental release of hazardous materials
into ground-water supplies. In our study, we outlined a
technique for predicting where local residents may be
placed at risk by drinking water from sources vulnerable
to contamination by industrial toxins.
280
-------
As stated earlier, the study area includes a major aqui-
fer. We noted industrial and related facilities that use and
store hazardous materials, and that are located over or
immediately adjacent to the area aquifers. We distin-
guished between sites where a previous spill had been
reported, those where a waste well is located, and all
other locations where a hazardous material is used or
stored. The results clearly indicated a significant poten-
tial for contamination.
To predict the number and locations of additional indus-
trial toxin sources that might appear over a 10-year
period following 1990 (the base year of the study), we
employed a procedure similar to that used in the air
pollution section of the study. We assumed that during
this period, the number of such sites would increase by
35 percent. The 35-percent increase represented an
arbitrarily selected figure approximately midway be-
tween estimates of 25- and 50-percent industry growth
in the study area. We used this increment to project
additional waste well sites and sites that would experi-
ence a spill sometime during the decade, as well as the
total number of new sites.
We projected the locations of the additional sites by
overlaying a 5,000-foot by 5,000-foot grid on the study
area and assigning the new sites to grid cells using a
random number generator. We considered only cells
lying completely or largely over the aquifers and also
falling within the area of projected industrial land use.
Each cell was assigned a relative contamination risk
factor based on the number of projected sites it con-
tained, with a multiplier of 3 applied to sites with a waste
well and a multiplier of 4 applied to sites assumed to
have had spills. We also assigned existing sites to grid
cells and scored them in the same manner.
To obtain more information, we used a simplified version
of DRASTIC maps prepared by the Ohio Department of
Natural Resources. DRASTIC is an acronym for Depth
to water, net Recharge, Aquifer media, Soil media, To-
pography, Impact of the vadose zone, and hydraulic
Conductivity of the aquifer. These factors contribute to
an index of the relative vulnerability of the aquifer to
pollution. Different shades on the map represented the
relative vulnerability of sections of the aquifers to
ground-water pollution. We assigned two points to aqui-
fers with a DRASTIC pollution potential index less than
180 and four points to aquifers with a higher index. This
allowed us to use the CIS to combine the DRASTIC
vulnerability map with the maps that showed risks of
pollution from the hazardous materials sites. The resul-
tant maps showed the existing and projected potential
risk of pollution in different areas of the aquifers.
The next step in the study related the above information
to public water companies that extract water from differ-
ent parts of the aquifers. Thus, we were able to associ-
ate risk levels with well locations as well as with the
areas served by water companies whose supplies may
be at risk. In this case, the projected risks of polluted
water supplies were identical to existing risk levels be-
cause no public water wells happened to be located in
areas where additional potential sources of pollution
were projected.
Of course, wide areas in the study area had no public
water supply. Residents in these areas may be at risk
depending on where they dig private wells. The maps
we generated showing areas at risk nonetheless provide
useful guides to potentially hazardous areas.
As in the construction of all such models, we needed to
make a number of assumptions, simplifications, and
value judgments. In this case, these included the pro-
jected number of new toxic sites and point scores as-
signed to hazardous materials sites and the various
areas in the DRASTIC maps. Also, for the sake of
simplicity, we projected no new well sites in preparing
one of the maps. This additional element should prob-
ably be included, assuming local water companies could
provide projected well locations. Use of a CIS, however,
makes it possible to explore the implications of adjust-
ment of any of these factors.
We must note two more significant omissions from the
model that the CIS cannot factor in. One was our inabil-
ity to identify from the data specific chemicals that each
sight might release and the relative effects of each. We
would have required more complicated techniques for
dealing with these variables and for projecting the travel
and dilution of plumes of contaminants in an aquifer. In
the interest of providing a simple, if relatively crude,
model capable of replication by a local planning agency
with a simple CIS, we elected not to propose use of
more sophisticated techniques.
Another obvious omission was the consideration of
water pumping and treatment measures that might miti-
gate risks of water contaminated by accidental release
of industrial toxins. Perhaps, with knowledge of the spe-
cific contaminants found in the water at any given time,
such mitigation might be possible. We elected not to
consider this factor for reasons of simplification but also
because this study aimed to provide planning agencies
the means to identify potential risks to local residents
and to prevent or minimize them through better land use
planning and regulatory measures.
Some Final Notes
As stated earlier, the purpose of the study was to pro-
pose a technique that planning agencies could use to
identify:
The locations of existing and projected patterns of
residential and industrial development in a multijuris-
dictional suburban area.
281
-------
The locations of existing and projected industrial,
storage, and disposal sites of hazardous materials.
Residential areas that might be placed at risk by the
accidental release of these hazardous materials into
the air or into ground-water supplies.
The relative levels of risks resulting from potential
exposure to more than one hazardous material at a
single site, or from multiple sites in the vicinity.
The technique we used in the study permitted projection
of relative risk levels. Projecting absolute risk levels is
impossible without data on the actual incidence of acci-
dental releases of toxins overtime in this or in similar
areas. A related question is whether it is possible to
meaningfully indicate the combined risk from airborne
releases of industrial toxins and drinking water contami-
nation to a particular residential area. The issue of
weighting relative risk factors is central here. Could we,
for example, weight the risk levels of exposure to air-
borne releases threeor possibly fourtimes higher
than water contamination risks? Or can we say that the
risks may be the same, but the danger from airborne
releases is three or four times greater?
Obviously, these would be futile exercises, especially
because the point scores in the separate mapping stud-
ies were arbitrarily assigned. The public should know,
however, which present or projected residential areas
carry some level of risk from both types of exposure.
Thus, after calculating point scores to derive relative risk
levels from waterborne pollutants, we multiplied the
scores by 4 to bring their maximum ranges into the same
order of magnitude. Otherwise, the effect of waterborne
pollutants would not be apparent. Consequently, we
created two maps to show the combined existing and
projected risks; therefore, we highlighted the combined
risks that residents face from both airborne and water-
borne hazardous materials.
The maps of airborne releases used in these combina-
tions showed TLV. Continuing exposure within the TLV
areas over an extended period can also have adverse
health effects. This study focused only on accidental
releases, however, and a single release is unlikely to
sustain continuous exposure. Of course, residential ar-
eas at risk from several sites might approach conditions
of sustained exposure. This situation is more analogous
to prolonged exposure to contaminated drinking water
supplies. We did not combine the maps of IDLH airborne
releases with the maps of areas at risk from ground-
water contamination because they are not analogous
conditions. Nonetheless, the IDLH risk maps in them-
selves reveal the conditions that local planning officials
should most seriously consider.
Replicating the procedures outlined in this study should
be technically and financially feasible for local planning
agencies. Armed with the results of such an investiga-
tion, their next step should be to establish the planning
and regulatory measures that would minimize both ex-
isting and projected levels of risk to area residents. The
attorney on our team has outlined a range of possible
measures, but detailing them would be the subject of
another paper.
282
-------
Integration of GIS and Hydrologic Models for Nutrient Management Planning
Clyde W. Fraisse, Kenneth L. Campbell, James W. Jones,
William G. Boggess, and Babak Negahban
Agricultural Engineering Department, University of Florida, Gainesville, Florida
Introduction
Recent evidence that agriculture in general, and animal
waste in particular, may be an important factor in surface
and ground-water quality degradation has increased the
interest in nutrient management research. The presence
of nitrogen and phosphorus in surface water bodies and
ground water is a significant water quality problem in
many parts of the world. Some forms of nitrogen and
phosphorus, such as nitrate N and soluble P, are readily
available to plants. If these forms are released into
surface waters, eutrophic conditions that severely impair
water quality may result. Advanced eutrophication (pH
variations, oxygen fluctuations or lack of it in lower
zones, organic substance accumulation) can cause
physical and chemical changes that may interfere with
recreational use and aesthetic appreciation of water. In
addition, possible taste and odor problems caused by
algae can make water less suitable or desirable for
water supply and human consumption (1).
Increases in nutrient loadings to water resources have
recently been observed in the southeastern United
States, where well-drained sandy soils with low nutrient
retention capacity and high water table conditions are
found in most coastal areas. Those increases were
associated statistically with nutrient sources such as
agricultural fertilizers and dense animal populations (2,
3). Repetitive occurrences of extensive blooms of blue-
green algae that threatened the overall health of Lake
Okeechobee, located in southern Florida, were attrib-
uted to an increase in nutrient loadings to the lake. The
South Florida Water Management District (SFWMD) re-
ported an increase of phosphorus concentrations in the
lake water from an annual average of 0.049 milligrams
per liter in 1973 and 1974 to a peak of 0.122 milligrams
per liter in 1988 (4).
Most water quality problems concerning phosphorus
result from transport with sediment in surface runoff into
receiving waters. Continuous high loadings from animal
waste on sandy soils with low retention capacity, however,
may contribute significant quantities of labile phospho-
rus to subsurface drainage. Ground-water aquifers may
also become polluted due to recharge of high loadings
of nitrogen. Drinking water with nitrate N concentrations
higherthan 10 milligrams per liter may lead to methemo-
globinemia in infants. Ground-water monitoring of the
Middle Suwannee River area in Florida has shown high
concentrations of nitrate nitrogen near intensive agricul-
tural operations. The U.S. Geological Survey has inten-
sively monitored dairy and poultry farms and has found
high nitrate levels below these operations compared with
nearby control wells (5).
Animal waste management has always been a part of
farming, but historically has been relatively easy due to
the buffering capacity of the land. In fact, land applica-
tion of animal waste at acceptable rates can provide
crops with an adequate level of nutrients, help reduce
soil erosion, and improve water holding capacity. As the
animal industry attempts to meet the food requirements
of a growing population, however, it applies new tech-
nologies that reduce the number of producers, but cre-
ate larger, more concentrated operations. That, in
addition to the decreasing amount of land available for
waste application, has increased the potential for water
quality degradation.
Successful planning of an animal waste management
system requires the ability to simulate the impact of
waste production, storage, treatment, and use on water
resources. It must address the overall nutrient manage-
ment for the operation, including other nutrient sources
such as supplemental fertilizer applications. Livestock
operations are highly variable in their physical facilities,
management systems, and the soil, drainage, and cli-
matic conditions that affect the risk of water pollution
from animal wastes (6). Linkage between geographic
information systems (GIS) and hydrologic models offers
an excellent way to represent spatial features of the-
fields being simulated and to improve results. In addition,
a GIS containing a relational database is an excellent way
283
-------
to store, retrieve, and format the spatial and tabular data
required to run a simulation model.
This paper examines some of the issues related to the
integration of hydrologic/water quality models and CIS
programs. In addition, the paper discusses the ap-
proaches used in the Lake Okeechobee Agricultural
Decision Support System (LOADSS), which was re-
cently developed to evaluate the effectiveness of differ-
ent phosphorus control practices (PCPs) in the Lake
Okeechobee basin. The paper also details a dairy
model, designed to simulate and evaluate the impacts
of alternative waste management policies for dairy op-
erations, that is currently under development.
Hydrologic Models and GIS
By using models, we can better understand or explain
natural phenomena and, under some conditions, make
predictions in a deterministic or probabilistic sense (7).
A hydrologic model is a mathematical representation of
the transport of water and its constituents on some part
of the land surface or subsurface environment. Hydro-
logic models can be used as planning tools for determin-
ing management practices that minimize nutrient
loadings from an agricultural activity to water resources.
The results obtained depend on an accurate repre-
sentation of the environment through which water flows
and of the spatial distribution of rainfall characteristics.
These models have successfully dealt with time, but
they are often spatially aggregated or lumped-parame-
ter models.
Recently, hydrologists have turned their attention to GIS
for assistance in studying the movement of water and
its constituents in the hydrologic cycle. GIS programs
are computer-based tools to capture, manipulate, proc-
ess, and display spatial or georeferenced data. They
contain geometry data (coordinates and topological in-
formation) and attribute data (i.e., information describing
the properties of geometrical objects such as points,
lines, and areas) (8). A GIS can represent the spatial
variation of a given field property by using a cell grid
structure in which the area is partitioned into regular grid
cells (raster GIS) or by using a set of points, lines, and
polygons (vector GIS).
A close connection obviously exists between GIS and
hydrologic models, and integrating them produces tre-
mendous benefits. Parameter determination is currently
one of the most active hydrology-related areas in GIS.
Parameters such as land surface slope, channel length,
land use, and soil properties of a watershed are being
extracted from both raster and vector GIS programs,
with a focus on raster-based systems. The spatial nature
of GIS also provides an ideal structure for modeling. A
GIS can be a substantial time saver that allows differ-
ent modeling approaches to be tried, sparing manual
encoding of parameters. Further, it can provide a tool for
examining the spatial information from various user-
defined perspectives (9). It enables the user to selec-
tively analyze the data pertinent to the situation and try
alternative approaches to analysis. GIS has been par-
ticularly successful in addressing environmental prob-
lems.
Approaches for Integrating GIS and Models
A significant amount of work has been done to integrate
raster and vector GIS with hydrologic/water quality mod-
els. Several strategies and approaches for the integra-
tion have been tried. Initial work tended to use simpler
models such as DRASTIC (10) and the Agricultural
Pollution Potential Index (11). In these cases, the mod-
els were implemented within the GIS themselves. These
studies attempted to develop CIS-based screening
methods to rank nonpoint pollution potential. The use of
more complex models requires that the GIS be used to
retrieve, and possibly format, the model data. The model
itself is implemented separately and communicates with
GIS via data files. Goodchild (12) refers to this mode as
"loose coupling," implying that the GIS and modeling
software are coupled sufficiently to allow the transfer of
data and perhaps also of results, in the reverse direc-
tion. Fedra (8) refers to this level of integration as "shal-
low coupling" (see Figure 1). Only the file formats and
the corresponding input and output routines, usually of
the model, must be adapted. Liao and Tim (13) describe
an application of this type, in which an interface was
developed to generate topographic data automatically
and simplify the data input process for the Agricultural
Nonpoint Source (AGNPS) Pollution Model (14), a water
quality model.
Shared Databases and Files
j
1
GIS
User Interface
i
MODEL
User Interface
Figure 1. Loose or shallow coupling through common files (8).
284
-------
Higher forms of connection use a common interface and
transparent file or information sharing and transfer be-
tween the respective components (see Figure 2). The
dairy model, currently under development, is an appli-
cation of this kind. It will link the Ground-Water Loading
Effects of Agricultural Management Systems (GLEAMS)
(15) model and CIS to evaluate potential leaching and
runoff of both nitrogen and phosphorus.
LOADSS is an extension of this type of application
because it includes an optimization module that enables
the system to select the best PCPs at the regional scale,
based on goals and constraints defined by the user.
Both applications use ARC/INFO's arc macro language
(AML), a high-level application language built into the
CIS. A subset of functions of a full-featured CIS, such
as creation of maps (including model output) and tabular
reports, as well as model-related analysis, are embed-
ded in the applications, giving the system great flexibility
and performance. Fedra (8) describes a deeper level of
integration that would merge the two previous ap-
proaches, such that the model becomes one of the
analytical functions of a CIS, or the CIS becomes yet
another option to generate and manipulate parameters,
input and state variables, and model output, and to
provide additional display options. In this case, software
components would share memory rather than files.
The choice between integrating a water quality model
with a raster or vector CIS depends on the importance
of spatial interactions in the process being studied and
the nature of the model itself. Some water quality mod-
els, such as GLEAMS, are field-scale models that pro-
vide edge-of-the-field values for surface runoff and
erosion as well as deep percolation of water and its
constituents. In this case, spatial interactions between
adjacent fields are ignored and a vector CIS can be
used to describe the system. Moreover, important fac-
tors in the simulation process, such as land use and
management practices, are normally field attributes and
thus, are better represented in a vector structure.
Other factors playing an important role in the hydrologic
process, such as field slope, aspect, and specific catch-
ment area, are hard to estimate in vector systems,
however. A raster-based CIS is better suited for handling
watershed models in which the routing process is impor-
tant and spatial interactions are considered. For those,
several algorithms for estimating important terrain attrib-
utes are often incorporated in commercially available
raster-based CIS programs.
LOADSS
LOADSS was developed to help address problems cre-
ated by phosphorus runoff into Lake Okeechobee. It was
designed to allow regional planners to alter land uses
and management practices in the Lake Okeechobee
Shared Databases and Files
GIS
MODEL
Common User Interface
Figure 2. Deep coupling in a common framework (18).
basin, then view the environmental and economic ef-
fects resulting from the changes. The Lake Okeechobee
basin coverage incorporates information about land
uses, soil associations, weather regions, management
practices, hydrologic features, and political boundaries
for approximately 1.5 million acres of land and consists
of close to 8,000 polygons.
The SFWMD, responsible for managing Lake Okeechobee,
has initiated numerous projects to develop effective con-
trol practices for reducing the level of phosphorus in
agricultural runoff as part of the Lake Okeechobee Sur-
face Water Improvement and Management (SWIM)
Plan. These projects, numbering more than 30, were
designed to develop information on the control and man-
agement of phosphorus within the lake basin and to
determine the costs and effectiveness of selected man-
agement options. Three types of control options are
being studied:
Nonpoint source controls, such as pasture management.
Point source controls, such as sewage treatment.
Basin-scale controls, such as aquifer storage and
retrieval.
After completing most of these research efforts, the
need arose for a comprehensive management tool that
could integrate the results for all three classes of PCPs.
In response to these needs, design and implementation
of a decision support system was initiated with the fol-
lowing objectives (16):
Organize spatial and nonspatial knowledge about
soils, weather, land use, hydrography of the lake ba-
sin, and PCPs under a GIS environment.
285
-------
Develop and implement algorithms for modeling non-
point source, point source, and basin-scale PCPs.
Develop and implement mechanisms for evaluating
the performance of the entire Lake Okeechobee basin
under different combinations of PCPs applied to the
basin.
Design and develop a user interface that would fa-
cilitate use of the system by noncomputer experts.
The goal in developing LOADSS was to create an infor-
mation system that would integrate available information
to help regional planners make decisions. LOADSS can
generate reports and maps concerning regional land
attributes, call external hydrologic simulation models,
and display actual water quality and quantity sampling
station data. LOADSS is a collection of different
components:
The regional scale CIS-based model used to develop
and manipulate regional plans for reducing phospho-
rus loading to Lake Okeechobee.
The Interactive Dairy Model (IDM) used to develop
field-level management plans for dairies and run the
Field Hydrologic and Nutrient Transport Model
(FHANTM) simulation model for nutrient transport
modeling.
An optimization module that enables the system to
select the best PCPs at the regional scale (currently
under development).
Although these components can run independently, they
are fully integrated in the LOADSS package and can
exchange information where necessary. A design sche-
matic of LOADSS is given in Figure 3.
Regional-Scale CIS-Based Model
LOADSS serves both as a decision support system for
regional planning and as a graphic user interface for
controlling the different components. One consideration
in the design of LOADSS was the size of the database
that was being manipulated. Because the land use da-
tabase consisted of nearly 8,000 polygons, running the
simulation models interactively would not be a feasible
option. Thus, the CREAMS-WT (17) runoff model was
prerun for different levels of inputs and management for
each land use, soil association, and weather region (18).
Depending on the land use and its relative importance
as a contributor of phosphorus to the lake, anywhere
from one (background levels of inputs to land uses like
barren land) to 25 (dairies, beef pastures) levels of
inputs were selected. Each set of inputs to a particular
land use was given a separate PCP identification code.
A CREAMS-WT simulation was performed for each
PCP, on each soil association and weather region. This
resulted in approximately 2,600 simulation runs. Annual
average results were computed for use in LOADSS.
CREAMS-WT provides an average annual estimate of
phosphorus runoff from each polygon. Phosphorus
assimilation along flow paths to Lake Okeechobee are
estimated as an exponential decay function of distance
traveled through canals and wetlands (4).
The imports, exports, and economics of each PCP are
based on a per production unit basis. Depending on the
type of polygon, the production unit can be acres (e.g.,
pastures, forests), number of cows (dairies), or millions
of gallons of effluent (waste treatment plants and sugar
mills). Developing a regional plan in LOADSS involves
assigning a PCP identification code to each one of the
polygons in the Lake Okeechobee basin. Accessing the
results of a regional plan involves multiplying the pro-
duction unit of each polygon by its appropriate database
import, export, or economic attribute and summing the
resulting values overall polygons in the Lake Okeechobee
basin. LOADSS runs in the ARC/INFO Version 6.1.1 CIS
software on SUN SPARC stations.
Interactive Dairy Model
Although the LOADSS level of detail is adequate for
regional planning, a more detailed model was necessary
to analyze individual dairies in the Lake Okeechobee
basin, as dairies were one of the large, concentrated
sources of phosphorus runoff into the lake. Thus, the
IDM was developed and incorporated into LOADSS.
IDM utilizes FHANTM to simulate phosphorus move-
ment in dairy fields. FHANTM is a modification of
DRAINMOD (19) with added functions to handle over-
land flow routing, dynamic seepage boundary, and sol-
uble phosphorus algorithms for P input, mass balance,
and transport (20).
Unlike in LOADSS, FHANTM is run interactively, as IDM
requires. Furthermore, in LOADSS, the user can only
select from a number of predefined PCPs, while in IDM,
the user has access to more than 100 input and man-
agement variables, all of which can take a range of
values. This allows for the development and evaluation
of detailed dairy management plans that otherwise
would be impossible at a regional scale. While LOADSS
only provides average annual results, IDM displays daily
time series simulation results. IDM uses the same as-
similation algorithm and can produce the same phos-
phorus budget maps and reports as LOADSS.
Optimization Module
A variety of factors must be considered in planning
nutrient management programs. Production and envi-
ronmental goals need to be balanced, and these goals
are often incompatible. Performing this exercise on a
regional scale, comprising many fields for which a variety
of land uses and management options can be assigned,
is a tremendously time-consuming, if not impossible,
286
-------
CIS Databases
LOADSS User Interface
CIS-BASED
USER
INTERFACE
SCREEN.
PRINTER OR
PLOTTER
MAPS
AND
REPORTS
PHOSPHORUS
& MATERIAL 8
BUDGETS
PHOSPHORUS,
MATERIAL &
ECONOMIC BUDGETS
Input Attribute
Databases
Spatial
Databases
DAIRY MANAGEMENT
PRACTICES
NONPOINT SOURCE
POP'S CREAMS-WT
INDIVIDUAL FIELD
MANAGEMENT
POINT SOURCE PCP'S
PROCESS ANALYSIS
FHANTM
SIMULATION MODEL
BASIN TREATMENTS
PROCESS ANALYSIS
Optimization Module
LOADSS
Models and
Analysis Tools
I DM and
Analysis Tools
Figure 3. LOADSS design schematic (16).
task. The optimization component of LOADSS, currently
under development, will determine the best combination
of agricultural, environmental, and regulatory practices
that protects and maintains the health of Lake
Okeechobee and also maintains the economic viability
of the region. The optimization process will provide an-
other method for modifying the PCPs assigned to indi-
vidual fields. Different optimization solution methods,
such as linear programming and integer linear program-
ming, will be available for solving the optimization prob-
lem that the user defines.
Dairy Simulation Model
The dairy model was expected to be fully functional by
the end of 1994. It is designed to be an additional tool
for answering questions related to the environmental
costs and impacts of dairy operations. A design sche-
matic of the dairy model is given in Figure 4. It differs
from the LOADSS/IDM model in the following aspects:
It is designed to be generic so that any dairy repre-
sented by a coverage for which relevant data, such
as topography, soil characteristics, weather, and field
boundaries, are available can be simulated.
The GLEAMS water quality model will be used for
simulating nutrient transport of nitrogen and phosphorus.
The user will be able to assign a larger variety of
crops and crop management practices to the individual
fields, including crop rotation.
GLEAMS (15) is a field-scale water quality model that
includes hydrology, erosion/sediment yield, pesticide,
and nutrient transport submodels. GLEAMS was devel-
oped to use the management oriented CREAMS (21)
model and incorporate more descriptive pesticide
subroutines and more extensive treatment of the flow of
water and chemicals in the root zone layers. The water
is routed through computational soil layers to simulate
the percolation through the root zone, but the volume of
percolation in each layer is saved for later routing in the
pesticide component. A minimum of three and a maxi-
mum of 12 layers with variable thickness may be used.
Soil parameter values are provided by soil horizon, and
287
-------
GIS Databases
TOPOGRAPHY
DAIRY MODEL User Interface
Spatial
Databases
Input Attribute
Databases
, - ! . "
DAIRY PLAN
»
NITROGEN * PHOSPHORUS
BALANCE
i
T
DAIRY MANAGEMENT
PRACTICES
L INDIVIDUAL FIELD
MANAGEMENT
CROP
MANAGEMENT
i
GLEAMS
SIMULATION MODEL
WEATHER
FILES
SOIL AND
CULTURAL FILES
Dairy Model
Analysis Tools
GLEAMS Simulation
Shell Controller
Figure 4. Dairy model design schematic.
the crop root zone may have up to five horizons. The
values for parameters, such as porosity, water retention
properties, and organic content, are automatically fitted
into the proper computational layers. Two options are
provided in the model to estimate potential evapotran-
spiration, the Priestly-Taylor method (22) and the Pen-
man-Monteith method (23). The nutrient component of
the model simulates land application of animal wastes
by creating appropriate nitrogen and phosphorus pools
for mineralization. It considers ammonia volatilization
from surface-applied animal waste by using a relation-
ship developed by Reddy et al. (24).
The graphic interface is designed to help the user plan
a balanced nutrient management program for the dairy
being simulated. First, total nutrient production and ac-
counting are estimated, based on information related to
the dairy management such as herd size, confinement
system, waste characterization, and handling. Figure 5
shows the general structure of the graphic interface and
a first version of the menu used to estimate the total
amount of nitrogen and phosphorus available for assign-
ment to the various fields. Nutrient losses during waste
storage and treatment vary widely depending on the
method of collection, storage, and treatment. Climate
can also have a great effect on the losses. Covering all
possible methods of storage and treatment is practically
impossible, especially in an application that is designed
to be generic and applied in any part of the country. A
simplification was made to overcome this problem: the
user must provide the percentage of original nitrogen
and phosphorus that is retained after waste storage and
treatment. The menus designed to enter information
related to the management of fields and crops are given
in Figure 6.
For each field, a sequence of crops can be defined in
the Field Management Table, and for each crop, the
sequence of practices or field operations is defined in
the Crop Management Table. Every time a waste appli-
cation operation is defined or a field is used as pasture
for a certain period, the corresponding amount of nutri-
288
-------
f^ii? AGRICULTURAL ENGINEERING DEPARTMENT- UNIVERSITY OF FLORIDA j
(Select a Dairy ^X0"01! Dairy tlanaeenent wJOlan Hanafer V^( Practice Henus »Xsi"ulation Results X Report Hanaeer ?Xprint "w 7X Utilities "XOuitJ 1
''»> FIELDS
(SELECT FIELD(S) J
(REHOVE FIELIKS)
(UNDO LflST SELECTION J
(i/IEM FIELD INFO )
ri& SIMULATION
(RUN SELECTED FIELDS )
(RUN (ILL FIELDS )
'' 3 MAP UTILITIES
HUP 1TFE:
I
MPP UIDO.S:
D ItOOIFY LEGEND
larSW
ECOMIH EODMOiml fccFRESH
' S' NUTRIENT'S' PRODUCTION TABLE
HERD SIZE « COIIFINEHEHT SYSTEM OfHRY : dairs.OS
Nunlier of RuHrage Munber of
cons : 100 ueitht 100
HH57E r.HimnnFRT?nTTnH a NUTRIENTS pRnnucnnH
Hitrneeii s/llll/.l.i.j) ; 0.4K l>l,o=|.l,.,r,,= (1 l,-j/HII/.l,..j) : 0.07
Hilrofien Phosphorus
Lbs/nil/ijcar lbs/!icor Lbs/flU/ucar Lbs/ucar
Confinen«nt 01 10071 13 1533
Ursine HI 10071 13 I!i33
TUMIL 1GO 2014B 2G 30CG
limiOLINO OF MIBTE PRODUCED UNDER COHFINEICHT
Percentage of oripinal nutrients retained after STDRHGE/TRERTMENT
NUTRiEiirs nr.r.nimTTiir, (pERCEHinnE)
nuailaltlc fnr flssignncnt TreatiMl/rtarkcted
Crajint; Land HpplicaLlon
HITROCEH 50 JO 22
PIIOCPIIORU!; 50 ' 33 17
(SBVE ma EXIT} (CIIHCEL)
' -a MESSAGES
5 : Drawing dairy nap |fZ|
i. IE : 1 la
Figure 5. The nutrient production table of the dairy model interface.
ents will decrease from the amount available for assign-
ment and the total available for future assignment will
be updated. Once the total amount of nutrients is as-
signed, the model can be run for the several fields in the
dairy and the results evaluated in terms of nutrient load-
ings to the edge of fields and ground water. Alternative
plans can be designed and saved for comparison and
selection of best management options. The best solu-
tions in terms of reducing nutrient loadings to surface
and ground water must also consider economic aspects.
A producer's decision about competing waste manage-
ment practices is ultimately economically motivated.
Thus, the system will eventually include a tool for eco-
nomic analysis of alternative management options.
Summary and Conclusions
The search for solutions to the many problems concern-
ing nutrient management that affect water resources
implies a continued demand for the development of
modeling systems that can be used to analyze, in a
holistic approach, the impact of alternative management
policies.
The development of LOADSS exemplifies how the inte-
gration of hydrologic models and a CIS can be used for
analyzing nutrient control practices at different scales.
The addition of optimization algorithms further enhances
the ability of policy- and decision-makers to analyze the
impact of alternative management practices and land
uses at the regional level.
The first part of LOADSS (Version 2.2) that includes the
CREAMS-WT regional-scale model and the IDM com-
ponents is fully functional and currently available at the
SFWMD. Preliminary results show that LOADSS be-
haves consistently with measured data at the lake basin
scale. Some of this, however, is due to offsetting errors
in model behavior at the subbasin scale, particularly in
subbasins that are adjacent to or very far from the lake.
Currently, projects are underway to further verify and
calibrate the model at the subbasin level to improve its
289
-------
Figure 6. Field and crop management tables of the dairy model interface.
performance at smaller scales (16). Initial results of the
optimization component are currently being evaluated.
The dairy model represents a different approach in inte-
grating water quality models and CIS in that it is de-
signed to be generic and focused mainly on the farm
level. It is primarily designed to help policy- and decision-
makers analyze the effects of alternative dairy waste
management practices on the farm level. The framework
can easily be adapted to handle different types of animal
wastes (such as beef cattle and poultry) and to simulate
the impact of other crop management practices such as
pesticide applications.
References
1. Novotny, V., and G. Chesters. 1981. Handbook of nonpoint pol-
lution: Sources and management. New York, NY: Van Nostrand
Reinhold Company.
2. Council on Environmental Quality (CEQ). 1989. Environmental
trends. Washington, DC: U.S. Government Printing Office.
3. U.S. Geological Survey (USGS). 1987. Analysis and interpreta-
tion of water quality trends in major U.S. rivers, 1974-81. Water
Supply Paper 2307. Washington, DC: U.S. Government Printing
Office.
4. SFWMD. 1989. Interim Surface Water Improvement and Man-
agement (SWIM) plan for Lake Okeechobee. West Palm Beach,
FL: South Florida Water Management District.
5. Reck, WR. 1994. GLEAMS modeling of BMPs to reduce nitrate
leaching in Middle Suwannee River area. In: Proceedings of Sec-
ond Conference on Environmentally Sound Agriculture, Orlando,
FL. American Society of Agricultural Engineers, pp. 361-367.
6. Fedkiw, J. 1992. Impacts of animal wastes on water quality: A
perspective from the USDA. In: Proceedings of the National
Workshop on Livestock, Poultry, and Aquaculture Waste Manage-
ment, Kansas City, Ml (1991). American Society of Agricultural
Engineers (ASAE) Pub. 03-92. pp. 52-62.
7. Woolhiser, D.A., and D.L. Brakensiek. 1982. Hydrologic system
synthesis. In: Haan, C.T., H.P. Johnson, and D.L. Brakensiek,
eds. Hydrologic modeling of small watersheds. ASAE Monograph
No. 5. St. Joseph, Ml: American Society of Agricultural Engineers.
pp. 3-16.
290
-------
8. Fedra, K. 1993. GIS and environmental modeling. In: Goodchild,
M., B. Parks, and L. Steyaert, eds. Environmental modeling with
CIS. New York, NY: Oxford University Press, pp. 35-50.
9. Tim, U.S., M. Milner, and J. Majure. 1992. Geographic information
systems/simulation model linkage: Processes, problems, and op-
portunities. ASAE Paper No. 92-3610. St. Joseph, Ml: American
Society of Agricultural Engineers.
10. Whittemore, D.O., J.W. Merchant, J. Whistler, C.E. McElwee, and
J.J. Woods. 1987. Ground-water protection planning using the
ERDAS geographic information system: Automation of DRASTIC
and time-related capture zones. In: Proceedings of the National
Water Well Association (NWWA FOCUS) Conference on Mid-
western Ground-Water Issues, Dublin, OH. pp. 359-374.
11. Petersen, G.W, J.M. Hamlett, G.M. Baumer, D.A. Miller, R.L. Day,
and J.M. Russo. 1991. Evaluation of agricultural nonpoint pollu-
tion potential in Pennsylvania using a geographic information
system. ER9105. University Park, PA: Environmental Resources
Research Institute.
12. Goodchild, M. 1993. The state of CIS for environmental problem-
solving. In: Goodchild, M., B. Parks, and L. Steyaert, eds. Envi-
ronmental modeling with GIS. New York, NY: Oxford University
Press, pp. 8-15.
13. Liao, H.H., and U.S. Tim. 1992. Integration of geographic infor-
mation system (GIS) and hydrologic/water quality modeling: An
interface. ASAE Paper No. 92-3612. St. Joseph, Ml: American
Society of Agricultural Engineers.
14. Young, R.A., C.A. Onstad, D.D. Bosch, and W.P. Anderson. 1989.
AGNPS: A nonpoint source pollution model for evaluation of ag-
ricultural watersheds. J. Soil and Water Conserv. 44(2):168-173.
15. Leonard, R.A., W.G. Knisel, and D.A. Still. 1987. GLEAMS:
Ground-water loading effects of agricultural management sys-
tems. Trans, of the Amer. Soc. of Agricul. Eng. (ASAE)
30(5):1,403-1,418.
16. Negahban, B., C. Fonyo, W Boggess, K. Campbell, J. Jones, G.
Kiker, E. Hamouda, E. Flaig, and H. Lai. 1993. A CIS-based
decision support system for regional environmental planning. In:
Proceedings of the Conference on Application of Advanced Infor-
mation Technologies: Effective Management of Natural Re-
sources, Spokane, WA. American Society of Agricultural
Engineers (ASAE). pp. 169-178.
17. Heatwole, C.D., K.L. Campbell, and A.B. Bottcher. 1987. Modified
CREAMS hydrology model for coastal plain flatwoods. Trans, of
the Amer. Soc. Agricul. Eng. (ASAE) 30(4):1,014-1,022.
18. Kiker, G.A., K.L. Campbell, and J. Zhang. 1992. CREAMS-WT linked
with GIS to simulate phosphorus loading. ASAE Paper No. 92-9016.
St. Joseph, Ml: American Society of Agricultural Engineers.
19. Skaggs, R.W 1980. DRAINMOD reference report. U.S. Depart-
ment of Agriculture, Soil Conservation Service. Fort Worth, TX:
South National Technical Center.
20. Tremwel, T.K., and K.L. Campbell. 1992. FHANTM, a modified
drainmod: Sensitivity and verification results. ASAE Paper No. 92-
2045. St. Joseph, Ml: American Society of Agricultural Engineers.
21. Knisel, W.G. 1980. CREAMS: A field-scale model for chemicals,
runoff, and erosion from agricultural management systems. Con-
serv. Res. Rep. No. 26. Washington, DC: U.S. Department of
Agriculture.
22. Priestly, C.H.B., and R.J. Taylor. 1972. On the assessment of
surface heat flux and evaporation using large-scale parameters.
Monthly Weather Rev. 100:81-92.
23. Jensen, M.E., R.D. Burman, and R.G. Allen, eds. 1990. Eva-
potranspiration and irrigation water requirements. Manuals and
reports on engineering practice. Amer. Soc. Civil Eng. No. 70.
24. Reddy, K.R., R. Khaleel, M.R. Overcash, and P.W. Westerman.
1979. A nonpoint source model for land areas receiving animal
waste: II. Ammonia volatilization. Trans, of the Amer. Soc. Agric.
Eng. (ASAE) 22(6): 1,398-1,405.
291
-------
Other GIS Applications
-------
Expedition of Water-Surface-Profile Computations Using GIS
Ralph J. Haefner, K. Scott Jackson, and James M. Sherwood
Water Resources Division, U.S. Geological Survey, Columbus, Ohio
Abstract
Water-surface profiles computed by use of a step-back-
water model such as Water Surface PROfile (WSPRO)
are frequently used in insurance studies, highway de-
sign, and development planning to delineate flood
boundaries. The WSPRO model requires input of hori-
zontal and vertical coordinate data that define cross-
sectional river-channel geometry. Cross-sectional and
other hydraulic data are manually coded into the WSPRO
model, a labor-intensive procedure. For each cross sec-
tion, output from the model assists in approximating the
flood boundaries and high-water elevations of floods with
specific recurrence intervals (for example, 100-year or
500-year). The flood-boundary locations along a series of
cross sections are connected to delineate the flood-prone
areas for the selected recurrence intervals.
To expedite the data collection and coding tasks re-
quired for modeling, the geographic information system
(GIS), ARC/INFO, was used to manipulate and process
digital data supplied in AutoCAD drawing interchange
file (DXF) format. The DXF files, which were derived
from aerial photographs, included 2-foot elevation data
along topographic contours with +0.5-foot resolution
and the outlines of stream channels. Cross-section
lines, located according to standard step-backwater cri-
teria, were digitized across the valleys. A three-dimen-
sional surface was generated from the 2-foot contours
by use of the GIS software, and the digitized section
lines were overlain on this surface. GIS calculated the
intersections of contour lines and cross-section lines,
which provided most of the required cross-sectional ge-
ometry data for input to the WSPRO model.
Most of the data collection and coding processes were
automated, significantly reducing labor costs and hu-
man error. In addition, maps at various scales can be
easily produced as needed after digitizing the flood-
prone areas from the WSPRO model into GIS.
Introduction and Problem Statement
Losses due to flood damage generally cost the Ameri-
can public hundreds of millions of dollars annually. In
1968, the National Flood Insurance Act established the
National Flood Insurance Program (NFIP) to help re-
duce the cost to the public and provide a framework to
help reduce future losses. The Federal Emergency Man-
agement Agency (FEMA) administers the NFIP. As listed
in Mrazik and Kinberg (1), the major objectives of the
NFIP are to:
Make nationwide flood insurance available to all com-
munities subject to periodic flooding.
Guide future development, where practical, away
from flood-prone areas.
Encourage state and local governments to make ap-
propriate land use adjustments to restrict develop-
ment of land that is subject to flood damage.
Establish a cooperative program involving the federal
government and the private insurance industry.
Encourage lending institutions, as a matter of national
policy, to assist in furthering program objectives.
Authorize the continuing studies of flood hazards.
Studies of flood-prone areas typically involve using
step-backwater computer algorithms (digital models) to
estimate river water-surface profile elevations and flood-
inundation patterns along the topography of the river
and its overbanks. FEMA recognizes the U.S. Geologi-
cal Survey's (USGS's) step-backwater model, Water
Surface PROfile (WSPRO), as a suitable computer
model for use in flood insurance studies (2, 3). Basic
data input for step-backwater models includes:
Estimates of flood discharge and initial water-surface
elevations.
Stream cross-sectional geometry.
Roughness coefficients for cross sections.
295
-------
Contracted opening geometry if bridges or culverts
are located along the study reach.
Obtaining meaningful model results typically requires
numerous stream cross sections referenced to a com-
mon elevation datum along a stream reach. The data-
collection efforts to obtain these cross-sectional data
require costly, labor-intensive fieldwork. Study efforts
along lengthy stream reaches may, however, involve the
generation of a contour map using aerial photogram-
metric mapping techniques. Processing the spatial data
may still require extensive labor to extract the cross-
sectional data needed for the WSPRO model.
The development of geographic information systems
(CIS) technology has greatly enhanced analyses of spa-
tial data such as topography. In an effort to improve the
quality of mapping and delineation of flood-prone areas
in Summit County, Ohio, the USGS developed a method
of using a CIS as a pre- and postprocessor of the input
and output data for the WSPRO model. This paper
describes the steps the USGS used to develop this
interface and discusses some difficulties encountered
during the process.
Approach
Several steps were taken that resulted in the delineation
of a flood-prone area in Summit County, Ohio. These
steps are shown in a flow chart (see Figure 1) and
described below.
Data were obtained for this study via aerial photography
during April 1990. These data include mappable fea-
tures at the given scale including topography at 2-foot
contour intervals, stream boundaries, roads, and build-
ings. The data are estimated to be vertically accurate to
+0.5 feet. The data were put into AutoCAD and were
prepared for delivery to the USGS on 3.5-inch floppy
disks in AutoCAD drawing interchange file (DXF) ASCII
format. ARC/INFO was used to convert the DXF file into
two separate data layers containing only the topography
and traces of stream banks within the study area.
A three-dimensional surface was generated from the
topographic data using the ARC/INFO software package
Triangulated Irregular Network (TIN). Cross-section
lines were digitized over the topography data layer. The
cross sections were placed according to standard step-
backwater criteria (4) and were generally:
Perpendicular to stream flow
At major breaks in streambed profiles
At minimum and maximum cross-sectional areas
At major changes in stream conveyance
Spaced about one cross-section width apart
Obtain
Aerial
Photography
Put Data in DXF Format
Process DXF Data With CIS (ARC/INFO)
Establish Relation of Attribute Data to Spatial Data
Generate Three-Dimensional Surface
Overlay Cross Sections and
Calculate Intersections
Verify Elevations With Topographic Controls
Generate Input Files for Model With
Cross-Section Data
Use WSPRO Model To Compute Water-Surface
Elevations for Each Cross Section
Plot Model Output on Topography Data Layer and
Connect Endpoints of Cross Sections
Figure 1. Flowchart of data conversion and processing for use
in the Water Surface PRQfile.
The cross-section lines were then overlaid on the three-
dimensional surface of topography, and CIS calculated
the intersections of the contour lines and cross sections.
The locations and elevations of these intersections were
output as an ASCII file and slightly modified for input into
the WSPRO model.
These CIS data were used along with the aforemen-
tioned required data as input to the WSPRO model.
Input for the model included estimates of the 100-year
flood discharge (5), stream cross-sectional geometry
(supplied by this work), and estimates of roughness
coefficients for cross sections. The WSPRO model was then
run, providing output in the form of water-surface eleva-
tions at specific distances along section lines correspond-
ing to the simulated elevation of a 100-year flood.
Points corresponding to the flood elevations along the
cross-section lines were plotted on the topography data
296
-------
layer and were connected manually (to delineate flood
boundaries) by interpolating the elevations with respect
to adjacent contours. A polygon of the flood "surface"
was generated and drawn on a map (see Figure 2).
Results
The supplied topographic data were of sufficient quality
and resolution to substitute for field-surveyed eleva-
tions; however, field surveys to verify the elevations
along the cross sections would augment this quality
control process (see Figure 1). Typically, a crew of two
individuals may take up to 4 days to survey and reduce
the field data for the study area chosen for this study.
Because aerial photography is commonly substituted for
land surveying, the most significant effort and source of
error may come from manually extracting elevations and
distances along cross sections for input into the WSPRO
0 500 1,000 1,500 2,000 2,500 Feet
I i i i i I
100 200 300 400 500 Meters
Legend
100-Year Flood-Prone
Area
Topographic Contours
(Contour Labels Omitted
for Clarity; Contour
Interval 2 Feet)
Cross-Section Lines
Stream-Bank Trace
Intersection of 100-Year
Flood-Prone Area and
Cross-Section Line
Figure 2. Watershed showing delineation of 100-year flood-prone area.
297
-------
model. Initial development of the method to use CIS for
this analysis took approximately 1 week to refine; how-
ever, future analyses would probably only take one per-
son 1 day to perform. This represents a significant cost
savings. Additionally, reducing the amount of human-in-
duced error can substantially improve the reliability and
accuracy of the computer-generated flood-prone area
data.
Because topography, stream traces, and other features
are supplied in the DXF file, these data can easily be
brought into CIS. Maps can be made that show these
features in relation to the predicted flood-prone area.
Maps showing a variety of features can be produced at
any scale, with accuracy limited only by the accuracy of
the source scale. Additionally, CIS can calculate the
intersection of map features that may lie within the
flood-prone area, such as buildings that may contain
hazardous materials. CIS can also overlay land use
data layers within the flood-prone area to define areas
that should not be developed or that have already been
overdeveloped in accordance with the aforementioned
NFIP objectives.
FEMA now requests that future flood-study mapping be
completed using CIS format, a common goal that both
the USGS and FEMA are working toward. These data
are important to land planners, flood-plain regulators,
and insurance companies that rely on accurate esti-
mates of flood-prone areas. By increasing the accessi-
bility of the data by using CIS, we can substantially
improve our ability to analyze spatial data efficiently.
Problems Encountered
Problems using data supplied in DXF format in conjunc-
tion with CIS resulted primarily from the fact that the
DXF data were prepared for the purpose of making a
topographic map, not a CIS data layer. The contour lines
were segmented; that is, where ends of segments met,
they were not physically connected to form a topologi-
cally viable data layer. The data layer needed to be
edited because CIS requires topology for spatial-data
processing. Additionally, in areas where the topographic
gradient was particularly steep, contour lines were omit-
ted. In both cases, an attempt was made to allow CIS
to establish a physical connection of contour lines, but
subsequent manual interpolation was also required.
This may have introduced error into the data set. If future
work requires the use of DXF data, the request for data
should specifically state that all topographic contours be
continuous.
AutoCAD stores data differently from CIS, so a relation-
ship needed to be established between the data file
containing elevations and the data file associated with
the lines that make up the topography data layer. Sev-
eral lines from the DXF file did not have any data asso-
ciated with them, thus necessitating the addition of
contour elevation data by context with the adjacent con-
tours that did have data. This step may also have intro-
duced errors, but quality-control measures to verify the
topographic contours and contour elevations could help
to minimize these errors.
Output from the WSPRO model is in the form of a series
of points along cross sections that were connected by
manual interpolation. This step also may introduce some
error, but the same process must be performed when
not using CIS.
Conclusions
This report documents an example of how CIS can be
used to facilitate step-backwater modeling of flood-
prone areas. The results of the study show that signifi-
cant savings may be expected in the form of reduced
labor requirements. Furthermore, FEMA now requires
the use of CIS to conduct flood-study mapping, thus
providing a means to conduct additional spatial analy-
ses more efficiently. As aerial photography and CIS
technology improve, although additional sources of error
may arise, the overall accuracy, reliability, and repro-
ducibility of the model input and results should also
improve.
References
1. Mrazik, B.R., and H.A. Kinberg. 1991. National flood insurance
program: Twenty years of progress toward decreasing nationwide
flood losses. Water Supply Paper 2375. U.S. Geological Survey.
2. Shearman, J.O., W.H. Kirby, V.R. Schneider, and H.N. Flippo.
1986. Bridge waterways analysis model. Research Report
FHWA/RD-86/108. Federal Highway Administration.
3. Shearman, J.O. 1990. Users manual for WSPRO, a computer
model for water-surface profile computations. FHWA-IP-89-027.
Federal Highway Administration.
4. Davidian, J. 1984. Computation of water-surface profiles in open
channels: Techniques of water-resources investigations of the
United States Geological Survey. In: Applications of Hydraulics,
Vol. 3.
5. Koltun, G.F., and J.W Roberts. 1990. Techniques for estimating
flood-peak discharges of rural unregulated streams in Ohio. In-
vestigations Report 89-4126. U.S. Geological Survey, Water Re-
sources.
298
-------
Small Is Beautiful: GIS and Small Native American Reservations
Approach, Problems, Pitfalls, and Advantages
Jeff Besougloff
Upper Sioux and Lower Sioux Indian Communities, Redwood Falls, Minnesota
Background
The Lower Sioux Indian Reservation
The Lower Sioux Indian Reservation covers 1,743 acres
in southwestern Minnesota bordering the Minnesota
River. The land base consists of several hundred acres
of prime, flat agricultural land, a large wetlands slough
complex, prairie pothole wetlands, bottom land wet-
lands, small lakes, and approximately 250 acres of tim-
ber and brush. The elevation ranges from Minnesota
River level to the adjacent bluffs several hundred feet
higher.
This rural reservation contains a moderate amount of
infrastructure, including paved and dirt roads, 12-acre
sewage lagoon serving a moderately sized casino, com-
munity water system composed of a tower and small
treatment plant for the 90 mostly single-family dwelling
homes, convenience store/gas station/gift shop, com-
munity center, small two-room schoolhouse, pottery works
with gift shop, warehouse, and church. The casino-
fueled economic boost to the community recently re-
sulted in improvements to infrastructure and plans for
additional projects.
Office of the Environment
The tribal government was formed under the Indian
Reorganization Act of 1934. The governing body is an
elected five-person tribal community council that admin-
isters several government departments and is responsi-
ble for all government activities.
Under a U.S. Environmental Protection Agency (EPA)
Region 5 multimedia grant, the Upper Sioux and Lower
Sioux Office of the Environment (OE) was formed in late
1992. This unique joint venture between two tribes and
EPA envisioned moving the tribal governments into com-
pliance with major federal environmental legislation. At
the present time, only the Lower Sioux are developing
a tribal geographic information system (GIS). Therefore,
this article is solely applicable to this community, al-
though adoption of an Upper Sioux Community GIS
would likely follow a similar lifeline.
Environmental Regulation in Indian Country
Reservations are subject to a bewildering array of envi-
ronmental regulations. Numerous meetings, publications,
projects, and court decisions are devoted to determining
what law does or does not apply on any particular
reservation. In very general terms, the following can be
stated: state environmental regulations do not apply,
federal regulations do apply, and tribal regulations may
apply. From a tribal environmental employee's point of
view, numerous environmental regulations (whether fed-
eral or tribal) do exist that apply to reservation activities
and land, and they require compliance.
The Problem and the Solution
The OE's responsibility is to bring the reservation into
compliance with the 14 major pieces of environmental
legislation administered through EPA and directly appli-
cable to tribes. The OE finds itself responsible for any
and all other applicable environmental regulations and
all other less-regulated environmental media. The OE
currently has a staff of one.
In addition to the responsibility of moving the tribes into
compliance with federal environmental regulations, the
OE also develops environmental infrastructure, insti-
tutes environmental programs, and performs grant writ-
ing. Lower Sioux programs currently include Clean
Water Act (CWA) and Safe Drinking Water Act (SDWA)
compliance, solid waste planning including develop-
ment and institution of a household recycling program,
wetlands regulations compliance, wetlands mapping
and restoration, National Environmental Policy Act
(NEPA) compliance and site assessments, basic hydro-
logical data gathering and mapping, radon testing and
mitigation, environmental education as necessary, SARA
299
-------
Title III compliance and planning, and a variety of related
tasks.
Contracts or grants are currently being administered
under several Bureau of Indian Affairs (BIA) programs,
U.S. Geological Survey (USGS) and U.S. Army Corps
of Engineers (COE) matching funds programs, two EPA
programs, one Federal Emergency Management
Agency (FEMA) program, one Administration of Native
Americans (ANA) program, one Great Lakes Intertribal
Council (GLITC) program, and a cooperative project
with the National Tribal Environmental Council (NTEC).
Needless to say, responsibilities of the OE are limited by
staff hours rather than need. As distressing as the res-
ervation's unaddressed environmental needs are,
equally distressing (prior to CIS development) was the
helter skelter manner in which the OE digested the data
and information flowing into the office. Because of its
broad responsibilities and the administrative problems
being encountered, the OE began to investigate devel-
oping a tribal CIS.
The Lower Sioux GIS
System Selection
The Lower Sioux GIS system is a networked PC station
through the Bureau of Indian Affairs Geographic Data
Service Center's (BIA GDSC's) two Sun MP690
SparcServers in Golden, Colorado. The GDSC is the
hub of BIA's GIS and remote sensing program, known
as the Indian Integrated Resources Information Pro-
gram (IIRIP). The purpose of the IIRIP is twofold: first,
make GIS and remote sensing technology available to
tribes and BIA personnel; second, transfer these tech-
nologies to tribal organizations.
Database development and management functions,
technical support, development of simplified user inter-
faces, remote sensing interpretation, and implementa-
tion of equipment directives are performed by
approximately 30 GDSC employees for approximately
230 GDSC users. User technical support is also avail-
able through BIA field offices, each of which has a
designated GIS coordinator. Simplified user interfaces
for specialized programs have been developed includ-
ing the Lightening Display System and the Land Title
Mapping System. Quality control is provided for non-
BIA-produced data that will be inputted.
The GDSC has standardized on the ARC/INFO family
of software produced by Environmental Systems Re-
search Institute (ESRI). GDSC has developed a number
of hardware/software configuration options depending
on tribal needs and financial resources and based upon
GDSC experience. The OE happily relies on this expe-
rience to avoid the familiar horror stories related to
equipment and software incompatibility.
Based upon GDSC configuration advice, the initial GIS
setup will be on the OE's existing Compaq PC using
Tektronix Terminal Emulation software (EM4105) and a
Multitech modem (MT932BA). The system can use the
OE's Hewlett Packard (HP) DeskJet 500, although a
significant upgrade, possibly to an HP Paint or HP Excel
Paint, is soon expected. Initial startup hardware and
software costs are minimal in this configuration. Costs
for the above equipment and introductory training are
less than $5,000.
GIS Users at Lower Sioux
Initial setup and data loading will be in the OE, and the
OE employee will receive introductory training on the
system. Because the OE is formed through a coopera-
tive agreement between two tribes, the Upper Sioux and
the Lower Sioux, the OE is centrally located between
the reservations. The system will probably be relocated
to the Lower Sioux Community Center within 1 year. A
tribal government employee will receive advanced GIS
training and be available for all tribal government depart-
ments and businesses.
Funding
In addition to tribal contributions, funding has come
through several sources and joint agreements with the
tribe and BIA, EPA, and ANA.
Training
The GDSC supplies no-cost training to tribes. The Geo-
graphic Data Service Center 1995 Training Catalog (no
federal document number available) offers eight formal
courses repeatedly throughout the year, a 5-week intern
program, and a cooperative student program. Courses
are held at the GDSC or by request at BIA field offices
and tribal locations.
The GDSC also produces the monthly The Service Cen-
ter Review (ISSN 1073-6190), a helpful compilation of
current issues, available resources, system bugs, and
other items of interest to GDSC users.
Data Collection and Input
Data collection can be divided into three categories:
aerial photography, portable global positioning system
(GPS) data, and ARC/INFO export files created under
contracted studies.
Aerial Photography
Surface features and topography will be obtained us-
ing aerial photography reduced to GIS format, then
downloaded to the GDSC. Coverages will consist of 62
categories of features on a scale of 1 inch = 100 feet
with 2-foot contour intervals.
300
-------
Global Positioning Systems
Use of a portable Trimble, Inc., GPS Pathfinder Pro XL
submeter GPS mapping system purchased with assis-
tance from an ANA grant will allow updating of surface
features and addition of nonsurface features as neces-
sary. It will also facilitate input of attribute data.
The GPS will also be used during field work by USGS
on the Lower Sioux hydrological mapping project to
obtain data that otherwise would not be put into
ARC/INFO export file for any reason (i.e., it might not be
directly related to the project at hand or outside the
agreed upon data to be converted to ARC/INFO export
file form but nevertheless is of importance to the OE).
The alternative is that this type of information never
makes it into the CIS and is lost.
ARC/INFO Export Files
Fortunately, most federal agencies that supply funding
to tribes for environmental work are well versed in CIS
applications and the need for CIS-ready data. The OE
now requires all information and mapping to be deliv-
ered as an ARC/INFO export file with data registered to
a real world coordinate system. Downloading of this
data to the GDSC mainframes allows for direct input of
the data. The OE has contemplated, but not acted on,
conversion of existing data for the CIS. This is an ex-
pensive and time consuming process that must be
weighed in comparison with recollecting the data. Ironi-
cally, the lack of reservation data therefore becomes a
benefit because time consuming and expensive data
conversion is unnecessary.
The Intertribal GIS Council
Information gathering, networking, and addressing
uniquely tribal problems were some of the accomplish-
ments at the first annual meeting of the Intertribal GIS
Council (IGC) held in June 1994. Vendors as well as BIA
regional office and GDSC representatives answered
questions and presented panels. This annual confer-
ence is likely to become a major benefit to the tribe as
it continually develops the GIS.
The Future
As the tribal government becomes more familiarwith the
GIS, its uses, and advantages, recognized governmen-
tal needs will likely drive the development of further
coverages. The OE also expects to access existing
governmental data of importance to the tribe in an effort
to expand the GIS database and is actively seeking
sources of such information.
Philosophical Caveat
Albert Einstein stated that, "The significant problems we
face cannot be solved at the same level of thinking we
were at when we created them." Some assume that GIS
is the next level of reasoning in the environmental pro-
fession because we can accomplish tasks more quickly,
more efficiently, with more variables accounted for, and
beyond what we could have hoped to accomplish prior
to GIS.
Essentially, what we have gained is speed and the
capacity to include additional data, which is not what
Einstein was referring to when he spoke of the next
level. Wisdom, in the sense of a higher level of
understanding, is the necessary ingredient to the
solution of current environmental problems; in other
words, movement beyond the paradigm that created
the problem. GIS may be the tool that pushes the envi-
ronmental professional to the next level of wisdom by
presenting the data and information in a manner that
allows the user to stand back and see more clearly on
a higher plane. But that level can be found only within
the environmental professional himself or herself and
not within GIS.
301
-------
A GIS-Based Approach to Characterizing Chemical Compounds in Soil and
Modeling of Remedial System Design
Leslie L. Chau, Charles R. Comstock, and R. Frank Keyser
ICF Kaiser Engineers, Inc., Oakland, California
Introduction: The Problem
The cost-effectiveness of implementing a computerized
geographic information system (CIS) for environmental
subsurface characterization should be based on long-
term remedial objectives. A CIS project was developed
to characterize soil contamination and to provide design
parameters for a soil vapor extraction remedial system,
as part of a $120-million remediation and "land sale"
project in California. The primary purposes of the CIS
were to efficiently combine and evaluate (model) dispa-
rate data sets, provide "new" and more useful informa-
tion to aid in short-term engineering decisions, and
support the development of long-term cleanup goals.
The project had a major change in scope early on, and
the schedule was expedited to allow for the develop-
ment of "land sale" options and for actual site redevel-
opment at the earliest opportunity. Characterization of
chemically affected soil would have been compromised
given the above circumstances without an ambitious
undertaking of concurrently developing and implement-
ing a CIS with three-dimensional (3-D) geostatistical
and predictive modeling capabilities.
The GIS Approach
Computer solutions included the use of a cross-platform
(DOS and UNIX) GIS to quickly and systematically in-
corporate spatial and chemical data sets and to provide
a distributed data processing and analysis environment
(see Figure 1). Networked, DOS-based relational data-
bases were used to compile and disseminate data for
the numerous investigatory and engineering tasks.
UNIX-based computer aided design (CAD) and model-
ing applications received data from databases, per-
formed quantitative analyses, and provided 3-D
computer graphics. Given the aggressive project sched-
ule, exclusive use of one platform would not be realistic
due mainly to the limited data modeling capacity and 3-D
graphics in DOS systems. On the other hand, the high
startup and operating costs of several UNIX worksta-
tions would render their exclusive use much less cost-
effective.
The hardware and software configurations were inte-
grated in a client/server Intergraph InterPro 6400 with
48 megabytes of memory. It is largely a 3-D CAD system
with add-on modules of geologic mapping and 3-D vis-
ual models capable of consolidating both environmental
and engineering parameters for analysis (see Figure 1).
Textual environmental and geologic data were extracted
by SQL queries from relational databases and were
transferred to mapping and modeling modules via
PC/TCP cross-platform linkage.
The GIS assisted in making short- and long-term deci-
sions regarding health-risk-based regulatory strategy
and engineering feasibility. Use of spatial statistical and
predictive models was part of a CIS-based decision-
making loop (see Figure 2). The process supported
concurrent activities in:
Data collection: field program.
Numerical models of remedial system configurations.
Development of cleanup goals from health risk
assessments.
Remedial design with CAD capability.
Site Background
In early 1993, the remedial investigation of the operable
unit for soil at a former aircraft manufacturing facility in
southern California was thought to be ready for remedial
alternatives feasibility study. ICF Kaiser Engineers, Inc.,
was awarded the contract to perform feasibility studies
on applicable soil cleanup technologies and to sub-
sequently design and manage the installation and early
operation of the selected technologies. After $700,000
was spent evaluating data collected by previous consult-
ants, it was decided that an additional $5 million worth
302
-------
DOS-Based
Token Ring Network
Site
Survey
Activities
Sampling
K
R*
RDBMS
(Paradox)
UNIX-Based
Data Visualization and Computer-Aided Engineering
-4
PC/TCP
Informix
RDBMS
Visualization
CADD
Laboratory
Analytical
Results
Health Risk
Assessment
Regulatory
Reporting/
Interaction
Modeling Module
3-D Kriging Spatial
Analysis
3-D Vapor Flow
Analysis
3-D Chemical Mass
Transport
Analysis
H
TCP/IP
High Impact 3-D
Solid Models
Volumetric
Analysis of
Static and
Dynamic Mass
Flow
Subsurface
Engineering
3-D Piping
Electrical
Treatment
System Housing
Accelerated Implementation To Establish Cle
Require Low Overhead Upfront Cost and Off-the
PC-Based
Data Visualization
Report Generation
anup Goals
Shelf Software
Regulatory and
Public
Interaction
Client
Deliverables
Extended Engineering Phase
Specialized CAD Software and Analytical Staff
Figure 1. Multiplatform GIS project.
Ongoing Site Investigation
Soil/Soil Vapor Sampling
Data Processing
1. Storage (RDBMS)
2. Dissemination
Modeling
1. Spatial Correlation
2. Fate and Transport
Conceptual
Treatment Alternatives
Health Risk
Assessment
Permitting
Environmental Planning
Site Redevelopment
Establishment of
Cleanup Goals
Final Remedial
Design
Figure 2. GIS-assisted decision tree.
of field activities were required to more definitively esti-
mate the volume of chemically affected soil and the
nature and extent of contamination at the facility. Be-
cause of the data gaps, the selection and design of
alternatives could not be addressed with a high degree
of certainty. Hence, computer assisted data processing
was crucial to speed up the feasibility study, accelerate
downstream work, and reduce the overall project sched-
ule to the minimum.
The site is environmentally complex, covering an area
of approximately 120 acres. As a result of nearly half a
century of aircraft production and development, soil be-
neath the facility is affected by fuel and heavy oil hydro-
carbons (TPH) commingled with volatile compounds,
mainly perchloroethylene (PCE) and trichloroethylene
(TCE) (see Figure 3). Ground water at 170 feet below
ground surface is affected by TCE and PCE, but it is not
part of the drinking water aquifer. The facility has been
demolished, and shallow contaminated soil has been
excavated and back-filled to an interim grade.
Methodology
Health-Risk-Based Cleanup Goals
Central to determining the volume and kinds of data to
be collected was the question of whether chemicals in
soil represented potentially unacceptable risks to human
303
-------
VAPOR DISTRIBUTION W
Figure 3. Aircraft manufacturing facility in California. Outline of demolished buildings located at the 120-acre site are shown as
surface features for reference. A geostatistical model of a 3-D kriged VOC soil vapor cloud in the subsurface was simulated
with Intergraph's MGVA. Views displayed are: 3-D isometric, vertical section of chemical isoplaths, and a nearly plan view.
Digital simulation also illustrates VOCs affecting ground water in a dispersive nature at a depth of nearly 170 feet bgs
(shown at bottom of isometric view).
health and to the environment, with the former being of
particular concern to construction workers onsite during
redevelopment.
Because site redevelopment was scheduled to begin in
the near term, data collection and CIS analysis concen-
trated on shallow depths (top 20 feet), with decreasing
sample density at greater depths. A health-risk-based
cleanup goal (HBCG) approach to collecting more data
was to establish cleanup goals for near-term remedia-
tion of the shallow soils as well as for long-term remedial
measures of contaminated soils at greater depths. Fur-
ther, various regulatory agencies had to approve the
estimated cleanup goals in a short time. Ongoing site
demolition and excavation schedules encouraged the
aggressive regulatory negotiations. The shallow cleanup
goals for volatile organic compounds (VOCs) and TPH
determined the volume of contaminated soil to be re-
moved. At greater depths, data gaps were minimized to
more definitively characterize the nature of TPH and
VOC contaminations and to facilitate the implementation
of long-term remedial objectives (i.e., in situ soil vapor
extraction).
In situ soil vapor and soil sampling composed the field
program, which provided data to map the subsurface
distribution of volatile organic compounds, including
TCE and PCE. Only in situ soil sampling was used for
characterizing TPH. The ratio of soil vapor to soil sam-
ples was 4:1. No previous soil vapor information was
available. ICF Kaiser has been refining the technique of
comparing results from paired soil vapor and soil sam-
ples in past and similar projects. Hydraulic probes were
used instead of drilling to acquire soil vapor samples at
shallow depths. This minimized waste and cost in the
field program significantly.
Risk Assessment and Spatial Analysis
Human health risk analyses were conducted for the
entire site, and risk factors were contoured and overlaid
304
-------
on maps of past usage and known soil contamination
areas. Before the risk modeling could proceed, chemical
and lithological data gathered in the past 7 years and
those acquired by ICF Kaiser populated the environ-
mental relational databases. Approximately 522 soil va-
por probes were located in 100-square-foot spacings
with additional probes in areas requiring better plume
definitions. The database contains approximately
15,000 xyz-records of soil and soil gas laboratory ana-
lytical results. This information in text and graphics form,
combined with site infrastructures and building outlines
with attributes of "past usage," were stored as map
layers, making up the CIS nucleus. Accuracy of site
maps was verified with aerial photographs when avail-
able. Data types combined for computerized evaluation
included known locations of contaminated soil, contami-
nated ground water, soil types, and site features. Com-
posite risk maps of the above data were analyzed for
data gaps at discrete depth intervals. This analysis was
performed while the field program was in progress
and hence gave guidance to optimize the locations of
additional data points and to minimize the number
of samples taken.
The MGLA/MGLM mapping module and the MSM ter-
rain modeling module tracked the earth excavation and
removal of contaminated soil. Excavation was largely
part of site demolition. It also expedited the removal of
TPH contaminated soils, however, because no other
short-term means of remediation are available for these
substances. Tracking of removed soils was essential
because concurrent field activities were occurring in site
demolition, data gathering, and risk modeling.
The CIS coordinated all three. Geologists and surveyors
provided terrain data from daily excavation activities,
which were transcribed into database formats. Maps
illustrated the locations of excavated soil and removed
chemicals in soil at various depths. Although TCE and
PCE were of foremost concern as health risks, all com-
pounds and some metals identified in soil were
screened for unacceptable risk. Terrain modeling (map-
ping) as part of health risk assessment may seem un-
usual, but results of estimated cleanup levels and
accurate locations of left-in-place contamination, mostly
soils at greater depths, were critical to the cost-effective-
ness and proper design of long-term remedial systems.
Characterization of Subsurface VOCs
In situ soil vapor extraction (SVE) of total volatile com-
pounds in dense nonaqueous, liquid, gaseous, and ad-
sorbed solid forms in the subsurface produced favorable
results that have been well documented in recent years.
ICF Kaiser proposed a very large-scale SVE system
(see Figure 4), perhaps the largest yet, as long-term
remedial technology for this former aircraft manufactur-
ing site. The primary design problem was speculating on
air flow capacity and operating time of the complex
system components. The SVE system comprises three
fundamental elements:
Front-end, in situ subsurface vents (totaling 193
corings).
Applied vacuum and air transport manifolds linking
the subsurface vents to the treatment compound (dis-
tance of one-quarter mile with over 100 manifolds).
A multivessel activated carbon treatment system.
To size the pipes, carbon vessels, and vacuum required
to achieve a certain rate of VOC removal, the total mass
and nature of sorption had to be known. Due to the
schedule-driven nature of this project, the SVE design
accounted for the time needed to accomplish the
cleanup goals.
To estimate the extent and total mass of VOCs in the
subsurface, soil vapor data were input to a 3-D kriging
algorithm (1) to produce a concentration continuum
model (see Figure 3). This solid model of predicted total
VOC concentrations took the form of a uniformly spaced
3-D grid-block that completely encased the site. Cell
sizes ranged from 10 to 20 cubic feet, depending on the
model run, number of data clusters, density of data
points in areas of clustered data, and the standard
deviation of variances for estimated values in all cells.
The Fortran program estimated a concentration value
for each cell based on the nearest field sample(s).
The validity of such "block kriging" models can be judged
by the size of the variances, smoothness, and agree-
ment with nearby field data. Because volume is a known
quantity in kriging, the total mass can be calculated by
incorporating soil bulk density or porosity, both of which
were less than abundant for this investigation. Render-
ings of kriged results in 2-D plan view contour maps,
cross-sectional maps, and 3-D "vapor cloud" (see Figure
3) were included in client reports and used in regulatory
presentations and public forums.
Remedial Design Layout
Final Extension of a Fully Integrated CIS
With the total mass and extent of VOCs derived from
3-D kriged results, the applied vacuum at individual vent
heads and the cumulative pressure (negative) neces-
sary to extract and transport VOC vapors from the sub-
surface to the treatment system can be estimated. We
performed 3-D air flow analysis by use of finite differ-
ence fluid flow models and chemical transport models.
The Fortran codes used to approximate compressible
flow and chemical transport were AIR3D (2) and VT3D
(3), respectively. Air flow simulations focused on maxi-
mizing vacuums at the shallow depths down to 20 feet
to expedite remediation of contaminated soils that were
305
-------
Figure 4. A rendering by the Intergraph 3-D Plant Design System of an in situ soil vapor extraction and treatment system. The
cutaway section located near the upper left portion of the figure exposes some of the 193 subsurface extraction vents
bottoming at 120 feet (bgs). These vents are located in a cluster for long-term extraction of the VOC vapor cloud presented
in Figure 3. Vents are connected to a system of parallel airflow manifolds (right side of figure), which runs one-quarter of
a mile to the treatment compound (foreground of figure).
not removed during site demolition and excavation. The
lower depths were also included in each simulation.
Transient mass transport models incorporate flow fields,
given by flow models, and predicted cleanup times
based on established HBCG cleanup goals. As VOC
concentrations in an operating SVE system fall below
cleanup levels in the top 20 feet, thus minimizing human
risk, available vacuums thereafter will be diverted to
vents at lower depths to be part of long-term extraction
scenarios. Models suggested that cleanup for the top 20
feet can be accomplished within 1 year.
Numerical models prescribed vacuum levels at each
vent head, which is the aboveground segment of a
subsurface SVE vent. The 193 vents are connected to
a system of parallel manifolds (see Figure 4) that trans-
port vapor to the treatment system. With the vacuums
known at vent heads, the size of manifolds and capacity
of vacuum blowers can be determined and integrated
into the overall system design. With 3-D Plant Design
module as part of the Intergraph CAD/GIS, manifold
layouts and treatment compound can be modeled in 3-D
and easily checked for pipe routing interferences. The
final layout of the SVE system was overlaid onto contour
maps of total VOC concentrations to check on accuracy
and completeness of vent locations and manifold layouts.
Conclusion
Maximized Visual and Analytical Responses
One goal of this project was to expedite regulatory
negotiations and gain early acceptance of cleanup
goals. The computerized data processing and visualiza-
tion contributed generously to the rapid understanding
of modeling results by expert regulators and the lay
public. Likewise, the CIS facilitated the response to
regulatory comments. Positive comments first came
306
-------
from the client's in-house review of model results and
the high impact 3-D color rendering of kriged VOC dis-
tributions in the subsurface (see Figure 3).
Analytically, benefits were derived from the efficiency of
electronic data access and the ability to "predict" the
presence of contaminant in areas with sparse field data.
The process of kriging involves the linear interpolation
and extrapolation of existing data. The resultant con-
taminant distribution is a "conservative" model that pro-
vided the best fit with field data and validated
conceptualized subsurface conditions. Further, models
provided conservative estimates of mass and extent of
PCE and TCE contaminations. Kriging also provided
information on the uncertainty of the predicted chemical
distribution, which is extremely useful for regulatory dis-
cussion and system design. The efficiency of computer
models allowed investigators to perform numerous
model runs with varied boundary parameters, such as
cell size and search radii, in the kriging process.
Accurate mapping of excavated soil and the removal of
most TPH source areas provided the incentive to criti-
cally assess the feasibility of a no-action remedial sce-
nario for these substances at greater depths. With
removal of many TPH source areas, 1 -D finite difference
models (4) were used to assess the mobility of TPH in
NAPL and adsorbed residual phase. Specifically, mod-
els assessed the likelihood of largely residual-phase
TPH affecting ground water and migrating upward to
affect indoor air volumes via gaseous diffusion. Results
were extremely favorable; models predicted negligible
likelihood of TPH affecting ground water or indoor air
volumes. Combined with CIS graphic evidence of spe-
cific areas of excavated soil and the absence of TPH
sources, regulatory agencies accepted the model re-
sults, and the no-action remedial alternative for TPH
was approved.
References
1. Deutsch, C.V., and A.G. Journal. 1992. Geostatistical software
library and user's guide. New York, NY: Oxford University Press.
2. U.S. Department of the Interior Geological Survey. 1993. AIR3D:
An adaptation of the ground-water flow code MODFLOWto simu-
late three-dimensional air flow in the unsaturated zone. Books and
open file reports. Denver, CO.
3. Zheng, C. 1994. VT3D: Numerical model for VOC removal from
unsaturated soil (draft). Bethesda, MD: S.S. Papadopulos and
Associates, Inc.
4. Rosenbloom, J., P. Mock, P. Lawson, J. Brown, and H.J. Turin.
1993. Application of VLEACH to vadose zone transport of VOCs
at an Arizona Superfund site. Groundwater monitoring and reme-
diation (summer), pp. 159-169.
307
-------
Polygon Development Improvement Techniques for Hazardous Waste
Environmental Impact Analysis
David A. Padgett
Austin Peay State University, Clarksville, Tennessee
Introduction
Recently, concern has arisen regarding the effect Super-
fund sites have on surrounding communities and, spe-
cifically, the distribution of those impacts on target
populations. In designing geographic information sys-
tems (CIS) applications for analyzing potential impacts
of hazardous wastes or waste sites on adjacent neigh-
borhoods, many challenges may be encountered. CIS
database design requires addressing questions of time,
space, and scale.
The U.S. Environmental Protection Agency (EPA) and
other federal agencies have conducted studies that in-
dicate that certain sectors of the population may be
more vulnerable to exposure to toxics than others. To
date, federal departments have enlisted in several CIS-
based research projects that attempt to delineate "geo-
graphic hot spots" of toxic contamination. Such CIS
applications at hazardous waste sites have typically
used polygons to represent data from census tracts
and/or municipal boundaries. In most cases, however,
census tract and other boundaries do not necessarily
jibe with community and neighborhood boundaries; there-
fore, the polygons representing characteristic data for
target populations may not be consistent with the actual
status of those populations.
The objective of this paper is to demonstrate CIS meth-
ods for producing, to the greatest degree possible, so-
cioeconomically and culturally homogenous polygons
for impact analysis of specific sensitive populations
and/or communities. The paper presents case studies
of community/neighborhood characterization problems
encountered in developing polygons during previous
field investigations involving lead (Pb) contamination,
toxic release inventory (TRI) sites, and solid/hazardous
waste sites. The paper attempts to demonstrate effec-
tive solutions and suggestions for improving polygon
development, including CIS data manipulations and
software applications. In addition, the paper provides
geographic and groundtruthing field methods to sup-
port and enhance the accuracy of remotely obtained
information. Finally, the discussion includes commu-
nity and geographic hot spot analyses for potential
public health impacts.
Background
In 1992, EPA established the Environmental Equity
Workgroup. Its members included personnel from the
Offices of Toxic Substances and Civil Rights, as well as
Policy, Planning and Evaluation. The workgroup con-
ducted an extensive study on environmental equity is-
sues. Their report offered several recommendations for
improving federal agency efforts in protecting minority
and low-income populations and recognized a need for
more spatial and demographic data. The final report,
titled Environmental Equity: Reducing Risk for All Com-
munities (1), was released in February 1992 and con-
cluded that "there is (sic) limited data on environmental
health effects by race; there are differences by race and
income in potential and actual exposures to some pol-
lutants." In response to the above findings, the work-
group offered the following recommendations (1):
EPA should establish and maintain information
which provides an objective basis for assessment of
risks by income and race, commencing with devel-
oping a research and data collection plan.
It (EPA) should revise its risk assessment proce-
dures to ensure . . . better characterization of risk
across population, communities or geographic ar-
eas. In some cases it may be important to know
whether there are any population groups at dispro-
portionately high risk.
The Agency for Toxic Substances and Disease Registry
(ATSDR) formed a community health branch to specifi-
cally examine the potential health impact of hazardous
waste sites upon people living in surrounding communi-
ties. The new branch's personnel direct ATSDR's minor-
308
-------
ity health initiative, which focuses upon health threats to
minority populations, including those from environ-
mental contaminants.
In addition, EPA established the Office of Environmental
Equity. The office's mission includes analyzing environ-
mental impacts upon minority populations, providing
technical assistance to disadvantaged communities,
and establishing environmental initiatives at minority
academic institutions (MAIs). The office serves as a
clearinghouse of environmental data and information for
groups and individuals involved in environmental equity
activities.
In 1993, Representative John Lewis (Democrat-Georgia)
introduced the Environmental Justice Act to Congress. The
act requires EPA and the Department of Health and Hu-
man Services (DHHS) to establish the geographic units for
determining environmental high-impact areas (EHIAs),
which are the 100 geographic areas found to have the
highest volumes of toxic chemical releases.
GIS Applications
CIS could potentially help address the above data and
information needs of EPA. The Agency specifically ac-
knowledges this in other recommendations (1):
EPA could further develop its enforcement prioritiza-
tion policy to target high-risk populations. Under this
scheme, the most exposed and highly susceptible
populations in each region would be targeted for
enforcement actions. Geographic Information Sys-
tem technology could be used to identify high-risk
populations.
Several recent environmental studies have employed
computer applications and spatial data. Goldman (2) used
GIS in a major study that graphically displayed counties
having high percentages of African-Americans, hazardous
wastes, and diseases. Mohai and Bryant (3) applied a
linear regression model to show a positive correlation
between increasing proportions of minority populations
and decreasing distances from hazardous waste sites in
Detroit. Lavalle and Coyle (4) conducted an extensive
analysis of computer databases that hold hazardous
waste law enforcement information for the past 10 years.
They found inequity in enforcement and remedial actions
in white communities versus nonwhite communities.
EPA has enlisted GIS for community environmental im-
pact projects at Regions II and III. EPAs Office of Health
Research (OHR) is investigating methods for linking
demographic data with TRI information to evaluate the
relationship between levels of hazardous waste re-
leases and exposure risks in minority communities. EPA
has also developed the TRI "risk screening" process,
which employs TRI, U.S. Census data, and GIS to iden-
tify TRI releases that may pose significant risk to human
health or the environment (5).
Both EPA (6) and the North Carolina Department of Envi-
ronment, Health, and Natural Resources (7) have recently
completed CIS-based environmental investigations. The
EPA study involved GIS analyses of TRI chemical releases
in the southeastern United States. The report included
numerous CIS-produced maps that show locations where
TRI releases may be affecting densely populated areas
and sensitive ecosystems. The North Carolina study
applied GIS in searching for sources of lead-poisoning
in children. Findings indicated a positive spatial correla-
tion between high lead-contamination risk communities
and those having certain socioeconomic characteristics,
such as low income, above-average African-American
population percentages, and above-average percent-
ages of residents receiving public assistance.
The ATSDR recently implemented a study using GIS to
evaluate and analyze the demographic characteristics
of populations near National Priorities List (NPL) sites.
According to the ATSDR, "As a result of our pilot tests,
we have determined GIS to be the best methodology for
identifying potentially impacted minority populations" (8).
Limitations of GIS
While GIS may be a viable tool for investigating environ-
mental inequity, it is not an absolute solution. Issues
involving hazardous waste impact assessments tend to
be very complex without the added dimension of racism
or discrimination. Efforts to determine a causal relation-
ship between the presence of minority communities and
environmental hazards must consider the questions of
time, scale, and place. Unfortunately, in many instances,
GIS applications may be unable to adequately illustrate
these three pertinent issues resulting in skewed or alto-
gether incorrect conclusions.
With respect to scale, among the immediate concerns
when applying GIS is selecting appropriate sizes for
polygons. As indicated above, EPA is in the process of
determining the scale for EHIAs. A polygon may be a
county or a census tract. Figure 1 illustrates problems
EH3
B
a 0.
m m " " H
o Q
90 Percent Nonminority
Figure 1. Problems of scale in GIS polygon design.
309
-------
of scale associated with polygon size selection. At the
county-size scale, a case of environmental inequity ap-
parently exists with the presence of a Superfund site in
the sample county because 80 percent of the county's
population belongs to a minority ethnic group. A closer
look, however, reveals that the population residing in the
immediate vicinity of the waste site is predominantly
nonminority.
Useful as CIS may be, its output in some cases may
display static conditions without considering human
movements over time. Figure 2 displays a common
situation associated with the "filtering" phenomenon, in
which a nonminority population moves out of an area
while increasing numbers of an ethnic minority group
moves into it. The figure shows that in 1950, a nonmi-
nority community surrounded a TRI site (i.e., an active
industrial site releasing toxic substances). By 1990,
the demographics of the neighborhood had changed
along with the status of the TRI site, which is now an
abandoned Superfund site. A CIS database probably
would contain only information on the community from
1990. Such an instance could suggest that some form
of environmental injustice exists given the presence of
the Superfund site within the minority community. Ac-
counting for the dynamics of time and human move-
ment, however, would show that the waste site preceded
the minority population and that, in actuality, the minority
community moved toward the site. This conflicts with the
prior notion that unsavory forces placed the site in the
minority community.
Figure 3, a schematic of polygons used in an investiga-
tion into sources of lead-poisoning in children, also dis-
plays the limitations of CIS with respect to time and
human dynamics, but at a lesser time interval. The study
area is divided into low- and high-risk areas for lead
contamination. The locations at which children with lead
poisoning were found, however, do not correspond with
the areal risk factors. In this case, the CIS is limited in
its ability to follow human movements on a daily basis.
For instance, a parent with a child who exhibits un-
healthy blood-lead concentrations may report the child's
home address as someplace within the low-risk poly-
gon. The child may attend school in the high-risk area,
however. The children's points of contact with lead may
not necessarily correspond with their home addresses,
resulting in an inaccurate graphic display.
With respect to place, CIS may be limited in its ability to
determine the specific borders of a socioeconomically
and demographically homogenous human population.
Tosta (9) and Coombes et al. (10) discuss the dilemmas
associated with neighborhood boundary delineation in
CIS applications. Figure 4 displays a schematic of a
census tract. The CIS database may list the tract's per
Figure 2. Example of changing community demographics with
time near a hazardous waste facility.
= Homes of Lead-Poisoned Children
Figure 3. Lack of correspondence between locations of lead-
poisoned children and high-contamination risk areas
due to daily dynamics of human movements.
Middle
Income
Community
Census Tract Schematic
Figure 4. Example of significant neighborhood-type variation
within a single polygon, possibly resulting in skewed
socioeconomic data.
310
-------
capita income as relatively low and may list the tract as
a low-income neighborhood. Further investigation may
find, however, that two very different socioeconomic
communities exist within the tract, one middle-class
and the other a public housing facility. Frequently,
middle-income, African-American communities have
low-income housing projects built adjacent to them. With
respect to the polygon in Figure 4, if an investigator
wanted to research health impacts of toxic wastes for
low-income households, using this polygon and others
like it would inaccurately depict communities within
them.
An additional problem in community definition is deter-
mining exactly what defines a minority community. The
most common indicator for discerning a minority com-
munity would be the existence of a clear majority of
some minority group as in Polygon A of Figure 5, or
where the minority group makes up more than 50 per-
cent of the population as in Polygon B. In some in-
stances, however, communities have received minority
status without the presence of the conditions in Poly-
gons A and B.
Previous investigations show a number of measures
used to identify minority communities and census tracts.
Greenpeace conducted a 1990 environmental justice
study that determined a community's status as minority
based upon the relationship between a community's
percentage of minority population and the selected mi-
nority group's national percentage (11).
Polygon C in Figure 5 depicts Greenpeace's minority
community definition. Taking the target ethnic group in
Polygon C as African-Americans and the hypothetical
extent of Polygon C as the United States (African-Ameri-
cans make up approximately 12 percent of the total U.S.
population), one may determine that Subpolygon C is a
minority polygon or community because its minority per-
centage is over twice that of the national percentage or
extent of the large population in Polygon C. The condi-
tion that Polygon C illustrates is also evident in a study
Subpolygon C
30 Percent Minority
Polygon A
80 Percent Minority
Polygon B
51 Percent Minority
Polygon C
12 Percent Minority
Figure 5. Problems in defining minority polygons.
by Mohai and Bryant (3). The authors claim that envi-
ronmental inequity exists in Detroit where they found
that on average, within a 1-mile radius of the city's
hazardous waste facilities, 48 percent of the population
is nonwhite.
Solutions With GIS and Supporting
Technological Methods
From the above discussion, GIS may appear very lim-
ited for use in environmental community impact investi-
gations, but GIS can actually be an extremely effective
tool if employed with appropriate supporting technology.
Preliminary Groundtruthing
To design GIS databases that reflect the true nature of
target groups and human dynamics, preliminary
groundtruthing may be necessary. In some cases, inves-
tigators make gross interpretations of suspected envi-
ronmental inequity without actually visiting the study
area. Without groundtruthing prior to final database de-
velopment, questions of time, scale, and human dynam-
ics may be left unanswered. The integrity of databases
produced this way and the antecedent conclusions
based upon them may fall into question. Thus, because
the nature of environmental and human health impact stud-
ies is complex, some on-the-ground work should precede
or at least accompany database construction efforts.
Cause-Effect Analyses
Epidemiological studies are increasingly employing
GIS. This use is important with respect to environmental
investigations because in many cases, proof of a corre-
lation between a waste site and community health prob-
lems may be necessary. Croner et al. (12) describe
statistics-supported GIS applications for linking "cancer
hot spots" with pollution sources. Without conclusive
evidence that waste sites and other environmental
hazards negatively affect health in socioeconomically
disadvantaged populations, claims of environmental in-
justice may be difficult if not impossible to prove.
In historical analyses of waste facility sitings, GIS may
be useful, along with the support of gravity models, in
investigating whether the sitings followed the prescribed
logic for such siting decisions. Noble (13) wrote that the
costliest aspect of waste facilities management is trans-
portation; therefore, siting decisions should favor loca-
tions in closest proximity to a selected community's
centroid of waste production. Using a GIS gravity model
with data for household wastes produced within a given
locality, the center of gravity of the volumes of wastes
produced could be located. Historical analyses of past
siting decisions may find that past siting decisions defied
logic. Instead of finding waste sites placed in environ-
mentally safe locations as close as possible to areas of
greatest refuse generation, analyses may find instead
that sites have been placed farther away in disadvantaged
311
-------
communities. Both the taxpaying and potentially af-
fected residents would pay for such unsavory siting
practices.
Conclusion
The potential for technological applications, including
CIS, in this arena is great. Increased involvement by
technologically trained environmental professionals is
imminent. Their future involvement, however, must fo-
cus on the scientific soundness of investigative meth-
ods, data integrity, and the equitable participation of
potentially affected citizens in any subsequent decision-
making processes.
References
1. U.S. EPA. 1992. Environmental equity: Reducing the risk for all
communities. EPA/230/DR-92/002. Washington, DC.
2. Goldman, B. 1991. The truth about where you live. New York,
NY: Times Books.
3. Mohai, P., and B. Bryant. 1992. Environmental racism: Reviewing
the evidence. In: Bryant, B., and P. Mohai, eds. Race and the
incidence of environmental hazards: A time for discourse. Boul-
der, CO: Westview Press.
4. Lavalle, M., and M. Coyle. 1992. Unequal protection: The racial
divide in environmental law. The Environmental Professional
153:S1-S12.
5. U.S. EPA. 1990. Toxics in the community: National and local
perspectives. EPA/560/4-90/017. Washington, DC.
6. Stockwell, J.R., J.W Sorensen, J.W. Eckert, Jr., and E.M. Car-
reras. 1993. The U.S. EPA geographic information system for
mapping environmental releases of toxic chemical release inven-
tory TRI chemicals. Risk Analysis 132:155-164.
7. North Carolina Department of Environment, Health, and Natural
Resources. 1993. Lead database evaluation and CIS modeling
project. PR-242592. Raleigh, NC: National Institute for Environ-
mental Health Sciences.
8. Harris, C.H., and R.C. Williams. 1992. Research directions: The
Public Health Service looks at hazards to minorities. EPA J.
181:40-41.
9. Tosta, N. 1993. Sensing space. Geo. Info. Sys. 39:26-32.
10. Coombes, M., S. Openshaw, C. Wong, and S. Raybould. 1993.
Community boundary definition: A GIS design specification. Re-
gional Studies 3:280-86.
11. Bullard, R.D. 1993. Environmental equity: Examining the evidence
of environmental racism. Land Use Forum (Winter), pp. 6-11.
12. Croner, C.M., L.W Pickle, D.R. Wolf, and A.A. White. 1992. A
GIS approach to hypothesis generation in epidemiology. In:
American Society for Photogrammetry and Remote Sens-
ing/American Congress on Surveying and Mapping/RT 92, Wash-
ington, DC (August). Technical papers, Vol. 3: GIS and
cartography, pp. 275-283.
13. Noble, G. 1992. Siting landfills and LULUs. Lancaster, PA: Tech-
nomic Publishing Company, Inc.
312
-------
Comparing Experiences in the British and U.S Virgin Islands in Implementing GIS
for Environmental Problem-Solving
Louis Potter
Office of the Chief Minister of the Government, British Virgin Islands
Bruce Potter
Island Resources Foundation, St. Thomas, U.S. Virgin Islands
The British and U.S. Virgin Islands:
Comparisons and Contrasts
British Virgin Islands
Three miles to the north and east of the U.S. Virgin
Islands (USVI) lie the British Virgin Islands (BVI), a
group of 36 islands and cays with a total area of 60
square miles. The four largest are Tortola (24 square
miles), Anegada (14 square miles), Virgin Gorda (8.5
square miles), and Jost Van Dyke (4.5 square miles).
Geologically, the BVI belong to the Greater Antilles, and
like the USVI and Puerto Rico, rise from the Virgin
Banks (or Puerto Rican Shelf). Rocks of the BVI, except
Anegada, consist of thick, steeply inclined, metamor-
phosed volcanic and sedimentary stratified series of
Cretaceous age, with dioritic and pegamitic intrusions.
Anegada is a 30-foot-high emergent coral limestone
platform, apparently from the Pleistocene age.
The BVI are a crown colony of the United Kingdom (UK)
with a total population in 1991 of 16,108. Most of the
population resides on Tortola (13,225 inhabitants). Be-
tween 1980 and 1991, population increased 46.6 per-
cent. The BVI have internal self-government with an
elected council headed by a chief minister. The UK
appoints a governor to represent the queen and to man-
age defense, internal security, external affairs, civil serv-
ice, and court administration.
The BVI economy is based mainly on tourism and serv-
icing international business. Sailing and diving are piv-
otal features in BVI tourism. The average number of
tourists per capita in the BVI is 221, compared with 119
for the USVI (1). The BVI 1990 per capita income was
recorded at $10,125. Prior to the 1970s, the BVI economy
was based on subsistence agriculture and remittances
from British Virgin Islanders who worked in the USVI.
U.S. Virgin Islands
The USVI are an unincorporated territory of the United
States, purchased from Denmark in 1917. The total popu-
lation in 1990 was about 101,000, divided among St.
Thomas (48,000), St. Croix (50,000), and St. John (3,500).
The tourism-dominated economy of the Virgin Islands gen-
erated a per capita Gross Territorial Product in 1990 of
approximately $11,000the highest in the Caribbean (2).
Geographically, geologically, and topographically, St.
Thomas (28 square miles) and St. John (20 square
miles) are similar; they are both largely volcanic, have
deeply indented coastlines, and lie on the Puerto Rican
Shelf. St. Thomas and St. John are close to the BVI. St.
Croix is a relatively large (84 square miles) and mostly
limestone island that lies on its own submarine ridge,
which rises more than 4,000 feet from the bottom of the
Caribbean Sea. St. Thomas and St. John are about
5 miles apart, and St. Croix is 48 miles south of them.
During the height of tourism development (from the
late 1950s through the mid-1970s), the USVI experi-
enced average annual compound population growth
rates of over 6 percent, as well as a doubling in real
incomes. This unprecedented paroxysm of growth is
still being assimilated by a population that differs
greatly from the 30,000 people who lived in a predomi-
nantly agricultural USVI in 1950. In 1990, the USVI
received a daily average of 37 visitors per square
kilometer. This compares with a visitor load of 23
visitors per day per square kilometer in the BVI, which
is also a high-density tourist destination (1).
Background to GIS Implementation
Activities
British Virgin Islands
The idea of geographic information systems (GIS) ap-
plications in the BVI first arose with a presentation about
313
-------
a proposed project for St. Lucia, made by Dr. Jan Ver-
meiren of the Organization of American States at the
Caribbean Conference of Planners in Kingston, Ja-
maica, in 1984 (3). The Town and Country Planning
Department recognized that it could use GISs analytical
and display properties to make presentations to the chief
minister and the BVI Executive Council. This proposal
fell on fertile ground, given a relatively long-held tradition
of support for the Town and Country Planning Depart-
ment by the United Nations Development Programme
(UNDP) and the British Development Division, dating
back to the early 1970s.
Subsequent to this inspiration, the Town and Country
Planning Department requested budget authority to de-
velop a land use database. This database would include
buildings, property boundaries, and constructed and
natural features of importance. The Finance Department
hoped this project would help combat growing competi-
tion to the postal services by independent package de-
livery services. They renamed the project the National
Addressing System, and the legislature provided
$200,000 to provide a physical address for each prop-
erty in the territory.
The Town and Country Planning Department conducted
a pilot project, focusing on Road Town, the capital. The
pilot project demonstrated that the hard copy land own-
ership or cadastral maps that the Survey Department
was then using were inadequate for accurately account-
ing for properties, even in the BVI's most developed
urban areas. Therefore, the Town and Country Planning
Department expanded the scope of the National Ad-
dressing System project to identify options for increas-
ing the accuracy of property ownership records,
including maps.
Because the Town and Country Planning Department
had little experience with computerized land information
systems, departmental managers sought support from
the UNDP office in Barbados. The UNDP had previously
assisted with the department's development control ap-
plications database. This mode of operation, in which
the BVI purchase services provided through the UNDP,
which acts as a "vetting agent" for consultants and other
technical assistance, continues to this day. An expert
from the United Nations Community and Housing Serv-
ices (UNCHS) Nairobi office provided the first such con-
sultation by exploring how the BVI might implement a
CIS. Over the next few years, three other experts
provided input, which the Town and Country Planning
Department gradually integrated into a picture of how to
use a CIS within the technical and financial limitations
of a small island government.
In the meantime, external conditions were improving the
chances for the success of the BVI program. British
Virgin Islanders were receiving formal and informal train-
ing in computer applications in general, and specifically
in AutoCAD drafting systems, which increasing numbers
of local architects and engineers are using. In addition,
the power of rugged microcomputer systems that could
withstand the harsh operating conditions of the Virgin
Islands was also improving, and local dealers were
increasing their skills in support of such systems.
U.S. Virgin Islands
In 1988, a proposal to develop a locally supported CIS
was being discussed in detail in the USVI (4). This
project resulted in a formal application from the govern-
ment of the USVI for financial assistance from technical
assistance funds provided by the Office of Territorial
and Insular Affairs of the U.S. Department of the Inte-
rior. The grant was awarded in March 1991. The pro-
posed project combined existing information from
several USVI government agencies to produce CIS
overlays, as shown in Table 1 (5).
In addition, according to the grant application, (which
was written by the Virgin Islands government and may
not have represented U.S. Geological Survey's
[USGS's] intentions) the National Mapping Division
(NMD) of the USGS agreed to digitize the eight USGS
1:24,000-scale quad sheets ("quadrangles") that cover
the USVI, including the following categories:
Roads and trails
- Power transmission lines
Hydrography
- Stream networks
- Shorelines
- Wetlands
- Mangroves
- Reefs
Table 1. GIS Overlays
Agency
GIS Overlays
DPNR
WAPA
VITEMA and emergency services
DPW
Zoning
Flood plain
Subdivision
Water distribution
Aquifer profiles
Electrical distribution
Critical routes
Critical facilities
Sewer line network
Transportation
Flood plain
DPNR = Department of Planning and Natural Resources
WAPA = [Virgin Islands] Water and Power Authority
VITEMA = Virgin Islands Emergency Management Agency
DPW = Department of Public Works
314
-------
Topographic contours
Political boundaries
(These categories were subsequently adjusted based
on discussion between the USGS, WAPA, and govern-
ment agencies. Documentation for these new cover-
ages is not available, but the general idea holds,
although no VITEMAdata were used, and the USGS did
not digitize the WAPA power system.)
The underlying notion was that the whole would be
bigger than the partseach organization would bring its
own corporate data and maps so all could share in the
digitized product.
Five objectives were identified for the project:
Provide the USGS in St. Thomas with a complete
microcomputer (CIS) workstation.
Develop a digitized
graphic maps.
database from USGS topo-
Contract for the digitizing.
Acquire digital data to load into CIS.
Enter data not in digital format.
The proposal required a 50-percent local cost contribu-
tion to the project, including a matching suite of hard-
ware for data maintenance, backup, and analysis. The
application stated that itemized costs of $305,000 "will
be provided by the Virgin Islands Water and Power
Authority (WAPA) and the Office of Territorial and Inter-
national Affairs."
The project application is ambiguous about the nature
of the hardware and operating systems that the CIS
requires. A list of hardware for USGS use refers to a
"microcomputer workstation." Later references to "ESRI
ARC/INFO workstation software," however, indicate to
sophisticated users that these applications require UNIX
workstations and UNIX language operating systems. In
general, UNIX is not used or supported in the Virgin
Islands. Some Virgin Island officials associated with the
CIS project feel they were not fully informed about the
CIS operating environment and the long-term support
costs to which they were committing.
A frustration for USGS in the project stems from the
somewhat limited role the agency has had in providing
high-quality data conversion services (digitizing). One
USGS staff member explained that unfortunately the
agency's mandate only extends to providing data; the
agency "can't get involved in applications."
The initial project proposal also was unclear about owner-
ship of the digitized data. The proposal spoke in general
terms: "All available digital data and attributes will be com-
plete, accurate, and up-to-date at the end of the project,
and will be available for use and transfer to the Depart-
ment of Planning and Natural Resources (DPNR) (5)."
Given the conditions and environmental constraints dis-
cussed, a number of specific differences developed be-
tween the CIS implementation processes of the BVI and
the USVI.
Initial System Planning Activities
British Virgin Islands
Although the BVI had no formal systems plan, consult-
ants worked with the Town and Country Planning De-
partment on four different occasions to provide insight
into some aspect of CIS applications. Sometimes, the
benefits from the consultations were neither the type
nor the quality the department originally expected. In
general, however, each provided some additional per-
spective on the possible benefits and perils of imple-
menting a CIS.
U.S. Virgin Islands
The USVI apparently made few initial system planning
efforts, although the U.S. Department of the Interior and
USGS have wide experience in the territory. (An in-
formed source claims a grant was made, possible by
EPA, for a preceding $50,000 CIS project, but no men-
tion of this has been found in the materials available for
this article, either as a proposal, or in terms of specific
products.) Possibly, this very familiarity led to a series of
unexamined assumptions and diminished communica-
tions about the exact terms of the assistance and serv-
ices that the USGS exchanged with several agencies of
the Virgin Island government.
One indicator of the lack of system planning activities in
the Virgin Islands is a proposal that the Virgin Islands
Emergency Management Agency (VITEMA) circulated
fora "Geographical Information Systems: Technical Op-
erators Meeting." This proposal, from an agency that
has always been one of the most important participants
in CIS activities, called for a "technical working group"
to examine existing database management systems in
the territory to develop a planning strategy for imple-
menting CIS.1 This proposal was dated September
28, 1992, 10 days before the USGS announced a
demonstration of the completed products of Phase I of
the "comprehensive geographic information system be-
ing developed for the USVI (6)."
1 Ward, R.G. 1992. Geographical information systems: Technical op-
erators meeting. Memorandum to Cyrille Singleton. VITEMA, St.
Thomas, U.S. Virgin Islands (September).
315
-------
Software Selected and Rationale
British Virgin Islands
In part because of the extended timeframe for the BVI
planning and initiation of the CIS system, the Town and
Country Planning Department never committed to a
specific system configuration until the last stages of the
planning process.
This process of "creative procrastination" had three syn-
ergistic results:
PC power increased (and prices decreased) to the
point where reasonably priced systems could perform
many of the compute- and data-intensive operations
demanded for graphics software mapping.
ARC/INFO released the ARC/CAD version of its CIS
software, which worked on PCs within the well-known
AutoCAD drafting software. Architect and engineering
offices in the BVI already used computer-aided design
(CAD) software, so upgrading to include CIS func-
tionality was relatively easy.
The fourth CIS consultant to work in the BVI was
experienced in implementing systems in the Carib-
bean and had special knowledge and access to early
versions of both ARC/CAD and Version 1.0 of
ARC/VIEW. These two tools, based on a last minute
proposal by International Development Advisory
Services (IDAS) of Miami, Florida, a private CIS sup-
port contractor, became the basis for the BVI CIS.
U.S. Virgin Islands
Workstation ARC/INFO 6.1 was selected as the basic
software for the USVI CIS project because it is the
USGS standard software. In the environment surround-
ing the USVI project, this seemed to be a sufficient
explanation, although there may have been other rea-
sons. Because of this decision, however, WAPAand the
DPNR lack any means of updating map or attribute data
files. Outside providers, such as the USGS's NMD, must
perform that service. The USGS has noted that the only
reason WAPA and DPNR lack these capabilities is be-
cause they (WAPA and DPNR) failed to provide the
matching suite of hardware and software specified in the
grant application.
The digitized water supply system offers an example of
the extra costs that such a condition creates. The USGS
built an ARC/INFO coverage by converting mapped data
from AutoCAD source files, which they then linked to
detailed attribute information about each element (e.g.,
pipe, valve, elbow) in the system. The USGS then used
ARC/INFO network software, purchased with project
funds, to build a network model that analyzes and dis-
plays the operation of the entire water distribution sys-
tembut no software in the Virgin Islands can run the
network model.
Hardware Platforms
British Virgin Islands
The BVI CIS was originally installed on a Compaq 486-
50, with a 21-inch screen. The Town and Country Plan-
ning Department soon learned, however, that data input
would be more efficient if two or three smaller machines
split the work, with the Compaq available for analysis
and data quality checking. The department upgraded its
existing office computers to handle the data entry. Users
already feel the need for networked applications to
share data more quickly. Plans for CIS expansion to
other offices, such as the Electricity Corporation, in-
crease the pressure for an extended local area network.
U.S. Virgin Islands
The system that the USGS used to build the USVI CIS
database was a Data General UNIX workstation with
one large digitizing tablet and one pen plotter. No match-
ing or comparable hardware are installed anywhere in
the USVI, as the original project proposal had foreseen.
Observers tend to agree, however, that the failure to
provide a specific hardware configuration is less signifi-
cant than the lack of committed, senior, full-time techni-
cal staff. This staff is required to operate the level of CIS
facility that the USGS envisioned.
Base Map Priorities and Layers
Constructed
British Virgin Islands
Building a map database is proving to be a long process
for the Town and Country Planning Department. This is
complicated by the failure of a key digitizing contractor
in Texas to provide property lines in a format conducive
to constructing accurate property polygons. Operators
in the Town and Country Planning Department have
increased their data entry efficiencies, however, and
most properties on the most densely inhabited islands
have now been digitized.
Producing demonstration data displays accounts for a
significant part of the cost of developing databases for
the early phases of the BVI CIS implementation. These
demonstrations aim to illustrate possible new applica-
tion areas for other agencies and departments of the BVI
that are interested in cooperating and sharing costs of
additional system development. For example, the Elec-
tricity Corporation and the National Disaster Prepared-
ness Agency need to map emergency services.
Converting the data (i.e., digitizing) in house in the BVI
has produced costs and benefits. The costs revolve
316
-------
around the steep learning curve for data entry proce-
dures and the constant distractions of responding
quickly to "outsiders" who may be important long-term
supporters of the CIS. The benefits include increasing
staff skills and the ability to build constituencies for the
program by promptly responding to real needs.
Coverage priorities for the BVI CIS include a national
addressing system, completion of the territorial land
use mapping, and accurate cadastral mapping (which
has major environmental planning and management
implications).
U.S. Virgin Islands
The USGS produced 44 coverages for the Virgin Islands
from a variety of sources. Table 2 shows the major
coverages and scales, by island.
Table 2. Major Coverages and Scales for the USVI
St. Croix St. John St. Thomas
STC water
distribution
1 :2,400
STC roads
1 :2,400
STC building
footprints
1 :2,400
STC shorelines
1 :2,400
STC DIG
boundaries
1 :24,000
STC DIG
roads
1 :24,000
STC DIG
hydrography
1 :24,000
STC DIG
hypsography
1 :24,000
STJ water
distribution
1 :2,400
STJ roads
1 :2,400
STJ building
footprints
1 :2,400
STJ shorelines
1 :2,400
STJ DIG boundaries
1 :24,000
STC DIG roads
1 :24,000
STJ DIG
hydrography
1 :24,000
STJ DIG
hypsography
1 :24,000
STT water distribution
1:2,400
STT roads
1:2,400
STT building
footprints
1:2,400
STT shorelines
1:2,400
STT DIG boundaries
1 :24,000
STT DIG roads
1 :24,000
STT DIG
hydrography
1 :24,000
STT DIG
hypsography
1 :24,000
In addition, the following National Park Service data
were converted and added to the data set but were not
produced by the USGS:
STJ NPS boundaries, 1:24,000
STJ NPS roads, 1:24,000
STJ NPS hydrography, 1:24,000
STJ NPS hypsography, 1:24,000
STJ NPS benthic communities, 1:24,000
STJ NPS historical sites, 1:24,000
STJ NPS vegetative cover, 1:24,000
Mapping for St. Thomas, St. John, and St. Croix identi-
fied a total of 10 coverages for each island. They are
based on information from the USGS ("quad sheets"
specifically demarking political boundaries, shorelines
and streams, topography, and roads) and higher preci-
sion WAPA mapping, which derives from 1986 aerial
photogrammetry, including left and right road bounda-
ries, building footprints, shorelines, and water supply
system data. The WAPA data are at 1:2,400 scale, an
order of magnitude more precise than the USGS base
map. St. John mapping consists of 10 added layers
based on data that the Virgin Islands National Park
(VINP) provided.
The original USVI project proposal referred to a two-phase
process of database development, shown below (5):
Phase
I. Base system
development
Agency extensions
(i.e., by USVI
agencies)
Tasks
Water distribution network
Power distribution network
Flood plain maps
Land use maps
Transportation networks
Emergency facility networks
Tax parcel/land value
The USGS announced that Phase I was completed in
October 1992 (6). Supposedly, the contents of these two
phases were subsequently adjusted to reflect a different
range of coverages, but the notion of a "Phase II" in
which local agencies would assume more operating
responsibilities was retained.
Ownership of, access to, and terms that govern the use
of this digital data are confused. The USGS says it is
unable to provide an authoritative catalog of the cover-
ages because "one has not been produced." WAPAsays
it has several diskettes of data in the safe but no equip-
ment to manipulate them. The VINP has learned that it
can use its own data as well as WAPA data converted
and attributed by the USGS, but the park does not
possess or use USGS digital line graph (DIG) data. To
personnel in USVI agencies, USGS statements have
clouded the question of access to the CIS information.
For example, one such statement announced that the
USGS Water Management Division cannot make the
digitized data available to Virgin Island government
agencies.
The DPNR apparently has no means of making direct
use of the digital data. First, DPNR has no hardware or
software that can use the data. Secondly, it has no
operators who can build the CIS systems to actually
apply the data to decision-making needs. The depart-
ment is said to be preparing a new CIS proposal for
training, hardware, and software for a new CIS system.
According to unconfirmed rumors, this system will be
based on MapGrafix, a Macintosh mapping system.
317
-------
In operational terms, the data seem to belong to WAPA,
which contributed major financial support and map re-
sources to the project. WAPA has been helpful in pro-
viding copies of the digital data to other groups and
agencies.
Environmental Problem-Solving in Local
Decision-Making
British Virgin Islands
In the BVI, the first priority of the CIS facilities is to
extend the National Addressing System and to improve
the property ownership system. This will both improve
postal services, as originally proposed, and provide bet-
ter information for important revenue and financial
analyses. Land use mapping and environmental impact
assessments are important second priorities for CIS
applications. Other features already developed for in-
terim studies and analyses include mapping of signifi-
cant coastal and wildlife features and environmentally
sensitive areas from the Anegada Development Plan
and mapping of important submarine habitats adjacent
to Virgin Gorda.
The BVI have concentrated on developing CIS applica-
tions to address strategically important issues in the
territory. Marine and coastal resources are vitally impor-
tant to the BVI economy. They embody historical and
cultural values, as well as maintain a high-quality envi-
ronment to support charter yacht-based tourism, which
is integral to the BVI economy. The Conservation and
Fisheries Department is working with the Town and
Country Planning Department to convert the country's
coastal atlas to digital form (see section on Principal
Users), as has been done on a demonstration basis for
the Anegada and Virgin Gorda mapping.
U.S. Virgin Islands
In the USVI, environmental decision-making generally
follows an adversarial, rather than a problem-solving
format. A combination of historical and cultural factors
have created the general assumption that the develop-
ment process creates winners and losers. In this envi-
ronment, information becomes an important tactical
weapon, making it difficult to gather support for activities
or programs that aim to make information more widely
accessible. Technology is more acceptable, and more
likely to receive leadership support, if it is justified on
technical, less "political" terms.
The USGS team made a presentation on the Virgin
Islands CIS in October 1992, after just completing the
digital coverages for Phase I of the CIS project. The
presentation emphasized that the CIS is intended to
provide decision-makers with easily accessible spatial
information (6).
Originally, the CIS was expected to benefit primarily the
territory's three coastal zone commissions in their as-
sessment of environmental effects of major develop-
ment proposals. The ground-water protection program
of the Division of Environmental Protection of the DPNR
is using CIS analyses produced by the Water Resources
Division of the USGS, employing the USVI CIS cover-
ages with added data (e.g., wells) that the Water Re-
sources Division is digitizing.
GIS Support Factors
According to the GIS support contractor for the BVI, a
successful GIS requires three key support elements:
GIS policy leadership
GIS technical leadership
Competent outside expert assistance
The following summarizes the comparative experience
of the two programs for these three key implementation
support factors:
Support
Factors
GIS
policy
BVI
Town and
Country Planning
Department
director led
project from chief
minister's office
GIS BVI technician
technician trained in the
United States
USVI
No GIS manager
in government
Outside
support
UNDP and IDAS
No GIS specialist
in government
(draftsman at
WAPA)
USGS technical
support and U.S.
Department of
the Interior
financial support
Principal Users: Planned and Actual
British Virgin Islands
The BVI Department of Finance is the first user of data
products from the GIS, based on the initial funding for
the addressing system. This system is based on detailed
parcel maps of the BVI so that the effective base map
resolution of the BVI system is 1:2,500. This is consid-
erably finer than the 1:24,000 scale of the USVI maps.
318
-------
The Town and Country Planning Department, however,
is working to recruit other users to the system, including:
Public Works.
Water and Sewerage Department for a systemwide
map (which may eventually spin off as a separate
system, given this department's long-term interest in
engineering-quality facilities management information).
Conservation and Fisheries.
British Overseas Development Administration, which
funded a coastal atlas for the BVI (7). At the sugges-
tion of the Town and Country Planning Department,
this mapping was developed in ARC/INFO. A pro-
posal has been made to convert the coastal atlas to
digital form for natural resource management appli-
cations, with a demonstration already developed
showing the distribution of sensitive marine commu-
nities around Virgin Gorda.
In addition to these uses, the CIS group is starting to
experiment with the use of remote sensing products in
CIS production, which would encourage the use of CIS
for natural resource change detection.
The Town and Country Planning Department's CIS spe-
cialist, Mikey Farara, is being reassigned to provide
networking support (including CIS distribution over the
network) for several government agencies. Meanwhile,
the CIS operation is adding a cartographer to assist in
tailoring CIS products to users' needs.
As enthusiasm for the CIS has blossomed in the BVI,
managing for realistic expectations and stressing the
investment costs that participating agencies can expect
have become problems for the Town and Country Plan-
ning Department.
U.S. Virgin Islands
Complex evaluation issues face the USVI's three
coastal zone commissions (one on each island).
Therefore, land use planning in general and coastal
zone permitting specifically were assumed to be im-
portant first users of the CIS. The DNPR, however,
had no process to prepare the Division of Comprehen-
sive and Coastal Zone Planning to implement this
system. In addition, the scale of permitting decisions
may be too fine forthe CIS base map. (See discussion
of scale below.)
WAPA is not using the CIS data. One senior manager
characterized their experience with the CIS project as
"paying a lot of money for a diskette of data that we keep
locked in the safe."
The VINP (part of the U.S. National Park Service) and
Biosphere Reserve have purchased a PC-based CIS
system and employed an analyst to implement it for the
park and adjacent areas on St. John and the surround-
ing seas. With this system, they plan to enter the USGS-
developed data into the database. In addition, the Virgin
Islands Resource Management Cooperative (a collabo-
ration of research and resource management organiza-
tions) makes the VINP CIS data and analytical capabilities
available to members, including government members.
At this time, the only major Virgin Islands government
user of the CIS is the ground-water protection group of
the DPNR's Division of Environmental Protection. Be-
cause they lack equipment or software to manipulate the
CIS data already available, they use the Water Re-
sources Division of USGS as a CIS contractor. This
arrangement has two problems:
High costs: Although the USGS "owns" the existing
digital data, and processing private contracts would
be complex, DPNR believes it could get similar serv-
ices at cheaper prices from other vendors.
Inappropriate scale: Environmental management
processes in the Virgin Islands (and in most other
small island states) require knowledge of property
ownership, implying maximum map scales of
1:5,000 to 1:10,000. The USGS quad sheet scale
of 1:24,000 is too coarse for many management
purposes. Costs of remapping areas of concern at
the higher resolution are high, and the problems of
maintaining multiple map resolutions and sources
are not trivial.
What GIS Can Do
Joseph Berry has proposed seven basic categories of
"What GIS Can Do for You" (8). These applications can
be related to the GIS products and proposals forthe BVI
and USVI, with special attention given to natural re-
source and environmental issues. Table 3 shows what
coverages that have been or are being developed for
the two systems can do.
Table 3 illustrates two contrasting issues separating the
two jurisdictions. The USVI have the data available to
perform a number of relatively complex analytical proc-
esses, especially in St. John. They have no capability to
actually execute any such studies, however. The BVI,
on the other hand, have proposed and often developed
pilot or demonstration applications for several GIS uses
but still need to develop the data resources to support
these on a territorywide basis.
Lessons Learned
The comparative experience of these two very distinct
GIS programs reinforces three basic lessons of informa-
tion system design and implementation:
Plan, don't assume: The prolonged, sometimes repe-
titious, planning process that evolved in the BVI
319
-------
Table 3. Coverage Capabilities
USVI
BVI
Questions:
Can you map it?
Where is what?
Where has it changed?
What relationships exist?
Where is it best?
What affects what?
What if...?
Analytical
Function
Mapping
Natural resource
management
Temporal
Spatial
Suitability
System
Simulation
Application
USGS and WAPA-based
coverages
DIG hydrography
DIG hypsography
Well inventory
STJ national park
coverages
DIG and WAPA
Boundaries and roads
1982 to 1989
STJ national park
coverages (limited
application)
STJ national park
coverages (limited
application)
None discussed
Status Application
Done Land use cadastral
Done Coastal atlas
Sensitive areas
Done Significant features
Population data
Done Land use
Done Sensitive areas
Done Land use updates population
Land use
Population data
Coastal atlas
Done Land use
Coastal atlas
Sensitive areas
Significant features
Done Land use
Coastal atlas
Sensitive areas
Significant features
Population data
Speculation, but no plans to
implement yet
Status
Done and
proposed
Proposed
and partial
Major
islands done
Proposed
Proposed
Partial
Proposed
Partial
Partial
Partial and
proposed
involved multiple consultants providing often conflict-
ing advice. This process served to educate policy-mak-
ers and managers in a much broader range of
possibilities and avoidable problems than were avail-
able to the USVI. A corollary to the need for careful
planning is the need to avoid making decisions or
commitments to specific systems before such deci-
sions are absolutely necessary. Especially in systems
involving high technology, premature decisions often
mean early obsolescence.
Implement in phases with early demonstration
products: Some issues, such as cadastral mapping
and scale, are so subtle to inexperienced users that
they need practice in real-life situations. If the USGS
had spotted the scale problems at an early stage in
the data conversion process, the USGS may have
been able to provide a better solution. Some USVI
critics claim the "1:24,000one size fits all" attitude
characterizes the federal approach.
Identify critical success factors for each situation: In
some environments (e.g., USVI), CIS is most attrac-
tive for its ability to provide enhanced powers of
analysis. In others, such as the BVI, it is seen as a
data integration tool and as a way to better inform
political leadership and the public. To ensure suc-
cess, major CIS implementations also need to meet
the three major support requirements:
- A political/senior management "chief
- A technical "chief
- Competent outside technical assistance
Finally, implementers should recognize that they have a
stake in open information sharing. They should seek
ways to redefine the decision-making process as a
nonzero sum game: more information should benefit all
parties. Of course, such changed attitudes require fun-
damental value shifts that take a long time to achieve
and may have high short-term costs.
References
1. McElroy, J.L. 1991. The stages of tourist development in small
Caribbean and Pacific islands. In: Proceedings of the International
Symposium on the Island Economies: Policy Models for Planning
Development, Lesbos, Greece (November).
2. Bureau of Economic Research, Virgin Island Department of
Commerce. 1994. Economic indicators. St. Thomas, U.S. Virgin
Islands (July).
320
-------
3. Potter, L. 1984. Program notes from the Caribbean Conference of
Planners, Kingston, Jamaica.
4. Potter, B., K. Green, and M. Goodwin. 1988. Management of natu-
ral resource information for the Virgin Islands National Park and
Biosphere Reserve: Special biosphere reserve report. St. Thomas,
U.S. Virgin Islands: Island Resources Foundation.
5. Government of the U.S. Virgin Islands. 1991. Application for tech-
nical assistance funds: Virgin Islands GIS system. Office of Terri-
torial and International Affairs, U.S. Department of the Interior
(April).
6. Parks, J.E. 1992. Invitation to a demonstration of the Virgin Islands
GIS. U.S. Geological Survey, St. Thomas, U.S. Virgin Islands (Oc-
tober).
7. University of Manchester, British Overseas Development Admini-
stration. 1993. British Virgin Islands coastal atlas. Project was
collaboration between graduate students at Manchester University,
the British government, and the government of the British Virgin
Islands.
8. Berry, J.K. 1994. What GIS can do for you. GIS World. Boulder,
CO (May).
321
-------
Application ofGIS for Environmental Impact Analysis in a Traffic Relief Study
Bruce Stauffer
Advanced Technology Solutions, Inc., Lancaster, Pennsylvania
Xinhao Wang
University of Cincinnati, Cincinnati, Ohio
Abstract
This paper presents an application of a geographic infor-
mation system (CIS) in a traffic relief study. Traffic conges-
tion has severely affected the environmental quality and
the quality of life for residents in the study area. A team of
planners, environmental specialists, historians, landscape
architects, traffic engineers, and CIS professionals organ-
ized to solve the problem. The team has evaluated the
environmental and socioeconomic impacts of highway
alignments from the very first step through every major
decision for the duration of the project.
The CIS professionals have played a crucial role in
maintaining constant and active interactions among
members of the project team, federal and state agen-
cies, and the public. CIS has helped to develop a natural
and cultural resource inventory, identify contamination
sources, assess environmental constraints, and evalu-
ate proposed highway alignment alternatives. CIS pro-
vides an ideal atmosphere for professionals to analyze
data, apply models, and make the best decisions. The
high-quality map products that CIS creates enhance the
quality of public presentations and reports. The authors
feel that, as this project has progressed, more people
have realized the benefit of using CIS.
Introduction
A traffic relief study, as one type of transportation
project, aims to resolve traffic congestion problems
through a combined strategy of upgrading existing
infrastructure, building new infrastructure, and con-
trolling traffic demand using congestion management
strategies (CMS). This type of study proceeds through
at least the following steps:
Problem identification
Data collection
Preliminary design
Environmental impact analysis
Final design
Construction
The process heavily involves federal, state, and local
government agencies, as well as the public. The goal of
the project is to develop an environmentally sound so-
lution to the traffic congestion, which also happens to
promote economic development and improve quality of
life for people in the local area and the region. Environ-
mental, social, and economic issues must be equally
addressed from the very first step through the final
design. Federal and state regulations generally require
an environmental impact statement (EIS) when con-
structing new infrastructure or upgrading existing road
systems. Preparing an EIS is a requirement for such a
project and demands a significant commitment of time,
money, staff, and technical resources.
A geographic information system (CIS) has the ability to
process spatially referenced data for particular pur-
poses. Along with the development of computer hard-
ware and software, CIS has progressed from pure
geoprocessing, to management of geographic informa-
tion, to decision support (1). This paper presents the
application of CIS in an ongoing traffic relief study in
Marshalls Creek, Pennsylvania. The CIS function in this
study has had various purposes:
Inventory data compilation
Spatial data analysis
Map production
Traffic modeling support
Public presentations
322
-------
The study shows that CIS can play an important and
innovative role in transportation studies.
Project Description
The study area is in the Pocono region, located in
northeastern Pennsylvania (see Figure 1). The Pocono
Mountains and Delaware Water Gap National Recrea-
tion Area possess a wealth of natural and cultural re-
sources. The area is famous for providing year-round
vacation activities. Attractions include fishing, canoeing,
and Whitewater rafting in the summer and downhill and
cross-country skiing, snowmobiling, snowboarding, and
ice fishing in the winter. The area includes quiet wood-
land trails past a rushing waterfall and scenic settings
for camping. In the fall, the area is ablaze with the
brilliant colors of foliage. Various scenic sights, recrea-
tional sites, and national historic sites make the area
ideal for attracting people to come for a day, a weekend,
or a longer vacation.
Although tourism brings people to the Pocono area and
promotes economic growth, it also brings a traffic con-
gestion problem to the community. In addition, the influx
of new home owners from New Jersey and New York
adds the problem of commuter traffic to the area. The
most troublesome section is in the vicinity of Marshalls
Creek where U.S. Route 209 intersects with Pennsylva-
nia (PA) Route 402. Two intersections are only about
500 feet apart. The traffic tieups can extend up to
3.5 miles on northbound Route 209, all the way back to
Interstate Highway I-80. Emergency response times on
U.S. Route 209 can be up to 20 minutes during peak
traffic. The heavy traffic volume results in traffic acci-
dents exceeding state averages on secondary roads as
motorists seek alternative routes to avoid congestion.
Through traffic traveling to and from New England using
U.S. Route 209 as a connector between I-80 and I-84
makes the problem even worse. The year-round outdoor
activities perpetuate the constant traffic problems that
have severely affected the quality of life for residents in
and around the Marshalls Creek area.
In response to the problems, the Pennsylvania Depart-
ment of Transportation (PennDOT) selected a project
team in February 1993 to conduct a traffic relief study in
the Marshalls Creek area. The project team consists of
individuals from seven firms and represents a wealth of
experience in the variety of disciplines necessary to
successfully complete this project. The team members
include land use and traffic planners, biologists, histori-
ans, traffic and environmental engineers, surveyors, and
CIS and global positioning system (GPS) professionals.
In addition to PennDOT, the funding agency, several
federal and state regulatory agencies periodically review
the development of the EIS to ensure that it meets
regulations. These agencies are the Federal Highway
Administration (FHWA), the U.S. Environmental Protec-
tion Agency (EPA), the U.S. Army Corps of Engineers,
the Pennsylvania Department of Environmental Re-
sources (DER), the Pennsylvania Historic Museum
Commission, and the Pennsylvania Fish and Game
Commission. Local planning commissions and citizen
representatives also actively participate in advisory ca-
pacities. A series of agency coordination meetings, pub-
lic meetings, and public information newsletters
coordinates the activities of all participants over the
course of the project.
Primary and Secondary Highways
Light Duty Roads
Phase I Project Boundary-
Streams, Ponds, and Reservoirs
5,000
Feet
10,000
Figure 1. Phase I project area.
323
-------
After collecting traffic data and performing traffic de-
mand modeling, the team realized that adopting strate-
gies to control traffic demand and upgrading existing
roads through widening and intersection improvements
would not suffice to meet demand projections for the
design year 2015. Consequently, the team has deter-
mined a new road is needed to alleviate congestion in
Marshalls Creek.
The study aims to identify alternatives to relieve traffic
congestion along U.S. Route 209, PA Route 402, and
Creek Road, and eliminate backups onto I-80 from U.S.
Route 209. The alternatives should also improve air
quality by reducing fuel consumption and vehicle emis-
sions and facilitate travel through Marshalls Creek for
local and through traffic. The improvements must com-
ply with federal and state regulations. The study team
must consider county and local government goals and
objectives so that the traffic capacity improvements will
be compatible with planned local development.
The project is being conducted in two phases. Phase I,
which is complete, was an investigation broad in scope.
It used inventory of secondary data to describe the
environmental characteristics of the area. A traffic de-
mand model identified the area for detailed study after
a preliminary analysis of a wide range of baseline data.
In Phase II, the team analyzes both primary and secon-
dary data and delineates alternative alignments or trans-
portation upgrading options that meet the need and
minimize impacts. Analysis of environmental and engi-
neering factors assists both in the determination of the
most practical alternative and in preparation of the final
EIS.
The nature of the study requires the analysis of a variety
of data at different scales by different professionals.
Through field investigation, the project team also con-
stantly updates and adds new data to the existing data-
base. The new data may be attribute data about some
geographic features or may be locational data. The
project team has found CIS to be an appropriate tool to
meet the challenge of better conducting the study.
The team has used CIS extensively in both Phase I and
Phase II studies. The two phases vary in data require-
ments, scales, and purposes of spatial analysis. With
the support of CIS, the team has been able to quickly
assemble data at adequate scales and present data in
formats that are familiar to different professionals. GIS's
data manipulation power distinguishes the different re-
quirements of the two phases and at the same time,
clearly depicts the linkage between the two phases. The
following sections describe CIS applications that have
helped facilitate the study and coordinate project team
members, public agencies, and citizens.
GIS Application
CIS contains powerful tools to process spatially refer-
enced data. These processes and their results are
meaningless, however, without a clearly defined objec-
tive. Many professionals point out the importance of
focusing GIS on practical problems. Using GIS is not an
end; it is a means to represent the real world in both
spatial and temporal dimensions. The benefits of using
GIS can be summarized in three aspects:
GIS helps to portray characteristics of the earth and
monitor changes of the environment in space and
time (2).
GIS helps us to more deeply understand the meaning
of spatial information and how that information can
more faithfully reflect the true nature of spatially dis-
tributed processes (3).
GIS helps us to model alternatives of actions and
processes operating in the environment (2), to antici-
pate possible results of planning decisions (4), and
to make better decisions.
This project demonstrates the advantages of applying
GIS to solve practical problems from the above three
aspects. An EIS requires extensive data about natural
resources, land uses, infrastructure, and distribution of
many interrelated socioeconomic factors. The accuracy
and availability of required data depend on the scope of
a study and the size of the study area. Our study shows
that GIS, with its data retrieval, analysis, and reporting
abilities, significantly improves the analysis. GIS helps
to collect data at various scales, store data, and present
data in forms that allow the project team to carry out the
study in an innovative way.
Phase I Study
Phase I of the traffic relief study was completed in 1993.
The goal of Phase I was to acquire understanding of the
general features of the area and to use a traffic demand
model for delineating an area for detailed study. The
study area is approximately 52 square miles. To provide
data for the preliminary analysis and the traffic demand
modeling, the team developed baseline data inventory
with GIS. Data were primarily secondary data that came
from several different sources in different formats. For
example, the U.S. Census Bureau 1990 population data
were in TIGER format, the U.S. Fish and Wildlife Service
National Wildlife Inventory (NWI) files in digital line
graph (DIG) format, the U.S. Soil Conservation Service
Monroe County Soils in DIG format, and the U.S. EPA
Monroe County Natural Areas in ARC/INFO format. The
majority of data sources were at scales between 1:15,000
and 1:24,000. With GIS tools, the team integrated these
baseline data into a common presentation scale and pro-
jection. This process ensured an effective and comprehen-
sive spatial analysis in the study area.
324
-------
The team arranged and stored data in layers according
to themes. Examples of the data layers included:
Road center lines.
Twenty-foot elevation contours.
Utility lines.
Water features.
Subdivision boundaries.
1990 Census population by Census Tracts and Blocks.
Flood plains.
Geological formations.
Public facilities, including schools, churches, and
cemeteries.
Political boundaries.
Hazardous waste locations.
Potential archaeological areas.
With these data layers, the team generated a series
of 17 thematic maps to describe the features of the
study areas. All maps were plotted on E-sized papers
(48 inches x 36 inches) with the same map layout. The
general reference map served as a base map for the
other themes. It included several data layers to provide
geographic references to the study area.
In addition to the base map, each individual theme map
showed only one theme at a time, such as soils, subdi-
visions, and wetlands. Some theme maps showed de-
rived data from the original data layers. For example, in
developing a slope theme map, the team first built a
three-dimensional surface from the 20-foot elevation
contours, then calculated slope in degrees and aggre-
gated areas based on a 10-degree interval. The slope
theme map showed the result from the data processing.
In addition, the project team created summary statistics
tables to help team members gain knowledge about the
study area. Table 1 is an example of the summary
statistics for land use categories.
Table 1. Phase I Statistical Summaries for Land Use
Categories
Land Use
Acres
Urban
Agricultural
Rangeland
Woodland
Water
Wetland
Transitional
10,628
1,183
12
20,957
1,394
1,428
422
During preparation of the Phase I inventory data and
summary statistics, traffic planners performed traffic de-
mand modeling to determine new road connections that
would provide a minimum acceptable level of service in
the year 2015. The modeling result was loaded into the
CIS and converted into the same format and projection
as other inventory data. Figure 2 displays the boundary
that the traffic demand model delineated and the actual
Phase II boundary. The two boundaries were not the
Primary and Secondary Highways
Light Duty Roads
Unimproved Roads
Trails
Phase II Project Boundary
Streams and Ponds
Minimum Acceptable Service Level
2,000
Feet
4,000
Figure 2. Phase II project area.
325
-------
same. The thick line enclosed a Phase II study area that
was delineated based on the traffic demand modeling
and the team's understanding of environmental and
other factors in the project area.
Phase II Study
The Phase II study is still ongoing. The area for the
Phase II study is much smaller than that for Phase I. It
is about 3.2 miles by 2.5 miles, or approximately
4 square miles.
The objective of the Phase II study is to conduct a detailed
analysis for delineating a full range of feasible highway
alignment alternatives. The alternatives must meet the
needs of relieving traffic congestion and minimizing its
impact on environmental and cultural resources.
Because the accuracy of the Phase I data was not
sufficient for the Phase II study, the project team has
collected data using different approaches to develop a
similar set of baseline data at a finer scale. The major
data source has been the photogrammetry data pro-
vided by PennDOT at a 1:2,400 scale. The data include:
Road cartways
Five-foot elevation contours
Utility lines
Water features
Buildings footprints
Bridges
The team directly digitized tax parcel boundaries from
Monroe County Tax Assessor's maps that range in
scales from 1:1,200 to 1:4,800. After digitizing each map
sheet separately, the team merged them together to
create a continuous parcel layer. The data layer has
been adjusted to fit with the PennDOT photogrammetry
data although the two data sets do not seem to match
exactly. In addition, digital orthophotographs at 5-foot
pixel resolution were also obtained for the project.
From these baseline data, the team has constructed
Phase II data layers in four different ways.
The first approach digitizes from compilations on project
base maps. For example, the team creates a land use
data layer from the digital orthophotographs and infrared
photography. The CIS group first plots the digital ortho-
photographs on a set of 1:2,400-scale map sheets.
Road cartway and water features are plotted on top of
the orthophotographs. Then land use specialists deline-
ate land use boundaries with fine color markers and
code land uses on maps according to the Anderson land
use classification. In the end, the CIS group digitizes the
land use boundaries from the compilations to create a
land use data layer. Similarly, the 100-year flood plain
data layer is delineated from compilations on project
base maps with 5-foot contours and digitized.
The second approach derives new data layers from
existing data. Buildings and structures are plotted at a
1:2,400 scale. Both historians and environmental engi-
neers use the plots in their field investigations. After
historians identify historic-eligible buildings on the plots,
the CIS group develops an attribute data file that links
the historic inventory data to the building geometry.
Similarly, field investigations identify buildings and struc-
tures associated with contamination sites. The system
stores types of contamination as building attributes. By
overlaying the building data layer with the tax parcel
data layer, the team can identify properties on which
historic buildings or contamination sites are located.
The third approach constructs data layers by refer-
encing Phase I data. For example, Phase II subdivision
boundaries are derived from the digitized tax parcel
boundaries by referring to the Phase I subdivision
boundaries. Phase I subdivisions were manually com-
piled at 1:24,000 using approximate location, which did
not align very well with the more accurate tax parcel
boundaries. Using a 1:7,200-scale plot that shows both
Phase I subdivision boundaries and the Phase II tax
parcel boundaries, planners can verify and indicate
properties associated with each subdivision. These
properties are dissolved to create new boundaries for
subdivisions that precisely fit with tax parcels. The same
approach is used to refine public parks and private
recreation areas.
The fourth approach obtains spatial data with GPS. The
GPS surveyors collect accurate locational data about
key features, such as boundaries of wetlands, site loca-
tions for hazardous waste, and locations of archaeologi-
cal field samples. GPS data also supplement existing
data, such as delineating footprints of new buildings to
update the PennDOT baseline data. The integration of
CIS and GPS provides the project team with accurate
and up-to-date data.
Phase II data layers have provided much richer informa-
tion for a detailed study of environmental features. They
are merged in many different combinations to show the
spatial distributions of different factors from different
perspectives. Table 2 lists some of the map themes
created for the Phase II study. All maps are plotted at a
1:7,200 scale.
In addition to using CIS as a data library and map
production tool, we use CIS to support decision-mak-
ing in two ways. First, the creation of a composite data
layer has revealed the impact of alternative align-
ments on several composite constraints. The compos-
ite data layer is an overlay of several inventory layers
and shows the various factors coincident at any loca-
tion, and the relative importance of these factors.
326
-------
Table 2. Selected Phase II Map Themes
Theme Description
General
reference
Parcels
Community
facilities
Flood plains
Land use
Subdivision
Slope
Wetlands
Historic
resources
Hazardous
wastes
Road networks, hydrographic features,
churches, municipal boundaries, and utilities
Tax parcel boundaries
Public parks and private recreation, cemeteries,
and public buildings
100-year flood plains
Current land uses
Approved subdivisions
Areas delineated with 5-degree slope intervals
Wetlands
Historic buildings and properties
Hazardous waste sites and related buildings
These constraints have been identified by citizens,
government agencies, and the project team. The major
composite constraints include wetlands, historic proper-
ties, steep slopes (slope greater than 15 degrees), pub-
lic parks and private recreation areas, 100-year flood
plains, potential archaeological areas, prime agricultural
land, subdivisions, and existing buildings and structures.
Two maps have been created from the composite con-
straints layer. One map shows the number of coincident
constraint layers that occurs in any one location (see
Figure 3). The other map shows the composite relative
importance of coincident features. Both maps present a
"sensitivity surface" view of the project area.
Secondly, the team uses CIS to perform interactive
summary statistics for each alternative alignment. The
project team analyzes the impact of the alignment alter-
natives on each individual constraint layer.
Two approaches define the impact areas for compari-
son. The first set impact areas are the areas enclosed
by the footprint that traffic engineers delineated for each
alternative. The second set impact areas are 300-foot
buffers on both sides of the alignment delineated by the
traffic engineers. The boundaries of the impact areas
overlay on constraint data layers. Figure 4 displays
wetlands crossed by alignment alternative ROW1B.
A set of summary statistics are calculated for each
alignment. In the end, we compare the statistics for each
alignment in a matrix (see Table 3). The matrix arranges
constraints as rows and alignment alternatives as col-
umns. The statistics include acres of selected features
within each impact area, such as wetlands or high-quality
watersheds, and total counts of features, such as his-
toric-eligible buildings. The summary statistics also in-
clude listings of building names for businesses or public
facilities within the impact areas.
The team has repeated the summary statistics several
times as alignments shift. This procedure ensures that
the final selected highway alignment minimizes environ-
mental impacts, best meets project needs, and is the
most cost-effective alignment to construct. The statisti-
cal matrix of impacts versus alignments is one of the
critical evaluation criteria for comparing alignments and
ultimately for selecting the final alignment.
Existing Roads
Pipeline
Phase II Project Boundary
One to Two Constraint Layers
Three to Four Constraint Layers
Five to Six Constraint Layers
o
1,000
Feet
2,000
Figure 3. Composite constraints.
327
-------
Table 3. Phase II Summary Statistics by Alternative
Constraints Alignment Alternatives
ROW1A ROW1B ROW2A ROW2B
ROWS
Wetlands
Acres
Count
2.76
13
3.27
14
3.25
17
1.8
15
13.64
17
High-Quality Watersheds
Acres 66.18 85.62 71.92 94.58 82.90
Hazardous Waste Parcels
Acres 99.50 116.77 98.25 120.68 215.38
Count 8 5 11 10 6
Parcels With Historic-Eligible Buildings
Acres 14.20 0.10 0.14 0.14 3.46
Count 31222
Historic-Eligible Buildings
Count 1011 1
Benefit of Using GIS
In recent years, many federal, state, and local agencies
have been actively acquiring and automating digital data
(5). These databases provide various types of informa-
tion at scales that are appropriate for a preliminary study
covering a large area.
A more detailed study, which usually covers a smaller
area, often requires more accurate data to describe the
spatial distribution of relevant factors. GIS is flexible,
allowing use of data at the scale and accuracy appropri-
ate to the study purpose. The team has found that this
feature helps improve the efficiency of the project with-
out sacrificing the accuracy. This study has required
two sets of scales ranging from 1:24,000 scale for
Phase I, which required projectwide socioeconomic
and environmental assessments, to 1:2,400 scale for
Phase II, which requires detailed analysis for design of
alternative alignments.
GIS has served as a digital database manager to as-
semble environmental, traffic, geographic, socioeco-
nomic, and other data into a centralized project
database. Data analyzed in this study originate from a
variety of sources, such as PennDOT, U.S. Geological
Survey (USGS), U.S. Census Bureau, Monroe County,
and field survey. They are in different formats, including
digital data in ARC/INFO, INTERGRAPH, and AutoCAD
formats, GPS data, digital images, paper maps, and
tabular data. Many of the public agencies and private
organizations involved in this project already have digital
data that GIS could easily use for this specific project.
This has helped to reduce the overall costs of data
collection and conversion.
This study demonstrates that GIS can support the infor-
mation needs of many disciplines within a common
framework and provide powerful, new tools for spatial
analysis. Aside from technological considerations, GIS
development initiates a higher-order systematization of
geographic thinking (3), which is crucial to the success
of a transportation project. In this project, GIS helps to
determine the total impacts of alternative alignments on
identified constraints. The team accomplishes this by
superimposing the alternative alignments on constraint
Existing Roads
Alternative ROW1B
Buffered Area Boundary
Pipeline
Phase II Project Boundary
Wetlands
Buffered Area of Alternative ROW1B
1,000
Feet
2,000
Figure 4. Wetlands crossed by an alternative alignment.
328
-------
data layers to determine the total amount of each con-
straint layer that each alignment encounters. This ap-
proach supports the analysis of multiple alignments
across the constraint surfaces for a variety of alternative
scenarios. The spatial analysis tools and statistical
function embedded in CIS prove to be very useful in
such study.
Digital data that the CIS stores are used to summarize
environmental, social, and economic data in many dif-
ferent ways. These functions include summing total
acreage, listing entities of special interest, and counting
numbers to provide useful baseline statistics for various
alignments. Within a day, CIS accomplished what would
have required several months of staff labor; CIS sum-
marized impacts of 11 alternative alignments on all con-
straint layers. In addition, CIS creates composite
constraints from individual data layers. A composite con-
straint data layer is created through a series of overlays
to illustrate geographic coincidence of inventory themes.
Conventionally, engineers in a project such as this first
delineate alternatives for alignments from the engineer-
ing perspective; they often consider factors such as
steep slopes and costs. Then, other professionals, such
as environmental specialists, historians, and planners,
evaluate the alternatives from each point of view. CIS
makes possible an early integration of environmental
and engineering activities, ongoing communication with
funding agencies and the public, and continual integra-
tion of a multidisciplinary team.
CIS helps to maintain high-quality data for the project.
It allows for error checking and quality control of multiple
data layers that would not be possible with conventional
mapping. The team always compares a new data layer
with other data to check for conflicts. Making check plots
allows for quick identification of errors and missing data.
For instance, in the process of assigning building-use
attributes to existing buildings, the CIS team first plotted
buildings on a map and created a table with building
identifiers. The field team then used the unique identifi-
ers shown on the plots when noting building names and
building uses in data collection tables. After relating the
data table with the building data layer, the team found
some buildings that did not have building use data.
Moreover, some buildings were assigned uses that were
out of range or seemed out of place, such as a residen-
tial building surrounded by several commercial build-
ings. The team highlighted the data for those buildings
and sent them to engineers for verification. Through this
data cleaning process, the team was able to obtain
complete building use information.
Data quality directly affects project quality. Without CIS,
this type of study often involves using and comparing
maps at different scales, which frequently introduces
serious errors. For each phase of this study, project
team members have used a fixed-base map scale for all
compilations. During data entry and data transformation,
the team has kept accurate registration between data
layers, ensuring data of the same resolution. In addition,
the team has used FREQUENCY, one of the tools that CIS
provides, to look for data values that are out of range as
well as missing data. CIS tools have also helped to
derive new relationships for features. For example, dis-
solving parcels has helped to create subdivision outlines,
or overlaying historic buildings with parcels has helped to
find the parcels on which they are located.
Planners, environmental specialists, historians, and
landscape architects on the project team are responsi-
ble for field data collection, verification, and if necessary,
compilation of field data into the standard project data-
base. Wherever possible, the team has used GPS to
eliminate the task of manual compilation and to improve
accuracy of locating data. A fundamental requirement in
applying the technology appropriately is to understand
its capabilities, requirements, and limitations. Because
several members manage inventory attributes, they
need to know how to maintain unique identifiers for
features so they can link up to the geometry. Because
AutoCAD data transfers occur routinely with engineers,
it is necessary to structure how the AutoCAD drawing
files can be organized, as well as how certain attributes
can be transferred by line color, layer name, or line
width. The CIS group coordinates closely with other
team specialists to identify quick, accurate, and cost-
effective methods of data collection, data analysis, and
presentation. In conjunction with the progress of the
project, specialists from different fields have become
familiar with the concept, requirements, and use of CIS.
They now feel comfortable discussing alternatives while
looking at results displayed on a computer screen.
CIS has created high-quality map products for public
presentations and reports. CIS has also been used in
several agency coordination meetings to display data
and alternative alignments. CIS has allowed for different
data layers to be displayed on the screen with specific
combinations of features at various scales. Public agen-
cies and citizens have been impressed by the clear and
friendly graphic response to their questions. They have
expressed interest in using the technology in future
projects.
Once the study is complete, digital data assembled in
this project will be an excellent resource for future pro-
jects in the study area. For these reasons, CIS provides
more cost-effective project support for gathering, man-
aging, and using data than that provided by paper and
mylar maps.
Summary
For this project, the importance of innovation based on
a solid scientific foundation cannot be overstated. In the
329
-------
current economic and regulatory climate, sound CIS
methods are emerging as the only convincing and cost-
effective means for locating, designing, and gaining
approval for major public and private infrastructure
projects.
The technology offers new and exciting tools for trans-
portation planning.
The methodology that this project team has used can be
successfully applied to other projects that require envi-
ronmental assessment. The team has found CIS to be
an extremely useful tool as users continue to learn its
capabilities and the multiple tools that it offers. The
regulatory agencies have repeatedly made favorable
comments on how CIS can offer interactive viewing in a
show-and-tell environment. This project is one of the first
EIS projects to use CIS in a PennDOT-funded project.
PennDOT appears convinced that CIS is an important
component for conducting EIS for highway studies.
More EIS projects probably will demand CIS services as
part of the project approach, in part, because many
federal and state regulatory agencies are increasingly
using CIS.
References
1. Fischer, M.M. 1994. From conventional to knowledge-based geo-
graphic information systems. Computer, Environment, and Urban
Systems 18(4):233-242.
2. Star, J., and J. Estes. 1990. Geographic information systems: An
introduction. Englewood Cliffs, NJ: Prentice Hall, Inc.
3. Bracken, I., and C. Webster. 1990. Information technology in ge-
ography and planning, including principles of CIS. London, Eng-
land: Routledge.
4. Burrough, P.A. 1986. Principles of geographic information system
for land resources assessment. Oxford, England: Clarendon
Press.
5. Hendrix, W.G., and D.J.A. Buckley. 1992. Use of a geographic
information system for selection of sites for land application of
sewage waste. J. Soil and Water Conserv. 47(3):271-275.
330
------- |