EPA/600/R-05/061
May 2005
Estimating and Projecting
Impervious Cover in the
Southeastern United States
by
Linda R. Exum
Sandra L. Bird
James Harrison1
Christine A. Perkins2
published by
Ecosystems Research Division
National Exposure Research Laboratory
U.S. Environmental Protection Agency
Athens, GA 30605-2700
1 U.S. Environmental Protection Agency
Region 4
Atlanta, Georgia
2 Computer Sciences Corporation
Athens, Georgia 30605-2700
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
Notice
The research described in this document was funded by the U.S. Environmental
Protection Agency through the Office of Research and Development and was conducted at the
Ecosystems Research Division of the U.S. Environmental Protection Agency, National Exposure
Research Laboratory in Athens, Georgia. Mention of trade names or commercial products does
not constitute endorsement or recommendation for use.
-------
Acknowledgments
We wish to thank Lourdes Prieto for her immense help in lending GIS technical
assistance, including helping determine impervious cover during the test data set development.
We also thank Stephen Alberty for his development of the GIS Cover Tool and helping
determine impervious cover during the test data set development.
Foreword
Complete identification and eventual prevention of urban water quality problems pose
significant monitoring, "smart growth" and water quality management challenges. Uncontrolled
increase of impervious surface areas (roads, buildings and parking lots) causes detrimental
hydrologic changes, stream channel erosion, habitat degradation and severe impairment of
aquatic communities. In conjunction with the U.S. Environmental Protection Agency, Region 4-
Atlanta, we provide a multiple data source estimation of imperviousness in the southeastern U.S.
These estimates demonstrate an inexpensive method of determining impervious cover with
known accuracy at the watershed and sub-watershed scales plus characterization of the change in
imperviousness over time. In addition, this report estimates future impervious cover in the
southeastern U.S. using the multiple data source technique. These estimates can guide in-situ
monitoring to confirm problems, aid listing of impaired waters under Section 303(d) of the Clean
Water Act and total maximum daily load (TMDL) development, provide reliable scientific
information to energize sound local planning and land-use decisions, and promote protection and
restoration of urban streams.
Rosemarie C. Russo, Ph.D.
Director
Ecosystems Research Division
Athens, Georgia
"When we see land as a community to which we belong, we may begin to use it with love and
respect." —Aldo Leopold
-------
Abstract
Urban/suburban land use is the most rapidly growing land use class. Along with
increased development inevitably comes increased impervious surface—areas preventing
infiltration of water into the underlying soil. The extensive hydro logical alteration of watersheds
associated with increased impervious cover is very difficult to control and correct relative to the
impact of urbanization on waterways. Development practices that reduce impervious area and
include preventative strategies to protect water quality are more effective and less costly than
remedial restoration efforts. Simple and reliable methods to estimate and project impervious
cover can help identify areas where a watershed is at risk of changing rapidly from a system with
relatively pristine streams to one with significant symptoms of degradation. In this study, a
method for estimating and projecting impervious cover for 12 and 14 digit HUCs over a large
area was developed and tested. These methods were then applied in EPA Region 4' s eight
southeastern states to provide the Region with a screening tool to guide monitoring and
educational efforts.
IV
-------
Table of Contents
Abstract iv
1. Background and Introduction 1
1.1 Stream Biotic Response to Impervious Cover 3
1.2 Using Impervious Cover as a Regional Indicator 5
1.3 Study Objectives 8
2. Test Data Set Development 10
2.1 Digital Orthophoto Quarter Quadrangles (DOQQs) 10
2.2 Data Collection System 12
2.3 Sampling System Design 14
2.4 Sample Size 15
2.5 Analyst Variability 16
2.6 Final Sampling Scheme 18
2.7 Results 18
3. Development of a Multiple Data Source Method for Regional Scale Estimates of Impervious Cover 26
3.1 Population Density Relationships 26
3.2 Use of Categorized Satellite Imagery 27
3.3 Multiple Data Source Approach 34
3.4 Comparison of NLCD only and Multiple Data Source (MDS) Approach 37
4. Impervious Cover in the Southeastern United States 42
4.1 Alabama 45
4.2 Florida 47
4.3 Georgia 49
4.4 Kentucky 51
4.5 Mississippi 53
4.6 North Carolina 55
4.7 South Carolina 57
4.8 Tennessee 59
5. Future Impervious Cover Projections for the Southeastern United States 61
5.1 The Nature of Errors in Population Projections 61
5.2 Impervious Cover Projection Method 63
5.2.1 Residential Component 63
5.2.2 Commercial/Industrial Component 65
5.2.3 Major Highway Component 67
5.3 Impervious Cover Projections 67
5.3.1 Alabama .'. 69
5.3.2 Florida 73
5.3.3 Georgia 77
5.3.4 Kentucky 81
5.3.5 Mississippi 85
5.3.6 North Carolina 89
5.3.7 South Carolina 93
5.3.8 Tennessee 97
5.4 Using the Impervious Cover Projections 101
6. Conclusions and Recommendations 102
References 104
Appendix 113
-------
List of Figures
Figure 1
Figure 1
Figure 2,
Figure 2
Figure 2
Figure 2
Figure 2
Figure 2
Figure 2
Figure 3
Figure 3
Figure 3
Figure 3
Figure 3.
Figure 3.1
'.5
..6
Figure 3.7
Figure 3
Figure 3
Figure 4
Figure 4
Figure 4
Figure 4
Figure 4
Figure 4
Figure 4
Figure 4
Figure 4
Figure 5
Figure
Figure
Figure
Figure
Figure
Figure
Figure
Figure
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Figure 5
Multiple Data Source Impervious Area for North Carolina Piedmont Benthic Site Watersheds 6
Percent Degraded Piedmont Sites vs. Total Impervious Area 7
Examples of Features in DOQQ scenes 12
Captured Screen from the Cover Tool Extension Software 14
Example of Difference (Unassigned Points) Between Analyst 1 and Analyst 2 15
Sample Size and Deviation vs. Grid Spacing 16
Comparison of Impervious Cover by Analyst 17
Impervious Cover Results from the DOQQ Interpretation for Frederick County, MD 20
Impervious Cover Results from the DOQQ Interpretation of 13 Atlanta Area HUCs 24
The three relationships between population density and %TIA 28
Land cover map of the eight Southeastern states using the NLCD92 30
Total acreage categorized as residential (combined high and low density) in the NLCD92 data 31
Impervious cover for Frederick County, MD watersheds measured from aerial photographs vs that
estimated from categorized satellite imagery 32
Impervious cover for 13 Atlanta, GA area HUCs measured from aerial photographs vs that estimated
from categorized satellite imagery 34
Impervious cover for Frederick County, MD watersheds measured from aerial photographs vs that
estimated from Multiple Data Sources 36
Impervious cover for 13 Atlanta area HUCs measured from aerial photographs vs that estimated from
Multiple Data Sources 37
Estimated 1993 %TIA for 1624 Georgia 12 digit HUCs 40
Estimated 1999 %TIA for 1624 Georgia 12 digit HUCs 41
Southeastern United States impervious cover for 2000 44
Alabama impervious cover for 2000 46
Florida impervious cover for 2000 48
Georgia impervious cover for 2000 50
Kentucky impervious cover for 2000 52
Mississippi impervious cover for 2000 54
North Carolina impervious cover for 2000 56
South Carolina impervious cover for 2000 58
Tennessee impervious cover for 2000 60
High Intensity Commercial/Industrial area (meters) vs. population for North Carolina 66
Alabama projected impervious cover out to 2025 70
Alabama Projected %TIA as % of Area out to 2025 71
Total River Miles in Alabama by %TIA Category out to 2025 72
Florida projected impervious cover out to 2025 74
Florida Projected %TIA as % of Area out to 2025 75
Total River Miles in Florida by %TIA Category out to 2025 76
Georgia impervious cover out to 2010 78
Georgia Projected %TIA as % of Area out to 2010 79
Total River Miles in Georgia by %TIA Category out to 2010 80
Kentucky impervious cover out to 2030 82
Kentucky Projected %TIA as % of Area out to 2030 83
Total River Miles in Kentucky by %TIA Category out to 2030 84
Mississippi impervious cover out to 2015 86
Mississippi Projected %TIA as % of Area out to 2015 87
Total River Miles in Mississippi by %TIA Category out to 2015 88
North Carolina impervious cover out to 2030 90
North Carolina Projected %TIA as % of Area out to 2030 91
Total River Miles in North Carolina by %TIA Category out to 2030 92
South Carolina impervious cover out to 2025 94
South Carolina Projected %TIA as % of Area out to 2025 95
Total River Miles in South Carolina by %TIA Category out to 2025 96
Tennessee impervious cover out to 2020 98
Tennessee Projected %TIA as % of Area out to 2020 99
Total River Miles in Tennessee by %TIA Category out to 2020 100
VI
-------
List of Tables
Table 2.1 Impervious Cover Interpretation of 1989 HUCs for Frederick County, Maryland 21
Table 2.2 Impervious Cover Interpretation of 1993 (Black & White) DOQQs and 1999 (Color) DOQQs of 13 12
digit HUCs in the Atlanta, Georgia Area 25
Table 3.1 Empirical relationships between population density and impervious area 27
Table 3.2 Impervious Cover for Frederick County, Maryland NLCD92 Land Cover Categories 33
Table 3.3 Percent Total Impervious Area (%TIA) Results for North Georgia Watersheds 38
Table 3.4 Evaluation of Impervious Cover Status of Georgia Watersheds/HUC's 39
Table 4.1 %TIA as a Percentage of the Total Land Area of Each Southeastern State Using the 2000
Census and the Multiple Data Source Approach 43
Table 5.1 Sources and Dates of Population Projections for Each Southeastern State 64
Table 5.2 2000 U.S. Census and State Population Projections for the Southeastern United States 68
Table 5.3 %TIA as a Percentage of the Total Land Area of Alabama out to 2025 71
Table 5.4 Total River Miles in Alabama by %TIA category out to 2025 72
Table 5.5 %TIA as a Percentage of the Total Land Area of Florida out to 2025 75
Table 5.6 Total River Miles in Florida per TIA category out to 2025 76
Table 5.7 %TIA as a Percentage of the Total Land Area of Georgia 79
Table 5.8 Total River Miles in Georgia per TIA category 80
Table 5.9 %TIA as a Percentage of the Total Land Area of Kentucky out to 2030 83
Table 5.10 Total River Miles in Kentucky per TIA category out to 2030 84
Table 5.11 %TIA as a Percentage of the Total Land Area ofMississippioutto2015 87
Table 5.12 Total River Miles in Mississippi per TIA category 88
Table 5.13 %TIA as a Percentage of the Total Land Area of North Carolina out to 2030 91
Table 5.14 Total River Miles in North Carolina per TIA category out to 2030 92
Table 5.15 %TIA as a Percentage of the Total Land Area of South Carolina 95
Table 5.16 Total River Miles in South Carolina per TIA category 96
Table 5.17 %TIA as a Percentage of the Total Land Area of Tennessee out to 2020 99
Table 5.18 Total River Miles in Tennessee per TIA category 100
vn
-------
1, Background and Introduction
Nonpoint source pollution (NFS), i.e., pollution from diffuse sources such as
urban/suburban areas and farmlands, is now recognized as the primary threat to water quality in
the United States (U.S. Environmental Protection Agency 1994). The pressure on water
resources due to urbanization is rapidly increasing as the U.S. population grows. Urban area in
the contiguous United States increased 26% and roads increased 2% from 1982 to 1992, while
rangeland and cropland/pasture each reduced 2%, respectively (USDA 1997). Between 1992
and 1997 the estimated urban area in the contiguous United States increased another 6 million
acres, or 11%, while grassland pasture and rangeland decreased by 11 million acres or another
2% (USDA 2003). In 1997 USDA identified a new subcategory named "rural residential" as
part of its miscellaneous uses, a category that includes marshes, swamps, bare rock areas, deserts
and transitional areas (USDA 2003). Miscellaneous uses also increased significantly between
1992 and 1997, due in large part to the increase of rural residential.
The U.S. population more than doubled from 133 million to 281 million people between
1945 and 2000, with the total households increasing to 106 million, a quarter of which consisted
of a single individual. Besides more land being converted to residential uses, especially for
homes, new residential areas also require land for schools, office buildings, shopping sites, and
other supporting commercial and industrial uses. The amount of urban land in the U.S. has risen
steadily from 15 million acres in 1945 to an estimated 66 million acres in 1997, converted
mostly from pasture, range and forest land (USDA 2003).
The pace of urban growth in the Southeastern United States is unprecedented. A recent
National Geographic map (Mitchell and Leen 2001) illustrates this extremely rapid
urban/suburban expansion using Department of Defense "city lights" data from two time
periods, 1993 and the "present." Huge areas of "sprawl" growth are particularly evident
throughout the Southeast and are most heavily concentrated in the area between Atlanta, GA and
Raleigh, NC. Based on National Resources Inventory data, developed land increased between
1992 and 1997 in the Southeast as follows: Alabama (16.2%); Florida (18.9%); Georgia
(27.4%); Kentucky (12.8%); Mississippi (16.2%); North Carolina (15.1%); South Carolina
(20.8%); and Tennessee (20.4%) (USDA 2000). A probability sample of landscape trends for
ecoregions of the mid-Atlantic and Southeastern United States documented an increase in urban
area for the Southern Piedmont from 12% to over 16% between 1972 and 2000, the most rapid
urban growth among the ecoregions sampled (Griffith, et al. 2003).
Rapid growth is expected to continue. Preliminary forecasts expect urban land in the
study area of the Southern Forest Resource Assessment to increase from 20 million acres in
1992 to 55 million acres in 2020, and to 81 million acres in 2040 (Wear and Greis, 2002). This
urban expansion will likely come at the expense of both agricultural and forest areas. Regions in
the Southeast likely to be most affected by future growth are the Piedmont, the Lower Atlantic
and Gulf Coastal Plains and the Southern Appalachians.
Fundamental social and economic forces govern conversion of land from uses of less
value to uses of greater value. Production of wealth drives much economic activity and growth.
In the Willamette River Basin (Oregon, USA), the dollar value of developed land relative to its
dollar value for dry land (non-irrigated) agriculture was 59 times for land prepared for homes,
253 times for land with single family homes, up to 552 times for land in commercial use, and
390 to 2535 times for industrial use (Hulse and Ribe 2000). This tremendous increase in land
valuation places intense economic pressure promoting development of land to urban use
whenever the demand exists.
-------
Urban growth produces many stresses on water quality. Often sanitary sewer
infrastructure is not properly maintained and capacity is insufficient. Combined and sanitary
sewer overflow, leaking sewer pipes and faulty septic systems lead to effluent inadvertently
reaching waterways. Sedimentation from construction activities, inadequate control of point
sources, polluted runoff and illicit discharges lead to a decline in water quality (Harrison, et al.
2001).
Arguably, the most difficult to control and correct relative to the impact of urbanization
on water courses is the extensive hydrologic alteration of watersheds , i.e., excessive (as well as
polluted) runoff from impervious surfaces and riparian area degradation. Along with increased
development inevitably comes increased impervious surface—areas preventing infiltration of
water into the underlying soil. Roadways, parking lots and rooftops account for the majority of
impervious area. It is estimated that there are more than 105,200,000 parking spaces in the U.S.,
with a ratio of off-street spaces to on-street spaces roughly two-to-one (NCDENR 2002).
Studies in some metropolitan areas indicate that there are seven times more parking spaces than
there are vehicles.
In addition to extremely deleterious ecological and water quality impacts, flooding is also
a devastating result of the urban hydrologic alteration (Inman 2000; Inman 1995), a stress that is
only sporadically regulated at the local level. Hydrologic (Poff, et al. 1997; Richter, et al. 1996)
and physical stresses (Gaff 2001), as well as chemical contamination, must be addressed to
protect and restore urban water resources.
Increased imperviousness causes a well-known cascade of damaging results to streams
(Wolman 1967 and Caraco , et al. 1998). Detrimental hydrologic changes cause more frequent,
higher peak flows (Jennings and Jarnagin, 2002) and lower water tables and base flows which
can influence both riparian (Groffman, et al. 2003) and aquatic communities. Due to lowered
base flows, streams have reduced resilience to recover from drought conditions. Watershed
runoff can increase by two to over five times normal for forested catchments as impervious area
increases from the 10 to 20% range to 75 to 100% respectively (Arnold and Gibbons 1996).
Altered high flow regimes also increase stream bank erosion and channel enlargement producing
significant sedimentation from the stream channel itself. The few available quantitative studies
of channel changes due to urbanization indicate that from one-half (1/2) to three-quarters (3/4) of
stream sediment load originates from channel erosion (Trimble 1997; Dartiguenave and
Maidment 1997; Corbett, et al. 1997) rather than upland sources. The resulting unstable channel
often evidences highly degraded aquatic habitat, largely due to unstable substrates. The end
result of these stresses is usually severe biological impairment and poor aquatic community
integrity. (See both Paul and Meyer 2001, and Center for Watershed Protection 2003 for
comprehensive reviews of impacts of impervious area on aquatic systems.)
Often, other ecological stresses compound hydrologic impacts from imperviousness.
Summer stream temperatures can be elevated due to runoff from pavement and structures,
placing additional stress on the biological communities. Riparian alterations regularly
exacerbate stream channel erosion and increase stream temperatures further. Additional habitat
degradation often ensues from reduced input of large woody debris (LWD), and from increased
stream crossings by roads, sewers and other structures that create barriers to fish movement
(Paul and Meyer 2001). Impervious surfaces channel pollutants directly into waterways,
preventing processing of these pollutants in soils. Higher pollutant loads, particularly oils, other
petroleum products and metals are typically associated with roadways, while biocides (pesticides
and herbicides) are generally associated with managed landscapes (Center for Watershed
Protection 2003).
-------
Effective storm water management practices implemented in a watershed to control
runoff volumes, flow rates and pollutant concentration can partially mitigate the impacts of
urbanization and increased imperviousness. Development practices that reduce effective
impervious area (E1A) and include other strategies to protect water quality are more effective
and less costly than remedial restoration efforts (Nichols, et al. 1999). EJA is that portion of the
total impervious area (TIA) that is directly connected to the stream drainage system. The E1A
includes streets, driveways, sidewalks adjacent to curbed streets, parking lots, and rooftops
hydraulically connected to the curb or storm sewer system. Empirical relationships between EIA
and TIA have been developed (Sutherland 1995). Rainfall on impervious areas that are not
directly connected hydraulically to the drainage collection system does not always result in
direct runoff and is not as damaging to the biotic integrity of the stream system.
Parcel based analyses of hydrologic and other impacts of impervious area are needed to
inform effective land use policies and local development regulations. Regression modeling
using six important aspects of parcel and street network design explained roughly 77% of
residential impervious cover variation in the Madison, Wisconsin area. This work pointed to
potentially effective policies to reduce imperviousness through zoning considerations such as lot
size, frontage, and front yard setbacks; through street and subdivision design practices such as
block size and intersection density; and through retrofit of existing residential driveways (~20%
of impervious area of parcels) with porous paving materials over time as resurfacing is needed
(Stone 2004).
The change from a watershed with relatively pristine streams to one with significant
symptoms of degradation can occur rapidly in high growth urban areas. Often this occurs before
an awareness by local planners develops on the need to consciously manage storm water
impacts. State storm water control mandates are often set well above the levels where instream
biotic degradation occurs. Impervious area estimates and projections are a potentially effective
tool for highlighting areas that are at-risk for aquatic resources degradation or where stream
system integrity is likely to decline in the near future if effective planning and management
programs are not implemented. These estimates and projections can also guide the selection of
monitoring locations by state and regional EPA officials, focus educational efforts in at-risk
areas, and aid wide-area planning.
1.1 Stream Biotic Response to Impervious Cover
Recent research has consistently shown strong relationships between the percentage of
impervious cover in a watershed and the health of the receiving stream. Booth and Jackson
(1994) suggest that 10% impervious watershed area "typically yields demonstrable loss of
aquatic system function," and that lower levels maybe significant to sensitive waters. In a
review of research on impervious cover, Schueler (1994) concluded that, despite a range of
different criteria for stream health, use of widely varying methods and a range of geographic
conditions, stream degradation consistently occurred at relatively low levels of imperviousness
(10% or greater). May, et al. (1997) found that indicators of stream health in the Puget Sound
Lowlands declined most rapidly from 5 to 10% impervious cover. A recent survey of Maryland
streams (Boward, et al. 1999) found that brook trout (Salvelinus fontinalis), a species very
sensitive to water temperature, were not present in any streams where the watershed was greater
than 2% impervious cover.
Fish IBI results for Ridge and Valley streams indicated poor or very poor fish
communities for catchments with greater than 7% urban land use (Snyder, et al. 2003). Ohio
urban gradient stream sites - excluding sites with allied stresses such as combined sewer
overflows, waste water treatment plants, sewer line problems and other habitat alterations -
-------
showed significant IBI declines with urban area greater than 13.8% and failed to meet Clean
Water Act goals where urban area exceeded 27.1% (Miltner, et al. 2004). Extensive loss of
mussel species (50 to 70%) occurred in Georgia streams experiencing impervious area expansion
(Gillies, et al. 2003). Tidal creek ecosystems in South Carolina experienced adverse physical
and chemical changes (hydrology, salinity, sediment, chemical contamination and fecal coliform
loading) above 10 to 20% imperviousness, with significant biological changes above 20 to 30%
impervious area (Holland, et al. 2004). For southeastern Wisconsin streams, fish communities
declined sharply between 8 to 12% connected imperviousness and were consistently poor above
12% impervious area (Wang, et al. 2001). Evaluation of 245 sites with biological data in
Montgomery County, Maryland required less than 10% impervious and greater than 60%
riparian tree cover to attain a stream health rating of good (Goetz, et al. 2003).
Scientists recognize that fish assemblages in developed watersheds are affected primarily
by nonpoint source anthropogenic stressors that result from land use development (Williams, et
al. 1989; Richter, et al. 1997; Wilcove, et al. 1998). Alteration of hydrologic regimes in terms of
the amount and variability of flow affect all aspects offish life history (e.g., Allan 1995).
Sedimentation can increase fish movement, interfere with fish feeding by reducing reactive
distance for sight-feeders and lower the abundance of insects available as food, and impair
reproduction of fishes with specific spawning habitat requirements (Newcombe and MacDonald
1991; Bergstedt and Bergersen 1997). Habitat destruction can isolate patches of suitable habitat
within a stream which reduces species' survival. Habitat destruction also changes the natural
mosaic of habitat conditions, thereby altering natural fish movement and migration patterns
(Reeves, et al. 1995).
This wide variety of stream response to imperviousness may likely be due to local slope,
soils, geology, land and storm water management practices and other factors. For example,
higher gradient sites in the Ridge and Valley show larger decreases in fish IB I with increasing
imperviousness than do lower gradient sites (Snyder, et al. 2003). Absent more specific local
models, Schueler's (1994) three imperviousness classes of impact provide a useful initial guide
to stream quality in the Southeastern United States:
Sensitive streams have OtolO% imperviousness and typically have good water quality,
good habitat structure, and diverse biological communities if riparian zones are intact and
other stresses are absent.
Impacted streams have 10 to 25% imperviousness and show clear signs of degradation
and only fair in-stream biological diversity.
Non-supporting streams have >25% impervious, a highly unstable channel and poor
biological condition supporting only pollutant-tolerant fish and insects.
A more extensive and updated review of this classification of impact corroborated these
original conclusions (Center for Watershed Protection 2003). While impervious cover alone is
not the sole causative agent for the decline of aquatic health in urbanizing areas (Miltner, et al.
2004), it contributes significantly to the decline and appears to serve as an integrative screening
indicator of urban hydrologic stress (Arnold and Gibbons, 1996).
While complete descriptions of the range of aquatic responses to imperviousness are not
available for all areas of the Southeastern United States, extensive biological sampling of benthic
macro invertebrates by the North Carolina Division of Water Quality covering the wide gradient
of impervious area throughout the Southern Piedmont ecological region (Griffith, et al. 2002)
provides the best existing data to begin building such relationships. Cursory descriptive
-------
examination of a portion of this data allows us to glimpse the potential for using existing and
new data to construct robust relationships valid for the entire Southeast.
Benthic data for over 300 Piedmont sites were kindly provided by Trish MacPherson of
the North Carolina Division of Water Quality (NCDWQ), along with point watersheds
delineated for those sites graciously shared by Dr. Halil Cakir and Dr. James Gilliam of North
Carolina State University. Their detailed, rigorous statistical examination of this data is
currently in preparation.
Figure 1.1 maps these North Carolina Piedmont watersheds by impervious class, in the
context of satellite based land use/land cover for that area. For 159 of these sites with non-
overlapping watersheds, Multiple Data Source (MDS - described in Section 3.3 of this report)
impervious area estimates were produced. The MDS imperviousness of these watersheds ranges
from 1% to 60%.
Figure 1.2 depicts simple box plots of the benthic biological condition response of
streams to increasing impervious area (using both 5% and 10% ranges) for that gradient of
Piedmont sites based on the North Carolina Biotic Index (NCBJ), a tolerance based metric used
for benthic community assessments and aquatic life use support determinations by NCDWQ
(North Carolina Department of Environment and Natural Resources 2003). Assuming NCBI
scores above 6.54 (worse than "fair" on the state's scale of: excellent, good, good-fair, fair, fair-
poor and poor) indicate degraded conditions, progressively greater fractions of degraded sites are
evident as impervious area increases. For watershed Total Impervious Area (TIA) greater than
10%: 62% (32/52) of sites are degraded; for TIA > 15%: 78% (25/32) of sites are degraded; for
TIA > 20%: 83% (19/23) of sites are degraded; and for TIA > 30%: 91% (10/11) of sites are
degraded. In contrast, for watersheds with TIA< 10%: 10% (11/107) of sites were degraded. The
figure also provides percentages and numbers of sites for individual 5% and 10% ranges of
impervious area.
1.2 Using Impervious Cover as a Regional Indicator
Impervious cover when used as an indicator of stream health is typically presented as a
percentage of the total land in an area that contains the impervious surfaces, or percent total
impervious area (%TIA). Several challenges exist in using impervious cover as a regional
indicator. First is simply defining impervious cover since it is not a single, unambiguous
quantity. Generally, paved surfaces and buildings fall unambiguously under the definition of
impervious surfaces. Ambiguity can exist, however, even for these categories since there is now
a pervious asphalt paving material that allows some infiltration. Other areas, such as dirt roads,
railroad yards and construction areas that may not be coated with manmade impervious
materials, are in many instances so heavily compacted as to be functionally impervious. Another
important distinction concerning impervious cover and its impact on stream health is between
connected and disconnected impervious surfaces. Connected impervious surfaces are networked
impervious surfaces (parking lots, roads, sidewalks, etc.) that are physically interconnected and
eventually flow directly into stream systems via storm sewers, ditches and culverts.
Disconnected impervious surfaces, such as rooftops, often deposit runoff onto vegetated
pervious areas. The water from these disconnected impervious surfaces flows through the
subsurface before reaching stream channel networks, mitigating some of the negative impact on
the receiving waters.
-------
I I North Carolina Level 4 Ecoreigons (Omemik)
Impervious Area Percent: Benthic Site Watersheds
IHJ 0 - 1 .999
p~*l 2 - 4.999
| |5- 9.999
| 1 10 - 19.999
HLCD Land Use/Land Cover (~ 1993)
Water
LOUD Density Residential
|^H High Density Residential
^B High Intensity Co mm ercia I /hdu stria I
I I Bare Rock/Sand
| ' j Q uarries/y in es/G ravel
Transitional
| "'' | Deciduous Forest
Evergreen Forest
['":-] Mixed Forest
| | Pasture/Hay
| ] Row Crops
p?"Fl Other Grasses
Woody Wetlands
Emergent Wetlands
No Data
30 0 30 60 90 Kilometers
US EPA
11/23/2004
Figure 1.1 Multiple Data Source Impervious Area for North Carolina Piedmont Benthic Site Watersheds
-------
Percent
Degraded ~~
\
Worst Condition
7-
Z
o1
h-
o
03
5 •
Best Condition
3-
\
\
* 9%
[
..............
tl'
[ 1
l — — l
,
y»(0-5%)
14%
*
J
Box and Whisker Plot '
35%
j
j i
Ed
I
(5-10%)
/^
(10-15%)
67% 1 75%
|
...r±q... 1 ,
L '
^P i ]
i
•
(15-20%) 1 (20-30%)
91%
[
*
(>30%)
f ;
Percent /" \ 2 3 4 s s
Impervious 79 28 20 T|A C|ass 9 12 11
NCBI
= 1554
Number of sites
Figure 1.2 Percent Degraded Piedmont Sites vs. Total Impervious Area
A second challenge in using impervious cover as a regional indicator is determining the
appropriate land area delineation to use in a regional coverage. For any single point in a stream,
the land area or watershed that drains water to that point in the stream affects the water quality at
that point. Delineating watersheds and defining %TIA for every stream mile is not a practical
approach. For this study, we have chosen to use 12 or 14 digit hydrological units (HUCs) based
on the U.S. Geologic Survey (USGS) hierarchical system.
The United States is divided hierarchically into successively smaller hydrologic units.
The USGS has prepared a national coverage of four nested levels identified by two to eight digit
codes (Seaber, et al. 1987). The first level of classification divides the U.S. into 21 major areas
containing either the drainage area of a major river or a combination of rivers. The second level
divides the nation into 222 subregions. The third level divides some of the subregions further
into a total of 352 hydrologic accounting units that are equivalent to or nest within the subunits.
The fourth level is the cataloging unit identified by an eight digit code. A total of 2150
cataloging units form this finest layer of the national coverage. Generally a cataloging unit is a
geographic area representing part or all of a surface drainage basin. These cataloging units are
typically referred to as eight digit hydrologic unit codes (HUCs).
-------
Individual states, in collaboration with the USGS and U.S. Department of Agriculture
(USDA) Natural Resources Conservation Service (NRCS), have delineated subunits of the 8
digit cataloging units into 11 digit and 12 or 14 digit units (depending on the particular states)
that are inappropriately referred to as "watersheds" and "subwatersheds", respectively. The
"subwatershed" delineations represent areas typically in the 5 to 50 sq mi range (although some
are larger or smaller). These small-scale subdivisions are more effective units for evaluating
potential impacts of impervious cover on small, perennial streams. They also provide decision-
makers with appropriate scale geographic frameworks of input for evaluating and managing
water resources at the local level.
There are, however, at least two major considerations in using these 12 and 14 digit HUC
coverages. First, they do not provide a consistent coverage across a multiple state region. This
problem is particularly obvious when discontinuities are observed along state boundaries. These
12 and 14 digit HUC delineations are, however, what individual states use for their water
resources planning and from that perspective are the appropriate mechanism for communication
between EPA Regional personnel and individual state governments.
A more subtle and insidious problem to keep in mind is that hydrologic units at any
hierarchical level are not synonymous with true watersheds. Omernik (2003) points out that
while true watersheds are areas within which surface water drains to a particular point, generally,
only 45 percent of HUCs meet this definition. In over half of the HUCs, the most downstream
points have greater drainage areas than those defined by the boundaries of the HUCs and thus are
not true watersheds. For such stream locations, impacts on instream resources occur due to
activities beyond a single, delineated HUC. That is, impacts on the stream are influenced by
activities in more than one of the HUCs.
A final challenge in use of impervious cover as an effective screening tool for identifying
at-risk streams is finding an easy and relatively accurate method for estimating it over a large
area. In addition, the ability to identify at-risk areas also requires the development of approaches
for estimating impervious cover that link projections of imperviousness to socioeconomic
projections.
1.3 Study Objectives
The objective of this study was to develop and test a method for estimating and
projecting impervious cover for 12 and 14 digit HUCs over a large area. This method was then
applied in EPA Region 4' s eight Southeastern states, providing the Region with a screening tool
to guide monitoring and educational efforts. These techniques will not replace the detailed
impervious cover information needed for planning and management of small watersheds, but
rather will give state and regional planners and managers an overview of potential areas of
concern so efficient monitoring and mitigation efforts can be initiated.
A major question then is with what degree of accuracy can impervious cover be
estimated for subwatershed areas in a region from data available throughout that region. To
answer this question, test data sets of impervious cover for Frederick County, Maryland and the
Atlanta, Georgia area were produced using an ESRI™ ArcView extension developed for use
with USGS aerial photography. Details on development of these test data sets are provided in
Section 2 of this report.
Existing wide area methods for estimating impervious cover were reviewed and tested
early in this effort. Multiple sources of data, including the U.S. Census Bureau 1990 and 2000
Census data, 1992 National Land Cover Data (NLCD) data, and highway information, were all
used to develop estimates of imperviousness. Section 3 discusses the media, methods and results
-------
of estimating impervious cover for the HUCs where test data were collected. Estimates of
impervious cover were then made for 12 or 14 digit HUCs for the eight Southeastern states in
Region 4 - Alabama, Florida, Georgia, Kentucky, Mississippi, North Carolina, South Carolina
and Tennessee - and presented in Section 4.
Finally, state population projections were added to the Multiple Data Source estimation
technique as the basis for projecting future impervious cover in the eight Southeastern states.
Projection methods and resulting projections of impervious cover are presented in Section 5.
-------
2. Test Data Set Development
The overall goal of this study is the development and application of a simple, reliable
method for estimating and projecting impervious cover in 12 and 14 digit HUCs for all the states
in EPA's Region 4. This task depends on the availability of a smaller area test data set to
determine if the region-wide estimation techniques developed adequately reflect what is on the
ground. Such a test data set should include watersheds with a range of %TIA from rural,
relatively undeveloped areas to high density urban watersheds. Multiple examples of low,
moderate, and intensely developed watersheds should be included in the sample. Ideally, sample
watershed data should be available from more than one geographic area. Section 2 describes the
method for developing this test data, describes the areas where the test data was measured and
results of the final measurements.
A number of approaches are used for measuring impervious cover. The most accurate
and costly are ground-based surveys. Ground-based methods are prohibitively expensive to use
where developing a data base from numerous watersheds as required in this study. The use of
manual interpretation of aerial photography is commonplace in accuracy assessments of
automated interpretation remote sensing techniques (Slonecker, et al. 2001) and for other
applications, including watershed management and tax assessment (Lee 1987; Kienegger 1992).
Manual interpretation of aerial photography was chosen for development of our test data sets
since it allows collection of data in a sufficient number of watersheds with an adequate degree of
accuracy.
Test data were collected from aerial photographs in two separate locations: 56, 14 digit
HUCs in Frederick County, Maryland covering 1728 sq km, and in 13, 12 digit HUCs in the
Atlanta, GA area covering 888 sq km. A data collection and storage system was developed that
allowed relatively rapid collection of the required data, plus allowed us to meet our data quality
objectives (DQO). In the quality assurance plan developed at the outset of the project, the DQO
was stated as +/- 10% of the %TIA, i.e. a 10 %TIA would be measured in the 9 to 11% TIA
range. In retrospect, for areas with a TIA of 10% and greater, this was an appropriate DQO. For
low impervious areas, however, this was an objective that was not only unreachable, but also
unnecessarily stringent given the use of the data, e.g. TIA data in the 1.6 to 2.2 % (about a +/-
20% variability) range is functionally indistinguishable. The final DQO was restated as +/- 10%
of the %TIA for areas with > 10 %TIA and as +/- 1 %TIA for areas with <10 %TIA.
An important decision in the initial phase of the study was whether to collect data in only
two categories, i.e., impervious vs. pervious cover, or to differentiate between different types of
impervious elements. While the multi-category data were not necessary to meet the most basic
needs of the study, it would have added significantly to the information data base and allowed us
to address additional research questions plus increased flexibility in the use of the data. A
decision to collect binary data was ultimately made on the basis of our DQOs and resource
constraints. The uncertainty associated with identifying types of impervious elements from the
aerial photography was high and the attempt to collect this data required a substantial increase in
analyst time.
2.1 Digital Orthophoto Quarter Quadrangles (DOQQs)
Manual analysis was done on digital orthophoto quarter quadrangles (DOQQs) obtained
from the USGS. DOQQs are digital versions of aerial photographs that have been orthorectified
so they represent true map distances and are available for any area of the country from the
USGS. The DOQQs have 1 m2 resolution, and their analysis can provide a high level of
accuracy in the determination of impervious cover at a subwatershed scale (Zandbergen, et al.
2000). The DOQQs for Frederick County, Maryland photographed in 1989 were single channel,
10
-------
gray-scale images with a small total variation in spectral characteristics. For the Atlanta, Georgia
area watershed, two sets of DOQQs were analyzed. The first, taken in 1993, was a black and
white (gray-scale) set of DOQQs similar to those used in the Frederick County, Maryland
analysis. The second set of DOQQs, taken in 1999, was color-infrared. The color-infrared
photography covered the same geographic location with the same resolution and was also
created by the USGS. An example of one of the Frederick County DOQQs illustrating several
pervious and impervious features is shown in Figure 2.1.
The proportion of area covered by a given type of surface feature can be estimated from
digital imagery using spectral or visual feature identification methods. Spectral feature
identification uses GIS software to automatically classify features while visual feature
identification involves classifying features manually by a human analyst. Spectral image
analysis involves using specialized GIS software to characterize each pixel in an image to
determine its spectral reflectance. Pixels with reflectance values within predefined ranges are
grouped together to form feature classes. Spectral analysis software is configured or "trained" to
recognize a surface feature based on the spectral characteristics it commonly exhibits. Image
analysis software allows the user to graphically select examples of each type of surface feature.
The programs then analyze the examples and search the entire image for areas that exhibit the
same spectral characteristics. Spectral analysis works well with multi-spectral color imagery and
when the surface features of interest are distinct and can be clearly defined. Features such as
rooftops can have a wide variety of spectral characteristics since roofing materials are available
in a broad range of colors. Spectral methods cannot identify the fact that a building or road
extends under tree canopy as can be done by a human analyst. While the spectral analysis
approach can be very efficient in terms of speed, for our analysis we were not confident that we
would be able to achieve an acceptable level of accuracy using automated methods.
Ground features can be identified and categorized efficiently and accurately by a human
analyst with the help of Geographic Information System (GIS) software. Overlaying ancillary
point, line or polygon data on top of a photographic image provides extra information that might
be useful in differentiating features. A user looking at a good quality photograph can
differentiate features using shape, spatial relationships and geographic context. For example, a
human can reason that a large rectangular feature in a rural area is more likely to be an
agricultural field than a parking lot (Figure 2.1). Even with the help of software tools and
ancillary data, visually identifying and categorizing features on aerial photography can be very
time intensive depending on the size of the area, the density of features, and the speed with
which features can be categorized. Visual identification can also be subjective and vary from
analyst to analyst. In addition, the possibility of missing very small impervious features, such as
sidewalks or even driveways, is very real. While the visual analysis of DOQQs appeared to be
our best option for developing the desired data base, software that allowed for efficient and
accurate collection of data and clear guidelines to maintain consistency between analysts were
important considerations for the success of this effort.
At the initiation of the analysis it was also very important to clearly state which features
we would categorize as impervious and pervious from the DOQQs. The features we designated
as impervious cover were commercial structures, parking areas, industrial areas, quarries,
constructions sites, railroad yards and railroads, residential structures, driveways, roads, paved
streets, dirt roads, highways (but not grassed medians) and airport runways. The features we
designated as pervious cover were vegetated or bare areas, agricultural fields, lawns, parks,
forests, grassed highway medians, water features (including swimming pools), lakes, ponds,
streams and swamps.
11
-------
Highway f'^ ",
,/' Highway ,.'*'»
• '„, Median '„• •
i ,' Agricultural ,'",'., "• ,
j« Field "•
;'~iPr,;;' ',,,,.,,£•'Road ,; » "', "M ,,,
Agricultural
Field
' Agricultural
Field
Figure 2.1 Examples of Features in DOQQ scenes.
2.2 Data Collection System
The amount of area covered by impervious surface can be measured directly by
delineating the extent of each impervious feature found on the DOQQ with a polygon. Because
of the spatial distribution, size and shape of impervious features, like rooftops and sidewalks, it
is time consuming to draw polygons that accurately delineate each feature. While delineating
each feature allows generation of a complete measure of the impervious cover of an area
including the location of the impervious cover within the watershed, our goal was to simply
estimate the fraction of impervious cover in the entire 12 or 14 digit HUC areas. Rather than
delineating individual impervious features for this study, we estimated impervious cover in HUC
areas using a point sampling technique. A grid of points was overlaid on the HUC area and the
%T1A (percent total impervious area) was estimated as the percentage of the points sampled in
the HUC classified as impervious. The selected software, sampling and analysis systems yielded
accurate and reproducible results and allowed efficient collection of data that was stored in a
georeferenced data format. Ground features were identified and categorized by human analysts
12
-------
with the help of Geographic Information System (GIS) software and with a "cover tool"
extension designed specifically for this data collection effort.
Both polygon and point sampling of impervious cover are limited in accuracy by the
ability to properly identify and resolve ground features. The limitations of this sampling are
variable based on both the clarity of the photographs and the nature of the ground cover.
Imperviousness in newly developed areas where photographic quality is high and landscaping
has not developed to obscure ground features can be identified with confidence, in older
neighbors where tree cover can obscure much of what is on the ground and mature shrubbery can
often obscure sidewalk and driveway edges, accuracy will inevitably be lower. Thus the
accuracy of our sampling system is limited by the characteristics of the media we are sampling.
The goal, however, was to develop an efficient sampling system that gave us accurate and
reproducible results within the limitations of the media being sampled. Lack of "ground truth"
data limited our ability to totally quantify the accuracy of our "air truth" data set.
The primary software design goal was to develop an efficient, flexible tool that provided
a framework for accurate and efficient land cover analysis. ArcView® GIS from Environmental
Systems Research Institute, Inc. (ESRI) was chosen as the development platform because it was
the U.S. EPA standard GIS software, was available and familiar to the analysts and provided an
object-oriented programming and development environment called Avenue® (ESRI 1996).
Avenue® scripts were written to add several new functions and controls for characterizing
impervious cover to the existing ArcView® user interface. Collectively, these new functions are
referred-to as the "Cover Tool." The Cover Tool functions fall into three categories: 1) sample
point generation, 2) land cover type assignment, and 3) quality assessment.
The sample point generation feature constructs a point coverage grid in ArcView at a
user-specified density overlaying a DOQQ. This feature was designed so the analyst could
configure the sampling density of a regular sampling grid by choosing the spacing between
points in both the vertical and horizontal directions. Alternately, the user can generate a random
coverage containing a specified number of points. A user-configurable sample point generator
was one of the original software requirements. It allows the analyst to test a range of grid
densities and configurations to find the configuration that minimizes the amount of time required
to analyze impervious cover while assuring that data quality objectives are met. Sample size
determination and sampling system design will be discussed subsequently.
Fast and accurate assignment of the land cover type was the primary requirement in the
design of the data collection software. An integrated point selection and cover type assignment
tool was designed to make this operation as efficient as possible. Analysts can select one or
more similar points and use function keys to rapidly assign a land cover type class to the selected
sample point(s). Alternatively, the analyst can click their secondary mouse button to display a
context-sensitive "popup" menu to change the cover type classification. Users can choose the
classification method that best suits their style, allowing them to work most efficiently. A
significant amount of an analyst's time during on-screen analysis is spent navigating across the
coverage. In order to navigate around an image, a control was designed to allow seamless
panning (i.e., changing the geographic display area). The pan control (Figure 2.2) provides
movement across a screen view width in the horizontal, vertical and diagonal directions and,
thus, provides a systematic way for analysts to locate and analyze sample points. As an added
benefit, the pan control allows the analysts to orient themselves and move efficiently across the
image in either rows or columns.
13
-------
Eiie Edft Kiev XHerne Image Analyse Graphic; Cover Type
a ran H
^ Ei EH
284,613:8 «
4,368,47746 t
^
Change the Cover Type foi the selected potrtfhj to "pervious",
Figure 2.2 Captured Screen from the Cover Tool Extension Software Showing the Pan Button
and Grid Method
To help ensure complete and reliable results, the cover tool includes reporting and
comparison features. The report feature calculates the percentage of pervious, impervious or
unassigned (i.e., not yet sampled) points, and lists preliminary and/or final analysis results. This
feature quickly summarizes land cover type percentages and helps the analyst determine if any
unclassified points remain. The comparison feature analyzes results from two independent
analysts and identifies individual points that are classified differently. After applying the
comparison tool, any sample point that is classified as "impervious" by one analyst and
"pervious" by a second analyst will be reclassified by the software as "unassigned" and reported
to the screen as shown in Figure 2.3. This allows a third, independent analyst to reclassify these
conflicting points to obtain the final results for the DOQQ.
2.3 Sampling System Design
After completion of a prototype version of the Cover Tool, a series of exercises to test the
software and refine the sampling system were conducted. The purpose of the exercises was to
identify potential sources of error and ensure the methods were efficient and reliable.
Two popular schemes for placing the point sample locations are random and systematic
point distribution. A GIS can employ the simple random sampling technique by placing a given
number of points at random locations within a specified geographic study area. Properly
designed random sampling schemes effectively reduce errors that can arise due to regular,
repeating features on the landscape and provide defensible results.
14
-------
€t ArcView GIS 3.2a
0ie Edit ^ew Iheme ,lmage Analysis 2faphr&v Cover Type i^indaw Help
ID MM fflSS Oi^Hlli
ra
Scale 1:1/4492
i •' pennous
-0- impervious
X unassigrifid
Ffed_^i img
Res 1:11,4.
•i(and_1
Figure 2.3 Example of Difference (Unassigned Points) Between Analyst 1 and Analyst 2.
Systematic point distribution can be an attractive alternative in cases where random
sampling is more difficult or time-consuming. With the systematic technique, a Cartesian grid
system with equally spaced points in the x and y dimensions (i.e., in rows and columns) is
applied to the study area. When using the systematic approach, it is important that the origin of
the grid be positioned randomly (Borgman and Quimby, 1988) to avoid personal bias. Lee (Lee,
1987) observed no systematic bias using regular versus random grids for sampling impervious
cover. During software testing, users found that a systematic sampling system in conjunction
with the pan tool provided a very efficient means of locating and classifying sample points. The
pan tool was used to move the photograph to the left and right along rows of sample points, or up
and down along columns of sample points. This helped orient users and seemed to increase
analysis speed. Both randomly and systematically spaced points were used and results compared
for two different DOQQs. One DOQQ was located in a rural area (Catoctin^se) while the other
was more urban (Fred^sw). Impervious cover results for random (4.81%) and systematic
(4.56%) point placement analyses on the Catoctin^se DOQQ were not significantly different,
X2(i N=469?) = 0.289, p=0.60, and were well within the data quality objectives. Impervious cover
estimated with random point placement on Fred^sw (13.1%) was slightly different from that
using systematic point placement (14.6%), yf(l N=4774) = 3.94, p<0.05. Analyst time required to
categorize the randomly spaced layout was greater than that with the regularly spaced grid, and
the analysts expressed a greater sense of fatigue categorizing the randomly spaced grid as well.
2.4 Sample Size
The primary factors used to determine an appropriate sampling point density are: 1) the
time available for sampling, and 2) the quality objectives. The optimal sampling density,
therefore, is the one that provides acceptable precision with the least effort. At the limit of an
infinite number of points, the point sampling becomes a continuous cover similar to the polygon
15
-------
delineation. Our goal was to find a sampling grid density that at a minimum would meet our
data quality objectives. Our goal was not to just minimally meet these quality objectives,
however, but would also exceed this minimal number and build in a margin of safety.
Impervious cover was analyzed on two representative DOQQs using a regular grid system. As a
test, sample points were positioned 50, 100, 200 and 400 m apart in both the x and y dimensions
on the Catoctin^se and Fred^sw DOQQs. Analysts then estimated the cover conditions on each
DOQQ. The deviations estimated in impervious percent cover relative to their 50 m estimates
were calculated for the two DOQQs and plotted against sample size to aid in determining the
optimal sampling density (Figure 2.4).
Sample Size and Deviation
vs. Grid Spacing
20
15
QJ '— •
M VI
m E
-5. 1 10
E °
(Fi I —
*J J •- r*
5
n
\
Y Sample Size ,D
Ji! ^
\ " * '
Deviation
m,, p-™ /
^Q-
\ /
\ n
V """^-.,
/ - """""•""-„ X
/'f li 1™s"~*i,., .X
^ "X,,|S— 1
\*.
fy '^*
-------
quarries were deemed imperious as well due to their heavily compacted nature. Actual surface
material and nature in these cases is often hard to determine from the aerial photography. In
addition, trees can interfere with the interpretation of ground features under the canopy, and the
analyst must interpolate what is under the canopy from surrounding features.
To quantify variation in cover type results by analyst, the same DOQQ was characterized
by six individuals (Figure 2.5). Each analyst used an identical sampling grid composed of 1,178
points spaced 200 m apart. The results were compared to determine if substantial bias existed
between analysts. Some analysts tended to interpret more area as pervious while others tended
toward impervious. Estimates of impervious cover for the test DOQQ ranged from 11% to 18%,
with an average estimated value of 14%. This range of results was outside that required to meet
our quality objectives (12.6% to 15.4%). The subjective judgement required and the resulting
analyst to analyst variability in the results appeared to be the area in the data collection most
likely to compromise our data quality standards. In the final development of our sampling
protocols, reducing these latter errors was the primary focus for our resource investment.
Comparison of Results
from Individual Analysts
6
5
4
f 5 3
0
E 2
O
4—
1
0
Analyst
• Analyst 1
• Analyst 2
• Analyst 3
D Analyst 4
D Analyst 5
P Analyst 6
Figure 2.5 Comparison of Impervious Cover by Analyst.
In order to control this error, sampling points overlaid on the DOQQs were characterized
by two independent analysts as either pervious or impervious. A third individual served as a
quality assurance checker. The quality assurance checker imported the results of the first two
analysts into a Cover Tool utility that automatically compared the two grids on a point-by-point
basis. Points with discrepancies in the categorization by the first two analysts were reviewed by
the quality assurance checker, who made the final determinations of assignment for these
contested points.
17
-------
2.6 Final Sampling Scheme
A regular sampling grid was used in the final analysis because test analysis found
categorization of the random grid much more time consuming and tiring than the regular grid.
The difference in results generated by the two sampling schemes was much lower than the
analyst to analyst variability. The time and energy saved was better spent on multiple analyses.
The final method selected relied on the three-analyst scheme described in the previous section.
Based on the sampling grid size results, a grid spacing of 200 m by 200 m was chosen. This
density yielded an average of nearly 800 sample points per 14 digit HUC in Frederick County,
Maryland. This sample size did not compromise our ability to meet our data quality
requirements. The number of sample points within the entire Frederick County study area at this
resolution totaled 43,816. At this resolution, approximately 3 to 4 hours was required per
analyst to categorize each DOQQ. For the 13 Atlanta, Georgia area HUCs, an average of
approximately 1700 points per HUC was sampled. A total of 22,206 points were sampled in the
Atlanta area from DOQQs taken at two different time periods.
Analysis procedures were developed to simplify land cover type assignment and ensure
the quality of estimates. Because analyst interpretation was identified as the major source of
sampling variability, training and validation procedures were designed to promote comparable
results. Each analyst received training in photographic interpretation that included graphic
examples of common pervious and impervious features. As a general rule, all analyses were
conducted at a scale appropriate for the resolution of the photographs. Analysts were
encouraged to zoom closer (i.e., increase the scale), however, to help classify hard-to-
differentiate points. Analysts were instructed to characterize each point as absolutely inside the
feature shown on the photograph. For example, analysts were warned against classifying a point
as impervious just because it fell "close" to a house.
To ensure the most reliable impervious cover estimates, two independent analysts
characterized each of the DOQQs. The DOQQs were randomly assigned to analysts so that no
individual analyzed a large, contiguous geographic region. A third individual served as a quality
assurance checker. This final individual imported the results of the first two analysts and
compared them on a point-by-point basis. This was accomplished using the Cover Tool's
custom comparison function to identify any classification discrepancies between the two
analysts.
Figure 2.3 illustrates results of the comparison function analysis of themes created by
analyst 1 and analyst 2. The Tool generates a third theme called "Cover Type" that shows points
highlighted as discrepancies between the two analysts. These discrepancies, symbolized by a
"X" in the screen view and labeled as unassigned, occurred when one analyst assigned the point
as pervious while the other assigned it as impervious. Figure 2.3 shows that on a point-by-point
basis for that DOQQ, the difference between analyst 1 and analyst 2 is 5.9% unassigned points.
Despite the difference, the total impervious cover estimates were 16.8% vs. 16.1% between the
first two analysts. A third quality assurance analyst examined only the points where there was a
discrepancy between analyst 1 and analyst 2. The final impervious cover for the DOQQ in
Figure 2.3 after the quality analyst review was 15.4%. A more detailed discussion of quality
assurance levels associated with this data collection scheme can be found in Bird, et al. (2000)
2.7 Results
The impervious cover for Frederick County, Maryland HUCs ranged from less than 1%
to 35% as illustrated in Figure 2.6. The highest intensity impervious area centered on the town
of Frederick, with the HUC containing most of the town having 23% TIA. Only three of the
Frederick County watersheds had impervious cover greater than 10%. The County mean value
18
-------
was 5.1% T1A, the median 4.6 % T1A. Table 2.1 contains the final %TIA interpretation data,
and lists all the HUCs completely or partially contained in Frederick County.
An ideal data set for testing the use of estimated impervious cover as an environmental
indicator would have more data points greater thanlO %T1A where stream impairment is
observed, than such data points contained in the Frederick County data. Atlanta, Georgia was
chosen for an additional data set with the chance of considerably more impervious cover. The
Atlanta area HUCs are in midtown, north Atlanta and in the Etowah River basin north of Atlanta.
Six of the thirteen Atlanta area HUCs shown in Figure 2.7 contained greater than 10 %T1A,
including one midtown watershed with a 50 %T1A. Data from both the 1993 and 1999
photography are summarized in Table 2.2. North Atlanta is a very high growth area, with one of
the HUCs there more than doubling in %T1A during that six year period.
19
-------
Measured Impervious Cover
Frederick County, Maryland using
1989 U.S.GS. Orthophoto Quadrangles
1.6
7 ~« \ "^Ifc
A~--f\,-C3"7 V<
2.6
3.5
\
2.1
2.6
2.6
,r 5,2
J""
,-, ,-/
1
7,3 j 4.6
2.7
5.4 {
(
,«i
7fl V i'
12.0 ; ';
4.9
•J'-, 3.9
-9,0 _,,-
y
.-^v
V, 3.6
% Impervious Cover
<3
\ 5.0 >'
-J L
i 1.5
-5.6
7.2
3.3
3-5
5-10
10-20
>20
U.S. Environmental Protection Agency
ORD/NERL/ERD Athens, Georgia
Figure 2.6 Impervious Cover Results from the DOQQ Interpretation for Frederick County, MD
20
-------
Table 2.1 Impervious Cover Interpretation of 1989 HUCs for Frederick County, Maryland
14 digit HUC
02070009040124
02070009040128
02070009030101
02070009030104
02070009030102
02070009040127
02070009060176
02070009060177
02070009060201
02070009060202
02070008010026
02070009060227
02070009060226
02070008010028
02070009060228
02070009050171
02070009060204
02070009060203
02070009060205
Impervious Cover
(% TIA)
1.6
3.4
2.1
4.7
7.8
2.6
2.5
3.7
3.3
2.8
2.1
7.8
3.5
2.6
2.6
4.2
5.2
4.3
3.7
Area
(Sq Mi)
0.9
4.4
11.0
13.2
5.5
5.8
21.6
18.0
4.7
6.6
10.8
16.5
6.9
15.2
18.3
8.0
9.4
3.5
18.3
HUC within County
Completely
Partially
Completely
Completely
Completely
Completely
Completely
Completely
Completely
Partially
Partially
Completely
Completely
Completely
Completely
Partially
Completely
Completely
Completely
21
-------
Table 2.1 Impervious Cover Interpretation of 1989 HUCs for Frederick County, Maryland
14 digit HUC
02070008010027
02070009050170
02070009060251
02070009050169
02070009060206
02070009050168
02070008010029
02070008010030
02070009060208
02070009060252
02070009060209
02070009070280
02070009070276
02070008010032
02070008010031
02070009060210
02070009070278
02070008010036
02070009070286
Impervious Cover
(% TIA)
3.5
2.3
6.0
6.6
6.4
4.9
5.4
4.5
7.6
7.3
7.0
4.6
2.7
8.0
4.6
23.0
3.9
3.7
5.6
Area
(Sq Mi)
7.3
7.3
14.0
8.3
12.3
3.7
17.3
12.8
8.1
19.1
17.8
21.1
16.3
10.2
14.1
28.3
15.1
15.9
12.3
HUC within County
Completely
Completely
Completely
Partially
Completely
Partially
Completely
Completely
Completely
Completely
Completely
Completely
Partially
Completely
Completely
Completely
Partially
Partially
Completely
22
-------
Table 2.1 Impervious Cover Interpretation of 1989 HUCs for Frederick County, Maryland
14 digit HUC
02070009070283
02070009080301
02070009080302
02070008010035
02070009080305
02070009080303
02070009080306
02070008010037
02070008010052
02070008010038
02070009080326
02070009080330
02070009080327
02070008010039
02070008010051
02070009080328
02070009080308
02070008020076
Impervious Cover
(% TIA)
3.0
12.0
14.8
6.1
4.9
9.0
5.0
8.8
5.2
3.6
5.6
3.3
7.2
3.5
4.9
1.8
1.5
0.0
Area
(Sq Mi)
15.8
20.0
5.1
6.5
19.4
13.6
17.1
17.4
24.0
10.6
7.1
17.3
6.6
4.0
5.7
4.2
11.5
0.7
HUC within County
Completely
Completely
Completely
Completely
Completely
Completely
Completely
Partially
Completely
Completely
Partially
Completely
Partially
Completely
Completely
Partially
Partially
Partially
23
-------
Measured Impervious Cover
for Atlanta Area HUCs
Using 1993 U.S.G.S.
Orothophoto Quadrangles
North Georgia!
% Impervious Cover
U.S. Environmental Protection Agency
ORD/NERL/ERD Athens, Georgia
Figure 2.7 Impervious Cover Results from the DOQQ Interpretation of 13 Atlanta Area HUCs
24
-------
Table 2.2 Impervious Cover Interpretation of 1993 (Black & White) DOQQs and
1999 (Color) DOQQs of 13 12 digit HUCs in the Atlanta, Georgia Area
12 digit HUC
031300010906
031300010907
031300011001
031300011002
031300011201
031300011202
031300011204
031501040301
031501040302
031501040303
031501040304
031501040305
031501040306
1993 Impervious
Cover (%TIA)
10.5
21.0
6.1
8.6
32.9
33.2
50.5
0.7
2.6
3.6
4.2
4.4
1.8
1999 Impervious
Cover (%TIA)
22.4
24.4
9.5
15.8
35.1
34.3
49.7
1.6
4.2
4.4
7.1
6.8
3.0
Area
(SqKm)
34.0
111.9
91.8
83.7
101.4
77.6
62.5
38.7
43.0
37.9
54.5
76.8
74.4
25
-------
3. Development of a Multiple Data Source Method for
Regional Scale Estimates of Impervious Cover
Regional (multi-state) scale estimates of impervious cover are not feasible using the
labor intensive methods discussed in Section 2. Regional scale estimates need to be based on
automated methodologies that are relatively rapid to implement. In order to achieve an
acceptable and consistent level of quality throughout the region, calculations should be based on
regionally available data of known and consistent quality. In this study an important feature of
the method to estimate current levels of impervious cover was the ability to be able to use the
same method as a basis to project future scenarios of impervious cover. Generally, the method
of choice should not be dependent on calibrations and preferably would provide a linkage to
demographics and other socioeconomic parameters to use as the basis for projections.
Our study considered three different approaches for performing wide-area estimates of
impervious cover. The first was based on the relationship of population density to impervious
cover. The second looked at the potential of using categorized satellite imagery as an estimation
approach. The third approach, and the one we adopted for the estimation and projection of
impervious cover throughout EPA Region 4 as detailed in the final chapters of this report, was
based on the use of Multiple Data Sources—block level census data, categorized land use/land
cover data and road networks. This section details each of these three approaches considered
and provides our evaluation of each.
3.1 Population Density Relationships
A number of relationships between population density and impervious cover have been
developed. City planners often use land-use zoning for rapid estimates of total impervious area.
Both population density and land-use zoning based estimation methods provide a means for
projecting an increase in impervious cover in a watershed, using either population growth or
build-out scenarios as the forcing function (Arnold and Gibbons 1996). Comprehensive land-use
zoning data is not available regionally, but population density is available from the U.S. Census
Bureau. Impervious cover is a result of human settlement, and thus, population density should
be a reasonable predictor of impervious cover arising from residential development and the
commercial areas that directly support them. Use of population density as a means to estimate
impervious cover is attractive since it provides a rapid technique for generating a quantitative
estimation of both present and projected land surface cover.
Stankowski (1972), Graham, et al. (1974) and the Greater Vancouver Sewerage and
Drainage District (GVS&DD 1999, Hicks and Woods 2000) developed empirical relationships
with different functional forms to relate population density (persons/mi2) to percent impervious
cover (%TIA). Table 3.1 shows %TIA as a function of population density developed in each of
these three studies. Stankowski developed his relationship using county scale data from New
Jersey with population densities ranging from 120 to 13,800 persons/mi2. The impervious cover
was estimated from land use data available from the state planning office. Graham, et al.
evaluated selected census tracts for the Washington, DC metropolitan region. Population
densities ranged from 350 to 53,300 persons/mi2. They developed impervious cover estimates at
the block level ranging from 14% to 98%. Test data of %TIA was developed using 1:50,000
aerial photography. GVS&DD developed their relationship based on data for the greater
Vancouver, British Columbia area using impervious cover estimated from land use zoning
categories.
26
-------
Table 3.1 Empirical relationships between population density and impervious area
Source Relationship
Stankowski(1972) %TIA = 0.0218 P1-206 - 0.100 logP
Graham et al. (1974) %TIA = 91.32 - 69.34 (0.9309P/640)
GVS&DD(1999) %TIA = 95 - 94 exp(-0.0001094 P)
Figure 3. la shows the %TIA predicted by the relationships developed by Stankowski,
Graham, et al. and GVS&DD. Data measured from aerial photographs are also included for the
Frederick County, Maryland watersheds as described in Section 2 and the census tract level data
from Washington, DC collected by Graham, et al. Whereas the Stankowski relationship
seriously under predicts %TIA at population densities greater than 1000 persons/mi2, the
Graham et al. relationship seriously over predicts %T1A for population densities under 500
persons/mi2. Although the GVS&DD relationship appears to provide the best fit overall, closer
inspection of the data for population densities under 2000 person/mi2 (Figure 3. Ib) indicates that
this function actually underestimates %TIA in this range. Not surprisingly, the greatest under
prediction occurred in watersheds (HUCs) with significant amounts of intensive
commercial/industrial and mining/quarrying land cover types. For the Frederick County data,
the most extreme error was for a watershed with 15 %T1A that was predicted by the GVS&DD
relationship to have only 4 %. The occurrence of this magnitude of potential error emphasizes
the limitation of relying solely on population data as an indicator of percent impervious surface
area. On average, the GVS&DD relationship underestimated impervious cover for the
Frederick County watersheds by 2 %T1A (sd = 2 %).
These three population-based approaches do not account for commercial, industrial or
mining contributions to impervious cover. In addition, they do not account for development
styles that can alter the per household level contribution to impervious cover. However,
population density is a good basis for screening level estimation of the residential contribution to
impervious cover. The exponential relationship of GVS&DD captures the general shape of the
relationship between population density and impervious cover, but generally underestimates the
impervious cover. This underestimate is not unexpected since commercial, mining,
manufacturing and some transportation contributions are not necessarily directly related to
population density.
3.2 Use of Categorized Satellite Imagery
While processing and categorization of satellite imagery is expensive and time-
consuming, use of categorized imagery is a rapid and relatively inexpensive method of
estimating impervious cover. Categorized land use and land cover systems derived from remote
sensing data define developed land cover classes based on the fraction of impervious cover in a
specified area (Anderson, et al. 1976; Vogelmann, et al. 1998a). Sleavin, et al. (2000) generated
percent impervious coefficients for generalized land use and land cover classes developed from
30 m Landsat Thematic Mapper imagery. While subpixel classification methods show promise
in the quantification of impervious cover (Ji and Jensen, 1999; Slonecker, et al. 2001; Yang, et
al. 2003), data sets developed using these methods are not yet available over large areas and
these methods generally do not attempt to estimate imperviousness in pixels with less than 20%
impervious cover.
27
-------
100
_ 80
IS
i 60
O
3
f 40
0*
a.
E
20
Graham * Graham Data
Stankowski T Frederick Data
GVS&DD
10 100 1000 10000
Population Density (personslsq mi)
100000
- GVS&DD
• Graham Data
T Frederick Data
0 500 1000 1500 2000
Population Density (persons/sq mi)
Figure 3.1 The three relationships between population density and %TIA
presented in Table 3.1 are shown in Part a (top figure above) along with data
collected for this study in watersheds in Frederick County, Maryland and by
Graham (1974) for census tracts in Washington, DC. Part b (bottom figure above)
shows the response of the GVS&DD (GVS&DD 1999, Hicks and Woods 2000)
relationship for population densities less than 2000 persons/sq mi compared to
data presented on a linear scale.
28
-------
The 1992 National Land Cover Data (NLCD 92) is a categorized land cover data set for
the continental United States developed for the Multi Resolution Land Characteristics
Consortium (Vogelmann, et al. 2001) that can be downloaded at no cost. It provides nationally
consistent land-use/land-cover based on 30 m Thematic Mapper data from the early 1990s plus a
variety of auxiliary data sources. A land cover map based on NLCD 92 is shown for the eight
Southeastern states in Figure 3.2. Once watershed boundaries and categorized imagery are
available in the same geographic projection, software such as ATtlLA (Ebert and Wade, 2000)
make watershed estimates of impervious cover using categorized imagery a very rapid
operation. For example, estimates of the imperviousness of several hundred watersheds can be
made within a day's time.
One difficulty with using the categorized land cover data for impervious cover estimation
is the fact that for a pixel to be categorized as even low density developed, it must be at least
30% impervious cover. Figure 3.3 shows the amount of developed residential land in different
lot size categories in Frederick County, Maryland based on property tax records (Maryland
Office of Planning 1999), and the total land area in the two residential cover classes from the
NLCD 92. The total amount of residential land identified by the NLCD 92 is consistent with
the acreage in residences on lots less than about '/2 acre. As noted previously, to be classified as
low density residential in the NLCD92, a 30 m cell must include at least 30% impervious cover.
Residential development with houses on lots greater than '/2 acre typically have less than 30% of
the area in impervious cover. Larger lot developments, consistent with these definitions of
developed land cover, are classified into one of the undeveloped categories in the NLCD 92.
The Frederick County, Maryland test data set was used to estimate the percentage of
impervious cover in each NLCD92 land cover category as well as the total area of impervious
cover in Frederick County that was in each land cover category. These estimates are shown in
Table 3.2. The percentage of the category estimated as impervious in Frederick County is the
percentage of the points sampled from the DOQQs falling in that land cover class that were
categorized by analysts as impervious. The sample size is the number of the DOQQ sampling
points that were located within the specific land cover class in the NLCD 92 coverage. The final
column of Table 3.2 is the percentage of the impervious cover points sampled from the DOQQs
in Frederick County that are located in the cells of each land cover type. Only 23% of the
DOQQ sampling points categorized as impervious in Frederick County are located in an area
categorized by NLCD 92 as developed. Over 50% are located in the agricultural categories.
While impervious cover certainly exists in rural agricultural areas, a significant portion of the
land classified as agricultural by the NLCD 92 is, in fact, low density residential development.
Frederick County is a suburban county, but the land cover data classifies much of the low
density development as agricultural land.
The percentage of imperviousness in each land cover class was then used with the
NLCD92 data for each of the Frederick County HUCs to estimate the percent impervious area in
each watershed. Since the impervious surface coefficients were derived from the data for the
whole county, on average the impervious cover estimates were expected to closely match the
measured data. The mean error between estimated and measured values was 0.2 %TIA (sd = 2).
The best fit regression line for the data plotted in Figure 3.4 has a slope of 0.522 and a y-
intercept of 2.74. Ideally, for the perfect model, the regression line would have a slope of 1.0
with an intercept of 0.0.
29
-------
/Atlantic Ocean
'; • '.;.| Low Density Residential
^| High Density Residential
^f High Intensity Commercial/Industrial
| | Bare Rock/Sand
LlLj Qua rries/Mnes/G ravel
Transitional
| ,| Deciduous Forest
B/ergreen Forest
[" "•••] Mixed Forest
jjj§ Shrubland
| | Orchards/vineyards
|^5j Grasslands/Herbaceous
| | Pasture/Hay
| | Row Crops
\«^\ Other Grasses
E~3 Woody Wetlands
| | Emergent Wetlands
| | Small Grains
| | Fallow
I | No Data
ijiTHw-cW'1";"*!
^1
Gulf of Mexico
sm^
"?*
0 50 100 150 200 250 300 Miles
U.S. Environmental FVotecti on Agency
Athe ns , Georgi a Novemb er 2004
Figure 3.2 Land cover map of the eight Southeastern states using the NLCD92
30
-------
35
30
25
20
15
10
HLCD residential < 5ac
< 2 ac
1ac
< 0.75ac
< 0.5ac
< 0.2 ac
Figure 3.3 Total acreage categorized as residential (combined high and low density) in the NLCD92 data (NLCD
residential) and by residential lot size category from property tax records for Frederick County, MD. The labels
for data from the property tax records indicate all the residential lots that are less than the indicated number of
acres per residence, e.g., <5 ac is the sum of all properties in the tax records that are on lots of less than 5 acres
per housing unit.
31
-------
o
U
01
Zi
o
25
20
15
£
0°
Q.
J! 10
Ul
LJJ
5 10 15 20
Measured Impervious Cover (%)
25
Figure 3.4 Impervious cover for Frederick County, MD watersheds measured from aerial
photographs vs that estimated from categorized satellite imagery and categorized coefficients
developed from county-wide data. This approach systemically over predicts imperviousness in
relatively underdeveloped watersheds and under predicts imperviousness in developed watersheds.
32
-------
Table 3.2 Impervious Cover for Frederick County, Maryland NLCD92 Land Cover Categories
Land Cover Category
low density residential
high density residential
commercial/industrial
quarries/mines/gravel
transitional barren
deciduous forest
evergreen forest
mixed forest
hay/pasture
row crops
other grasses
woody wetland
herbaceous wetland
Percentage
of Total
Area in
Land Cover
Class
2.2
0.2
0.4
0.3
0.1
25.5
1.6
7.6
53.8
6.0
0.1
0.9
0.3
Sample
Size
990
76
156
117
29
11159
697
3400
23497
2663
33
368
138
Percentage
of the
Category
Impervious
42
77
57
62
17
2
4
5
5
8
9
3
1
Percent of
Impervious Area
in Frederick
County Accounted
for by Category
17
2
4
3
0
9
1
7
48
9
0
0
0
The impervious surface coefficients derived from the Frederick County, Maryland data
were also used to estimate impervious area in the 13 Atlanta area HUCs. Figure 3.5 shows the
measured vs estimated %TIA for the Atlanta area HUCs. Once again, the mean error is low (0.6
%T1A; sd = 3.4). In this case, the regression line slope is still less than 1.0, 0.812, although not
as flat as the slope through the Frederick County data, with a y intercept of 1.96. The Atlanta
area data set is more heavily influenced by more developed watersheds, areas where the satellite
imagery is expected to perform best. Nevertheless, there is still an underestimate of
imperviousness at the lower end and over estimate of impervious cover in the more developed
watersheds although not as pronounced as for Frederick County.
33
-------
60
50
g
8 40
0)
Q.
E
0)
-i— •
m
E
to
LJJ
30
20
j£W
0
10 20 30 40 50
Measured Impervious Cover (%)
60
Figure 3.5 Impervious cover for 13 Atlanta, GA area HUCs measured from aerial photographs vs that
estimated from categorized satellite imagery and category coefficients developed from Frederick County,
Maryland data. This approach systematically over predicts imperviousness in relatively underdeveloped
watersheds and under predicts imperviousness in developed watersheds.
Jennings et al. (2004) addressed this issue by developing three sets of coefficients of
imperviousness for each NLCD92 land cover category (a total of 42 coefficients ) based on the
percent of developed land in a watershed area. This multiple coefficient approach resulted in an
approximately 2% absolute %TIA error, but does not have the systematic over and under
prediction bias that using single coefficients per land cover category produced.
3.3 Multiple Data Source Approach
In the Multiple Data Source (MDS) approach, three different data types—population
density from block level census data, the commercial-industrial and quarrying-mining land cover
category from NLCD 92 and interstates and major US highway coverages—were combined to
estimate impervious cover. The MDS uses the different data types to represent components of
34
-------
imperviousness most appropriate to the specific data source. In the MDS approach, population
density is used as an indicator of impervious cover generated by residential development.
Categorized satellite imagery from the NLCD 92 is used to evaluate the contribution of
commercial and industrial areas—areas that are clearly identified from satellite imagery. Road
networks from the National Transportation Atlas (USDOT, 2001) data are used as a source for
major highways to estimate impervious cover contributed by major highways that are not related
to local residential development.
The residential contribution to imperviousness was estimated based on population
density using the GVS&DD method (Hicks and Woods 2000, GVS&DD 1999) discussed in
detail in Section 3.1. We used U.S. Census 2000 block level data to estimate population density
in individual HUCs. Both population data and vacant housing was used to develop an effective
population density in the watershed. Many areas of the Southeast, specifically the coastal and
mountain areas, have high rates of vacation and seasonal housing which is not reflected in the
resident census count. The number of vacant dwellings multiplied by the average persons per
household for the state in the 2000 Census was added to the residential population for each HUC
to calculate an effective population. The effective population divided by the HUC area was used
in the GVS&DD formula to calculate the residential %TIA.
The categories pulled from the NLCD 92 for the MDS approach were #23/Commercial/
Industrial/Transportation and #32/Quarries/Strip Mines/Gravel Pits. The two NLCD 92
categories add information on the contributions to imperviousness from major manufacturing,
commercial and quarrying areas that can be detected by satellite imagery. These latter categories
are assumed to be 90% impervious (Caraco, et al. 1998). By definition the commercial-
manufacturing category is 80% or greater impervious in the NLCD 92 classification.
Impervious area due to major highways was calculated based on the total length of
interstate and other major US highways arcs (USDOT, 2001) in a watershed (HUC), times the
number of lanes for an individual road arc multiplied by an assumed lane width of 12 ft. Where
highway arcs overlap with the NLCD categories we extracted, the road arcs were removed to
prevent double-accounting.
Total %TIA for the HUC was calculated by summing the impervious area contributed by
major highways, commercial and mining in each HUC, dividing by the total HUC area and
multiplying by 100 to convert to percentage, and adding to the %TIA calculated for the
residential component from the GVS&DD equation. Calculations were performed using
ArcView 3.2 and a detailed step-by-step procedure and Avenue scripts used in the computations
are included as an Appendix in this report.
Estimates of impervious cover based on combining Multiple Data Sources are illustrated
in Figure 3.6 that compares the estimated impervious cover using the combined data set to the
measured values for Frederick County, Maryland. The straight line indicates a one-to-one match
between the estimated and measured %TIA values. Overall, this technique underestimated
impervious cover by 1 %TIA with an average, absolute error of 1 %TIA. This estimate was
obtained without fitting to the test data set. For Frederick County as a whole, the residential area
calculated from population density contributed 65% of the imperviousness, commercial/
industrial land cover from the NLCD contributed 25% of the calculated imperviousness, the
major highways contributed 6%, and quarrying and mining contributed 4%.
35
-------
25
O
w ,, -
15
- 10
LU
10
15
20
Measured Impervious Cover (%'
25
Figure 3.6 Impervious cover for Frederick County, MD watersheds measured from aerial photographs vs
that estimated from Multiple Data Sources, including U.S. Census population density, manufacturing and
industrial areas from categorized satellite imagery, and major highway networks from U.S. Department of
Transportation. Overall, this approach under estimated impervious cover by 1 %TIA.
The Multiple Data Source approach was then applied to the 13 North Georgia HUCs
(mentioned previously in Section 2). Comparison of estimated and measured %T1A values are
shown in Figure 3.7 with the straight line once again showing the one-to-one match. Overall, the
impervious area showed an overestimate of 2 %T1A (sd = 2). The greatest overestimate was 8
%TIA in one of the central Atlanta watersheds. Impervious cover was generally over estimated
somewhat greater in the higher impervious area watersheds (HUCs). The Multiple Data Source
approach was generally able to accurately reflect the wide range of %T1A values in this data set.
36
-------
60
£ 50
0)
o
O 40
Q.
_£
"°
CD
-1—•
OJ
.1
"S
LJJ
0
0
10 20 30 40 50
Measured Impervious Cover (%)
60
Figure 3.7 Impervious cover for 13 Atlanta area HUCs measured from aerial photographs vs that estimated
from Multiple Data Sources, including U.S. Census population density, manufacturing and industrial areas
from derived satellite imagery and major highway networks from U.S. Department of Transportation. For this
data set, impervious cover was over estimated on average by 2 %TIA.
3.4 Comparison of NLCD only and Multiple Data Source (MDS) Approach
Table 3.3 contains the measured impervious cover in the 13 Atlanta area watersheds
from both the 1993 and 1999 DOQQs, along with estimations from NLCD only data for 1993
and from the MDS method for 1993 and 1999. Since the NLCD is based on 1993 data, the 1993
set of aerial photography was excellent for evaluating and comparing the two estimation
methods. The two time windows were informative relative to change detection, since two of the
North Atlanta watersheds doubled in impervious cover during this time period. Both the NLCD
only and the MDS approach provided reasonable %T1A estimates for urbanized watersheds
(HUCs). For low impervious area watersheds, the MDS approach underestimated the
impervious area somewhat, similar to the Frederick County results. The NLCD data only
method underestimated impervious area even more significantly than the MDS method. Since
the MDS method relied on updated population data for the 1999 estimates, but only the 1993
37
-------
commercial/industrial area land cover contribution, there was a somewhat greater underestimate
for 1999 using the MDS method. By contrast, the MDS approach appeared to slightly over-
estimate the imperviousness in the very-developed, mid-town Atlanta watersheds (HUCs).
Table 3.3 Percent Total Impervious Area (%T1A) Results for North Georgia Watersheds
HUC number
031300011204
031300011202
031300011201
031300011002
031300011001
031300010907
031300010906
031501040301
031501040302
031501040303
031501040304
031501040305
031501040306
DOQQ
1993
52.1
35.8
33.8
8.6
6.1
21.0
10.5
1.6
3.4
3.7
3.6
5.4
2.0
NLCD
1993
44.9
31.6
31.8
6.5
3.4
20.7
11.3
1.9
1.9
1.9
2.0
2.0
1.8
Multiple Data
Sources- 1993
54.9
36.6
41.0
9.7
6.2
24.6
14.9
1.6
1.7
2.5
3.6
3.8
1.7
DOQQ
1999
49.1
32.3
34.1
15.8
9.5
24.4
22.4
2.0
5.1
5.5
7.9
8.4
3.9
Multiple Data
Sources- 1999
58.1
38.0
44.8
13,8
7.9
27.6
23,9
1.7
1.9
2.9
4.4
4.9
1.7
Impervious cover was subsequently estimated for 1624, 12 digit watersheds (HUCs)
wholly contained within the state of Georgia, using both the simple NLCD-only approach and
the MDS approach. The use of NLDC data with the ATtlLA landscape factor extension tool
provided a very rapid analysis and identified most of the potentially degraded watersheds (Table
3.4). The NLCD-only method identified 69 watersheds as having over 10% TIA whereas the
MDS approach identified 80. The NLCD-only method under estimated the number of
watersheds in the at-risk, 5 to 10 % TIA, range. For 1993, the MDS approach identified 117
HUCs in the 5 to 10% impervious class versus 76 for the NLCD only approach~35% fewer.
38
-------
Table 3.4 Evaluation of Impervious Cover Status of Georgia Watersheds/HUC's.
Impervious
Cover
Class
(% TIA)
0-5
5- 10
10-25
>25
NLCD Data
Only (1993)
(#of
watersheds)
1479
76
58
11
Multiple Data
Sources
(1993)
(#of
watersheds)
1427
117
62
18
Multiple Data
Sources
(1999)
(#of
watersheds)
1395
137
67
25
Change
(1993-1999)
from lower to
higher class*
-32
+32
+12
+7
High %TIA
Growth Rate
> 0.2 %TIA/
year (# of
watersheds)
12
19
36
13
HUC's (62 - 7) in the 10 to 25 class; however, during the same time period 12 HUC's moved from the 5 to 10 class
to the 10 to 25 class for a total of 67 (1999 MDS). The calculation for movement from the 0 to 5 class to the 5 to 10
class is similar.
Thus, the NLCD-only approach appears to have the most serious limitations for
identifying imperviousness in the 5 to 10% range. This range, particularly in areas with
significant growth, likely incorporates the most critical areas where prevention of storm water
problems might be most effective. Figure 3.8 identifies for 1993 the specific Georgia HUCs
categorized by MDS as 'of concern' (i.e. >5 %TIA) that were not identified by the NLCD-only.
It is important to remember that the MDS approach may underestimate these HUCs somewhat as
well.
Between 1993 and 1999, we estimated that a total of 51 HUCs changed to a higher risk
impervious cover category. Figure 3.9 shows that the majority of these watersheds were in the
Atlanta area. The largest change was 32 HUCs moving from the 0 to 5% class to the 5 to 10%
class. Appreciable imperviousness changes were also evident in the higher impervious classes,
with 12 HUCs moving from the 5 to 10% range to the 10 to 25% range and 7 HUCs from the 10
to 25% to the >25% range. For 1999, we estimated that there were a total of 229 HUCs of
concern, i.e. HUCs that are currently impaired or likely to be in the near future (14% of 1624):
92 (~6%) for likely existing impairment (imperviousness above 10%), and 137 (~8%) for likely
impairment in the near future (5 to 10% impervious range) if appropriate planning and
management is not undertaken. The expected result is increasing storm water hydrologic,
pollutant and habitat degradation stress on the streams in these areas.
39
-------
MDS Differences
01/05
20 40 60 SO Miles
U.S. Environmental Protection Agency. Ecosystems Research D if is ion. Athens, Georgia
Figure 3.8 Estimated 1993 %TIA for 1624 Georgia 12 digit HUCs. Fifty-two (52) HUCs
identified as at-risk (5-10% impervious) or potentially degraded (>10% impervious) using Multiple
Data Sources (MDS), but not identified using the land cover data alone, are outlined.
40
-------
Changes 1993 to 1 999
*Q SO
U S. Eiilroime ttal Plot etc t Aget
-------
4. Impervious Cover in the Southeastern United States
This section contains the estimated impervious cover results for U.S. EPA Region 4 in
2000 using the Multiple Data Source approach described in Section 3. Headquartered in
Atlanta, Georgia, U.S. EPA Region 4 includes Alabama, Florida, Georgia, Kentucky,
Mississippi, North Carolina, South Carolina and Tennessee. The Southeastern U.S. is one of the
fastest growing areas of the country, with Florida, Georgia and North Carolina in the top ten
growth states.
The major centers of population in the South are expanding, putting stress on all its
ecosystems, but this sprawl is especially harmful to coastal areas, wetlands and mountains. Just
10% of the earth's land surface holds the overwhelming majority of the earth's population along
or near coasts, and the United States is no exception. The Southeastern United States is growing
rapidly along its coasts. Over 20 million people live in 99 coastal counties along the U.S. Gulf
Coast (some of which are outside our study area). Florida's burgeoning population is clearly
evident, with the highest density along its eastern coast, but with significant population
expansion from Tampa southward along the western coast.
Using the Multiple Data Source Approach, Figure 4.1 shows the estimated impervious
cover by 12 or 14 digit HUC (depending on the individual states) for the year 2000 for the eight
Southeastern states. It is easy to see the growth around the cities and interstate corridors,
especially the Interstate 85 corridor from Atlanta to the Raleigh-Durham-Chapel Hill, North
Carolina area. The urban intensity along the Florida east coast is also particularly evident.
Table 4.1 summarizes estimated %TIA for 2000 by state for five %TIA categories,
providing a quick reference for each category for each Southeastern state. Streams in watersheds
with > 20 %TIA are seriously degraded, and even the most intensive remediation efforts are
likely to only partially restore functionality of those water bodies. While streams in the 10 to 20
%TIA category are also likely to suffer significant degradation from urbanization, remediation
efforts can potentially restore functionality to these streams. Streams in watersheds with 5 to 10
%TIA suffer only modest degradation due to urbanization and can benefit substantially if careful
planning and management of water resources is undertaken at that point in the development of
the watershed. Table 4.1 clearly illustrates the extremes in the extent in urbanization in the
Southeast. Whereas, only 0.1% of Mississippi's land area is contained in HUCs with >20 %TIA,
7.0% of Florida's HUCs are in this largely degraded category.
The maps for individual states (Figures 4.2 to 4.9) include labels of the Metropolitan
Statistical Areas (MSAs) in the state. An MSA is a statistical definition by the U.S. Census
Bureau to account for decentralized settlement and economic activity. It not only includes
urbanized areas and outlying urban places, but also surrounding counties that are integrated with
these urban centers as measured by substantial amounts of daily commuting, even if many of
these surrounding areas have densities far too low to be classified as urban. A MSA must
include at least one city with 50,000 inhabitants and a total metropolitan population of 100,000
or more. Approximately 82% of the U.S. population is contained in these MSAs (Kaiser
statehealthfacts.org, 2004).
42
-------
Table 4.1 % TIA as a Percentage of the Total Land Area of Each Southeastern State Using the 2000
Census and the Multiple Data Source Approach
State
Alabama
Florida
Georgia
Kentucky
Mississippi
North
Carolina
South
Carolina
Tennessee
Total #
of
water-
sheds
1414
1365
1865
1241
1114
1601
1031
1093
Area
in sq
miles
52197
58373
58754
40407
49409
52662
31145
42139
>20% TIA
# water-
sheds /
sq miles
127
402
1167
4071
497
1381
11 7
315
27
67
447
1008
397
513
287
811
%area
0.8
7.0
2.4
0.8
0.1
1.9
1.7
1.9
10 -20% TIA
# water-
sheds 7
sq miles
297
990
1047
4174
61 7
2369
247
905
157
612
101 7
2720
487
1232
407
1287
%area
1.9
7.2
4.0
2.2
1.2
5.2
4.0
3.1
5-10% TIA
# water-
sheds 7
sq miles
767
3112
1607
7053
1577
4614
91 7
2827
377
1427
1777
5769
1067
3181
767
2917
% area
6.0
12.1
7.9
7.0
2.9
11.0
10.1
6.9
2-5% TIA
# water-
sheds 7
sq miles
397 7
15426
3377
13818
5597
17660
4707
16244
2227
10138
628 7
20479
3627
11740
383 7
15885
% area
29.6
23.7
30.1
40.2
20.5
38.9
37.7
37.7
<2% TIA
# water-
sheds 7
sq miles
900 7
32267
648 7
29257
1039 7
32730
645 7
20116
838 7
37165
651 7
22686
4767
14547
566 7
21239
% area
61.8
50.1
55.7
49.8
75.2
43.1
46.7
50.4
(Figures may not add up to 100% due to rounding up or down.)
43
-------
70 140 210 280 Mies
U.S. Environmental Protection Agency
Athens, Georgia November 2004
Figure 4.1 Southeastern United States impervious cover for 2000. Impervious cover as %TIA (percent total
impervious area) by 12- and 14 digit HUC using the Multiple Data Source approach. Data sources used in the
calculation include 1993 NLCD commercial and industrial, 2000 Census data and U.S. DOT data for interstates
and other major highways.
44
-------
4.1 Alabama
The population of Alabama in 2000 was 4.4 million people, ranking it 23rd most
populous among the 50 states (U.S. Census Bureau, 2004). The state's population increased
10.1% in the decade 1990 to 2000, below the U.S. average of 13.1%. (U.S. Census Bureau,
2004). The metropolitan population for Alabama is 71%, below the national rate of 82% (Kaiser
statehealthfacts.org, 2004).
Figure 4.2 illustrates the impervious cover estimated by 12 digit HUC for Alabama in
2000. Impervious cover calculations used the Multiple Data Source approach described in
Section 3. The state's most extensive impervious cover is in and around its largest city and a
long-time industrial center in the south, Birmingham. The cities of Huntsville, Montgomery (the
capital), Mobile and to a lesser degree Tuscaloosa, Florence, Gadsen and Anniston also
contribute to urbanization in the state. For the most part, Alabama is very rural and contains less
impervious cover than many of the other Southeastern states. As in every state, there is more
impervious cover around the interstate highways, but the intensity is less than other Southeastern
states.
Alabama has 1414, 12 digit HUCs, 12 of which are >20% T1A, or 0.8% of its total land
area. Five of those > 20% are >30%, and are in the Birmingham and Mobile MSAs. Only one
watershed, in Birmingham, is greater than 40%TIA, at 44.7%. Alabama has 29 watersheds in
the 10 to 20% TIA category, or 1.9% of its area; 76 watersheds in the 5 to 10% T1A range, 6%
of its area; 397 watersheds in the 2 to 5% TIA range, or 29.6% of its area; and 900 watersheds <
2% TIA, 61.8% of its area. Although Alabama has one of the lowest %TIA in the Southeast
with only 41 watersheds and 8.7% of the state >10% TIA, the fragile coastal area around Mobile
has 10 HUCs greater than 10%.
45
-------
Huntsville MSA
Birmingham MSA
Montgomery MSA
Gulf of Mexico
60 Miles
U .S . Environmental Protection Agency
Athens, Georgia November 2004
Figure 4.2 Alabama impervious cover for 2000. Impervious cover as %TIA (percent total impervious area)
by 12 digit HUC calculated using the Multiple Data Source approach. Data sources used in the calculation
include 1993 NLCD commercial and industrial, 2000 Census data and U.S. DOT data for interstates and other
major highways.
46
-------
4.2 Florida
Florida is the fourth most populous state in the union with the official estimate as of
April 1, 2003 at approximately 17.1 million, a 6.8% increase in the first three years of this
century. The U.S. Census Bureau ranked nine counties in Florida among the 100 fastest growing
counties in the nation during April 1, 2000 to July 1, 2003 (U.S. Census Bureau, 2004). During
the 1990's Florida's population grew by 23.5%, during the 1980's by 32.7%. Net migration
continues to be the predominate pattern of growth for Florida, with 10.8% of the state's growth
due to natural increase and 89.2% due to net migration (Florida OCEDR 2004).
Although one of the largest and fastest growing populations in the United States, Florida
still contains low density areas of rural land dominated by agricultural uses and vast areas of
wetlands. The majority of the population crowds in and around the biggest cities and along the
coastal areas, providing typical sprawl problems and concerns. The metropolitan population
(percentage of the population located in MSAs) for Florida is 96%, the third largest of the 50
states and well above the U.S. average of 82% (Kaiser statehealthfacts.org, 2004). Florida is the
only state in the Southeast with a metropolitan population greater than the U.S. average.
Figure 4.3 illustrates the impervious cover by 12 digit HUC estimated for Florida in 2000
by the Multiple Data Source approach. Florida has 1365 12 digit HUCs, 116 of which are >20
%TIA, or 7% of its area. Twenty-nine of those >20 %TIA HUCs are >40%, accounting for
~1.6% of Florida's area, with 20 watersheds in the Miami-Ft. Lauderdale area between 45.2 and
89.7 %TIA. Out of the 87 remaining HUCs in the >20 to 40 %TIA range, all but 21 are in
coastal areas or sprawl from a coastal area, accounting for -5.3% of Florida's area. In the 10 to
20 %TIA range, Florida has 104 HUCs, accounting for 7.2% of its area. In the 5 to 10 %TIA
range, Florida has 160 HUCs, accounting for 12.1% of its area. Florida has the highest total of
impervious cover in the top 3 categories %T1A of our study, accounting for 26.3% of its area. In
the 2 to 5 %TIA range, Florida has 337 HUCs, a vast majority in the interior, accounting for
23.7% of its area. Florida has 648 watersheds <2 %TIA, the vast majority in Northern Florida,
the Everglades and the interior area just north of the Everglades that accounts for 50.1% of its
area. There are no watersheds below 2 %TIA on the eastern coast of Florida. Much of
Florida's coastal cities are becoming interconnected as the population grows, endangering the
fragile wetlands, coastal areas and various other water ecosystems around the state.
47
-------
Jacksonville MSA
Pensacola M SA
Panama City MSA
% TIA
Cape Coral MSA
Naples MSA
40
80
120
160 Miles
U.S.Environmental Protection Agency
Athens, G eorgia November 2004
Orlando MSA
Tampa-St. Petersburg-
Clearwater M SA
Atlantic
Ocean
Fort Pierce-
Port St. Lucie MSA
_West Palm
Beach-Boca
Raton MSA
Miami-
Fort Lauderdale
MSA
Figure 4.3 Florida impervious cover for 2000. Impervious cover as %TIA (percent total impervious area) by 12 digit HUC calculated
using the Multiple Data Source approach. Data sources used in the calculation include 1993 NLCD commercial and industrial, 2000 Census
data and U.S. DOT data for interstates and other major highways.
48
-------
4.3 Georgia
Georgia's 2000 population at 8.2 million ranks 10th in the nation. Twenty counties in
Georgia ranked in the 100 fastest growing counties in the nation during April 1, 2000 to July 1,
2003. Five counties were in the national top ten with growth rates above 20% in that three year
period (U.S. Census Bureau, 2004). The state's population increased 26.4% in the decade 1990
to 2000, more than double the U.S. average of 13.1%. (U.S. Census Bureau, 2004). The
population for Georgia in MSAs is 72%, below the national rate of 82% (Kaiser
statehealthfacts.org, 2004). Most of Georgia's urban population is in the 28 county metropolitan
statistical area (MSA) of Atlanta. This huge MSA spreads out in all directions, making
congestion and environmental concerns serious issues for the state.
Figure 4.4 illustrates the impervious cover by 12 digit HUC estimated for Georgia in
2000 by the Multiple Data Source approach. There are 1865, 12 digit HUCs in Georgia, with 49
watersheds with %T1A >20% that account for 2.4% of the area of the state. Forty of these 49
HUCs are in the Atlanta MSA. The incredible sprawl facing Atlanta area residents is readily
evident in Figure 4.4. Ten of the 49 >20 %TIA HUCs have >40 %T1A. There are 61 watersheds
in the 10 to 20 %T1A category, accounting for 4% of Georgia's area. The 5 to 10 %TIA range
contains 157 watersheds, accounting for 7.9% of Georgia's area. Georgia has 14.3% of its area
in the top three %T1A categories of this study. Scattered throughout the state are 559 watersheds
in the 2 to 5 %TIA range, accounting for 30.1% of Georgia's area. The remaining 1039
watersheds are in the <2 %T1A range, the majority of these are in the rural south and eastern
Georgia accounting for 55.7% of its area.
49
-------
Atlanta MSA
Augusta
Aiken (SC) MSA
Columbus
MSA
M aeon
MSA
Savannah
MSA
f
Atlantic
Ocean
% TIA
<2
2-5
5-10
10-20
>20
0
30
GO
90 Miles
U.S. Environmental Protection Agency
Athens, Georgia November2QQ4
Figure 4.4 Georgia impervious cover for 2000. Impervious cover as %TIA (percent total impervious area) by
12 digit HUC calculated using the Multiple Data Source approach. Data sources used in the calculation include
1993 NLCD commercial and industrial, 2000 Census data and U.S. DOT data for interstates and other major
highways.
50
-------
4.4 Kentucky
The 2000 Census lists Kentucky's population at slightly more than 4 million with a 9.6%
increase in the decade 1990 to 2000, well below the U.S. average of 13.1%. Migration is the key
component of growth in the Kentucky population, affecting the composition as well as the size
of the population. Kentucky has an aging population, with the rate of decadal natural increase of
it's population decreasing by 70% since 1960 to a current value of only 6% (Price,, et al. 2004).
The population in MSAs for Kentucky is 45%, nearly half the national rate of 82% (Kaiser
statehealthfacts.org, 2004). Kentucky's population is centered in the north central part of the
state, with the cities of Louisville, Lexington and suburban overflow population of Cincinnati,
Ohio, accounting for the majority of the metropolitan population.
Huge tracks of national forest and the eastern edge of the Eastern Kentucky Coal Field
(and Cumberland Plateau), called the Pottsville or Cumberland Escarpment and formed from
weathering of resistant sandstones and conglomerates, dominate the eastern portion of the state.
The escarpment is stepped in south-central Kentucky because several thick, resistant sandstone
layers are separated by less resistant shales. The manner in which the sandstones weather and
are eroded along the escarpment results in sheer cliffs, steep-walled gorges, rock shelters,
waterfalls, natural bridges and arches. The Eastern Kentucky topography and the karst
topography of middle Kentucky make intensive urban development difficult.
Figure 4.5 illustrates the impervious cover by 12 digit HUC estimated for Kentucky in
2000 by the Multiple Data Source approach. There are 1241, 12 digit HUCs in Kentucky, only
11 of which are >20% T1A and account for only 0.8% of the state's area. These 11 urban HUCs
are found exclusively in Louisville and Hamilton (Cincinnati, Ohio suburb). The 10 to 20 %T1A
range contains 24 watersheds and accounts for 2.2% of Kentucky's area. There are 91
watersheds in the 5 to 10 TIA% range, accounting for 7% of Kentucky's area. Only 10% of
Kentucky land area is in our top three %TIA categories where water quality impacts due to
urbanization are a concern. Scattered throughout the state are 470 watersheds in the 2 to 5 %TIA
range, accounting for 40.2% of Kentucky's area. The remaining 645 watersheds in Kentucky are
in the <2 %TIA range, accounting for 49.8% of its area.
51
-------
Cincinnati-Ham ilton
OH/KY/IN MSA
Lexington M SA
Evansville-
Henderson MSA
Louisville M SA
40
80
120 Miles
U .S . Environmental Protection Agency
Athens, G eorgia Novem ber 2004
Figure 4.5 Kentucky impervious cover for 2000. Impervious cover as %TIA (percent total impervious area) by 12 digit HUC calculated using the
Multiple Data Source approach. Data sources used in the calculation include 1993 NLCD commercial and industrial, 2000 Census data and U.S. DOT
data for interstates and other major highways.
52
-------
4.5 Mississippi
The 2000 Mississippi population is listed at approximately 2.8 million, making it the
least populated state in the Southeastern region. A 10.5% increase in population in the decade
1990 to 2000 puts Mississippi's growth rate below the U.S. average of 13.1% (U.S. Census
Bureau, 2004). The metropolitan population (percentage of population in MSAs) for Mississippi
is 34%, making it fourth lowest in the nation (Kaiser statehealthfacts.org, 2004). Mississippi's
urban population is centered around it's largest city, Jackson, the overflow from Memphis along
its northern border, and the booming coastal area of Gulfport and Biloxi.
Figure 4.5 illustrates the impervious cover by 12 digit HUC estimated for Mississippi in
2000 by the Multiple Data Source approach. Of the 1114, 12 digit HUCs in Mississippi, only 2
are >20% T1A, accounting for only 0.1% of its area. One of these >20 %TIA watersheds is
located in the Jackson MSA with 38.6 %T1A, and the other is located along the Gulf Coast with
22.1 %TIA. Even the next category, the 10 to 20 %T1A range, only contains 15 watersheds,
accounting for 1.2 % of Mississippi's area. There are 37 watersheds in the 5 to 10 %TIA range,
accounting for 2.9 % of Mississippi's area. In the 2 to 5 %T1A range, Mississippi has 222
watersheds, accounting for 20.5% of Mississippi's area. The remaining 838 watersheds in
Mississippi are in the <2 %TIA range, accounting for 75.2% of its area. Mississippi has the
lowest impervious cover of all the Southeastern states, with only 4.2% of its area in the top three
categories of %T1A. Unfortunately, approximately 90% of the area in the fragile ecosystems of
the Gulf Coast are in the top three categories of impervious cover (>5%) of this study.
53
-------
Memphis
(TN-AR-MS)
MSA
A
20 40 60 80 Miles
Jackson MSA
U .S . Environmental Protection Agency
Athens, Georgia November 2004
Biloxi--
Gulf Port--
Pascagoula M SA
Gulf of Mexico
Figure 4.6 Mississippi impervious cover for 2000. Impervious cover as %TIA (percent total impervious area)
by 12 digit HUC calculated using the Multiple Data Source approach. Data sources used in the calculation
include 1993 NLCD commercial and industrial, 2000 Census data and U.S. DOT data for interstates and other
major highways.
54
-------
4.6 North Carolina
The North Carolina 2000 population was slightly more than 8 million, with a 21.4%
increase in the decade 1990 to 2000, well above the U.S. average of 13.1%. Five counties in
North Carolina ranked in the 100 fastest growing counties in the nation during April 1, 2000 to
July 1, 2003 (U.S. Census Bureau, 2004). The metropolitan population (percentage of the
population living in MSAs) for North Carolina is 70%, below the national rate of 82% (Kaiser
statehealthfacts.org, 2004). The majority of the North Carolina population is concentrated in the
middle of the state, in clusters around their major metropolitan statistical areas-Charlotte-
Gastonia-Rock Hill, Raleigh-Durham-Chapel Hill, Greensboro-Winston-Salem-High Point and
Fayetteville. The western portion of North Carolina is mountainous and the eastern portion is
coastal, making North Carolina attractive for second homes, retirement and seasonal housing.
North Carolina was fourth in the nation in adding persons 65 and older to its numbers April 1,
2000 to July 1, 2003 (U.S. Census Bureau, 2004).
Figure 4.7 illustrates the state's impervious cover by 12 digit HUC estimated for 2000
using the Multiple Data Source approach. This figure clearly shows the extensive development
around its MSAs in the central part of the state and along interstates 85 and 40. There are 1601,
14 digit HUCs in North Carolina, 44 of which are >20 %T1A, accounting for 1.9% of its area.
There are 101 watersheds 10 to 20 %T1A accounting for 5.2% of North Carolina's area. The 5
to 10 %TIA range contains 177 watersheds, accounting for 11% of its area. There are 628
watersheds in the 2 to 5 %TIA range, accounting for 38.9% of North Carolina's area. The
remaining 651 watersheds are in the <2 %T1A range, accounting for 43.1% of its area, the lowest
amount for any Southeastern state. Although North Carolina does not have an extremely high
number of watersheds in the highest %TIA category, it does have a relatively high percentage of
its land area, 18.1%, in the top three %T1A categories, making it second highest in the Southeast.
55
-------
C ha rlotte-
G astonia--
Rock Hill MSA
Asheville MSA
Greensboro--
Win sto n-S alem --
High Point MSA
% T IA
BB<2
f;?:-y 2 - 5
5-10
10- 20
^B > 20
R ale igh -
D urham --
Chapel Hill MSA
W ilm in gton
MSA
Fayetteville MSA
Atlantic
0 ce an
30
60
90
120 150 Miles
U .S . Environm ental P rotection Agency
Athens, Ge orgia N ovem ber 2004
Figure 4.7 North Carolina impervious cover for 2000. Impervious cover as %TIA (percent total impervious area) by 14 digit HUC calculated using
the Multiple Data Source approach. Data sources used in the calculation include 1993 NLCD commercial and industrial, 2000 Census data and U.S.
DOT data for interstates and other major highways.
56
-------
4.7 South Carolina
The South Carolina 2000 population was 4 million, with a 15.1% increase in the decade
1990 to 2000, just slightly above the U.S. average of 13.1%. The metropolitan population
(percentage of the population in MSAs) for South Carolina is 75%, below the national rate of
82% (Kaiser statehealthfacts.org, 2004). South Carolina's population is much like the other
Southern states, expanding out from its biggest metropolitan area, the Anderson-Greenville-
Spartanburg MSA, along the Interstate 85 corridor. Other areas of growth in South Carolina
include its capitol, Columbia, and the coastal port of Charleston.
Figure 4.8 illustrates the South Carolina impervious cover by 14 digit HUC estimated for
2000 using the Multiple Data Source approach. South Carolina has 1031, 14 digit HUCs, 39 of
which are >20% TIA, representing 1.7% of its land area. Eighteen of those >20 %T1A are >30
%T1A and are located in Columbia, the Greenville-Spartanburg-Anderson MSA and in the
Atlantic coastal area. South Carolina has 48 watersheds in the 10 to 20% TIA range,
representing 4% of its land area; 106 watersheds in the 5 to 10% TIA range, or 10.1% of its area.
South Carolina's total area in our top three %TIA categories is 15.8%, third highest in the
Southeast. South Carolina's 362 watersheds in the 2 to 5% TIA range account for 37.7% of its
area, and the 476 watersheds <2 %TIA account for 46.7% of its area.
57
-------
A
Greenville-
Spartanburg--
Anderson MSA
Atlantic
Ocean
Colum bia M SA
Charleston-
North Charleston MSA
% TIA
20
40
60
80 Miles
U .S . E n vironm ental P rotection Agency
Athens, Georgia Novem ber 2004
Figure 4.8 South Carolina impervious cover for 2000. Impervious cover as %TIA (percent total impervious area)
by 14 digit HUC calculated using the Multiple Data Source approach. Data sources used in the calculation include
1993 NLCD commercial and industrial, 2000 Census data and U.S. DOT data for interstates and other major
highways.
58
-------
4.8 Tennessee
The Tennessee 2000 population was 5.7 million, with a 21.4% increase in the decade
1990 to 2000, about 60% greater than the U.S. average of 13.1%. The metropolitan population
(percentage of the population residing in MSAs) for Tennessee was 69%, below the national rate
of 82% (Kaiser statehealthfacts.org, 2004). The majority of the population in Tennessee is
centered around Memphis in the southwestern corner of the state, Nashville in central Tennessee,
and in the eastern portion of the state from Chattanooga to Knoxville and along a corridor
stretching along Interstates 75, 40 and 81 to Bristol, Tennessee at the Virginia border.
Figure 4.9 illustrates the estimated Tennessee impervious cover by 12 digit HUC for
2000 using the Multiple Data Source approach. Tennessee has 1093, 12 digit HUCs, 28 of which
are >20% T1A, or 1.9% of its area; 40 watersheds 10 to 20% T1A, or 3.1% of its area; 76
watersheds 5 to 10% T1A, or 6.9% of its area; 383 watersheds 2 to 5% T1A, or 37.7% of its area;
and 566 watersheds < 2% TIA, or 50.4% of its area. Tennessee has 11.9% of its total land area
in the top three %T1A categories of this study.
59
-------
A
J oh ns o n C ity-
K in gs p ort--
B risto I M SA
Kn oxville MSA
C h atta n oo g a MSA
40
40
80 Miles
U.S. Environm en ta I Protection Agency
Athe ns . ft eorg ia N O¥e m b er 2004
Figure 4.9 Tennessee impervious cover for 2000. Impervious cover as %TIA (percent total impervious area) by 12 digit HUC calculated using the
Multiple Data Source approach. Data sources used in the calculation include 1993 NLCD commercial and industrial, 2000 Census data and U.S. DOT
data for interstates and other major highways.
60
-------
5, Future Impervious Cover Projections
for the Southeastern United States
According to the Census Bureau population projections, several states in the Southeastern
United States will be among the fastest growing of the country in the next three decades. EPA's
Region 4 contains two states, Florida and Georgia, that are expected to have the third and fourth
largest population increase between 1995 and 2025 in the nation (Campbell, 1997). Florida is
expected to add over six million additional residents while Georgia and North Carolina are
expected to add over two million. Mississippi and Kentucky are projected to show the slowest
net population growth, adding just under 1A million individuals. Coastal areas will generally
show the greatest growth, with slower growth in the inland areas.
Along with this population growth will inevitably come an increase in impervious cover.
In this section projections of impervious cover in the!2 and 14 digit HUCs are made based on
county scale population projections obtained from the individual states combined with the
Multiple Data Source impervious cover estimation method discussed in Sections 3 and 4. These
projections represent a single scenario of a possible distribution of impervious cover in the eight
Southeastern states and should be used in this context. Included is a discussion of error
estimates of population projections upon which the imperviousness is based to give the reader a
sense of the magnitude of error associated with these projections.
5.1 The Nature of Errors in Population Projections
Estimates of potential error for population projections are not done routinely among the
community of demographers (Smith 1987). Such estimates are particularly desirable for this
impervious estimation study, however, since population is the key driving component of the
impervious area projections and estimates. Several existing population projection error studies
were surveyed to establish a basis for potential errors associated with impervious estimates and
projections. These studies spanned several spatial scales including: state, county and sub-
county areas. Population projection error estimates at all these scales are summarized here and
used in to illustrate the likely bounds of potential impervious area estimates based largely on
population projections.
Smith and Sincich (1992) carried out a comprehensive retrospective evaluation of state
level population projection errors using an array of simple extrapolation approaches compared to
each other and to more complex models. They evaluated the accuracy of state population
projections for the I960' s, 70' s and 80' s using: linear extrapolation (LINE); exponential
extrapolation (EXPO); a moving average time series model (ARIMA); ratio techniques including
shift-share (SHIFT), which assumes that a state's share of national population changes "...by the
same annual amount during the projection horizon as the average annual change during the base
period," and share of growth (SHARE), which assumes that "...each state's share of national
population growth during the projection horizon is the same as during the base period." Other
more complex models evaluated included those of the Census bureau, the National Planning
Association, and the Bureau of Economic Analysis. The measures of error that they evaluated
included: mean absolute percent error (MAPE), root mean square percent error (RMSPE), 90th
percentile of absolute percent errors (90PE), Mean Algebraic Percent Error (MALPE),
Percentage of positive errors (%POS)
The two best simple models were LINE and SHARE, with SHARE performing slightly
better. The simple models performed as well or better than the more complex approaches (Smith
& Sincich 1992). Typical MAPE ranges for these techniques were: 5 to 8% for 10 year
61
-------
projections, and 10 to 12% for 20 year projections. Thirty (30) year projections were not
evaluated. Here we assume that a reasonable MAPE for 30-year projections at the State level
might be in the range 15 to 18%. This "guess", however, is based only on limited experience
and best professional judgement and no hard data. Clearly, additional research is needed to
establish potential error for long term population projections over 20 years.
The U.S. Bureau of the Census and others, especially users of population projections, are
becoming increasingly interested in consistent, regular, sound evaluation of population
projection error. Campbell and others evaluated error in state population projections using
Census 2000 counts. They found that short term projections up to 5 years might expect MAPE's
of about 0.5% per year at the state level (Campbell, et al. 2002). These values are consistent
with the results of Smith and Sincich (1992) discussed above.
At the county scale a similar evaluation process found "...mean absolute errors of around
15% for 10-year projections and around 30% for 20-year projections" (Smith 1987). Smith
evaluated projections vs. census enumerations for 2,971 counties over the years 1950 through
1980. He observed that migration patterns dominated both increases and decreases in population
and that extremely high growth rates tend to moderate over time. He also established several
general characteristics of population projection accuracy:
• forecast errors increase with length of projection horizon,
• larger errors are expected for places with high growth rates,
• larger errors are expected for small (low population) places, and
• there is no way to predict whether errors will be positive or negative.
Since the errors in population projections depend on population (size of place) and on the
rate of population growth, "... population forecast errors are frequently going to be large,
especially for small and/or rapidly growing places" (Smith 1987). Recognizing this, Smith
suggested that population projection refinements could be based on separating counties by rate
of population growth (during the base period), by size of place (population) or both.
At the census tract scale a study of three diverse Florida counties found a MAPE in the
range of 17 to 20% for 10-year projections (Smith and Shahidullah 1995). These authors found
larger errors for smaller tracts, ~ 30 to 35% for <2,500 population over 10 years, and larger
errors for larger absolute growth rates. They also tested and found the smallest errors for a
composite (COMP) method where the estimation approach was tailored to population growth
characteristics of each census tract. The COMP errors for census tracts by growth rate for a 10-
year period were as follows:
Growth Rate COMP error
<10% -20%
10 to 25% -10%
25 to 50% -20%
>50% -30%
Using over 40,000 sub-county areas (municipalities, townships, etc.) nationwide, Harper
and others evaluated 10 year errors for projections using sub-county housing unit data to
distribute county populations to sub-county areas (Harper, et al. 2003). They found that error
(MAPE and MAPLE) depended on both sub-county area size (population) and growth rate.
62
-------
MAPE for all areas was 12.4% ranging from 4.0% (population 50,000 to 100,000) to 35.1%
(population < 100).
Projections of urban expansion tailored to specific locales can incorporate considerable
additional complexity and potentially more accuracy relative to the simpler approaches based on
population projections taken here. One example of a more detailed local projection of urban
growth is that for the Charleston region of South Carolina (Allen and Lu 2003) which predicts a
change in urban area from 250 square miles in 1994 to 868 square miles in 2030. This effort
utilized multiple approaches including logistic regression, rule based suitability (for transition
probabilities) and focus groups as the basis for an integrated future urban growth model.
Comprehensive accuracy assessments of these complex localized approaches have not been done
and no good guidelines exist to evaluate improvement in projection accurate as a function of this
increased complexity.
5.2 Impervious Cover Projection Method
Impervious cover projections based on an Multiple Data Source approach are analogous
to those described for current condition, but incorporate projected population growth as
described below. Additional details of calculation and data processing can be found in the
Appendix.
5.2.1 Residential Component
Although the U.S. Census Bureau prepares state level population projections for the
entire nation, individual states prepare projections of population at the county scale level. Both
methodology and available time period for projection vary from state to state. Some states in
Region 4 have official population projections until 2030, while other states have them only until
2010 or 2015. Population projections by county for the eight Southeastern states in Region 4
were obtained from each individual state. Table 5.1 summarizes information on county scale
projections for each state including the projection time period, date projections were made and
organization from which projections were obtained.
Since we are interested in projections for the 12 and 14 digit HUCs, the environmentally
significant subdivision, the coarser political (county) scale projections need to be apportioned to
the finer HUC scale. The first step was to apportion the growth to the 2000 census block level.
With this done the population could be apportioned to the individual blocks as described
previously for current condition estimation of impervious cover. Two different approaches were
considered to distribute county scale projections to the block level: (1) based on the most recent
growth at the block level, i.e. proportional to block level population change between 1990 and
2000, or (2) based on the most recent block level population density. Both of these approaches
have arguable advantages and disadvantages. The first approach assumes that the recent
historical level of growth will continue over the next 30 years while the second assumes that
population growth will be proportional to the current population. Neither of these methods can
be considered the "right" way to apportion the growth, but both can form the basis for making
reasonable projections.
Basing the distribution on growth patterns in the past ten years identifies recent boom
areas of a county. This method is likely to overestimate future growth for the very rapidly
growing areas of the counties and to underestimate growth for areas that are currently
experiencing slower growth and to continue contraction of populations unrealistically in some
63
-------
Table 5.1 Sources and Dates of Population Projections for Each Southeastern State
State
Alabama
Florida
Georgia
Kentucky
Mississippi
North
Carolina
South
Carolina
Tennessee
Source
Univ. of
Alabama
Univ. of
Florida
Gov's. Ofc.
of Planning
& Budget
Univ. of
Louisville
MS
Institutes of
Higher
Learning
NC Ofc. of
State Budget
& Mgmt.
SC State
Budget &
Control Bd.
Univ. of
Tennessee
Date of
Projections
August
2001
July 2003
June 2002
July 2003
March
2002
June 2003
October
2002
March
1999
Years Available
2005
X
X
X
X
X
X
X
2010
X
X
X
X
X
X
X
X
2015
X
X
X
X
X
X
X
2020
X
X
X
X
X
X
2025
X
X
X
X
X
2030
X
X
blocks that have had declines in population. Local peak growth or contraction areas will shift
over less than the 25 year period for most of the areas. A technical difficulty in trying to use this
approach is that census blocks were realigned between the 1990 and 2000 Census so determining
the growth individual blocks presents difficulties. A commercial product which offers 1990
Census data mapped onto the 2000 block configuration was evaluated for use in this process.
Unfortunately, this product shows some unrealistic expansion and contractions in some areas.
Blocks around large military bases seemed to be particularly problematic but other pockets of
problems also seem to exist with this data set. Some block populations are zeroed out and other
show unrealistically high increases in population between the two census periods. Doing
adequate quality control with this approach for a multiple state area did not appear feasible. This
approach may be possible to use for smaller multiple county areas. Due to the quality control
issues with method one, projected growth in a county was distributed based on the 2000
population in the blocks. Distributing growth proportional to the 2000 population will tend to
underestimate growth in rural areas of a county and overestimate growth in urban areas.
64
-------
The population projection errors at the HUC level due to their relatively small size will
be larger than county scale projection errors no matter how the population is distributed. For
example, the average of the HUC population in Georgia in 2000 (based on 8.2 million people
and 1865 HUCs) is only 4400 and there are an average of 12 HUCs per county in the state.
Based on the error in population projection analysis [Smith (1987), Smith and Shahidullah
(1995)] discussed in the previous section, smaller high growth rate HUCs could easily have
projection errors in the 50% range even for a 10 year projection horizon.
The projected population density in each HUC was calculated for the projection time
periods available from individual states. The projection window ranged from Georgia with
projections out only to 2010, to Kentucky and North Carolina with projections out to 2030.
Projections of HUC populations were made at five year intervals based on the individual state
projections. The residential component of impervious cover for each time period was then
calculated based on the Hicks equation as described in Section 3.
5.2.2 Commercial/Industrial Component
The critical assumption for the High Intensity Commerial/industrial (H1CI) area future
projections was to maintain in a particular HUC the HIC1 (in square miles) 710,000 population
ratio constant for future periods. This assumption would provide at least a reasonable estimate
of future H1C1 change due to population growth since the commercial/industrial contribution to
%T1A contains a major component that is proportional to the population density in the nearby
area, i.e. commercial areas that serve the residents of an area. Figure 5.1 illustrates the
relationship between population and the High Intensity Commercial/Industrial class for North
Carolina HUC's for the single time period -- 1993 -- when the categorized land cover data was
available.
The above assumption reflects both the linear nature and the dispersion of the relation-
ship between H1CI area and population. The implications of this assumption for the projection
are that the historical pattern of commercial growth with respect to population for any particular
area (HUC) will continue into the projection periods. Thus, areas with high HICFpopulation
ratios are projected to experience high commercial growth with population increases, areas with
low Hid/population ratios are projected to experience low commercial growth with population
increases, and those areas with intermediate ratios will be in between. Actual future change of
H1C1 area with population for any individual HUC could differ from this assumption, but lacking
historical change information for the HICl/population ratio, we consider this "space for time"
substitution approach for the assumption to be objective and sensible based on the available data.
As future coincident commercial/industrial area and population data become available, including
an updated NLCD for 2000, future censuses and other appropriate data, the actual reliability of
this assumption should be tested and evaluated.
65
-------
NC Commercial Area to Population Relationship
Y= 571753 + 169.051X
R-Sq=72.9 W
O
O
30000000 —
20QQQOQQ —
1DODOODO —
D —
* * * •
I
1DDDDD
POPULATION
Regression
95% Cl
r
200000
Figure 5.1 High Intensity Commercial/Industrial area (meters) vs. population for North Carolina 12 digit
HUC's
For the majority of HUCs, the projection was made as follows:
ComTlA%[period] = (YComm area[ 19931 * Bpop[period1) / HUC area) * 0.90 * 100
Where:
Comm_area[1993]
and
Bpop [period]
HUC^area
0.90
ComTIA% [period]
= NLCD H1C1 of HUC for 1993
= population of HUC for 1993
(interpolated between 1990 & 2000)
= population of HUC for projected period
(2000, 2010,2020 & 2030)
= area of HUC in square miles
= assumed fraction of T1A for H1C1 class of NLCD
= commercial/industrial component of TIA for projected period
However, a major industrial component in this category can be relatively independent of the
surrounding residential population, i.e., large manufacturing or transportation facilities maybe
located at a distance from local population centers. In addition, if population declines in an area
the impervious area does not tend to decline.
66
-------
Two constraints were implemented to maintain realistic ranges and distribution of
variability in future projections of the HICI (High Intensity Commercial/Industrial) TIA
component. First, decreases of HICI (High Intensity Commercial/Industrial area) for future time
periods are not considered likely and are not allowed, i.e.:
IF HICI(future period) 3.0
....THEN HICI(future period) = HICI( 1993) + Median HICI Change
Where: Median HICI Change
= 0.6 square mile * ((Population(period2) - Populati on (period!))/! 0,000
This constant is based on the median HICI area per 10,000 population ratio of all HUCs
statewide in North Carolina. (The mean ratio was 0.7.)
5.2.3 Major Highway Component
The only roadways included in this component were interstate highways and major US
highways. This is a minor component in the impervious cover estimation and very little new
connector highway construction is proposed. Projected construction information is in a variety
of formats and not easily obtainable from individual states. Since updated values will have a
negligible effects on projections, estimates of current status described in Chapter 4 was used in
projection estimates.
5.3 Impervious Cover Projections
The West and the South are projected to have the greatest net population change over the
next three decades in the nation. All of the states in the Southeast are expected to increase in
population in the following decades with Florida and Georgia leading the growth. Florida is
expected to replace New York as the third (behind California and Texas) most populous state by
2020 (Campbell, 1997). Population projections from each of the eight Southeastern states in
EPA Region 4 are show in Table 5.2 and form the basis for the impervious cover projections for
each state of the eight states detailed below.
67
-------
Table 5.2 2000 U.S. Census and State Population Projections for the Southeastern United States
in Thousands
State
Alabama
Florida
Georgia
Kentucky
Mississippi
North
Carolina
South
Carolina
Tennessee
2000
Population
4,447
15,982
8,186
4,042
2,845
8,049
4,012
5,689
2005
Population
4,645
17,499
~
4,183
2,991
8,784
4,155
5,798
2010
Population
4,839
18,978
9,592
4,321
3,118
9,491
4,388
6,063
2015
Population
5,028
20,387
~
4,447
3,227
10,227
4,618
6,327
2020
Population
5,211
21,807
~
4,563
~
10,966
4,850
6,593
2025
Population
5,386
23,178
~
4,663
~
11,712
5,077
~
2030
Population
--
--
--
4,744
--
12,448
--
--
68
-------
5.3.1 Alabama
In the 2000 Census Alabama was the 22nd most populous state with 4.4 million people
and is not expected to change its ranking significantly among the 50 states and District of
Columbia in the next 25 years (U.S. Census Bureau, 2001 a and b). Overall, growth will be
moderate with the population projected to grow by just under 1 million people reaching 5.4
million people by 2025, an increase of 21% over a 25 year period (Campbell, 1997).
Although overall population increase is projected to be less than 1% per year over the
next two decades, some counties are projected to grow much more rapidly while some will show
a decline in population. Shelby County, located in the Southeastern portion of the Birmingham
MSA is projected to add over 120,000 people, a percentage increase of 85% or over 3% annual
growth rate. Baldwin County located on the Gulf Coast on the eastern side of the Mobile MSA
is projected to add over a 100,000 people with a 77% increase over the 25 year period. Five
other counties (Autauga, Blount, Elmore, Lee, and St.Clair) are expected to show an increase in
population of over 50% by 2025. These counties are located on the outskirts of existing
population centers including Birmingham, Montgomery and Auburn. At the other end of the
spectrum, several rural counties distant from major population centers are projected to decline in
population. The population projections by county for the state of Alabama were obtained from
the Center for Business and Economic Research, University of Alabama. The projections,
available out to 2025, were published in August 2001. Projections in this series are based on
trends between the 1990 and 2000 censuses.
Figure 5.2 shows the impervious cover projections for Alabama watersheds calculated
using the Multiple Data Source approach described in Section 4 and Section 5.2. The number of
HUCs and percentage land area in each impervious cover class are summarized in Table 5.3 with
the pattern of percent area by TIA class shown in Figure 5.3 The number of stream miles in
each %TIA category is summarized in Table 5.4 with the pattern of stream miles by TIA class
shown in Figure 5.4.
The increase in impervious cover between 2000 and 2025 can be seen in the far south
Gulf Coast and around the other urban centers. Figure 5.3 shows a drop in the area of HUCs
with <2% impervious and increases in the other four classes. In 2000, 2.7 % of the land area
(1409 mi2) was in HUCs with %TIA > 10% (areas where stream quality is likely degraded). By
2025, 3.5% of the land area (1827 mi2) was in HUCs with %T1A > 10%. By 2025, 3.0 % (2294
mi) of Alabama streams are projected to be in HUCs with %TIA > 10% while 91 % (70,116 mi)
are in areas not immediately threatened by urbanization with < 5% impervious area.
69
-------
2010
0 40 80 120 160
U.S. Environmental Protection Agency Athens, Georgia November 2004
Figure 5.2 Alabama projected impervious cover out to 2025. Impervious cover as
%TIA (percent total impervious area) by 12 digit HUC calculated using the Multiple Data
Source approach. Data sources used in the calculation include 1993 NLCD commercial and
industrial cover, 2000 Census data, county level population projections from University of
Alabama and U.S. DOT data for interstates and other major highways.
70
-------
Table 5.3 %TIA as a Percentage of the Total Land Area of Alabama out to 2025
Year
2000
2010
2020
2025
>20% TIA
#
HUCs
12
15
19
21
%
area
0.8
1.0
1.3
1.4
10-20% TIA
#
HUCs
29
31
31
31
%
area
1.9
2.0
2.1
2.1
5-10% TIA
#
HUCs
76
78
90
92
%
area
6.0
6.2
7.0
7.0
2-5% TIA
#
HUCs
397
430
435
443
%
area
29.6
31.8
32.2
32.7
<2% TIA
#
HUCs
900
860
839
827
% area
61.8
59.0
57.5
56.8
Total num
Total area
berofHUCs: 1414
52,197.5 sq mi
QUO.
SD-
jj
Figure 5.3 Alabama Projected %TIA as % of Area out to 2025
71
-------
Table 5.4 Total River Miles in Alabama by %TIA category out to 2025
Year
2000
2010
2020
2025
>20% T1A
#
HUCs
12
15
19
21
Total number ot
Total river miles
river
miles
420
579
783
882
HUCs: 1
77,389
10-20%T1A
#
HUCs
29
31
31
31
river
miles
1225
1314
1329
1412
5-10%TlA
#
HUCs
76
78
90
92
river
miles
4320
4452
5049
4977
2-5%TlA
#
HUCs
397
430
435
443
river
miles
21654
23472
23815
24216
<2% T1A
#
HUCs
900
860
839
827
river
miles
49769
47572
46411
45900
414
60000
5HBD •
40000 .
.& 30000 .
1
o
H
20000 •
10000 .
20 +
Figure 5.4 Total River Miles in Alabama by %TIA Category out to 2025
72
-------
5.3.2 Florida
Florida is currently the 4th most populous state in the U.S. with nearly 16 million people
counted in the 2000 census (U.S. Censure Bureau, 2001). According to the U.S. Census Bureau,
Florida is projected to replace New York as the 3rd most populous state in the union and by 2025
the population is projected to grow to 20.7 million people (Campbell, 1997). The University of
Florida Population Program, on the other hand, projects even more rapid growth for the state
with a total of 23.2 million people by 2025 (Smith and Nogle, 2003). These projections refer
solely to permanent residents and do not include tourists or seasonal residents, a major category
in some areas of Florida. Florida is a highly diverse state and growth is not distributed evenly.
While some areas have grown rapidly, others have grown slowly or even declined in population.
Rapid growth is expected along the northeastern Atlantic coast with St. Johns (St.
Augustine area) and Flagler County (south of St. Augustine) doubling in population between
2000 and 2025. Sumter County (southwest of Ocala) is projected to double in population while
Marion County located west of Ocala will nearly double as well. Collier County on the Gulf
Coast south of Fort Meyers is also expected to double in population. Miami-Dade is expected to
add 800,000 more people, with its population reaching 3.1 million by 2025. Counties
surrounding Orlando are expected to grow rapidly as well. Some rural areas of the state such as
the area south and east of Tallahassee will experience slow or no growth.
The population projections by county for the state of Florida were obtained from the
Bureau of Economic and Business Research, Warrington College of Business Administration,
University of Florida, Gainesville, Florida. The projections, available out to 2025, were
published July 2003. These county scale projections served as the basis for the projection of
future impervious cover for the state of Florida
Figure 5.5 shows the impervious cover estimations and projections for Florida
watersheds calculated using the Multiple Data Source approach described in Section 4 and
Section 5.2. The number of HUCs and percentage land area in each impervious cover class are
summarized in Table 5.5 with the pattern of percent area by T1A class shown in Figure 5.6. The
number of stream miles in each %TIA category is summarized in Table 5.6 with the pattern of
stream miles by T1A class shown in Figure 5.7.
In 2000, 14.2 % of the land area (8289 mi2) was in the 220 HUCs with %T1A > 10%
(areas where stream quality is likely degraded). By 2025, 21.3% of the land area (12434 mi2)
was projected to be in the 298 HUCs with %T1A > 10%. By 2025, 26% (15341 mi) of Florida
streams are projected to be in HUCs with %T1A > 10% while 63% (36598 mi) are in areas not
immediately threatened by urbanization with < 5% impervious area. Between 2000 and 2025,
4900 more miles of streams, a 45% increase, will be located in watersheds likely to suffer
serious degradation due to development (>10 %T1A) unless advanced planning and mitigation
efforts are undertaken soon. These watersheds are located along both the Atlantic and Gulf
coastal areas and in the central Florida expanding east and southwest of Orlando.
73
-------
2000
010
% TIA
<2
}::g^, 2 - 5
5- 10
10-20
>20
0 60 120 180 Miles
U.S. Environmental Protection Agency Athens, Georgia November 2004
Figure 5.5 Florida projected impervious cover out to 2025. Impervious cover as %TIA (percent total impervious
area) by 12 digit HUC calculated using the Multiple Data Source approach. Data sources used in the calculation
include 1993 NLCD commercial and industrial cover, 2000 Census data, county level population projections from
University of Florida and U.S. DOT data for interstates and other major highways.
74
-------
Table 5.5 % TIA as a Percentage of the Total Land Area of Florida out to 2025
Year
2000
2010
2020
2025
>20% TIA
#
HUCs
116
133
142
152
%
area
7.0
7.9
8.5
9.6
10-20% TIA
#
HUCs
104
127
144
146
%
area
7.2
8.6
10.1
11.7
5-10% TIA
#
HUCs
160
142
142
147
%
area
12.1
11.4
11.4
9.9
2-5% TIA
#
HUCs
337
351
352
342
%
area
23.7
25.2
25.0
24.1
<2% TIA
#
HUCs
648
612
585
578
% area
50.0
47.0
45.1
44.7
Total number of HUCs: 1365
Total area: 58,376.2 sq mi
D2000 02010
• 2CEO IZG5
<2
2D+
Figure 5.6 Florida Projected %TIA as % of Area out to 2025
75
-------
Table 5.6 Total River Miles in Florida per TIA category out to 2025
Year
2000
2010
2020
2025
>20% TIA
^
HUCs
116
133
142
152
Total number ot
Total river miles
river
miles
5346
5787
6090
7048
HUCs: 1
57,953
10-20% TIA
#
HUCs
104
127
144
146
river
miles
5101
6398
6885
8293
5- 10% TIA
#
HUCs
160
142
142
147
river
miles
8034
7025
7870
6011
2-5% TIA
#
HUCs
337
351
352
342
river
miles
12008
13357
12536
12209
<2% TIA
ft
HUCs
648
612
585
578
river
miles
27462
25384
24569
24389
365
3QQOO
2SOOQ -
330GQ -
1SOCD -
loom -
SQQD -
D2QOQ D2010
• 2020 "2025
20 +
Figure 5.7 Total River Miles in Florida by %TIA Category out to 2025
76
-------
5.3.3 Georgia
According to the U.S. Census Bureau (Campbell, 1997), by 2025 Georgia is projected to
be the 9th most populous state (it was ranked 10th in the 2000 census) and it is projected to rank
4th largest in net growth between 1995 and 2025. Between 1990 and 2000, the Georgia
population increased from 6.5 million to 8.2 million (U.S. Census Bureau, 2001). The Atlanta
MSA added 1.2 million people between 1990 and 2000 (U.S. Census Bureau, 2003) accounting
for approximately 70% of the net population growth in the state. The Atlanta MSA experienced
a 38.4% rate of growth while the state as a whole grew by 26.4%.
The State of Georgia, Governor's Office of Planning and Budget estimates the state's
population will grow to 9.6 million by 2010, an addition of 1.4 million people between 2000 and
2010. Counties in the Atlanta MSA will continue to dominate the growth in the state in this ten
year period. While population growth will plateau in the central Atlanta counties of Fulton and
DeKalb, counties in the outer ring are projected to grow rapidly with several counties including
Cherokee, Forsyth, Henry and Newton projected to grow by more than 50% in a ten year period.
The population projections by county for the state of Georgia were obtained from the
Planning, Research, & Evaluation Division, Governor's Office of Planning and Budget, Atlanta,
Georgia. The projections, currently available only to 2010, were published June 2002. These
county scale projections served as the basis for the projection of future impervious cover for the
state of Georgia.
Figure 5.8 shows the impervious cover estimations and projections for Georgia
watersheds calculated using the Multiple Data Source approach described in Section 4 and
Section 5.2. The number of HUCs and percentage land area in each impervious cover class are
summarized in Table 5.7 with the pattern of percent area by TIA class shown in Figure 5.9. The
number of stream miles in each %T1A category is summarized in Table 5.8 with the pattern of
stream miles by TIA class shown in Figure 5.10.
Watersheds in Georgia with increasing impervious cover will be located primarily in the
greater Atlanta metropolitan area. In 2000, 6.2% of the land area (3760 mi2) was in the 110
HUCs with %T1A >10% (areas where stream quality is likely degraded). By 2010, 7.5% of the
land area (4407 mi2) was projected to be in the 133 HUCs with %T1A >10%. In Georgia,
HUCs in the 5 to 10 %TIA category will increase from 157 to 169 between 2000 and 2010.
Streams in watersheds in the 5 to 10 %TIA category are vulnerable to degradation if any
additional growth occurs, although not necessarily severely degraded at that level of %TIA.
By 2010, 7.1% (5026 mi) of Georgia streams are projected to be in HUCs with %TIA >
10% while 84% of the streams (59590 mi) are in areas not immediately threatened by
urbanization with < 5% impervious area. Between 2000 and 2010, 799 more miles of streams
will be located in watersheds likely to suffer serious degradation due to development (>10
%TIA) unless advanced planning and mitigation efforts are undertaken.
77
-------
2000
% TIA
SO 100 Miles
U.S. Environmental Protection Agency Athens, Georgia November 2004
Figure 5.8 Georgia impervious cover out to 2010. Impervious cover as %TIA (percent total
impervious area) by 12 digit HUC calculated using the Multiple Data Source approach. Data sources
used in the calculation include 1993 NLCD commercial and industrial cover, 2000 Census data,
county level population projections from Georgia Governor' s Office of Planning & Budget and U.S.
DOT data for interstates and other major highways.
78
-------
Table 5.7 % TIA as a Percentage of the Total Land Area of Georgia
Year
2000
2010
>20% TIA
#
HUCs
49
58
%
area
2.4
2.8
10-20% TIA
#
HUCs
61
75
%
area
4.0
4.7
5-10% TIA
#
HUCs
157
169
%
area
7.9
8.3
2-5% TIA
#
HUCs
559
579
%
area
30.1
31.2
<2% TIA
#
HUCs
1039
984
%
area
55.7
53.0
Total number of HUCs: 1865
Total area: 58,754.1
20+
Figure 5.9 Georgia Projected %TIA as % of Area out to 2010
79
-------
Table 5.8 Total River Miles in Georgia per TIA category
Year
2000
2010
>20% TIA
#
HUCs
49
58
Total number of
Total river miles
river
miles
1464
1735
HUCs: 1
70,966
10-20% TIA
#
HUCs
61
75
river
miles
2763
3291
5- 10% TIA
#
HUCs
157
169
river
miles
5833
6350
2-5% TIA
#
HUCs
559
579
river
miles
22671
23599
<2% TIA
#
HUCs
1039
984
river
miles
38235
35991
865
3Q+
Figure 5.10 Total River Miles in Georgia by %TIA Category out to 2010
80
-------
5.3.4 Kentucky
According to the U.S. Census Bureau (Campbell, 1997), Kentucky is projected to be one
of the slowest growing states in the nation between 1995 and 2025. A landlocked state at the
northern edge of Region 4, Kentucky's growth patterns are likely to be more similar to its
midwestern neighbors to the north and west than the booming South Atlantic states. The
University of Louisville, Urban Studies Institute (2003) projects an increase in Kentucky's
population from 4.0 million in 2000 to 4.7 million in 2030. Growth in the state will occur
primarily around Lexington, Louisville, and in the suburban areas south of Cincinnati, OH.
Population projections by county for the state of Kentucky were obtained from the
Kentucky State Data Center, University of Louisville, Urban Studies Institute, Louisville,
Kentucky http://ksdc.louisville.edu/kpr/pro/pro2002.htm. The projections, available out to 2030,
were published July 2003.
Figure 5.11 shows the impervious cover estimations and projections for Kentucky
watersheds calculated using the Multiple Data Source approach described in Section 4 and
Section 5.2. The number of HUCs and percentage land area in each impervious cover class are
summarized in Table 5.9 with the pattern of percent area by TIA class shown in Figure 5.12.
The number of stream miles in each %T1A category is summarized in Table 5.10 with the
pattern of stream miles by TIA class shown in Figure 5.13.
Inspection of Figure 5.11 shows relatively little change between 2000 and 2030. In 2000,
3.0 % of the land area (1212 mi2) were in the 35 HUCs with %TIA >10% (areas where stream
quality is likely degraded). By 2030, 3.7% of the land area (1495 mi2) was projected to be in the
42 HUCs with %TIA >10%. HUCs in the 5 to 10 %TIA category will increase from 91 to 105
between 2000 and 2030. Streams in watersheds in the 5 to 10 %TIA category are vulnerable to
degradation if any additional growth occurs, although not necessarily severely degraded at that
level of %TIA.
By 2030, 2.9% (1464 mi) of Kentucky streams are projected to be in HUCs with %TIA >
10% while 89% of the streams (43828 mi) are in areas not immediately threatened by
urbanization with < 5% impervious area. In 2030, only 273 more miles of streams than in 2000
are likely to be located in watersheds where they will suffer serious degradation due to
development (>10 %TIA) unless advanced planning and mitigation efforts are undertaken. An
additional 3877 stream miles are projected to be in areas with 5 to 10% TIA. Streams in
watersheds in the 5 to 10 %TIA category are vulnerable to degradation if any additional growth
occurs although not necessarily severely degraded at that level of %TIA.
81
-------
2000
0 30 60 90 Miles
U.S. Environmental Protection Agency Athens, G eorgia November 2004
Figure 5.11 Kentucky impervious cover out to 2030. Impervious cover as %TIA (percent total
impervious area) by 12 digit HUC calculated using the Multiple Data Source approach. Data sources used
in the calculation include 1993 NLCD commercial and industrial cover, 2000 Census data, county level
population projections from University of Louisville and U.S. DOT data for interstates and other major
highways.
82
-------
Table 5.9 % TIA as a Percentage of the Total Land Area of Kentucky out to 2030
Year
2000
2010
2020
2030
>20% TIA
#
HUCs
11
11
12
13
%
area
0.8
0.8
0.9
1.0
10-20% TIA
#
HUCs
24
26
29
29
%
area
2.2
2.4
2.6
2.7
5-10% TIA
#
HUCs
91
96
99
105
%
area
7.0
8.2
8.2
9.1
2-5% TIA
#
HUCs
470
483
499
506
%
area
40.2
40.5
42.4
42.4
<2% TIA
#
HUCs
645
625
602
588
% area
49.8
48.1
45.9
44.9
Total number HUCs: 1241
Total area: 40,407.4
2 +
Figure 5.12 Kentucky Projected %TIA as % of Area out to 2030
83
-------
Table 5.10 Total River Miles in Kentucky per TIA category out to 2030
Year
2000
2010
2020
2030
>20% TIA
#
HUCs
11
11
12
13
Total number ot
Total river miles
river
miles
302
302
356
371
HUCs: 1
49,169
10-20% TIA
$
HUCs
24
26
29
29
river
miles
889
970
1070
1093
5- 10% TIA
#
HUCs
91
96
99
105
river
miles
3230
3483
3509
3877
2-5% TIA
#
HUCs
470
483
499
506
river
miles
18904
19374
20011
20075
<2% TIA
#
HUCs
645
625
602
588
river
miles
25846
25041
24223
23753
241
Figure 5.13 Total River Miles in Kentucky by %TIA Category out to 2030
84
-------
5.3.5 Mississippi
Mississippi is currently the 31st most populous state with a population of 2.8 million in
the 2000 Census. During the period 1990 to 2000 Mississippi was ranked 33rd by numeric
population change and was ranked 24th by percentage population growth (U.S. Census Bureau,
2001). According to U.S. Census Bureau projections for the period 1995 to 2025, Mississippi
will be one of the slowest growing states in Region 4 (only Kentucky is projected to grow
slower) and is ranked 34th nationally by percentage population change (Campbell, 1997).
The population projections by county for the state of Mississippi were obtained from the
Center for Policy Research and Planning, Mississippi Institutions of Higher Learning, Jackson,
Mississippi, www.ihl.state.ms.us. The projections, available out to 2015, were published March
2002. While overall growth in Mississippi is projected to be moderate, growth in two counties is
projected to have an annual average growth of approximately 2%. In the northeast corner of the
state, Desoto County located in the Memphis MSA is projected to have the highest growth rate
in the state. Madison County, a suburban county in the Jackson MSA, has the second highest
projected growth rate. Moderate growth is also projected for the counties along the Gulf Coast
(Hancock, Harrison, and Jackson). Declining populations are projected along the western border
of the state south of the influence of Memphis (Washington, Issaquena, Sharkey, Humphreys).
Figure 5.14 shows the impervious cover estimations and projections for Mississippi
watersheds calculated using the Multiple Data Source approach described in Section 4 and
Section 5.2. The number of HUCs and percentage land area in each impervious cover class are
summarized in Table 5.11 with the pattern of percent area by T1A class shown in Figure 5.15.
The number of stream miles in each %T1A category is summarized in Table 5.12 with the
pattern of stream miles by T1A class shown in Figure 5.16.
Inspection of Figure 5.14 shows little change between 2000 and 2015. The number of
HUCs with >20% TIA is projected to increase from 2 to 4 HUCs while the number in the 10 to
20 %TIA category is projected to increase to 17 from 15 (Table 5.11). These changes are
located along the Gulf Coast, south of Memphis, and near Jackson.
In 2000, 1.3 % of the land area (642 mi2) was in the 17 HUCs with %TIA >10% (areas
where stream quality is likely degraded). By 2015, 1.9% of the land area (938mi2) was projected
to be in the 21 HUCs with %TIA >10%. HUCs in the 5 to 10 %TIA category will increase from
37 to 42 between 2000 and 2015. Streams in watersheds in the 5 to 10 %TIA category are
vulnerable to degradation if any additional growth occurs, although not necessarily severely
degraded at that level of %TIA.
By 2015, 1.7% (1494 mi) of Mississippi streams are projected to be in HUCs with %TIA
>10% while 89% of the streams (81610 mi) are in areas not immediately threatened by
urbanization with < 5% impervious area. In 2015, 376 more miles of streams than in 2000 are
likely to be located in watersheds where they will suffer serious degradation due to development
(>10 %TIA) unless advanced planning and mitigation efforts are undertaken. By 2015, a total
of 2652 stream miles are projected to be in areas with 5 to 10% TIA.
85
-------
2000
2010
0 30 60 90 Miles
U.S. Environmental Protection Agency
Athens, Georgia November 2004
A
Figure 5.14 Mississippi impervious cover out to 2015. Impervious cover as %TIA (percent total impervious
area) by 12 digit HUC calculated using the Multiple Data Source approach. Data sources used in the
calculation include 1993 NLCD commercial and industrial cover, 2000 Census data, county level population
projections from Mississippi Institutes of Higher Learning and U.S. DOT data for interstates and other major
highways.
86
-------
Table 5.11 % TIA as a Percentage of the Total Land Area of Mississippi out to 2015
Year
2000
2010
2015
>20% TIA
#
HUCs
2
3
4
%
area
0.1
0.3
0.4
10-20% TIA
#
HUCs
15
18
17
%
area
1.2
1.6
1.5
5-10% TIA
#
HUCs
37
40
42
%
area
2.9
3,0
3,2
2-5% TIA
#
HUCs
222
244
249
%
area
20.5
23.7
24.0
<2% TIA
#
HUCs
838
809
802
% area
75.2
71.6
71.1
Total number of HUCs: 1114
Total area: 49,409.3
80.0
00
5-10
F* ram, TIA
ID-3D
20+
Figure 5.15 Mississippi Projected %TIA as % of Area out to 2015
87
-------
Table 5.12 Total River Miles in Mississippi per T1A category
Year
2000
2010
2015
>20% T1A
#
HUCs
2
3
4
Total number ot
Total river miles
river
miles
92
165
259
HUCs: 1
85,756
10-20%T1A
#
HUCs
15
18
17
river
miles
1026
1328
1235
5-10%TlA
#
HUCs
37
40
42
river
miles
2428
2497
2652
2-5%TIA
#
HUCs
222
244
249
river
miles
17605
20671
20988
<2% T1A
#
HUCs
838
809
802
river
miles
64604
61094
60622
114
fOOCO
SHJQ-
SXfll-
= JfflBO-
S 3QOQO-
3MOO-
20 +
Figure 5.16 Total River Miles in Mississippi by %TIA Category out to 2015
88
-------
5.3.6 North Carolina
North Carolina is currently the 11th most populous state with just over 8.0 million people
in the 2000 census. Between 1990 and 2000 North Carolina's population increased 21.4%,
adding 1.4 million and making it the 9th fasting growing state in the nation on a percentage basis
and 6' ranked by numeric population change (U.S. Census Bureau, 2001). Between 1995 and
2025 the U.S. Census Bureau projects North Carolina to be ranked 7th of the 50 states in net
increase in population (Campbell, 1997).
The population projections by county for the state of North Carolina were obtained from
North Carolina Office of State Budget and Management, Raleigh, North Carolina
http://www.osbm.state.nc.us/osbm/index.html. The projections, available out to 2030, were
published June 2003. Overall, North Carolina is predicted to have a net gain of 4.4 million
people between 2000 and 2030, a 55% increase in population over the 30 year period. Several
counties located in three areas of the state will show an increase in population greater then 70%
during the 30 year period. A set of counties north, south and east of Raleigh (Franklin, Wake,
Johnston, Harnett, Hoke and Sampson) are projected to show growth rates from 74% to 123%.
Four high growth counties surround Charlotte (Mecklenburg, Cabarrus, Iredell and Union), all
with a projected population increase in excess of 80%. Southern coastal counties of Fender and
Brunswick are predicted to show growth rates of 85% and 78%, respectively. Declining popula-
tions are projected in a few counties in northeastern North Carolina including Hertford, Bertie,
Edgecombe and Washington with other counties in the area showing lower than average growth.
Figure 5.17 shows the impervious cover estimations and projections for North Carolina
watersheds calculated using the Multiple Data Source approach described in Section 4 and
Section 5.2. The number of HUCs and percentage land area in each impervious cover class are
summarized in Table 5.13 with the pattern of percent area byTIA class shown in Figure 5.18.
The number of stream miles in each %T1A category is summarized in Table 5.14 with the
pattern of stream miles by T1A class shown in Figure 5.19. Inspection of Figure 5.19 shows
watersheds with increasing impervious cover located throughout the state with the greatest
increase located around Raleigh and Charlotte, along the 1-85 corridor between these two cities,
and along a corridor between Raleigh and Fayetteville. Watersheds increasing in impervious
cover are also in evidence along the southern coastline.
Between 2000 and 2030 the number of HUCs with >20% T1A is projected to double
from 44 to 89 HUCs, while the number in the 10 to 20 %TIA category is projected to increase to
130 from 101. In 2000, 7.1% of the land area (3739 mi2) was in the 145 HUCs with %TIA >
10% (areas where stream quality is likely degraded). By 2030, 11.7% of the land area (6161
mi2) was projected to be in the 219 HUCs with %TIA >10%. HUCs in the 5 to 10 %TIA
category will increase from 177 to 231 between 2000 and 2030. Streams in watersheds in the 5
to 10 %TIA category are vulnerable to degradation if any additional growth occurs, although not
necessarily severely degraded at that level of %T1A.
By 2030, 11.6% (7931mi) of North Carolina streams are projected to be in HUCs with
%TIA >10%, while 73% of the streams (50096 mi) are in areas not immediately threatened by
urbanization with < 5% impervious area. In 2030, 3272 more miles of streams than in 2000 are
likely to be located in watersheds where they will suffer serious degradation due to development
(>10 %TIA) unless advanced planning and mitigation efforts are undertaken. By 2030, a total
of 10251 stream miles are projected to be in areas with 5 to 10% TIA. Streams in watersheds in
the 5 to 10 %T1A category are vulnerable to degradation if any additional growth occurs,
although not necessarily severely degraded at that level of %T1A.
89
-------
m 2-5
'0-20 2020
U.S. Environmental Protection Agency Athens, Georgia November 2004
Figure 5.17 North Carolina impervious cover out to 2030. Impervious cover as %TIA (percent
total impervious area) by 14 digit HUC calculated using the Multiple Data Source approach. Data
sources used in the calculation include 1993 NLCD commercial and industrial cover, 2000 Census
data, county level population projections from North Carolina Office of State Budget & Control
Board and U.S. DOT data for interstates and other major highways.
90
-------
Table 5.13 % TIA as a Percentage of the Total Land Area of North Carolina out to 2030
Year
2000
2010
2020
2030
>20% TIA
#
HUCs
44
62
78
89
%
area
1.9
2.9
3.9
4.4
10-20% TIA
#
HUCs
101
115
120
130
%
area
5.2
6.1
6.6
7.3
5-10% TIA
#
HUCs
177
204
226
231
%
area
11.0
13,0
14.2
14.7
2-5% TIA
#
HUCs
628
642
638
642
%
area
38.9
39.3
39.6
39.8
<2% TIA
#
HUCs
651
578
539
509
%
area
43,1
38.7
35.8
33,8
Total number of HUCs: 1601
Total area: 52,662.7
20+
Figure 5.18 North Carolina Projected %TIA as % of Area out to 2030
91
-------
Table 5.14 Total River Miles in North Carolina per TIA. category out to 2030
Year
2000
2010
2020
2030
>20% TIA
#
HUCs
44
62
78
89
Total number of
Total river miles
river
miles
1213
1800
2442
2864
HUCs: 1
68,278
10-20% TIA
#
HUCs
101
115
120
130
river
miles
3446
4139
4574
5067
5-10% TIA
#
HUCs
177
204
226
231
river
miles
7684
9259
9846
10251
2-5% TIA
#
HUCs
628
642
638
642
river
miles
27845
28123
28515
28668
<2% TIA
#
HUCs
651
578
539
509
river
miles
28090
24957
22901
21428
601
SDOQ
3GDOQ-
23XD-
2DOQ-
13)00-
1DDOO-
3XD-
2-5
5-10
hunt IK
B-2D
30 +
Figure 5.19 Total River Miles in North Carolina by %TIA Category out to 2030
92
-------
5.3.7 South Carolina
In the 2000 census, South Carolina was ranked 26th by population and between 1990 and
2000 was ranked 19th by numeric population change and had a 15.1% growth in the population
(U.S. Census, 2001). Between 1995 and 2025, the U.S. Census Bureau projected South Carolina
to be the 20lh ranked by percent increase in population (Campbell, 1997).
The population projections by county for the state of South Carolina were obtained from
South Carolina State Budget and Control Board, Office of Research and Statistics, Health and
Demographics Division, http://www.ors2.state.sc.us/pODulation/Droiections.asp. The
projections, available out to 2025, were published October 2002. The state of South Carolina
projects an increase of approximately 1.0 million people between 2000 and 2025 with a total
population of 5.1 million projected by 2025. The highest growth in the state is in the coastal
counties with Horry (Myrtle Beach) and Beaufort showing the briskest pace of growth with a
projected increase of 58% and 61%, respectively, over the 25 year period. Growth in inland
areas is primarily in counties along the 1-20 and 1-85 corridors.
Figure 5.20 shows the impervious cover estimations and projections for South Carolina
watersheds calculated using the Multiple Data Source approach described in Section 4 and
Section 5.2. The number of HUCs and percentage land area in each impervious cover class are
summarized in Table 5.15 with the pattern of percent area by T1A class shown in Figure 5.21.
The number of stream miles in each %T1A category is summarized in Table 5.16 with the
pattern of stream miles by T1A class shown in Figure 5.22.
Inspection of Figure 5.20 shows watersheds with increasing impervious cover located
throughout the state with the greatest increase located along the coast, particularly in the
northern coastal areas near Myrtle Beach. In the northern inland area increased impervious
cover can be seen along the 1-85 corridor and in the area south of Charlotte, NC. Increases are
also evident in the central part of the state in Columbia area.
Between 2000 and 2025, the number of HUCs with >20% TIA is projected to increase
from 39 to 59 HUCs while the number in the 10 to 20 %TIA category is projected to increase to
56 from 48. In 2000, 5.7 % of the land area (1775 mi2) were in the 87 HUCs with %TIA >10%
(areas where stream quality is likely degraded). By 2025, 8.2% of the land area (2554 mi2) was
projected to be in the 115 HUCs with %TIA >10%. HUCs in the 5 to 10 %TIA category will
increase from 106 to 114 between 2000 and 2025. Streams in watersheds in the 5 to 10 %TIA
category are vulnerable to degradation if any additional growth occurs, although not necessarily
severely degraded at that level of %TIA.
By 2025, 8.3% (2926 mi) of South Carolina streams are projected to be in HUCs with
%TIA >10% while 80% of the streams (28284 mi) are in areas not immediately threatened by
urbanization with < 5% impervious area. In 2025, 866 more miles of streams than in 2000 are
likely to be located in watersheds where they will suffer serious degradation due to development
(>10 %TIA) unless advanced planning and mitigation efforts are undertaken. By 2025, a total of
4122 stream miles are projected to be in areas with 5 to 10% TIA.
93
-------
2000
2020
0 30 60 90 120 Miles
U.S. Environmental Protection Agency Athens, Georgia November 2004
Figure 5.20 South Carolina impervious cover out to 2025. Impervious cover as %TIA (percent total impervious
area) by 14 digit HUC calculated using the Multiple Data Source approach. Data sources used in the calculation
include 1993 NLCD commercial and industrial cover, 2000 Census data, county level population projections from
South Carolina State Budget & Control Board and U.S. DOT data for interstates and other major highways.
94
-------
Table 5.15 % TIA as a Percentage of the Total Land Area of South Carolina
Year
2000
2010
2020
2025
>20% TIA
#
HUCs
39
45
54
59
%
area
1.7
2.0
2.6
2.9
10-20% TIA
#
HUCs
48
52
54
56
%
area
4.0
4.5
4.8
5.3
5-10% TIA
#
HUCs
106
107
108
114
%
area
10.1
10.4
10.8
11.5
2-5% TIA
#
HUCs
362
366
380
378
%
area
37.7
38.8
39.6
39.1
<2% TIA
#
HUCs
476
461
435
424
% area
46.7
44.4
42.3
41.4
Total number of HUCs: 1031
Total area: 31,144.8
Figure 5.21 South Carolina Projected %TIA as % of Area out to 2025
95
-------
Table 5.16 Total River Miles in South Carolina per TIA category
Year
2000
2010
2020
2025
>20% TIA
#
HUCs
39
45
54
59
river
miles
553
653
906
986
10-20% TIA
#
HUCs
48
52
54
56
river
miles
1507
1714
1811
1940
5-10% TIA
#
HUCs
106
107
108
114
river
miles
3734
3819
3952
4122
2-5% TIA
#
HUCs
362
366
380
378
river
miles
13697
14230
14438
14190
<2% TIA
#
HUCs
476
461
435
424
river
miles
15841
14916
14226
14094
Total number of HUCs: 1031
Total river miles: 35,332
16110
160DQ-
Figure 5.22 Total River Miles in South Carolina by %TIA Category out to 2025
96
-------
5.3.8 Tennessee
In the 2000 Census, Tennessee was ranked 16th in the nation based on total population
with 5.7 million people. Tennessee was ranked 12th in the nation based on total population
increase between 1990 and 2000 adding 0.8 million people, a 16.7% increase in population (U.S.
Census Bureau, 2001). According to the U.S. Census Bureau, Tennessee is expected to be 13lh
in the nation ranked by net increase in population between 1995 and 2005 and 19th in the nation
based on percent population increase for this time period (Campbell, 1997).
The population projections by county for the state of Tennessee were prepared by the
Center for Business and Economic Research, College of Business Administration, The
University of Tennessee, Knoxville, Tennessee http://bus.utk.edu/cber/census/tnpopdat.htm.
The projections, available out to 2020, were published March 1999. Based on these state
projections, Tennessee's population will increase to 6.6 million by 2020. Rapid growth is
projected to focus primarily in the north central portion of the state in counties surrounding
Nashville. Nashville and eight surrounding counties are projected to account for over 40% of
the increase in population in Tennessee during the projection period.
Figure 5.23 shows the impervious cover estimations and projections for Tennessee
watersheds calculated using the Multiple Data Source approach described in Section 4 and
Section 5.2. The number of HUCs and percentage land area in each impervious cover class are
summarized in Table 5.17 with the pattern of percent area by T1A class shown in Figure 5.24.
The number of stream miles in each %T1A category is summarized in Table 5.18 with the
pattern of stream miles by T1A class shown in Figure 5.25.
Inspection of Figure 5.23 shows watersheds with increasing impervious cover located
primarily in counties surrounding Nashville. Increases in impervious cover are also evident in
the eastern part of the state along the 1-40 corridor from Chattanooga to Knoxville and Johnson
City.
Between 2000 and 2020, the number of HUCs with >20% T1A is projected to increase
from 28 to 36 HUCs, while the number in the 10 to 20 %T1A category is projected to increase to
48 from 40. In 2000, 5.0 % of the land area (2106 mi2) was in the 68 HUCs with %TIA >10%
(areas where stream quality is likely degraded). By 2020, 6.5% of the land area (2739 mi2) was
projected to be in the 84 HUCs with %T1A >10%. HUCs in the 5 to 10 %T1A category will
increase from 76 to 98 between 2000 and 2020. Streams in watersheds in the 5 to 10 %TIA
category are vulnerable to degradation if any additional growth occurs, although not necessarily
severely degraded at that level of %TIA.
By 2020, 5.6% (3610 mi) of Tennessee streams are projected to be in HUCs with %T1A
>10% while 85.6% of the streams (54743 mi) are in areas not immediately threatened by
urbanization with < 5% impervious area. In 2020, 867 more miles of streams than in 2000 are
likely to be located in watersheds where they will suffer serious degradation due to development
(>10 %T1A) unless advanced planning and mitigation efforts are undertaken. By 2020, a total of
5628 stream miles are projected to be in areas with 5 to 10% TIA.
97
-------
2000
A
2010
2020
% TIA
60
120
180 Miles
U.S. Environmental Protection Agency
Athens, Georgia November 2004
Figure 5.23 Tennessee impervious cover out to 2020. Impervious cover as %TIA (percent total impervious area) by
12 digit HUC calculated using the Multiple Data Source approach. Data sources used in the calculation include 1993
NLCD commercial and industrial cover, 2000 Census data, county level population projections from University of
Tennessee and U.S. DOT data for interstates and other major highways.
98
-------
Table 5.17 % TIA as a Percentage of the Total Land Area of Tennessee out to 2020
Year
2000
2010
2020
>20% TIA
#
HUCs
28
32
36
%
area
1.9
2.2
2.4
10-20% TIA
#
HUCs
40
43
48
%
area
3.1
3.5
4.1
5-10% TIA
#
HUCs
76
94
98
%
area
6.9
8.8
9.3
2-5% TIA
#
HUCs
383
392
412
%
area
37.7
38.2
40.0
<2% TIA
#
HUCs
566
532
499
% area
50.4
47.2
44.0
Total number of HUCs: 1093
Total area: 42,139.4
Figure 5.24 Tennessee Projected %TIA as % of Area out to 2020
99
-------
Table 5.18 Total River Miles in Tennessee per TIA category
Year
2000
2010
2020
>20% TIA
#
HUCs
28
32
36
river
miles
994
1143
1226
10-20% TIA
#
HUCs
40
43
48
river
miles
1740
1995
2384
5- 10% TIA
#
HUCs
76
94
98
river
miles
4151
5392
5628
2-5% TIA
#
HUCs
383
392
412
river
miles
24025
24280
25492
<2% TIA
#
HUCs
566
532
499
river
miles
33072
31173
29251
Total number of HUCs:
Total river miles: 63,981
093
urn-
33000'
+2Q
Figure 5.25 Total River Miles in Tennessee by %TIA Category out to 2020
100
-------
5.4 Using the Impervious Cover Projections
As discussed in Section 5.1, population projections for small (12 or 14 digit HUC scale)
and rapidly growing places will be quite large. The population projections alone in HUCs in high
growth areas can have errors in the 30 to 50% range for a 20 year time horizon. Impervious cover
calculations for future periods will add additional errors in addition to the very substantial errors
associated with small scale population projections. A good assessment of impervious cover
projections can only come as a retrospective analysis similar to what has been done at multiple
scales for population projections discussed in Section 5.1.
The projections of impervious cover are not meant to be used as definitive forecasts of the
future state of a specific watershed, but rather as a plausible scenario to identify where to look for
potential impairments and begin timely prevention efforts. These projections can give state and
local planners and resource managers a reasonable assessment of the magnitude of the problems
the states need to prepare for in the upcoming decades and can be used to guide monitoring to
identify problems as they begin to surface. Stream remediation is very expensive. The North
Carolina Ecosystem Enhancement Program stream reported costs for restoration/rehabilitation of
urban streams in 2004 to be $201.00 per foot of stream length (Jurek 2004). Well-timed and
targeted prevention and management actions can avoid the need for at least some of these very
expensive remediation expenditures in the future. Spatial tools, including the impervious cover
projections for the Southeastern states presented in this section, can aid in targeting these
prevention and management activities.
101
-------
6. Conclusions and Recommendations
Complete identification and eventual prevention of urban water quality problems pose
significant monitoring and water quality management challenges. The purpose of the methods
and analyses discussed in this report was to provide tools to assist decision makers in meeting
these challenges. The Multiple Data Source (MDS) wide area impervious estimation and
projection techniques can assist in meeting these challenges by providing: 1) cheap estimates of
impervious cover at the watershed and sub-watershed scales; 2) a region-wide approach to
screening for waters likely impaired or threatened by urban storm water; and 3) projections of
change in imperviousness over time.
The point sampling approach to aerial photo interpretation of imperviousness supplied an
essential, cost-effective, independent assessment for both the MDS and the NLCD only
estimation techniques and identified the appropriate uses for these two approaches. The use of
the NLCD data with the ATtlLA tool identifies most watersheds that are likely suffering severe
impairment from urbanization and allows a very rapid assessment. Unfortunately, this tool is not
as robust in identifying watersheds whose condition may be in a borderline category and
vulnerable to impairment in the near future. The MDS technique provides a more reliable
method for identifying watersheds impaired by urbanization compared to the use of
land-use/land-cover alone, especially for watersheds in the 5 to 10% impervious range where
prevention of storm water problems is critical.
The region-wide MDS impervious area estimates provide a screening tool for designing
water quality monitoring programs by identifying areas for priority monitoring for urban and
urbanizing watersheds. State monitoring programs have limited resources and thus cannot sample
everywhere. This landscape screening process provides workable, defensible methods to:
extrapolate condition estimates to waters lacking in-stream data; identify suspected problem areas
(likely impaired waters); and efficiently target additional monitoring to confirm problems.
The current (year 2000) impervious area estimates of this study identify specific
watersheds/HUC' s where existing adverse impacts due to impervious surfaces are likely (the 10
to 20% and >20% impervious classes). Some urban streams in these watersheds are listed as
impaired through Section 303(d) of the Clean Water Act and are subject to TMDL development.
Many potentially degraded waters are not yet listed, however, primarily due to a lack of
systematic monitoring approaches to identify urban water quality problems. Using the results
presented in this study, streams in watersheds/HUCs with imperviousness exceeding 10% that are
not already listed under the 303(d) impaired waters listing process for sediment and biological
integrity impairment should be prioritized for monitoring to ascertain if they are in fact impaired.
Prevention is critical. Stream channels de-stablized by excessive urban storm water runoff
from impervious surfaces continue to erode for many decades (or longer) (Hammer 1972), have
little potential to recover naturally and can be restored only with great difficulty and expense
(Rosgen 1994). Successful rehabilitation and restoration of streams in urbanized watersheds will
require complex, holistic approaches and should follow the sequence of: 1) hydrology, 2)
channel and habitat, 3) riparian zones, and 4) aquatic biological communities, recommended by
the National Research Council (National Research Council 1992, and Brosnan, et al. 1999). The
future impervious area projections of this study highlight the high growth areas of the Southeast,
and the specific watersheds/HUC's where this growth will be most likely to occur. These are the
very areas where effective storm water management and prevention of urban storm water impacts
are likely to be most cost effective. These same areas should be carefully considered to receive
increased attention for storm water education for local leaders and the public, and to institute
state-of-the-art storm water management practices. HUCs currently within the 5 to 10%
102
-------
imperviousness range and projected to experience growth in the next decade need to be the
highest priority focus of educational efforts and proactive storm water management actions to
prevent water quality degradation.
A three to four fold increase of urban area in the Southeast over the next 40 years need not
result in the widespread destruction of our streams, a resource vital to every community's quality
of life. If we focus now on the importance of imperviousness to future stream health, we can
avoid totally unnecessary storm water degradation of streams, and put those waters already
impacted back on the road to recovery.
Additional research will be needed to describe and explain differences in sensitivity to
impervious cover and hydrologic storm water stress in different areas. A number of geographic
frameworks should be tested to evaluate the variation in response to hydrologic stress from
impervious areas including: ecoregions and subecoregions (McMahon, et al. 2001); hydrologic
landscapes (Winter 2001); and average hydrologic response (Woodruff and Hewlett 1970).
Relationships to in-stream response need to be refined for the Piedmont ecological region and
developed using existing and new data for other areas of the Southeast. This might be done
region by region (such as the Blue Ridge, Piedmont, and upper/lower coastal plains in North
Carolina) using appropriate biological data sets encompass ing the full gradient of imperviousness,
or by using more generalized techniques such as multi-variate analyses incorporating critical
physical factors likely to drive stream channel and sediment processes (and thus influence habitat
and biological responses) such as relief, slope, soil properties (erodability in particular) and
riparian vegetation. A potential advantage of the latter approach is that resulting response models
might be applicable to much broader regions. The USGS' s series of "urban gradient" studies,
which are gathering both landscape and in-stream data for a variety of urban areas around the
nation, as well as existing state biological data networks, should provide useful data for building
reliable empirical response models. Such efforts could provide valuable information to help
understand variations in response to imperviousness and other urban stresses. Since some
impervious areas are not directly connected to streams and other waters, work is also needed to
incorporate cost-effective estimates of effective impervious area into storm water planning
(Sutherland 1995 and Alley and Veenhuis 1983).
Tools to estimate impervious area and in-stream response attack just one of many stresses
associated with urban expansion (Karr 1999). Practical screening tools are also needed for
nutrient and upland sediment loading (Jones, et al. 2001, and Wickham, et al. 2002), bacterial
contamination (Mallin, et al. 2000) and for pesticide/herbicide contamination.
103
-------
References
Allan, J.D. 1995. Steam ecology. Chapman & Hall, London, England.
Allan, J.D. 2004. Landscapes and Riverscapes: The Influence of Land Use on Stream
Ecosystems. Annual Review of Ecology, Evolution, and Systematics. 35:257-284.
Allen, J. and K. Lu. 2003. Modeling and prediction of Future Urban Growth in the Charleston
Region of South Carolina: a GIS-based Integrated Approach. Conservation Ecology. 8(2):2.
[online] URL: http://www.consecol.org/vol8/iss2/art2
Alley, W.A. and I.E. Veenhuis. 1983. Effective impervious area in urban runoff modeling.
Journal of Hydrological Engineering. ASCE. 109(2):313-319.
Brosnan, T., B.K. Ferguson, R. losco, G. Moll, T. Schueler, R. Sotir, and R. Watson. 1999.
Mitigation of urban runoff impacts on Atlanta streams. Proceedings of the 1999 Georgia Water
Resources Conference. March 30-31,1999. Kathryn J. Hatcher, ed. Institute of Ecology,
University of Georgia. Athens, G A. pp. 158-161.
Anderson, J.F., 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. Department of the Interior,
Geological Survey. Professional Paper 28.
Arnold, C.L., Jr. and C.J. Gibbons. 1996. Impervious Surface Coverage: The Emergence of a
Key Environmental Indicator. Journal of the American Planning Association 62(2): 243-258.
Bergstedt, L.C. and E.P. Bergersen. 1997. Health and movements offish in response to sediment
sluicing in the Wind River, Wyoming. Canadian Journal of Fisheries and Aquatic Sciences 54:
312-319.
Bird, S.L., S.W. Alberty and L.R. Exum. 2000. Generating High Quality Impervious Cover
Data. Journal of Quality Assurance. 8:91-103.
Booth, D.B. and C.R. Jackson. 1994. Urbanization of aquatic systems - degradation thresholds
and the limits of mitigation. Effects of Human-Induced Changes on Hydrologic Systems -
American Water Resources Association. June 1994. pp. 425-434.
Borgman, L.E. and W.F. Quimby. 1988. Sampling for Tests of Hypothesis When Data Are
Correlated in Space and Time. In Principles of Environmental Sampling, edited by L. H. Keith.
Salem, MA: American Chemical Society.
Boward, D., P. Kazyak, S. Stranko, M. Kurd and A. Prochaska. 1999. From the mountains to the
sea: The state of Maryland's freshwater streams. U.S. Environmental Protection Agency. Office
of Research and Development. EPA/903/R-99/023. 54pp.
Brosnan, T., Ferguson, B.K., losco, R., Moll, G., Schueler, T., Sotir, R., and Watson, R. 1999.
Mitigation of urban runoff impacts on Atlanta streams. Proceedings of the 1999 Georgia Water
Resources Conference. March 30-31, 1999. Kathryn J. Hatcher, ed. Institute of Ecology,
University of Georgia. Athens, GA. pp. 158-161.
104
-------
Brown, M.T. and M.B. Vivas. 2004. Landscape Development Intensity Index. Environmental
Monitoring and Assessment. In Press.
Campbell, P. 1997. Current Population Reports, Population Projections: States, 1995-2025. U.S.
Census Bureau, U.S. Department of Commerce, Economics and Statistics Administration. P25-
1131.
Campbell, P.R. 2002. Evaluating forecast error in state population projections using Census
2000 counts. Population Division Working Paper Series No. 57. U.S. Bureau of the Census.
14pp.
Caraco, D., R. Claytor, P. Hinkel, H.Y. Kwon, T. Schueler, C. Swann, S. Vysotsky and J.
Zielinske. 1998. Rapid Watershed Planning Handbook: A comprehensive guide for managing
urbanizing watersheds. Center for Watershed Protection. Ellicott City, Maryland.
Center for Watershed Protection. 2003. Impacts of Impervious Cover on Aquatic Systems.
Watershed Protection Research Monograph No. 1. Ellicott City, MD.
Chaloud, D.J., C.M. Edmonds and D.T. Heggem. 2001. Savannah River Basin landscape
analysis. U.S. Environmental Protection Agency. Office of Research and Development.
Washington, DC. EPA/600/R-01/069.
Corbett, C.W., M. Wahl, D.E. Porter, D. Edwards and C. Moise. 1997. Nonpoint source runoff
modeling: A comparison of a forested watershed and an urban watershed on the South Carolina
coast. Journal of Experimental Marine Biology and Ecology. 213(1):133-149
Dartiguenave, C.M.and D.R. Maidment 1997. Water quality master planning for Austin. Center
for Research in Water Resources (CRWR). University of Texas at Austin, Bureau of Engineering
Research, Austin, Texas. CRWR Online Report 97-6.
http://www.crwr.utexas.edu/reports/pdf/1997/rpt97-6.pdf
Ebert, D.W. and T.G. Wade. 2000. Analytical Tools Interface for Landscape Assessments
(ATtlLA) User Guide: Version 2.0. Office of Research and Development. U.S. Environmental
Protection Agency. Las Vegas, NV. 23pp.
Environmental Systems Research Institute (ESRI). 1996. Using ArcView GIS. ESRI, Inc.,
Redlands, CA.
Florida Office of Economic and Demographic Research, "Demographic Information for
Members and Staff", accessed February 2004, www.state.fl.us/edr/.
Gaff, W.L. 2001. Damage control: Restoring the physical integrity of America's rivers. Annals
of the Association of American Geographers. 91:1, 1-27.
Georgia Office of Planning & Budget (OPB). June 2002. Policy, Planning and Technical
Support, State of Georgia OPB, Atlanta, Georgia.
Gillies, R.R., J.B. Box, J. Symanzik, and E.J. Rodemaker. 2003. Effects of urbanization on the
aquatic fauna of the Line Creek watershed, Atlanta: a satellite perspective. Remote Sensing of
Environment. 86:3. 411-422.
105
-------
Goetz, S.J., R.K. Wright, A.J. Smith, E. Zinecker and E. Schaub. 2003. IKONOS imagery for
resource management: Tree cover, impervious surfaces, and riparian buffer analyses in the
mid-Atlantic region. Remote Sensing of Environment. 88:1-2. 195-208.
Graham, P. H., L. S. Costello and H.J. Mallon. 1974. Estimation of Imperviousness and Specific
Curb Length for Forecasting Storm water Quality and Quantity. Journal Water Pollution Control
Federation. 46(4):717-725.
Griffith, G.E., J.M. Omernik, J.A. Comstock, M.P. Schafale, W.H. McNab, D.R. Lenat, T.F.
MacPherson, J.B. Glover and V.B. Shelburne. 2002. Ecoregions of North Carolina and South
Carolina (color poster with map, descriptive text, summary tables and photographs). Reston,
Virginia. U.S. Geological Survey (map scale 1:1,500,000).
Griffith, J.A., S.V. Stehman and T.R. Loveland. 2003. Landscape Trends in Mid-Atlantic and
Southeastern United States Ecoregions. Environmental Management. 32(5):572-588.
Griffith, J.A., E.A. Martinko, J.L. Whistler and K.P. Price. 2002. Interrelationships among
landscapes, NDV1, and stream water quality in the U.S. Central Plains. Ecological Applications.
12(6): 1702-1718.
Groffman, P.M., J.J. Bain, L.E. Band, K.T. Belt, G.S. Brush, J.M. Grove, R.V. Pouyat, 1C.
Yesilonis and W.C. Zipperer. 2003. Down by the riverside: urban riparian ecology. Frontiers in
Ecology and the Environment. 1 (6): 315-321.
GVS&DD (Greater Vancouver Sewerage and Drainage District). 1999. Assessment of current
and future GVS&DD area watershed and catchment conditions, url: <
http://www.gvrd.be.ca/sewerage/stormwater_reports_1997_2002/assessment/main_rpt/assessmen
t_watersheds_catchments.pdf > last accessed March 7, 2005.
Hammer, T.B. 1972. Stream channel enlargement due to urbanization. Water Resources
Research. 8:1530-1540.
Harper, G., Coleman, C., and Devine, J. 2003. Evaluation of 2000 subcounty population
estimates. Population Division Working Paper Series No. 70. U.S. Bureau of the Census. 24pp.
Harrison, J., P. Hayward and B. Walker. 2001. Lullwater Fork Improvement Project (Atlanta,
GA USA): Integrating Innovative Urban Watershed, Hydrology and Planning Approaches with
"the Usual" Monitoring. Proceedings of the 6th National Volunteer Monitoring Conference:
Moving Into the Mainstream. April 26-29, 2000. Austin, TX. US Environmental Protection
Agency. EPA 841-R-01-001. pp. 129-134.
Harrison, J.E. 1998. Key water quality monitoring questions: Designing monitoring and
assessment systems to meet multiple objectives. Proceedings of the National Water Quality
Monitoring Council National Conference. Reno, NV. pp. Ill-175-187.
Hicks, R.W.B. and S. D. Woods. 2000. Pollutant Load, Population Growth and Land Use.
Progress: Water Environment Research Foundation. 11: p. 10 url: <
http://www.werf.org/press/fallOO/OOflplut.cfm> last accessed March 10, 2005.
106
-------
Holland, F., D.M. Sanger, C.P. Gawle, S.B. Lerberg, M.S. Santiago, G.H.M. Riekerk, L.E.
Zimmerman and G.I. Scott. 2004. Linkages between tidal creek ecosystems and the landscape
and demographic attributes of their watersheds. Journal of Experimental Marine Biology and
Ecology. 298(2):151-178.
Hulse, D. and R. Ribe. 2000. Land Conversion and the Production of Wealth. Ecological
Applications. 10(3):679-681.
Inman, E.J. 2000. Lagtime relations for urban streams in Georgia. U.S. Geological Survey.
Water Resources Investigations Report 00-4049. Atlanta, GA. 12 pp.
Inman, E.J. 1995. Flood-frequency relations for urban streams in Georgia - 1994 update. U.S.
Geological Survey. Water Resources Investigations Report 95-4017. Atlanta, GA. 27 pp.
Jennings, D.B. and S.T. Jarnagin. 2002. Changes in anthropogenic impervious surfaces,
precipitation and daily streamflow discharge: a historical perspective in a mid-atlantic
subwatershed. Landscape Ecology. 17(5):471-489.
Jennings, David B., S.T. Jarnagin and D.W. Ebert. 2004. A Modeling Approach for Estimating
Watershed Impervious Surface Area from National Land Cover Data 92. Photogrammetric
Engineering & Remote Sensing. 70(11): 1295-1307.
Ji, M. and J.R. Jensen. 1999. Effectiveness of subpixel analysis in detecting and quantifying
urban imperviousness from Landsat Thematic Mapper Imagery. Geocarto International.
14(4):33_41.
Jones, K.B., A.C. Neale, M.S. Nash, R.D. Van Remortel, J.D. Wickham, K.H. Riitters and R.V.
O'Neill. 2001. Predicting nutrient and sediment loadings to streams from landscape metrics: A
multiple watershed study from the United States Mid-Atlantic region. Landscape Ecology. 16:
310-312.
Kaiser Statehealthfacts.org, Kaiser Family Foundation, accessed December 2004,
http://www.statehealthfacts.kff.org/cgi-bin/healthfacts.cgi?
Karr, J.R. 1999. Urbanization and stream biology: The dangers of basing permit decisions on
total impervious area. Northwest Biological Assessment Workgroup. 10th Annual Meeting. Port
Angeles, Washington. November 3-5,1999.
Kienegger, Erwin H. 1992. Assessment of a Wastewater Service Charge by Integrating Aerial
Photography and GIS. Photogrammetric Engineering & Remote Sensing. 58(11): 1601-1606.
Lee, K.H. 1987. Determining Impervious Area for Storm Water Assessment. Paper read at
American Society of Photogrammetry and Remote Sensing/American Congress on Surveying and
Mapping Annual Convention, 1987, at Baltimore, Maryland.
Mallin, M.A., K.E. Williams, E.G. Esham and R.P. Lowe. 2000. Effect of human development
on bacteriological water quality in coastal watersheds. Ecological Applications. 10(4):
1047-1058.
Maryland Office of Planning. 1999. MdProperty View, Maryland Office of Planning, Planning
Data Services Division, Baltimore, MD 21201.
107
-------
May, C.W., R.R. Horner, J.R. Karr, B.W. Mar and E.B. Welch. 1997. Effects of Urbanization on
Small Streams in the Puget Sound Lowland Ecoregion. Watershed Protection Techniques. 2:
483-493.
McMahon, G., S.M. Gregonis, S.W. Waltman, J.M. Omernik, T.D. Thorson, J.A. Freeouf, A.M.
Rorick and I.E. Keys. 2001. Developing a spatial framework of common ecological regions for
the conterminous United States. Environmental Management. 28(3):293-316.
Miltner, R.J., D. White and C. Yoder. 2004. The biotic integrity of streams in urban and
suburbanizing landscapes. Landscape and Urban Planning. 69:87-100.
Mississippi Institutions of Higher Learning. March 2002. "Mississippi Population Projections",
Center for Policy Research and Planning, Jackson, Mississippi.
Mitchell, J.G. and S. Leen. 2001. Urban Sprawl. National Geographic. 200:1, July 2001. 48-
73. http://magma.nationalgeographic.com/ngm/data/2001/07/01/html/ft 20010701.3.html#
National Research Council. 1992. Restoration of Aquatic Ecosystems: Science, Technology and
Public Policy. National Academy Press. Washington, DC. 576pp.
Nash, M.S. and D.J. Chaloud. 2002. Multivariate Analyses (Canonical Correlation and Partial
Least Square (PLS)) to Model and Assess the Association of Landscape Metrics to Surface Water
Chemical and Biological Properties Using Savannah River Basin Data. U.S. Environmental
Protection Agency. Office of Research and Development. Washington, DC. EPA/600/R-02/091.
Newcombe, C.P. andD.D. MacDonald. 1991. Effects of suspended sediments on aquatic
ecosystems. North American Journal of Fisheries Management. 11: 72-82.
Nichols, D.B., B.K. Ferguson, S. Seinberg and M.A.A. Akers. 1999. Development ordinances to
protect streams. Proceedings of the 1999 Georgia Water Resources Conference. March 30-31,
1999. Kathryn J. Hatcher, ed. Institute of Ecology, University of Georgia. Athens, GA. pp. 151-
154.
North Carolina Department of Environment and Natural Resources. 2003. Standard operating
procedures for benthic macroinvertebrates. Biological Assessment Unit. Raleigh, NC. 21pp.
North Carolina Department of the Environment and Natural Resources (NCDENR). 2002. North
Carolina Division of Water Quality, Water Supply Watershed Protection, "Streamlines", Volume
6, Issue 3, http://h2o.enr.state.nc.us/wswp/SL/v6i3.pdf
North Carolina Office of State Budget and Management. June 2003. North Carolina State Data
Center, Raleigh, North Carolina.
Office of Management and Budget (OMB). December 27, 2000 Federal Register Notice,
Standards for Defining Metropolitan and Micropolitan Statistical Areas, Vol. 65, No. 249,
accessed 2004 at
http://www.census.gov/population/www/estimates/00-32997.pdf
Omernik, J. 2003. The Misuse of Hydrologic Unit Maps for Extrapolation, Reporting and
Ecosystem Management. Journal of the American Water Resources Association. 39(3): 563-573.
108
-------
Paul, M.J. and J. Meyer. 2001. Streams in the urban landscape. Annual Review of Ecology and
Systematics. 32:333-65.
Poff, N.L, J.D. Allan, M.B. Bain, J.R. Karr, K.L. Prestegaard, B.D. Richter, R.E. Sparks and J.C.
Stromberg. 1997. The natural flow regime: A paradigm for river conservation and restoration.
BioScience. 47(ll):769-784.
Potter, K.M., F.Q. Cubbage and R.H. Schaberg. 2005. Multiple-scale landscape predictors of
benthic macroinvertebrate community structure in North Carolina. Landscape and Urban
Planning. 71:77-90.
Price, M., M. Scorbee and T. Sawyer. "Kentucky Migration: Consequences for State Population
& Labor Force," 2004. University of Louisville, Urban Studies Institute.
Reeves, G.H., L.E. Benda, K.M. Burnett, P.A. Bisson, and J.R. Sedell. 1995. A disturbance-based
ecosystem approach to maintaining and restoring freshwater habitats of evolutionary significant
units of anadromous salmonids in the Pacific Northwest. American Fisheries Society Symposium
17: 334-349.
Richter, B.D., J.V. Baumgartner, J. Powell and D.P. Braun. 1996. A method for assessing
hydrologic alteration within ecosystems. Conservation Biology. 10:1163-1174.
Richter, B.D., D.P. Braun, M.A. Mendelson and L.L. Master. 1997. Threats to imperiled
freshwater fauna. Conservation Biology. 11:1081-1093.
Rosgen, David L. 1994. A classification of natural rivers. Catena. 22:169-199.
Schueler, T. 1994. The Importance of Imperviousness. Watershed Protection Techniques. 1:
100-111.
Seaber, P.R., P.P. Kapinos and G.L. Knapp. 1987. Hydrologic Unit Maps: U.S. Geological
Survey Water Supply Paper 2294, 63 p.
Sleavin, W.J., S. Prisloe, L. Giannotti, and D.L. Civco. 2000. Measuring impervious surfaces for
non-point source pollution modeling, 2000 American Society for Photogrammetry and Remote
Sensing (ASPRS) Annual Convention, Washington, DC.
Slonecker, E.T., D.B. Jennings and D. Garofalo. 2001. Remote Sensing of Impervious Surfaces:
A Review. Remote Sensing Reviews. 20:227-255.
Smith, S.K. 1987. Tests of forecast accuracy and bias for county population projections. Journal
of the American Statistical Association. 82(400):991-1013.
Smith, S.K. and J.M. Nogle. "Florida Population Studies: Population Projections by Age, Sex
and Race for Florida and its Counties, 2002-2025." July 2003. Vol 36, No. 3, Bulletin 136.
Smith, S.K. and Shahidullah, M. 1995. An evaluation of population projection errors for census
tracts. Journal of the American Statistical Association. 90(429):64-71.
Smith, S.K. and Sincich, T. 1992. Evaluating the forecast accuracy and bias of alternative
population projections for states. International Journal of Forecasting. 8:495-508.
109
-------
Smith, S.K. and Sincich, T. 1990. The relationship between the length of the base period and
population forecast errors. Journal of the American Statistical Association. 85(410):367-375.
Snyder, C.D., J.A. Young, R. Villella and D.P. Lemarie. 2003. Influences of upland and riparian
land use patterns on stream biotic integrity. Landscape Ecology. 18:647-664.
South Carolina State Budget & Control Board. October 2002. Office of Research & Statistics,
Health & Demographics Division, Columbia, South Carolina.
Stankowski, S.J. 1972. Population Density as an Indirect Indicator of Urban and Suburban
Land-Surface Modifications, U.S. Geological Survey: B219-B224.
Stone, B., Jr. 2004. Paving over paradise: How land use regulations promote residential
imperviousness. Landscape and Urban Planning. In Press.
Sutherland, R.C. 1995. Methodology for estimating the effective impervious area of urban
watersheds. Technical Note 58. Watershed Protection Techniques. 2(1)282-284.
Trimble, S.W. 1997. Contribution of stream channel erosion to sediment yield from an
urbanizing watershed. Science. 278:1442-1444.
University of Alabama. August 2001. Center for Business and Economic Research, Culverhouse
College of Commerce, Tuscaloosa, Alabama.
University of Florida. July 2003. Bureau of Economic and Business Research, Warrington
College of Business Administration, Gainesville, Florida.
University of Louisville. July 2003. Kentucky State Data Center, Urban Studies Institute,
Louisville, Kentucky.
University of Tennessee. March 1999. Center for Business and Economic Research, College of
Business Administration, Knoxville, Tennessee.
U.S. Census Bureau. 2001. Census 2000 Redistricting Data (P.L. 94-171) Summary File and
1990 Census. Census 2000 PHC-T-2. Ranking Tables for States: 1990 and 2000, Table 1: States
Ranked by Population 2000, internet release date April 2001.
U.S. Census Bureau. April 2001. Census 2000 Redistricting Data (P.L. 94-171) Summary File
and 1990 Census. Census 2000 PHC-T-2. Ranking Tables for States: 1990 and 2000, Table 2:
States Ranked by Numeric Population Change: 1990 to 2000, internet release date April 2001.
U.S. Census Bureau. State and County QuickFacts, 2004.
http://quickfacts.census.gov/qfd/states/01000.html
U.S. Census Bureau. Metropolitan and Micropolitan Statistical Area Definitions, 2004.
http://www.census.gov/population/www/estiniates/metroarea.html or U.S. Census Bureau,
Metropolitan Area Page Size Maps, 2004,
http://www.census.gov/geo/www/mapGallerv/mapage.html
no
-------
U.S. Department of Agriculture. 1997. ARE1 Updates on Agricultural Resources and
Environmental Indicators, Economic Research Service, Natural Resources and Environment
Division, Washington, DC, Number 3, June 1997, 4 pages.
http://ers.usda.gov/publications/arei/97upd/upd97-3.pdf
U.S. Department of Agriculture. 2003. "Briefing Room: Land Use, Value, and Management:
Major Uses of Land", Economic Research Service, Washington, DC, internet document updated
10/270/2003. http://ers.usda.gov/Briefing/LandUse/maiorlandusechapter.htm
U.S. Department of Agriculture. 2000. Summary Report: 1997 National Resources Inventory
(revised December 2000), Natural Resources Conservation Service, Washington, DC, and
Statistical Laboratory, Iowa State University, Ames, Iowa, 89 pages.
http://www.nrcs.usda.gov/technical/nri/! 997/summary_report/
U.S. Department of Transportation. 2001. National Transportation Atlas Databases, National
Highway Planning Network (NHPN)Version 3.0. Bureau of Transportation Statistics, Federal
Highway Administration, Office of Intermodal and Statewide Programs (HEPS-20), Washington,
DC.
USEPA. 2000a. BASINS Technical Note 6: Estimating Hydrology and Hydraulic Parameters for
HSPF. U.S. Environmental Protection Agency, Office of Water, Washington, DC. EPA/823/R-
00/012.
USEPA. 2000b. Better Assessment Sciences Integrating Point and Nonpoint Sources (BASINS).
U.S. Environmental Protection Agency, Office of Water, Office of Science and Technology,
Washington, DC.
USEPA. 1994. The Quality of Our Nation's Water: 1992. U.S. Environmental Protection Agency,
Office of Water, Washington, DC. EPA/84l/F-94/002.
Vogelmann, I.E., T. Sohl and S.M. Howard. 1998a. "Regional Characterization of Land Cover
Using Multiple Sources of Data." Photogrammetric Engineering & Remote Sensing. 64(1): 45-
57.
Vogelmann, I.E., T. Sohl and S.M. Howard. 1998b. Regional Land Cover Characteristics using
Landsat Thematic Mapper Data and Ancillary Data Sources. Environmental Monitoring and
Assessment. 51:415-428.
Vogelmann, I.E., S.M. Howard, L. Yang, C.R. Larson, B.K. Wylie and N. Van Driel. 2001.
Completion of the 1990s National Land Cover Data Set for the Conterminous United States from
Landsat Thematic Mapper Data and Ancillary Data Sources. Photogrammetric Engineering and
Remote Sensing. 67:650-652.
Wear, David N. and J.G. Greis. 2002. Southern Forest Resource Assessment Final Report.
Southern Research Station, USDA Forest Service, and Southern Region, USDA Forest Service.
http://www.srs.fs.usda.gov/sustain
Wickham, J.D., R.V. O'Neill, K.H. Riitters, E.R. Smith, T.G. Wade and K.B. Jones. 2002.
Geographic targeting of increases in nutrient export due to future urbanization. Ecological
Applications. 12(1):93-106.
in
-------
Wilcove, D.S., D. Rothstein, J. Dubow, A. Phillips and E. Losos. 1998. Quantifying threats to
imperiled species in the United States. BioScience. 48:607-615.
Williams, I.E., I.E. Johnson, D.A. Hendrickson, S. Contreras-Balderas, J.D. Williams, M.
Navarro-Mendoza, D.E. McAllister and I.E. Deacon. 1989. Fishes of North America:
endangered, threatened, or of species concern. Fisheries 14: 2-20.
Winter, T.C. 2001. The concept of hydrologic landscapes. Journal of the American Water
Resources Association. 37(2):335-349.
Wolman, M.G. 1967. A cycle of sedimentation and erosion in urban river channels. Geografiska
Annaler. 49A(2-4):385-395.
Woodruff, J.F. and J.D. Hewlett. 1970. Predicting and mapping the average hydrologic response
for the Eastern United States. Water Resources Research. 6(5): 1312-1326.
Yang, L., C. Huang, C. Homer, B. Wylie and M. Coan. 2003. An approach for mapping large-
area impervious surfaces; synergistic use of Landsat 7 ETM.+ and high spatial resolution imagery.
Canadian Journal of Remote Sensing. 20(2):230-240.
Zandbergen, P., J. Houston and H. Schreier. 2000. Comparative Analysis of Methodologies for
Measuring Watershed Imperviousness. Watershed Management 2000 Conference, Institute for
Resources and Environment, University of British Columbia, Vancouver, British Columbia,
Canada, July 2000.
112
-------
Appendix
The following procedures were used to calculate the current and projected %TIA for the
eight Southeastern United States: Alabama, Florida, Georgia, Kentucky, Mississippi, North
Carolina, South Carolina and Tennessee. All the GIS work was done in Environmental Systems
Research Institute, Inc.(ESRI)' s Arc View 3.2 and Arclnfo Workstion 9. The scripts in Section
XIll are in the ArcView scripting language Avenue. Source data used in these processing steps
include the following and are referred to by data type number (1 -6) in the stepwise procedures.
1. NLCD 92-21 category land cover classification with 30 m spatial resolution raster
(gridded) coverage (http://edc.usgs.gov/products/landcover/nlcd.html). Data was provided on CD
by USEPA Region 4 for the eight Southeastern states from regional data archives.
2. National Transportation Atlas (NTA) available on CD from U.S. Department of
Transportation (USDOT (2001) as a shapefile. Class 1 (interstates and major highways) roads for
the eight Southeastern states were saved as a shapefile for use in analysis.
3. U.S. Bureau of Census 1990 and 2000 block level population and population density
data. Block level data as a shapefile was provided by USEPA Region 4 from regional data
archives for the eight Southeastern states.
4. U.S. Bureau of Census 2000 block level vacant housing data. Data was downloaded
from U.S. Bureau of Census website (www.census.gov) and processed as described in Section V
below.
5. 12 or 14 digit HUC shapes files obtained from individual Southeastern states
6. County level population projections obtained from individual Southeastern states.
A state scale version for each of these data sets was developed and stored in state specific
directories. The following procedures were used to calculate values for each of the eight
individual states.
I. To calculate road area:
Step 1: remove road arcs that run through commercial industrial and mining cells to prevent
double-accounting:
A. To obtain only the high intensity commercial grid cells (value=23) from data set 1,
from the grid prompt in:
GRID: outgrid = select(mrlc grid, 'value = 23')
B. Convert the value23 grid to a polygon coverage by:
ARC: gridpoly value23 value23_poly
C. Convert road shapefile (data set 2) to a coverage using the shapearc command in ARC.
D. After setting up the ArcEdit: environment, ensure that the following two commands
execute:
AE: nodesnap off
AE: intersectarcs all
then "get" the poly arcs into the road coverage to make intersections at all polygons.
Save and then select all arcs with lanes = 0 and delete.
E. Convert the roads coverage back to a shapefile (newroadshape).
F. In ArcView, use the select by theme option and select the road arcs in newroadshape
that are completely within the value = 23 polygons. Delete the selected arcs.
G. Repeat steps A, B, C, and F to remove roads in grid cells classified as mining value =
32. Replace the number 23 with 32 in each of these steps.
Step 2: calculate road area:
A. Select the 2-lane arcs (Lanes = 2) from newroadshape and create a new shapefile from
the selected two lane arcs.
113
-------
B. Select-by-theme the hue polygons that are intersected by the 2-lane road arcs. Reverse
the selection to select hue polygons that do not contain road arcs and calculate the
2-lane area to -9999. Then run the rd_clip.ave script.
C. Repeat steps A & B using 4-lane roads.
D. The rd^clip.ave script will clip the road arcs with the hue polygon and calculate the
new road arc lengths in meters. The road lengths are then converted to road area
in square miles using the following formula:
road length x 3.2808 x 24 (or 48) x 0.00000003587006
Where 3.2808 is the conversion constant for meters to feet and 24 is the
assumed width, in feet, of a two-lane road (48 is the width, in feet, for a
four-lane road) and 0.00000003587006 is the conversion constant for
square feet to square miles.
E. The field total_rd_area is calculate by adding the 2-lane and 4-lane area fields. The
null values (-9999) will be re-calculated to -1111 by the script and then
selected and recalculated after script completion to 0 since they contain no
highway segment.
II. To calculate industrial and mining area:
1. Obtain the high intensity commercial grid cells from data type 1 in Arclnfo using
GRID;
GRID: outgrid = select(, Value = 23')
2. Then convert the outgrid into a polygon coverage:
ARC: gridpoly outgrid value23jpoly
3. Convert value23 j>oly to a shapefile.
4. In ArcView select the hue polygons (Data type 5) that completely contain the features
of value23 j>oly.shp
5. Reverse the selection and calculate the selected records' commercial area field to
- 9999 and run the fixed_loop.ave. The cell count will be calculated to the area
field. Converted to square meters, then convert to sq. miles. The other records
(-9999) will be re-calculateded to -1111 and should be re-calculated to 0 since they
contain no commercial property.
6. For mining, repeat above steps using the value 32 instead of 23.
7. The areas in square miles are calculated to the fields indus^area (or comm^area) and
mining^area respectively.
III. Populations for the years 2000 and 1990 and state population projections:
1. The FIPs code was calculated for the block shapefile (data set 2) by adding a field
([FIPS]) as a string with a width of five and calculating it equal to the [Areakey]
field.
2. A new field was added to the block shapefile called [orig^area] and the value was
calculated (Shape.ReturnArea).
3. The block shapefile was exported as a geodatabase and then converted back to a
shapefile.
4. The avenue script clip_blocks.ave was run, which clips the blocks by the HUCs (data
type 2).
5. The clipped blocks are then merged back together using the GeoProcessing Wizard in
ArcView.
6A new field, [new^area] is added and calculated. A [factor] field is also added and is
calculated by [new^area] / [orig^area].
7. A selection is run on records with a factor > 1.01 and those are deleted. These values
are believe to exist because of problems within the original block shapefile due to
manipulation by the original creators to account for changes between 1990 and
114
-------
2000. Records that have a zero population for both 1990 and 2000 are also deleted
as they will not play a factor in any population estimates.
8. New population fields are added for 1990 and 2000 to the block shapefile and are
calculated by multiplying the original populations by the factor value to get the
corrected population for the split blocks (newjop2000 and new jop 1990).
9. An excel table containing the Census 1990 & 2000 figures per county along with the
state's population estimates for any or all of the following years: 2010, 2015,
2020, 2025, and 2030 was joined by FIPS code field to the block shapefile.
10. Block population estimates are calculated in the block shapefile by:
block jopX = new jop2000 / County_pop2000 * Countyj>opX,
where X = year estimated.
11. The block population estimates are summed and calculated to the HUCs by running
the calc_pop.ave script. Minor edits are made to the script for each year's
estimate.
IV. Population Density:
1. All population values divided by square miles.
V. Vacant Housing Data:
Step 1: import data from Census website (data type 4):
A. Download each state's files (for example, Alabama's data would be named
al00037ufl.zip (data) and algeoufl.zip (geography)) from the Census Bureau's FTP
site and unzip each file.
B. Dowloaded the Summary File template file for MSAccess97 from the Census
Bureau's website. This file enables the import of the data and geographic files into
Access because it contains the field names and structure for the data. This site also
includes instructions for importing the files into MSAccess.
C. Change the *00037.ufl and *geo.ufl file extensions to .txt and follow the import
procedures into MSAccess (instructions located at Census website.)
D. Save each database as a .dbf.
E. Import the .dbf files into ArcView.
F. Perform a join on the fields contained in both the geo and data files.
G. Export the joined table as a new .dbf file to permanently preserve join.
H. Create field as a string and calculated to match in Region 4
block data. Populate the records by calculating the new field as
state+county+tract+block (concatenate the values). Join the two tables (census
data + block data).
Step 2: Calculate number of homes per type of vacancy:
A. Calculate the following, per HUC (data type 5): total housing units, occupied units,
vacant units, rental homes, homes for sale, rented or sold homes that are vacant,
seasonal housing, migrant housing and homes vacant for other reasons. The
calculations are completed using the calc_vcn.ave script.
B. Density for vacant housing is based on number of homes per square mile.
Step 3: Calculate estimated population for vacant housing:
A. Calculate population equivalent using the vacant housing information and average
household size by state (based on data from US Census Bureau website) for each
HUC. Calculation was a weighted average based on owner-occupied and renter-
occupied—
Average population per household for North Carolina = 2.48798
Average population per household for Alabama = 2.49575
Average population per household for Florida = 2.4601
Average population per household for Georgia = 2.645
115
-------
Average population per household for Kentucky = 2.46824
Average population per household for Mississippi = 2.62845
Average population per household for South Carolina = 2.52884
Average population per household for Tennessee = 2.48572
Average person per household multiplied by number of vacant homes. The total
vacant and seasonal population fields are based only on vacant and seasonal
housing data. The total vacant and seasonal population density field are based on
the vacant and seasonal populations + the 2000 population per square mile.
VII. %TIA values based on population:
%TIA based on the Hicks (GVSS&DD) calculations use the following formula:
(94 x (1 - (2.7183 A (-0.00010938 x PopulationDensityValue)))) + 1
VIII. Extrapolated 1993 Population Value:
The population value for 1993 was derived using:
1990 Population + (((2000 Population - 1990 Population) /10) x 3)
IX. Projected high intensity commercial industrial (HICI) area:
1. HiciX = commercial area x PopulationX / 1993 Population
Where X = projected year
2. Query: HiciX < commercial area; if yes, calc HiciX = commercial area
3. Add a temporary field, calculated as:
commercial area/ 1993 population x 10,000
4. Evaluate the temporary field to determine if any values are greater than 3.
5. If there are values greater than 3, add a temporary field, calculated as:
0.6 x (PopulationX - 1993 Population) /10,000
6. Use the value from Step 5 and add it to the commercial area to determine
HiciX for that record.
7. Reevaluate HiciX < commercial area and recalculate HiciX = commercial
area if necessary.
X. Commercial Total Impervious Area:
% Commercial T1A is calculated as:
(HiciX / total area HUC) x 0.90 x 100
XI. Percent Total Impervious Area:
Percent T1A is calculated as:
% road area + % mining area + % Commercial T1A + Hicks calculation
The field TlA2000_vct using the above formula with the Hicks value (Step VII) bein^
calculated from the total vacant population density figure (from Section V, Step 3.)
XII. Projected Population with Vacant:
1. Projected population for 2010, 2015, etc with projected vacant population added:
Year X projected HUC population * (1 + total vacant population / 2000
population)
2. The density is then calculated by dividing by area in square miles.
116
-------
XIII. Scripts:
• Script 1: Fixed_loop.ave
'loop_grd_clp.ave
'Christine Perkins, CSC
'this scrip loops through a polygon coverage, selects each polygon
'using an index number,
'clips the grid and returns the count value to be calculated into the
'polygon attribute table.
theProject = av.Get Project
theView = theProject,FindDoc("Viewl")
PolyThm = theView.FindTheme ( "huc!2_al . shp" )
GridThm = theView.FindTheme("mining_grid")
GridThm.SetActive(true)
PolyFTab = PolyThm.GetFTab
theField = PolyFTab.FindFieId("Index")
theField2 = PolyFTab.FindField("mining_area")
PolyFTab.SetEditable (true)
theSel = PolyFTab.GetSelection
theValue = 0
for each rec in PolyFTab
theValue = theValue + 1
QueryString = "[Index] =" + theValue.AsString
PolyFTab.Query(QueryString,theSel,#VTAB_SELTYPE_NEW)
PolyFTab.UpdateSelection
newVal = PolyFTab.ReturnValue(theField2,rec)
if (newVal < 0) then
PolyFTab.SetValueNumber(theField2,rec,-llll)
else
theGrid=theView.GetActiveThemes.Get(0).GetGrid
'Get bounds of clipping area as a rectangle
thePolyThmExtent = PolyThm.getselectedextent
if (thePolyThmExtent .IsEmpty) then thePolyThmExtent =
PolyThm.ReturnExtent end
'Get parameters for the new grid
theFtab = PolyThm.GetFTab
theProj = theView.GetProjection
theCell = theGrid.GetCellSize
theExtent = theGrid.GetExtent
ae = theView.GetExtension(AnalysisEnvironment)
ae.SetExtent(#ANALYSISENV_VALUE, thePolyThmExtent)
ae.SetCellSize(tANALYSISENV_VALUE, theCell)
' Activate the settings for the analysis envirnonment as returned
' by the above 3 lines of code.
ae.Activate
'the actual extraction occurs here
tempGrid = Grid.MakeFromFtab(theFtab,theProj,nil,{theCell,theExtent})
newGrid = (tempGrid.IsNul1).Con (tempGrid, theGrid)
' rename data set
aFN = av.GetProject.GetWorkDir.MakeTmp("mine", "")
newGrid.Rename(aFN)
117
-------
' check if output is ok
if (newGrid.HasError) then return NIL end
' create a theme
gridThm = theme.make(newGrid.GetSrcName)
' set name of theme
gridThm.SetName("Hue" + theValue.AsString)
' add theme to the specifiedview
theView.addTheme(gridThm)
' Resets the analysis environment to the maximum of inputs (i.e. the
default)
aRect = Nil
ae = theView.GetExtension(AnalysisEnvironment)
ae.SetExtent(#ANALYSISENV_MAXOF, aRect)
ae.SetCellSize(#ANALYSISENV_MAXOF, aRect)
gridThm.invalidate (true)
newerGrid = gridThm.GetGrid
newVTab = newerGrid.GetVTab
theFieldest = newVTab.FindField ("count")
answers = newVTab.ReturnValue (theFieIdest,0)
PolyFTab.SetValue(theField2,rec,answers)
theProject.save
end
end
• Script 2: rd_clip.ave
'clip_themes.ave
'Christine Perkins, May 2001
'this script loops through a polygon-type shapefile, selects each
'polygon,
'clips and creates a road shapefile based on the poly boundaries,
'the new road lengths, and plugs the length into the still-selected
'polygon
'uses the extensions Batch Clip and Clip Theme, available online.
'Basic setup
theProject = av.GetProject
theView = theProject.FindDoc("Viewl")
PolyThm = theView.FindTheme("Hucl2_al.shp")
PolyThm.SetActive(true)
LineThm = theView.FindTheme("Four lanes.shp")
LineThm.SetActive(true)
PolyFTab = PolyThm.GetFTab
LineFTab = LineThm.GetFTab
'Finds necessary fields and sets up tables
theField = PolyFTab.FindField ("Index ")
theField2 = PolyFTab.FindField("four_lane")
PolyFTab.SetEditable (true)
theSel = PolyFTab.GetSelection
'Starts the loop
theValue = 0
for each rec in PolyFTab
118
-------
theValue = theValue + 1
Q'UeryString = "[Index] =" + theValue .AsStr ing
PolyFTab.Query (QueryString,theSel,#VTAB_SELTYPE_NEW)
PolyFTab.UpdateSelection
newVal = PolyFTab.ReturnValue(theField2,rec)
if (newVal < 0) then
PolyFTab.SetValueNumber(theField2,rec,-1111)
else
'not really needed
'theView = av.GetActiveDoc
thePrj = theView.GetProjection
activeThemes = theView.GetActiveThemes
' -- get the FTab for the theme to clip, and if two themes are selected,
' -- the FTab of the theme containing the clipping polygons
if (activeThemes.Count = 1) then
sourceTheme = activeThemes.Get(0)
sourceFTab = sourceTheme.GetFTab
clipFTab = nil
else
themel = activeThemes.Get ( 0)
ftabl = themel.GetFTab
theme2 = activeThemes.Get (1)
ftab2 = theme2.GetFTab
' -- if only one theme is a polygon theme, then it is the clipping
theme
sourceFTab = nil
if (ftabl.GetShapeClass.GetClassName = "Polygon") then
if (ftab2.GetShapeClass.GetClassName <> "Polygon") then
sourceTheme = theme2
sourceFTab = ftab2
clipFTab = ftabl
end
else
if (ftab2.GetShapeClass.GetClassName = "Polygon") then
sourceTheme = themel
sourceFTab = ftabl
clipFTab = ftab2
end
end
' -- get the output file name
outFileName = FileName.GetCWD.MakeTmp("roadest", "shp")
'outFileName = FileDialog.Put(outFileName,"*.shp","Specify the output
shapefile " )
'if (outFileName = Nil) then
'return nil
' end
shapeType = sourceFTab.FindField ("Shape").GetType
if (shapeType = #FIELD_SHAPELINE) then
outclass = POLYLINE
elseif (shapeType = #FIELD_SHAPEMULTIPOINT) then
outclass = MULTIPOINT
119
-------
elseif (shapeType = #FIELD_SHAPEPOINT) then
outclass = POINT
elseif (shapeType = #FIELD_SHAPEPOLY) then
outclass = POLYGON
else
'MsgBox.Error("Invalid shape field type.",theTitle)
' return nil
end
' -- check if shapes should be projected
'doProjection = false
'if (thePrj <> nil) then
'doProjection = MsgBox.YesNo ( "Output shapes in projected
coordinates?",theTitle,true)
' end
' — create one large polygon from the input polygons
if (clipFTab <> nil) then
clipPoly = av.Run("View.ClipThemeUnionFTab",clipFTab)
else
clipPoly = av.Run("View.ClipThemeUnionGraphics",theView)
if (thePrj <> nil) then
clipPoly = clipPoly.ReturnUnprojected (thePrj)
end
end
oldSelection = sourceFTab,GetSelection,Clone
if (sourceFTab.GetSelection . Count = 0) then
sourceFTab.GetSelection.SetAll
end
'if (clipOption = "inside") then
sourceFTab.SelectByPolygon(clipPoly,#VTAB_SELTYPE_AND)
processSelection = sourceFTab.GetSelection.Clone
sourceFTab.SetSelection (oldSelection)
sourceFTab.UpdateSelection
' — create the new shapefile
outFTab = FTab.MakeNew(outFileName, outclass)
outFTab.SetEditable (true)
outFields = sourceFTab.GetFields.DeepClone
outFields.Remove(0)
outFTab.AddFields (outFields)
outFields = outFTab.GetFields
outShapeField = outFTab.FindField ("shape")
sourceFields = sourceFTab.GetFields
sourceShapeField = sourceFTab.FindField ("shape ")
' — process the features
selCount = processSelection.Count
c = 0
av.ShowMsg("Clipping ...")
av.SetStatus (0)
120
-------
for each sRec in
s = sourceFTab.Returnvalue(sourceShapeField,sRec)
'if (clipOption = "inside") then
if (s.IsContainedin(clipPoly).Not) then
end
'if ( doProjection) then
' s = s.ReturnProjected (thePrj)
' end
oRec = outFTab.AddRecord
outFTab.SetValue(outShapeField,oRec,s)
for each i in 1. . (sourceFieIds.Count - 1)
v = sourceFTab.Returnvalue(sourceFieIds.Get(i),sRec)
outFTab.SetValue (outFields.Get (i) ,oRec,v)
end
c = c + 1
av.SetStatus ( (c / selCount) * 100)
end
outFTab.SetEditable (false)
av.ClearMsg
av.ClearStatus
' — display in a view if requested
viewList = List.Make
for each d in av.GetProject.GetDocs
if (d.Is(View)) then
viewList.Add(d)
end
end
''add new theme
newTheme = FTheme.Make(outFTab)
theView.AddTheme(newTheme )
theView.GetWin.Activate
end
' 'calculate length of road segments
newTab = newTheme.GetFTab
newTab.SetEditable (true)
lengthField = Field.Make("Lengths",#FIELD_DECIMAL,16,3)
newTab.AddFields({lengthField})
'calc each segment's length and attribute the record
for each rec in newTab
theShape = newTab.Returnvalue(newTab.FindFieId ("shape"),rec)
x = theShape.ReturnLength
newTab.SetValue(lengthField,rec,x)
end
newTab.SetEditable (false)
'add lengths from new shapefile
total = 0
theLength = newTab.FindField ( "Lengths ")
for each rec in newTab
lengths = newTab.ReturnValue(theLength,rec)
121
-------
total = total + lengths
end
'plug in length value into polygon record
PolyFTab.SetValue(theField2,rec,total)
theProject.save
end
end
Script 3: Clipjblocks.ave
'TITLE: clip_blocks.ave
'AUTHOR: Christine Perkins, CSC
'PURPOSE: clips a block shapefile with a HUC shapefile and adds the
'clipped blocks to the ArcView project
theProject = av.GetProject
theView = theProject.FindDoc("Viewl")
SRCTheme = theView.FindTheme("Blocks_geo.shp")
interTheme = theView.FindTheme("Arc_hucl2.shp")
interTheme.SetActive (true)
interFTab = interTheme.GetFTab
SRCFTab = SRCTheme.GetFTab
interFTab.SetEditable (true)
theField = interFTab.FindField("Index")
theSel = interFTab.GetSelection
theValue = 0
for each rec in interFTab
theValue = theValue + 1
QueryString = "[Index] =" + theValue.AsString
interFTab.Query(QueryString,theSel,#VTAB_SELTYPE_NEW)
interFTab.UpdateSelection
' Specify the output shapefile. . .
outFName = FileName.GetCWD.MakeTmp("clip", "shp")
shapeType = SRCTheme.GetFTab.FindFieId("Shape").GetType
if (shapeType = #FIELD_SHAPELINE) then
outclass = POLYLINE
elseif (shapeType = #FIELD_SHAPEMULTIPOINT) then
outclass = MULTIPOINT
elseif (shapeType = #FIELD_SHAPEPOINT) then
outclass = POINT
elseif (shapeType = #FIELD_SHAPEPOLY) then
outclass = POLYGON
else
MsgBox.Error("Invalid shape field type.", "Merge Themes")
exit
end
'Set the variables
OutputFTab = FTab.MakeNew ( outFName, outclass )
SRCfields = List.Make
= List.Make
122
-------
for each f in SRCTheme.GetFTab.GetFields
if (f.GetName = "Shape") then
continue
else
fCopy = f.Clone
SRCfields.Add(fCopy)
end
end
'add the fields to the output file
if (SRCFields.Count > 0) then
OutputFTab.AddFields ( SRCFields )
end
outshpfld = OutputFtab.FindField ( "Shape")
Themel = SRCTheme
ftabl=Themel.GetFTab
shpfldl = ftabl.FindField ("Shape")
therecsl = ftabl.GetSelection
theoldsel = ftabl.GetSelection.Clone
if (therecsl.Count=0) then
therecsl=ftabl
end
Theme2 = InterTheme
ftab2=Theme2.GetFtab
shpfld2 = ftab2.FindField ("Shape")
therecs2 = ftab2.GetSelection
if (therecs2 . Count = 0) then
therecs2=ftab2.GetSelection.SetAll
ftab2.UpdateSelect ion
therecs2 = ftab2.GetSelection
end
OutputFtab.SetEditable(False)
OutputFtab.SetEditable (True)
totalshape = ftab2.ReturnValue(shpfid2, therecs2.GetNextSet (-1 ) )
for each apshape in therecs2
totalshape = totalshape.ReturnUnion(ftab2.ReturnValue(shpfId2,
apshape))
end
'Start processing each record in the selected overlay polys
'Get the polygon shape and select all records within that shape
theSRCshape = totalshape
if (theView.GetProjection.isNul1) then
Themel.SelectbyShapes({theSRCshape}, #VTAB_SELTYPE_NEW)
else
pshp = theSRCShape.ReturnProjected(theView.GetProject ion)
Themel.SelectbyShapes ({pshp}, #VTAB_SELTYPE_NEW)
end
123
-------
'For each selected record
recordCount = 0
for each Selrec in ftabl.<
recordCount = RecordCount +1
av.ShowMsgt"Splitting Shapes. . .")
av.SetStatus ( (recordCount / ftabl.GetSelection . Count) * 100)
'Get the shape of the record
SelectedShape = ftabl.ReturnValue(shpfldl,Selrec)
'If the output is a line
if (outshpfId.getType = #FIELD_SHAPELINE) then
'If the line is wholly within the polygon (no intersection) then
if (SelectedShape.IsContainedln(theSRCShape)) then
aLineshp = SelectedShape
'Else split the line using the polygon
else
aLineShp = SelectedShape.Linelntersect ion(theSRCshape)
end
'Add the new record
theoutrec=outputftab.AddRecord
' Set the shape value
outputFtab.SetValue(outshpfId, theoutrec, alineshp)
'Set the field values
for each afield in SRCfields
oldfield=ftabl.FindField(afield.GetName)
' Due to field name renaming between INFO and dBASE,
' some of the fields may not be found. In those cases,
' leave the value blank.
if (oldf ieldonil) then
oldvalue=ftabl.ReturnValue(oldfield, selrec)
outputftab.SetValue(afieId, theoutrec, oldvalue)
end
end
'This enters into the polygon loop
else
shplntersect = SelectedShape.Returnlntersect ion(theSRCshape)
if (shpintersect.IsEmpty) then
cont inue
end
' Geometric operations (such as Returnlnters.ection) return
multipoints
' instead of points, so if we are trying to write out points,
convert
' from multipoints
if (outshpfId.getType = #FIELD_SHAPEPOINT) then
shplntersect = shpintersect.AsList.Get ( 0)
end
theoutrec=outputftab.AddRecord
outputFtab.SetValue(outshpfId,theoutrec,shplntersect)
for each afield in SRCfields
oldfield=ftabl.FindField(afield.GetName)
' Due to field name renaming between INFO and dBASE,
' some of the fields may not be found. In those cases.
124
-------
' leave the value blank.
if (oldf ieldonil) then
oldvalue=ftabl.ReturnValue(oldfield, selrec)
outputftab.SetValue(afieId, theoutrec, oldvalue)
end
end
end
av,PurgeObjects
end
'Set editing OFF
OutputFTab.SetEditable (false)
theView.GetGraphics.EndBatch
ftabl.SetSelection (theoldsel)
ftabl.UpdateSelection
if (OutputFTab.HasError) then
MsgBox.Error ( "The out FTab has an error","")
exit
end
mergeTheme = FTheme.Make( OutputFTab )
theView,AddTheme( mergeTheme )
end
• Script 4: Calc_pop.ave
Title: calc pop.ave
'Author: Christine Perkins, CSC
'Purpose: calc values of pop90, pop2000, bpoplO, bpop20, bpopSO, etc
'per hue; where bpop = block population per HUC
'These have already got the factor (in block shapefile) applied.
theProject = av.GetProject
theView = theProject.FindDoc("Viewl")
SRCTheme= theView.FindTheme("Merged_blks1.shp")
interTheme = theView.FindTheme("Setl harrison. shp")
interTheme.SetActive(true)
interFTab = interTheme.GetFTab
SRCFTab = SRCTheme.GetFTab
newHUCField = interFTab.FindField ( "Bpop30 ")
interFTab.SetEditable (true)
SRCFTab.SetEditable (true)
theField = interFTab.FindField("Index")
theSel = interFTab.GetSelection
newPopField = SRCFTab.FindField ( "BpopSO ")
theValue = 0
for each rec in interFTab
theValue = theValue + 1
QueryString = "[Index] =" + theValue.AsString
interFTab.Query(QueryString,theSel,#VTAB_SELTYPE_NEW)
interFTab.UpdateSelection
SRCTheme.SelectbyTheme(interTheme,#FTAB_RELTYPE_ISCOMPLETELYWITHIN,0,#VT
AB_SELTYPE_NEW)
SRCFTab.UpdateSelection
125
-------
theBitMP = SRCFTab.<
newPopVal = 0
for each rec in theBitMP
newpop90Val = SRCFTab.ReturnValue(newPopField,rec)
newPopVal = newPopVal + newPop90Val
end
interFTab.SetValue(newHUCField,rec,newPopVal)
end
Script 5: Calc_vcn.ave
'Title: calc_vcn.ave
'Author: Christine Perkins, CSC
'Purpose: calc values of vacant properties
theProject = av.GetProject
theView = theProject.FindDoc("Viewl")
SRCTheme= theView.FindTheme("Test_clip.shp")
interTheme=theView.FindTheme("Problem.shp")
interTheme.SetActive(true)
interFTab = interTheme.GetFTab
SRCFTab = SRCTheme.GetFTab
newHUCField = interFTab.FindField("other_vcnt")
interFTab.SetEditable (true)
SRCFTab.SetEditable(true)
theField = interFTab.FindField ("Index" )
theSel = interFTab.GetSelection
theFactorField = SRCFTab.FindField ("Fact or ")
newPopField = SRCFTab.FindField("other_vcnt")
theValue = 0
for each rec in interFTab
theValue = theValue + 1
QueryString = "[Index] =" + theValue.AsString
interFTab.Query(QueryString,theSel,#VTAB_SELTYPE_NEW)
interFTab.UpdateSelection
SRCTheme.SelectbyTheme(interTheme,#FTAB_RELTYPE_ISCOMPLETELYWITHIN,0,#VT
AB_SELTYPE_NEW)
SRCFTab.UpdateSelection
theBitMP = SRCFTab.GetSelection
newPopVal = 0
for each rec in theBitMP
newpop90Val = SRCFTab.ReturnValue(newPopField,rec)
theFactorVal = SRCFTab.ReturnValue(theFactorField,rec)
newVal = newpop90Val * theFactorVal
newPopVal = newVal + newPopVal
end
interFTab.SetValue(newHUCField,rec,newPopVal)
end
126
------- |