EPA-600/3-77-105
September 1977
Ecological Research Series
NONPOINT SOURCE - STREAM NUTRIENT LEVEL
RELATIONSHIPS: A Nationwide Study
% PRO^
Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Corvallis, Oregon 97330
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ECOLOGICAL RESEARCH series. This series
describes research on the effects of pollution on humans, plant and animal spe-
cies, and materials. Problems are assessed for their long- and short-term influ-
ences. Investigations include formation, transport, and pathway studies to deter-
mine the fate of pollutants and their effects. This work provides the technical basis
for setting standards to minimize undesirable changes in living organisms in the
aquatic, terrestrial, and atmospheric environments.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/3-77-105
September 1977
NONPOINT SOURCE—STREAM NUTRIENT LEVEL
RELATIONSHIPS: A NATIONWIDE STUDY
By
James M. Omernik
Special Studies Branch
Corvallis Environmental Research Laboratory
Corvallis, Oregon 97330
CORVALLIS ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CORVALLIS, OREGON 97330
-------
DISCLAIMER
This report has been reviewed by the Corvallis Environmental Research
Laboratory, U.S. Environmental Protection Agency, and approved for publi-
cation. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
i i
-------
Effective regulatory and enforcement actions by the Environmental Pro-
tection Agency would be virtually impossible without sound scientific data
on pollutants and their impact on environmental stability and human health.
Responsibility for building this data base has been assigned to EPA's Office
of Research and Development and its 15 major field installations, one of
which is the Corvallis Environmental Research Laboratory (CERL).
The primary mission of the Corvallis Laboratory is research on the
effects of environmental pollutants on terrestrial, freshwater, and marine
ecosystems; the behavior, effects and control of pollutants in lake systems;
and the development of predictive models on the movement of pollutants in
the biosphere.
This report relates phosphorus and nitrogen levels in streams to the
nonpoint source influences which are present in their drainage areas and
also demonstrates the geographic trends in stream nutrient levels in the
contiguous United States. As such, the information provided herein should
be of interest and utility to water quality managers.
A. F. Bartsch
Director, CERL
-------
ABSTRACT
National Eutrophication Survey (NES) data collected from a nationwide
network of 928 nonpoint-source watersheds were studied for relationships
between macro-drainage area characteristics (particularly land use) and
nutrient levels in streams. Both the total and inorganic forms of phos-
phorus and nitrogen concentrations and loads in streams were considered.
For both nationwide and regional data sets, good correlations were
found between general land use and nutrient concentrations in streams. Mean
concentrations were considerably higher in streams draining agricultural
watersheds than in streams draining forested watersheds. The overall
relationships and regionalizes of the relationships and interrelationships
with other characteristics are illustrated cartographically and by regres-
sion techniques.
Two methods are provided for predicting nonpoint source nutrient levels
in streams; one utilizing mapped interpretations of NES nonpoint source data
and the other, regional regression equations and mapped residuals of these
regressions. Both methods afford a limited accountability for regional
characteri sties.
This report covers a period from June, 1972 to December, 1975; work was
completed as of September, 1977.
iv
-------
CONTENTS
Pac[e
Foreword iii
Abstract iv
List of Figures vii
List of Tables x
Attachments xi
Acknowledgements xi i
Sections
I Conclusions 1
II Introduction 4
III Applications 6
IV Background 11
History and Objectives 11
Literature 12
Selection of Study Watersheds 14
Data Collection Methods 15
Drainage Area Measurement and Land Use Identification. . . 15
Land Use Percentage Computation 17
Animal Unit Density Computations 17
Geology Identification 18
Slope and Drainage Density Computations 22
Soil pH Calculations 22
Precipitation 22
Nutrient Concentration and Export Estimates. 23
V Results and Discussion 24
Areal Distributions of Data 24
Overall Land Use--Nutrient Runoff Relationships 25
Category Definitions 25
General Analysis 27
Regional ity 37
Regression Analyses and Predictive Capability 50
Contributing Land Use Types 50
Contributing Land Use--Nutrient Concentration
Relationships 52
Contributing and Forest Land Use—Nutrient
Concentration Relationships 61
Mapped Interpretations 71
Individual Relationships 73
Nutrient Runoff--Soi1s Relationships 73
Nutrient Runoff—Geology Relationships 74
Nutrient Runoff—Drainage Density Relationships 77
v
-------
CONTENTS (Continued)
Page
Anomalies 78
Follow-up Studies 80
VI References 82
VII Appendix 87
vi
-------
No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Page
9
13
28
29
32
33
35
36
38
39
42
43
45
46
LIST OF FIGURES
An application of NES mapped nonpoint source stream nutrient
concentration interpretations
Distribution of individual NES nonpoint source study watersheds.
Relationships between general land use and total phosphorus
and orthophosphorus concentrations in streams
Relationships between general land use and total nitrogen and
inorganic nitrogen concentrations in streams
Relationships between general land use and stream exports of
total phosphorus and orthophosphorus
Relationships between general land use and stream exports of
total nitrogen and inorganic nitrogen
Frequency polygons of mean total phosphorus and mean orthophos-
phorus concentrations in streams by overall land use category. .
Frequency polygons of mean total nitrogen and mean inorganic
nitrogen concentrations in streams by overall land use category.
Regional relationships between general land use and total
phosphorus and orthophosphorus concentrations in streams . . . .
Regional relationships between general land use and total
nitrogen and inorganic nitrogen concentrations in streams. . . .
Regionalities of total phosphorus concentrations in NES sampled
streams draining forested watersheds
Regionalities of total phosphorus concentrations in NES sampled
streams draining agricultural watersheds
Regionalities of orthophosphorus concentrations in NES sampled
streams draining forested watersheds
Regionalities of orthophosphorus concentrations in NES sampled
streams draining agricultural watersheds
vii
-------
No
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
47
48
49
51
53
56
58
60
64
65
66
68
69
70
LIST OF FIGURES (Continued)
Regionalities of total nitrogen concentrations in NES sampled
streams draining forested watersheds
Regionalities of inorganic nitrogen concentrations in NES
sampled streams draining forested watersheds
Regionalities of total nitrogen concentrations in NES sampled
streams draining agricultural watersheds
Regionalities of inorganic nitrogen concentrations in NES
sampled streams draining agricultural watersheds
Scattergram of "contributing" land use types related to total
phosphorus concentrations in streams
Scattergram of "contributing" land use types related to
orthophosphorus concentrations in streams
Scattergram of "contributing" land use types related to total
nitrogen concentrations in streams
Scattergram of "contributing" land use types related to
inorganic nitrogen concentrations in streams ,
Differences in predictive characteristics between regional
land use-phosphorus concentration models
Differences in predictive characteristics between regional
land use-nitrogen concentration models ,
Areal distribution of residuals from regional predictive
models for total phosphorus concentrations in streams
Areal distribution of residuals from regional predictive
models for orthophosphorus concentrations in streams
Areal distribution of residuals from regional predictive
models for total nitrogen concentrations in streams
Areal distributions of residuals from regional predictive
models in inorganic nitrogen concentrations in streams . . . .
Geologic classification and mean annual stream phosphorus
concentrations and exports from a nationwide set of 586
nonpoint source watersheds. Data grouped by overall land
use category (attachment included in back cover jacket)
viii
-------
LIST OF FIGURES (Continued)
Geologic classification and mean annual stream nitrogen
concentrations and exports from a nationwide set of 586
nonpoint source watersheds. Data grouped by overall land
use category (attachment included in back cover jacket)
ix
-------
LIST OF TABLES
No. Page
1 Computation of mean annual nonpoint' source total nitrogen
concentration in the Willamette River at Salem, Oregon 8
2 Animal nutrient production 17
3 Total phosphorus compositions of rock types 21
4 Percent of mean inorganic nitrogen to mean total nitrogen
concentrations in streams by region 40
5 Estimated mean total phosphorus concentrations 54
6 Estimated mean total phosphorus concentrations for nationwide
and regional models 55
7 Estimated mean total nitrogen concentrations 57
8 Estimated mean total nitrogen concentrations for nationwide
and regional models 59
9 Regional stream nutrient concentration predictive models 62
10 Mean annual phosphorus concentrations in streams vs. phosphorus
compositions of predominant rock types 76
x
-------
ATTACHMENTS
Total phosphorus concentrations in streams from nonpoint sources (map).
Total nitrogen concentrations in streams from nonpoint sources (map).
Inorganic nitrogen concentrations in streams from nonpoint sources
(map).
Geologic classification and mean annual stream phosphorus concentrations
and exports from a nationwide set of 586 nonpoint source watersheds.
Data grouped by overall land use category (Figure 29).
Geologic classification and mean annual stream nitrogen concentrations
and exports from a nationwide set of 586 nonpoint source watersheds.
Data grouped by overall land use category (Figure 30).
xi
-------
ACKNOWLEDGEMENTS
This study would not have been possible without the volunteer manpower
supplied by the National Guard of each state. The Guardsmen were respon-
sible for collecting, preserving, and shipping the monthly samples from each
designated stream site. Their contribution is sincerely appreciated.
Also gratefully acknowledged are the efforts of Robert R. Payne (Coor-
dinator, National Eutrophication Survey, Washington, D.C.), who worked with
each State water pollution control agency in initiating the survey, and Lt.
Col. Louis R. Dworshak (Coordinator of Military Resources, Washington,
D.C.), who arranged for the participation of the National Guard in each
state.
Most of the data compilation (land use photo interpretation, drainage
area and slope measurement, etc.) was accomplished by the following persons:
June Fabryka, Madeline Hall, Thomas Jackson, Rose McCloud, Martha McCoy,
Theodore McDowell, Nola Murri, Michael Ness, James Sachet, Leta Gay Snyder
and Larry Warnick. Many of these individuals also assisted in graphics
compilation and various aspects of basic research. Particularly noteworthy
were Madeline Hall's research on phosphorus composition of rock types,
Theodore McDowell's work on the utility of other nonpoint source predictive
methods, and James Sachet's efforts on the graphics.
The negatives of the colored stream nutrient concentration maps were
prepared, under the direction of Dr. A. Jon Kimerling, by the Cartographic
Services of the Department of Geography, Oregon State University.
_Dr; Don A. Pierce was primarily responsible for the construction of the
prediction models. Dr. Pierce and Dr. Dale H. P. Boland, David Deckebach
and Donna West provided assistance with computer programming and statistical
data manipulation.
Many of the staff at the Corvallis Environmental Research Laboratory
contributed input to this study through logistic and analytical support,
constructive suggestions, and technical editing. A special appreciation is
extended to Marvin 0. Allum for his editorial comments and, in particular,
his careful scrutiny of all of the nutrient data for errors and inconsis-'
tencies. Daniel F. ^Krawczyk and his staff are gratefully acknowledged for
conducting the chemical analysis on the thousands of samples involved in.
this study. Also deserving of recognition is Dr. Norbert S. Jaworski for
his support during the formative stages of this study. Lastly, the author
is especially indebted to Dr. Jack H. Gakstatter for his continual solid
support and invaluable guidance.
xi i
-------
SECTION I
CONCLUSIONS
The analysis of watershed characteristics and stream nutrient runoff
data for a nationwide set of 928 nonpoint-source (NPS) type watersheds
indicates that:
1. There were good correlations between general land use and NPS
nutrient concentrations in streams. Streams draining agricultural
watersheds had, on the average, considerably higher nutrient
concentrations than those draining forested watersheds. Nutrient
concentrations were generally proportional to the percent of land
in agriculture and inversely proportional to the percent of land
in forest. Mean concentrations of both total phosphorus and total
nitrogen were nearly nine times greater in streams draining agri-
cultural lands than in streams draining forested areas.
2. In general, inorganic nitrogen made up a larger percentage of
total nitrogen concentrations in streams with watersheds having
larger percentages of agricultural land. The inorganic nitrogen
component increased from about 18% in streams draining forested
areas to almost 80% in streams draining agricultural watersheds.
The inorganic (orthophosphorus) portion of the total phosphorus
component stayed roughly at the 40% to 50% level regardless of
land use type.
3. Several regional patterns were found in the relationships and
interrelationships between macro-watershed characteristics and
stream nutrient concentrations. The most noteworthy of these
were:
a. Mean annual phosphorus concentrations in streams draining
forested watersheds in the west were generally twice as high
as those in the east.
b. In the east, mean inorganic nitrogen concentrations in
streams draining forested watersheds were 2.3 to 3.3 times
higher than in the western or central parts of the country.
The inorganic-N concentrations were particularly high in New
York, Pennsylvania, and western Maryland.
c. Total and inorganic nitrogen in streams draining agricultural
watersheds were considerably higher in the heart of the corn
belt than elsewhere.
1
-------
4. Differences in nutrient loads in streams associated with different
land use categories were not as pronounced as differences in
nutrient concentrations. Both mean total phosphorus and mean
total nitrogen export from agricultural lands were about 2.9 times
greater than that from forested lands; mean inorganic nitrogen
export was over 13 times greater. Differences in magnitude between
the relationships of concentration to land use and export to land
use appear to be due mainly to differences in areal stream flow
from different land use types, and to a lesser degree, to differ-
ences in the mean annual precipitation patterns and mean slope of
study areas.
5. Relationships between nutrient levels in streams and "contributing"
land use types (percent of drainage area in agricultural land plus
the percent in urban land use [% agriculture plus % urban]) cor-
related better than those considering only one land use type.
Separate regression analyses of % agriculture plus % urban against
both the total and inorganic forms of phosphorus and nitrogen
concentrations were performed. Regional equations from these
analyses offer a limited predictive capability. However, slightly
more complicated regional equations, which consider both the
positive correlation between "contributing" land use percentage
and stream nutrient concentrations and the negative correlation
between percent of watershed in forest and stream nutrients,
afford a somewhat better predictive capability, particularly for
the western and central parts of the United States.
6. Because of the general availability of stream flow data and the
difficulty of incorporating two factors with differing spatial and
temporal variations into one model, stream nutrient export models
were not constructed. It appears that the most accurate method of
predicting export values or stream loads is to use the appropriate
model to determine stream nutrient concentrations, obtain flow
data from the U.S. Geological Survey Surface Water Records for the
particular area(s) of interest, and make the necessary calculations
7. Qualitative refinement of the simple prediction models for total
phosphorus, orthophosphorus, total nitrogen and inorganic nitrogen
concentrations are provided in maps of the residuals of each
model. These maps indicate where, in the United States, nutrient
concentrations can be expected to be greater, equal to, or less
than those predicted by the models.
8. Three colored maps of NPS-related concentrations of total phos-
phorus, total nitrogen, and inorganic nitrogen in streams were
compiled from interpretations of the NES data and relationships
between the data and macro-watershed characteristics. They provide
planners, managers, and other users with a broad national overview
of the NPS stream-nutrient-level relationships. In addition, a
quick and relatively accurate method utilizing these maps is pro-
vided for predictions of NPS stream nutrient concentrations in
2
-------
watersheds, basins, or regions where detailed information necessary
for more accurate predictions is unavailable.
9. Using a geological classification scheme based on origin and the
National Eutrophication Survey nutrient data, no clear relationships
were found between geology and phosphorus or nitrogen in streams.
Another classification scheme based on phosphorus composition of
major rock types revealed only slight relationships, too slight to
provide assistance in the compilation of predictive-models.
10. No good correlations were found between drainage density (stream
lengths per unit area of watershed) and stream nutrient levels.
3
-------
SECTION II
INTRODUCTION
With increasing population densities, higher standards of living, and
increasing demands on natural resources in our country, we are becoming more
and more aware of the insidious effects our existence, life styles, and
demands are having on the water quality in our lakes and streams. Public
Law 92-500, enacted in 1972, was a result of this awareness and our interest
in understanding just how we are affecting the quality of our surface waters
and what our alternatives might be for doing something about it, should we
choose to do so. One of the key areas addressed by Public Law 92-500,
contained in Section 208, deals with the impacts of nonpoint sources on
surface waters. The area of nonpoint source (NPS) impact is extremely
difficult to understand, much more so than point-source impact, because of
the spatial, as well as temporal, relationships and interrelationships
involving the various water quality parameters and the various nonpoint
sources, both natural and anthropogenic.*
The primary purpose of this study is to help clarify one aspect of the
water quality—NPS impact area, that of phosphorus and nitrogen levels in
streams. Although recently considerable effort has been spent on gaining a
better understanding of the subject, and particularly on developing predic-
tive methods, most of this work has been site specific. That which has been
directed toward providing regional and national perspectives and related
predictive methods has been largely theoretical and/or based on empirical
data from different sources, taken at different times, and using different
sampling and laboratory methods. Understandably, the results of these
studies have been varied, and application of the results by planners and
managers in their problem assessment tasks has been cumbersome and frequently
without sound basis.
The relationships and predictive tools provided in this paper are based
on tributary data collected from a nationwide set of 928 nonpoint-source
watersheds in which land uses and other NPS characteristics were generally
typical of the areas they represented. The stream sampling was conducted
monthly forgone year. Sampling and laboratory methods were uniform through-
out the entire data set. For the purpose of this study, NPS watersheds are
defined as thosewithout municipal and industrial waste discharges and
animal feedlots identified as point sources.
*As used in this report anthropogenic refers to influences or impacts of
man's activities on natural systems.
4
-------
It is felt that the relationships and predictive methods presented in
this study will be helpful in giving planners and managers a spatial picture
of nutrient levels which can be attributed to nonpoint sources in their
geographical areas of interest as well as logical, easy to use, general
predictive tools for use in areas where extensive hard data are unavailable
for more exacting procedures. The results presented should also be useful
for interpreting the applicability of site-specific study results.
5
-------
SECTION III
APPLICATIONS
For those tasked with assessing the impacts of nonpoint sources on
nutrient levels in streams, the avenues of approach are generally dictated
by one or more of the following: (1) availability of resource materials,
such as stream nutrient data, land use maps, and aerial imagery; (2) availa-
bility of time and personnel; and (3) availability of funds. The best
possible situation, of course, is to have enough stream nutrient data to
make the necessary assessments quickly and at little expense. If, on the
other hand, a planner lacks sufficient stream nutrient data to make the
assessments but has enough time and money to conduct a sampling program,
collecting the data through stream sampling would be the logical route. The
methods outlined in this study are intended for use in those situations
where adequate stream nutrient data are unavailable and time and/or money
constraints preclude a sampling program. In this section, examples are
given to indicate when and how the suggested methods might be used.
One of the most likely situations a planner is apt to encounter is a
lack of stream nutrient data to make assessments of NPS impacts as well as a
lack of time and money to conduct a sampling program. However, for the
particular geographical area of interest, it is probable that there are
either good quality land use maps or recent usable aerial imagery from which
the general land use can be determined. Even if the land use has to be
determined from available aerial photography, the cost (in terms of time and
funds) probably is a small fraction of that needed to complete a stream
sampling program. The suggested approach for this type situation would be
use of the regional regression models shown in Table 9 (page 62) together
with the residuals of the model.
A simple example of this is provided here, considering the estimation
of total nitrogen stream levels attributable to nonpoint sources in the
Willamette River watershed upstream from Salem, Oregon. Relatively accurate
percentages of the forested, agricultural, and urban land use can be obtained
from land use maps and/or aerial imagery; but for purposes of this example,
percentages were estimated from more general sources. Forest was estimated
to cover about 49% of the watershed; agriculture, 37%; and urban, 1%. Using
the total nitrogen concentration model for the western region (Table 9; page
62) and the regional characteristics of values used to compile the model
(Figure 27; page 69), the prediction for the sample watershed is made as
follows:
6
-------
Log10 (NCONC) = -0.03665 + 0.00425 (38) - 0.00376 (49)
Log10 (NCONC) = -0.03665 + 0.16150 - 0.18420
Log]0 (NCONC) = -0.05935
However, judging from the values shown in Figure 27, one would expect the
predicted values in this geographical area to be at least 0.5 standard
deviations below that predicted by the model. Since the standard deviation
is the Loglf) value of the multiplicative standard error, Log,, of 1.75 =
0.24304 ana 0.5 standard deviations is 0.12152. Therefore,
Log]0 (NCONC) = -0.05935 - 0.12152
Log10 (NCONC) = -0.18089
TNC0NC = 0.659 mg/1
The mean annual export (load) of total nitrogen attributable to non-
point sources is computed using U.S. Geological Survey (USGS) flow records
(which if not available for a given location can be estimated based on USGS
flow per unit area records for similar adjacent or nearby watersheds) and
the equation given on page 23 of this report. Total nitrogen export for the
sample watershed is computed as follows:
Annual T-N Load = 0.659 mg/1 x 674 m3/sec. x 31,536
Annual T-N Load = 14,007,200 kg.
Using the area for the sample watershed (18,900 km2), the export per
unit area is computed to be 741 kg. total nitrogen per square kilometer per
year.
The procedures for predicting inorganic nitrogen, total phosphorus, and
orthophosphorus stream levels are similar. Moreover, the procedures for
predicting the nutrient levels attributable to nonpoint sources up-gradient
from any point on any stream are similar. Only general land use (as defined
in this study) and stream flow (if export predictions are desired) data need
to be computed from other sources for each watershed. It should be stressed
that these procedures are estimative; interpretations of their accuracy are
given in Section V of this paper. Logically, accuracy will be better in
Parts of the United States where densities of NES data were the greatest
and/or where relationships between NES data and NPS watershed characteristics
appear most significant. In areas where surface runoff is minimal or topo-
graphic watersheds are difficult or impossible to define, the procedures
will be less accurate.
Another situation a planner is likely to encounter is also one of
lacking stream nutrient data and sufficient time and money to carry out
a sampling program, but in this case either being unable to obtain good
quality land use data or lacking time or money necessary to obtain aerial
imagery (if available) and interpret it for land use identification. For
7
-------
situations such as this, quick and relatively accurate assessments generally
can be made using the colored maps in the back cover jacket of this report.
Illustration of how the maps can be interpreted to make these assessments is
provided below by again using as an example the assessment of total nitrogen
levels from nonpoint sources in the Willamette River watershed upstream from
Salem, Oregon. The general procedures are as follows:
1. Outline the topographic watershed of the Willamette River at Salem
on a 1:500,000 0SGS topographic map and a 1:3,168,000 USGS map of
the U.S. (1:500,000 USGS topographic maps are available for all
the states and the 1:3,168,000 map is a standard USGS wall map on
which major water courses and cities are shown). The delineation
should be performed on the 1:500,000 scale first so it may be used
as a guide when determining the watershed on the 1:3,168,000
seale.
2. Enlarge the Willamette River watershed portion of the total
nitrogen concentrations colored map to the 1:3,168,000 scale using
a reducing/ enlarging projector such as a Lacy-Luci Visualizer, a
Kail Reflecting Projector or a Map-0-Graph Projector. State
outlines should be used for registration; i.e., to insure proper
scale alignment of the watersheds.
3. At the enlarged scale (1:3,168,000), draw the watershed outline
and total nitrogen concentration map units (Figure 1). Then
using a transparent equidistant dot pattern, determine the percent
of the total watershed each map unit comprises. This method is
explained in the Data Collection Methods sub-section. For very
small areas, transparencies of 8, 10, and 12 line/inch grids can
be used with the line intersections constituting the "dots".
4. The weighting process used to compute the mean annual total
nitrogen stream concentration attributable to nonpoint sources is
shown i n Table 1.
TABLE 1. COMPUTATION OF MEAN ANNUAL NONPOINT SOURCE TOTAL NITROGEN CONCEN-
TRATION (mg/1) IN THE WILLAMETTE RIVER AT SALEM, OREGON
Map Map Unit Mean of % of Watershed
Unit Range Range* Within Map Unit
1 <0.500 0.275 x 54.2 = 0.149
2 0.501 to 0.700 0.592 x 5.3 = 0.031
4 0.901 to 1.100 0.996 x 7.0 = 0.070
6 1.401 to 1.700 1.543 x 20.1 = 0.310
7 1.701 to 2.000 1.844 x 0.7 = 0.013
8 2.001 to 3.000 2.450 x 12.7 = 0.311
0.884
^Geometric mean (of two measures, the geometric mean is the square root of
their product), unless for first or last map unit in which case average
values are approximated from NES watersheds most representative of the
particular map unit(s). Geographic coordinates of all NES stream sampling
sites are in ST0RET, an EPA computer-based water quality system.
8
-------
Nonpoint Source Total Nitrogen Stream Concentration
Map Units in the Willamette River Watershed Upstream
from Salem, Oregon
Willamette River
Salem
watershed
boundary
Figure 1. An application of NES mapped nonpoint source stream nutrient
concentration interpretations.
9
-------
This system is also useful for areas where the predominant land use
types are neither agriculture nor forest; however, close attention should be
paid to the map's reliability inset relative to the location of one's subject
area. The figure obtained by the first predictive method, the mathematical
model, was about 25% lower than that obtained using the maps. This may be
due in part to the generalized method by which land use data were estimated
for use in the model or it may be that a fairly large amount of this water-
shed is in the cleared-unproductive land use category, a category not directly
considered by the models.
10
-------
SECTION IV
BACKGROUND
HISTORY AND OBJECTIVES
The initial planning for the National Eutrophication Survey (NES)
visualized a detailed watershed land use study for each of the approximately
750 lakes to be done parallel to the field sampling program. The idea
stemmed from a desire to better understand the relationship between lake
trophic state and watershed land use. It was hoped that the "fruits" of
this effort would be the development of a quick, relatively accurate method
of assessing nutrient loadings to lakes based on analysis of land use in
their watersheds.
The original concept pictured identification and mensuration of overall
land use types through aerial photo and topographic map interpretation of
the entire watershed of each lake included in the NES. For many reasons,
including the unavailability of usable photo and/or map coverage for many
watersheds or parts of watersheds, the original concept was considerably
modified. The project finally evolved into a study of over 900 nonpoint
type drainage areas associated with NES tributary sampling sites, most of
which are within watersheds of lakes that have been studied by the NES.
In its final format, the basic objectives of the NES land use study
were to: (1) investigate and gain a better understanding of the relation-
ships and interrelationships between nonpoint watershed characteristics and
stream nutrient levels, (2) develop a means for predicting stream nitrogen
and phosphorus levels based on land use and related geographical character-
istics, and (3) investigate and define possible regionalities in macro-
watershed characteristics—stream nutrient level relationships and provide
some accountability for these regionalities in the predictive methods.
Because the project was an appendage of the massive stream and lake sampling
program conducted by NES and included a large number of watersheds repre-
senting a nationwide variety of climatic and geographic conditions, it
afforded a unique opportunity to view the land use--stream nutrient levels--
eutrophication relationship on a national scale and to develop a predictive
system/or systems to reflect geographical or regional differences.
By reason of its ties with the NES field sampling program which was
accomplished in three phases, the NES land use study followed the same
pattern. One hundred and thirty three drainage areas were selected for land
use analysis in the area in which tributary sampling began in the summer of
1972. Three hundred and forty drainage areas were selected in the NES study
area where sampling was initiated in 1973; and, in the remainder of the
11
-------
conterminous United States where sampling began in 1974, 455 drainage areas
were defined. Figure 2 illustrates the distribution of the individual study
drainage areas and the overall areas covered by each of the three phases.
Upon completion of data compilation for each of these phases, a report
was to be written to present the data collected to date and to present some
analyses of these data. National Eutrophication Survey Working Paper No. 25
(U. S. Environmental Protection Agency, 1974b) presented data compiled
through the first phase. "The Influence of Land Use on Stream Nutrient
Levels" (Omernik, 1976) presented data compiled for the first two phases.
This report comprises an analysis of the entire data set.
LITERATURE
Recently with the surge of interest in the effects of nonpoint sources
on water quality, numerous literature reviews and/or assessments of the
state-of-the-art have been published on the relationships of watershed
characteristics to NPS nitrogen and phosphorus levels in streams (Weimer,
McGuire, and Gasperino, 1976; Wanielista, Yousef, and McLellon, 1977; Uttor-
mark, Chapin, and Green, 1974; McElroy et al. 1976; U.S. Environmental
Protection Agency, 1976; U.S. Environmental Protection Agency and U.S.
Department of Agriculture, 1976b; Dornbush, Anderson, and Harms, 1974;
Loehr, 1974; Dillon and Kirchner, 1975; Likens and Bormann, 1974; and Ryden,
Syers, and Harris, 1973).
Many of the publications that have assessed the state of the art in the
NPS nutrient field have also as an objective the development of methods for
quantifying the relationships. One of the most recent of these was compiled
by the Midwest Research Institute (McElroy et al., 1976). They based their
system on the Universal Soil Loss Equation (USLE), relating NPS nutrient
loadings of streams and lakes to soil erosion and sediment transport.
However, application of the USLE for estimating nutrient yields appears to
be beyond its original design (McDowell, 1976). Moreover, the factors
employed in the nutrient loading functions part of the system (which include
nutrient concentrations in soil, nutrient availability, and nutrient enrich-
ment ratios) require data which are unavailable for much of the United
States. The suggested alternatives to be used where data are unavailable,
generally call for extrapolation from dated, extremely small scale maps
(Parker et al. , 1946) and some measured values from a few experimental,
generally site-specific studies for which no theoretical approaches to broad
scale prediction have been attempted (U. S. Environmental Protection Agency,
1976; Viets, 1975; Stoltenberg and White, 1953; Massey, Jackson, and Hays,
1953; and Kilmer [as cited by McElroy et al., 1976]). The USLE was designed
to predict soil loss from sheet and rill erosion; it was not designed to
estimate the amount of sediment entering a stream system, and it does not
account for deposition within watersheds (Wischmeier, 1976). Therefore, the
applicability of the USLE to areas larger than a field of a few acres, e.g.,
a 100 km2 watershed with different land uses and topographical variations,
is extremely questionable.
Two earlier studies that reviewed the literature on NPS nutrient
relationships, as well as attempted to provide predictive procedures, were
12
-------
*• £ A
30-
DISTRIBUTION OF
N.E.S. NONPOINT
SOURCE WATERSHEDS
Each of the 928 dots
represents a stream
sampling site and its
associated drainage area.
'72/73, and 74
refer to the years stream
sampling began in each
group of states.
Figure 2. Distribution of individual NES nonpoint source study watersheds.
-------
published by Uttormark, Chapin, and Green (1974) and Weimer, McGuire, and
Gasperino (1976). The authors gathered data from many of the investigations
that, by and large, based results on information collected from a small
number of drainage areas within specific geographic regions. In attempting
to develop systems for estimating nutrient runoff from land use based on
coefficients developed entirely from the literature, these reviewers have
summarized their findings by presenting ranges of values and midpoints or
averages. Generally, these ranges are quite wide and the midpoints, or
other indicators of central tendency, do not vary from one land use type to
another as appreciably as one might expect.
It would seem that uniformity in procedure, which is largely lacking
from one study to another, would limit the validity of comparing the results
of one with another or using combined results to establish nutrient loading
coefficients. More importantly, there is an insufficient quantity and an
inadequate distribution of data points available from literature sources to
study the regional aspects of nutrient runoff or to be able to distinguish
between regional and overall relationships. Examples of the latter point
are shown later in this paper in the sections dealing with the effects of
geology and drainage density on nutrients in streams.
SELECTION OF STUDY WATERSHEDS
In general, criteria for selecting tributary sites for study drainage
areas were:
A. Absence of identifiable point sources.
B. Availability of usable aerial photography (preferably
in scales of from 1:40,000 to 1:80,000) and/or existing
general land-use data.
C. Availability of accurate topographic maps for drainage
area delineation.
D. Sufficient relief for clear definition of drainage area
1imits.
E. Sufficient precipitation to facilitate sampling and provide
significant surface runoff.
F. Need to encompass a variety of geographic and climatic
areas and obtain, where possible, land-use homogeneity
within subdrainage areas.
A few exceptions to these criteria were necessary to accomodate study
of particular types of areas. Within the 1973 study area, several heavily-
mined watersheds and several predominately urban watersheds (but without
apparent industrial or municipal wastewater treatment facilities) have been
included.
14
-------
It should be noted that an overriding selection constraint was that
tributary sites for study drainage areas had to be among those already
selected for inclusion in the NES or had to be selected within a reasonable
distance of NES lakes to accomodate sampling by the National Guard. At the
time selections were being made of drainage areas within the 1972 study
area, tributary sites had already been selected and the field sampling was
underway. Moreover, at that time the major thrust of the Survey was on
point-source impact to lakes, particularly municipal wastewater treatment
plant discharges. Generally, only problem lakes were selected for study,
and many of these were in watersheds having problem-type land uses. Hence,
it was somewhat difficult to find drainage areas without point sources and
obtain adequate coverage of all land use types.
At the time tributary sites were being selected for 1973 NES lake
studies, emphasis was still on point sources. However, tributary sampling
had not yet begun in several of the states which allowed selection of addi-
tional sites where their inclusion was warranted by land use homogeneity and
other factors suitable for land use and nutrient runoff analyses.
Selection of tributary sites for land use study drainage areas in the
1974 area (west of the Mississippi) was made under more ideal conditions.
By that time, water research mandates had been revised by passage of Public
Law 92-500. These revisions resulted in a broadening of Survey objectives
to include assessment of relationships of nonpoint sources to lake nutrient
levels. In addition, lake selection criteria were modified to no longer
include just problem lakes, but lakes representative of the full range of
water quality. This presented a better balance of lake watersheds and land
use types as well as lake types. In several instances where it was necessary
to achieve a better geographic distribution of drainage areas or obtain a
better balance of land use categories, stream sampling sites were selected
outside NES lake watersheds. Figure 2, page 13, reveals a somewhat uneven
areal distribution of data points. Major gaps in geographical coverage of
study areas are in (1) the low lying Atlantic and Gulf coastal plains, (2)
the Great Plains west of the 100th meridian, and (3) the Intermontane Basin
and Range region. Selection of stream sampling sites in these areas was
generally precluded because one or more of the above mentioned criteria
could not be met. The most common problem in the west was intermittent
stream flows resulting from aridity. In the east, the density of point
sources and insufficient relief for definition of watershed boundaries were
major problems.
DATA COLLECTION METHODS
Drainage Area Measurement and Land Use Identification
Following tributary site selection, individual drainage areas were
delineated on U. S. Geological Survey (USGS) topographic maps and their
areas were determined with a compensating polar pianimeter or an electronic
Planimeter. General land use identifications were made using recent aerial
Photography and/or land use maps. Agencies or companies from which the
Photos or maps were obtained were:
15
-------
1. U. S. Department of Agriculture, Agricultural Stabilization
and Conservation Service. (487)*
2. U. S. Department of Agriculture, Forest Service. (136)
3. National Aeronautics and Space Administration, Lyndon B. Johnson
Space Center, Houston, Texas. (98)
4. U. S. Department of Agriculture, Soil Conservation Service. (66)
5. National Aeronautics and Space Administration, Ames Research
Center, Moffett Field, California. (49)
6. New York Land Use and Natural Resources Inventory, Cornell Univer-
sity, Ithaca, New York. (59)
7. U. S. Department of Interior, Geological Survey. (36)
8. State of Wisconsin, Department of Transportation. (18)
9. Mark Hurd Aerial Surveys, Inc., Minneapolis, Minnesota. (15)
10. Minnesota Land Information System, Minnesota State Planning Agency
(15)
11. State of Washington, Department of Natural Resources. (10)
12. Environmental Remote Sensing Applications Laboratory, Oregon State
University, Corvallis, Oregon. (9)
13. U. S. Department of Interior, Bureau of Land Management (photos
obtained through Environmental Photographic Interpretation Center,
EPA Warrenton, VA). (6)
14. Department of Air Force, Headquarters Rome Air Development Center,
Griffiss Air Force Base, New York. (5)
15. Arizona State Highway Department. (1)
16. Cartwright Aerial Surveys, Inc., Sacramento, California. (1)
Finished land use maps were generally unavailable for most of the
country at the time the study was conducted. The New York Land Use and
Natural Resources Inventory and the Minnesota Land Information System
comprised the only sources for usable land use maps. All the other sources
listed were for aerial imagery.
Land use categories included: (1) forest, (2) cleared-unproductive,
(3) rangeland, (4) agriculture, (5) urban, (6) wetland, and (7) "other"
*The number in parenthesis following each source indicates the approximate
number of study watersheds covered either completely or in part by that
particular source's photos or maps.
16
-------
(including barren, extractive, and open water). The land use categories
roughly correspond in level of classification to Level I of the recently
developed USGS Land Use Classification System (Anderson et al., 1976).
Land Use Percentage Computation
For each drainage area, percent coverage of each land use type was
compiled by use of equidistant dot pattern overlays. The dot patterns were
placed non-selectively over the map overlays on which land use units had
been outlined. To determine the land use percentage for a given drainage
area, the number of dots that fell on each land use category was totaled,
each total was multiplied by 100, and the resulting products were divided by
the number of dots falling in the drainage area. Dot pattern densities
varied from one drainage area to another depending on the overall size of
the drainage area; generally, the larger the area the less dense the dot
pattern. At least 400 dots per drainage area, but for convenience preferably
less than 800, were needed for a valid determination of percent coverage.
Animal Unit Density Computations
It is generally accepted that animal wastes are major contributors of
the nitrogen and phosphorus in agricultural land runoff (Holt, Timmons, and
Latterell, 1970; Holt, 1971; Robbins, Howells, and Kriz, 1971). Early in
this study, it seemed some mechanism should be developed to analyze this
aspect of agricultural runoff. Because of shifts in agricultural land use,
particularly from season to season and year to year, it seemed impractical
and perhaps impossible to accurately separate pasture from cropland. A more
expeditious method was to determine overall animal densities (i.e., animal
units per acre of subdrainage area).
For the most part, animal unit densities for each drainage area were
computed from U. S. Census of Agriculture figures, other literature sources,
and personal communications (Johnson and Mountney, 1969; Miner and Willrich,
1970; Miner, 1971; U.S. Department of Commerce, Bureau of Census, 1972a;
Wisconsin Statistical Reporting Service, 1973; Anonymous, 1974; Arscott,
1975; Harper, 1975; Hohenboken, 1975; and Miner, 1975). The quantities of
total nitrogen and total phosphorus produced annually by common farm animals
were also compiled from these sources (Table 2).
TABLE 2. ANIMAL NUTRIENT PRODUCTION (kg/yr/animal)
Total P Total N
Cattle 17.60 57.49
Hogs 3.23 9.68
Sheep 1.47 10.06
Poultry
Layers 0.16 0.42
Broilers 0.09 0.39
Turkeys 0.39 0.84
17
-------
These data, together with Census of Agriculture figures for each
county, were used to compile animal unit densities per drainage area using
the following equations:
1 D C + (0.184-H)+(0.084-S)+(0.0093-P,)+(0.0011-P,)+(0.0222-P.)
a _ 3 3 IDT.
*p- rc 1 ;
and
1 D C + (0.169-H)+(0.175-S)+(0.0073-P1)+(0.0015-P. )+(0.0147-P.)
^ _ 3 3 I D L
*c 1 ;
where: A^ = Animal units per square kilometer for Total P;
An = Animal units per square kilometer for Total N;
Ac = Total agricultural land (by county) in square kilometers;
D = Percent of subdrainage area in agriculture;
d
C = Total cattle and calves (by county);
3
H = Total hogs and pigs (by county);
S = Total sheep and lambs (by county);
P-j = Total layers (by county);
Pb = Total broilers (by county); and
P^ = Total turkeys (by county).
For drainage areas located in more than one county, weighted unit
densities were determined based on the amount of agricultural land in each
county. The coefficients were adjusted to reflect average animal weights
relative to an average weight for cattle and calves (same sources as for
Table 2). It should be noted that coefficients for poultry take into con-
sideration average life spans and broods per year.
Geology Identification
Some recent works on NPS nutrients in streams have given as much or
more emphasis to the effects of geology than to the effects of land use
(Dillon and Kirchner, 1975; Thomas and Crutchfield, 1974; Stone, 1974;
Betson and McMaster, 1975; and Likens and Bormann, 1974).
None of these studies approached the geology—stream nutrient level
relationships subject from a national perspective. All were based on data
which were completely, or for the most part, collected from one specific
basin or geographical region. In addition, there appears to be little
18
-------
agreement as to which aspect of geology is related to stream nutrient levels.
Dillon and Kirchner noted significant differences in phosphorus loads in
streams from watersheds of sedimentary origin versus those of igneous origin,
and that a furthur breakdown of the igneous classification to plutonic
(intrusive) versus volcanic (extrusive) seemed to be important. Thomas and
Crutchfield, dealing exclusively with sedimentary watersheds, concluded that
the presence of medium- to high-phosphate limestone resulted in higher
phosphorus and nitrogen stream concentrations than where these geologic
types were absent and others such as sandstone and shale were present.
Stone studied one watershed with limestone and one without limestone and
concluded that this geologic characteristic was unrelated to stream phosphate
concentration but that the anthropogenic aspects, such as human and animal
densities and land use, were significantly related. Using these studies as
a guide and with an aim of providing as many combinations of possible rela-
tionships as possible given the source materials available for the entire
data set, the following general breakdown was used to classify NES study
subdrainage areas:
1. Sedimentary-mixed
2. Sedimentary-mixed, without limestone
3. Sedimentary-1imestone
4. Sedimentary-sandstone
5. Sedimentary-shale
6. Sedi mentary-undi fferenti ated
7. Sedimentary-clay
8. Sedimentary-glauconite, phosphate rock, or carbonate bearing
9. Sedimentary-marine sediments
10. Sedimentary-megafossils
11. Igneous-mixed
12. Igneous-plutonic
13. Igneous-plutonic-acidic
14. Igneous-plutonic-basic
15. Igneous-volcanic
16. Igneous-volcanic-acidic
17. Igneous-volcanic-basic
19
-------
18. Metamorphic-mixed
19. Metamorphic-derived
from
sedimentary-mixed
20. Metamorphic-derived
from
sedimentary-mixed without limestone
21. Metamorphic-derived
from
igneous
22. Metamorphic-derived
from
igneous-plutonic
23. Metamorphic-derived
from
igneous-plutonic-basic
24. Metamorphic-derived
from
igneous-plutonic-acidic
25. Metamorphic-derived
from
igneous-volcanic
26. Metamorphic-derived
from
igneous-volcanic-basic
27. Metamorphic-derived
from
igneous-volcanic-acidic
28. Mixed
29. Composition Unknown
Where drainage areas included two of the above classifications, com-
binations were shown with the predominant type first. Sources for these
data were mostly state and federal geologic maps of varying dates and scales,
although most were of individual states at scales ranging from 1:250,000 to
1:500,000.
An abreviated version of the above breakdown, based on geologic origin,
was used in the earlier NES report on the eastern 473 watersheds (Omernik,
1976). In that report, it was concluded that on the basis of that particular
classification scheme and the NES eastern data set, there was no apparent
significant effect of geology on either phosphorus or nitrogen levels in
streams. It was suggested that, insofar as phosphorus levels in streams
were concerned, a more appropriate classification scheme be used based on
the mineral composition of rocks rather than entirely or primarily on origin.
Based largely on the data generated by Goldschmidt (1958) and Van Wazer
(1961), a list of phosphorus composition percentages for the most common
rock types was prepared (Table 3).
20
-------
TABLE 3. TOTAL PHOSPHORUS COMPOSITIONS OF ROCK TYPES
Rock Type Total P Composition (%)
Sedimentary
Limestone 0.020
Sandstones 0.040
Shales 0.080
Red clay 0.140
Sedimentary-mixed (averaged) 0.070
Igneous
Rhyolite 0.055
Granite 0.087
Andesite 0.123
Syenite 0.133
Monzonite 0.139
Diorite and Dacite 0.144
Gabbro 0.170
Basalt 0.244
Igneous rock5" 0.118
Igneous-plutonic (averaged) 0.134**
Igneous-volcanic (averaged) 0.141**
*An average of the values given by Goldschmidt (1958) and
Van Wazer (1961).
**Averages of values for plutonic or volcanic types given in this
table. No data are available from the literature on igneous-
plutonic or igneous-volcanic per se.
Using the best available geologic maps for each of the study watersheds,
the major rock type, or types, were identified and their general extent
computed. Then the percent of total area of each watershed underlain by
each rock type was multiplied by percentage for that particular rock type
(Table 3). By weighting on the basis of areal extent of each rock type,
this procedure was continued until an overall phosphorus composition had
been determined. Metamorphic rocks were assigned the values for rocks from
which they were derived. Although information on metamorphic rocks is
gather scarce compared to that available for igneous and sedimentary rocks,
it appears that metamorphism produces no detectable change in the distribu-
tion of elements unless the rocks have been permeated by other materials
during metamorphism (Krauskopf, 1967). Figures for sedimentary-mixed,
igneous rock, igneous-plutonic and igneous-volcanic were included in Table 3
to accommodate computation of values for watersheds where source materials
of more detailed levels of classification were unavailable.
21
-------
Slope and Drainage Density Computations
For each drainage area, mean slope was calculated using an equidistant
dot pattern overlay. Less dense patterns were used for this work than were
used for computing land use percentages. For this procedure, 40 to 80 dots
per drainage area were used. The overlays were placed randomly over the
topographic maps on which the drainage areas had been outlined. Then, using
an appropriate slope indicator, the percent of slope for the points under
each dot was calculated. The data were then totaled and divided by the
number of points falling in the drainage area. The slope indicators used
were transparent templates indicating percent of slope from distances between
map contours, adjusted for map scale and contour interval.
Average drainage densities for 291 of the study areas were computed by
dividing the total length of "blue-line" streams (both perennial and inter-
mittent) by the watershed area. For purposes of this study, drainage density
is defined as total mapped stream lengths per unit area of watershed and is
expressed in kilometers per square kilometer. These data were compiled
through the use of an electronic planimeter capable of both linear and areal
mensuration. In all cases, the most recent U.S.G.S. 1:24,000 or 1:62,000
scale topographic maps were utilized. Where both scales were available for
a given area and the map compilation dates were similar, the 1:24,000 scale
was chosen. Because of the time involved in determining these data, the
compilations were made only for those watersheds which were in the predomi-
nantly (>90% and >75%) forested and predominantly agricultural land use
categories.
Soi1 pH Calculations
Early in the course of this project, it was found that even a good
approximation of mean surface soil pH for each of the over 900 drainage
areas was not available except through time-consuming work with numerous
large-scale maps from widely scattered sources and contact with local soils
scientists. The time and expense prohibited this approach, at least for the
present. The best alternative source was a collection of estimates of
surface soil pH ranges for map units appearing on the National Atlas soils
map (Smith, 1975; and U. S. Geological Survey, 1970). Each drainage area
was identified with a midpoint of the pH range for the soils map unit pre-
dominant within it. The limitations of this method are covered in the
discussion on nutrient runoff-soils relationships section of this report.
Precipitation
Mean annual precipitation data were compiled from National Oceanic and
Atmospheric Administration (N0AA) climatological data publications. In
computing these data, consideration was given to the locations of climato-
logical data collection site(s) in relation to each of the study drainage
areas, especially with regard to elevation and exposure, and adjustments
were made accordingly.
22
-------
Nutrient Concentration and Export Estimates
Explanations of the following procedures are given in NES Working Paper
No. 1 (U. S. Environmental Protection Agency, 1974a); and NES Working Paper
No. 175 (U. S. Environmental Protection Agency, 1975):
1. Tributary sampling methods and handling.
2. Analytical methods (stream samples).
3. Nutrient loading estimates.
4. Stream flow estimates.
Mean annual concentrations were computed as the arithmetic mean of all
the values for a given sampling site for the year of sampling. Where mean
annual values were given for a group of watersheds they represent the arith-
metic mean of the arithmetic mean values for all of the data points within
that group.
It should be noted that nutrient exports reported herein were computed
using "normalized" flow data (long-term averages) from USGS and drainage
area measurements as determined by NES. For a few tributary sampling sites
where USGS did not provide flow estimates, they were calculated by NES from
runoff patterns in adjacent, overlapping, or nearby areas for which USGS
estimates were provided. Loadings for all drainage areas in this study were
estimated according to the following equation:
Annual Load = (C)(F)(31,536)
where: C = Mean annual concentration in milligrams per liter, and
F = Mean normalized annual stream flow in cubic meters per second.
The factor 31,536 is used to obtain loads in kilograms per year. The
annual loads were then divided by the area of their respective watershed in
square kilometers.
23
-------
SECTION V
RESULTS AND DISCUSSION
This section discusses the analysis of land use, other drainage area
characteristics, and stream nutrient runoff data compiled for the entire set
of drainage areas within all three groups of states covered by the NES. The
raw data are presented in Appendix A, and the distribution of the data
points (comprising the 1972, 1973 and 1974 areas) are illustrated in Figure
2 (page 13).
AREAL DISTRIBUTIONS OF DATA
After compiling the data presented in Appendix A, several types of
these data were sorted into classes and plotted on maps of the United States
using a graduated color scheme. It was theorized that this might aid in
uncovering correlations, regional patterns, and (by comparison with maps of
other macro aspects such as physiographic regions, fertilizer usage, human
population densities, domestic animal densities, geology, soils, and climate)
possible covariants. Maps were compiled for total P concentrations, total N
concentrations, ortho-P concentrations, inorganic N concentrations, total P
export, total N export, land use categories, flow per unit area per year,
and mean slope. The expense of publishing these colored maps precluded
their inclusion in this report. Less costly black and white versions, such
as those used in the previous report on the eastern data set, do not ade-
quately illustrate the patterns. Nevertheless, the gross patterns illus-
trated by the colored maps of value distributions indicate certain correla-
tions and these are worthy of mention.
The map depicting the various overall land use categories of the study
watersheds revealed what might be expected from knowledge of general land
use patterns in the United States. Generally, in parts of the country where
the land use is mostly agricultural, so were the study areas. Conversely,
in parts of the country where the land use is mostly forested, so were the
study areas. In the mixed-land use regions, such as west-central New York,
south-central Kentucky and northeastern Texas, the predominant land use in
the study areas varied from mostly forest to mostly agriculture, with a
scattering of predominantly urban and/or cleared unproductive. Generally,
it was evident that land use classifications in the NES nonpoint-source
watershed data set were typical of these found in the geographical area they
represent. Considering the objectives of this study this seems quite note-
worthy.
The map illustrating the areal distribution of mean total phosphorus
concentrations disclosed a pattern roughly similar to that of the land use
24
-------
map, and hence a possible correlation. Comparison of the total P and land
use maps revealed several groups of drainage areas with notably lower phos-
phorus concentrations than might be expected from land use alone. These
areas were in southwestern Missouri, east-central Maryland, Pennsylvania,
and parts of central and eastern Ohio and New York (particularly the Finger
Lakes area).
Two groups of drainage areas with somewhat higher phosphorus concen-
trations than might be expected from an association with land use were (1)
the predominantly agricultural region extending from central and eastern
North Dakota to southeastern Nebraska and (2) a region extending from
eastern Texas through Louisiana into western Mississippi.
The map of total nitrogen concentrations also had a pattern similar to
the map of land use. Some noteworthy differences were: (1) particularly
high total nitrogen concentrations in the Midwest (especially the "corn
belt") and the middle Atlantic region, from Maryland and Delaware through
New Jersey and southwestern Connecticut; and (2) very low values in northeast
Texas, Western Louisiana, southwest Arkansas, and in the mixed farming areas
of the central and southwestern Appalachians and Piedmont portion of the
Southeast.
Examination of the areal distributions of both total P and total N
export values revealed some similarities in pattern to that of land use but
far less than the likenesses between nutrient concentrations and land use.
The distribution maps of slope and flow (discharge/unit area) values were
constructed to study the possible relationships of these factors to the
differences in correlation between nutrient concentration and land use and
nutrient expert and land use. Analyses of these data will follow in other
sections of this paper.
OVERALL LAND USE-NUTRIENT RUNOFF RELATIONSHIPS
Category Definitions
Individual drainage areas were assigned overall land use categories
according to the following criteria:
1. Forest; negligible extents of other types
a. > 90% forest (including forested wetland)
b. <2% agriculture
c. < 1% urban
d. < 5% rangeland
2. Mostly forest; minor extents of other types
a. > 75% forest
25
-------
b. < 7% agriculture
c. <2% urban
d. < 2% rangeland
e. not included in > 90% forest category
Mostly forest; major extents of other types
a. > 50% forest
b. not included in > 90% forest or > 75% forest categories
Agriculture; other types negligible
a. > 90% agriculture
b. <3% urban
Mostly agriculture; minor extents of other types
a. > 75% agriculture
b. <7% urban
c. not included in > 90% agriculture category
Mostly agriculture; major extents of other types
a. > 50% agriculture
b. not included in > 90% agriculture or > 75% agriculture
categories
Rangeland
> 75% rangeland
Mostly rangeland; remainder predominantly agriculture and/or
urban
a. > 50% rangeland
b. > 20% agriculture and urban
c. not included in other mostly rangeland categories
Mostly rangeland; remainder largely forest, cleared unproducti
and/or wetland
a. > 50% rangeland
26
-------
b. > 20% forest, cleared unproductive, and/or wetland
c. not included in other mostly rangeland categories
10. Cleared unproductive
> 75% cleared unproductive
11. Mostly cleared unproductive
> 50%, but < 75% cleared unproductive
12. Urban
> 40% urban
13. Mixed; not included in any other category
For ease in referencing, the categories which were differentiated from
others mainly by differences in the coverage percentages of their major land
uses were named after those differentiating percentages (e.g., > 75% forest,
> 50% forest, etc).
General Analysis
The relationships between these overall land use categories and nutrient
runoff are illustrated in Figures 3 and 4. It should be emphasized that
these graphs represent mean annual nutrient values for all watersheds in
each land use category from the entire 45 state area and do not reflect
regional relationships. For example, one should not conclude from Figure 3
that total phosphorus concentrations in streams draining > 50% agricultural
watersheds in Vermont and New York will average about 0.085 miligrams per
liter. Analysis of stream nutrient data for this geographical region indi-
cates average values are somewhat lower. By the same token, higher concen-
trations are generally found in streams draining > 50% agriculture watersheds
in Iowa. Obviously, other interrelated factors such as agricultural prac-
tices, slope, soils, climate, etc., are important.
However, Figures 3 and 4 illustrate some significant overall land use-
stream nutrient concentration relationships. Nutrient concentrations are
substantially higher in streams draining agricultural lands than in streams
draining forested areas. The concentrations are generally proportional to
the percent of land in agricultural use. Both total phosphorus and total
nitrogen concentrations are roughly nine times higher in streams draining
predominantly agricultural (>90% agriculture) watersheds than in streams
draining predominantly forested (>90% forest) watersheds. However, ignoring
the land use categories represented by small data sets (N < 50), the rate of
increase in nutrient concentration with increased amounts of agricultural
land usage appears to be more constant with regard to phosphorus than nitro-
gen. Over half of the nine-fold difference in total nitrogen concentration
occurs as agricultural use increases from at least 75% of the land area to
90% of the land area.
27
-------
N
68
77
295
5
16
103
12
17
10
144
11
72
74
Land Use
vs.
Mean Total Phosphorus and Mean
Orthophosphorus Stream Concentrations
Data from 904 "Nonpoint source-type" watersheds
distributed throughout the United States
orthophosphorus concentration
.034
total phosphorus concentration
>90% Forest
- 7 5% Forest
> 50% Forest
> 75% Cleared
Unproductive
> 50% Cleared
Unproductive
Mixed
>50% Range:
Remainder predominantly
forest
> 75% Range
> 50% Range:
Remainder predominantly
agriculture
> 50% Agriculture
>40% Urban
>75% Agriculture
£90% Agriculture
.08 .10 .12
Milligrams per Liter
Figure 3. Relationships between general land use and total phosphorus and
orthophosphorus concentrations in streams.
-------
68
77
295
5
16
103
12
17
10
144
11
72
74
>90% Forest
i 75% Forest
i 50% Forest
> 75% Cleared
Unproductive
> 50% Cleared
Unproductive
Mixed
^ 50% Range:
Remainder predominantly
forest
>75% Range
£50% Range:
Remainder predominantly
agriculture
> 50% Agriculture
> 40% Urban
£ 75% Agriculture
>90% Agriculture
Land Use
vs.
Mean Total Nitrogen and Mean
Inorganic Nitrogen Stream Concentrations
Data from 904 "Nonpoint source-type" watersheds
distributed throughout the United States
inorganic nitrogen concentration
a839
t
1.009
1.297
1.383
1.536
4.233
total nitrogen concentration
3.0 4.0
Milligrams per Liter
Figure 4. Relationships between general land use and total nitrogen and
inorganic nitrogen concentrations in streams.
-------
The cleared unproductive, range, and urban land use categories were
represented by relatively few watersheds, and therefore the interpretation
of their mean nutrient runoff values is somewhat restricted. The cleared
unproductive categories included abandoned farmland, areas of recent timber
harvests, and natural grasslands and savanna or open low scrubland which was
apparently not being used for domestic animal grazing. One would expect the
anthropogenic effects on stream nutrient levels in these watersheds to be
considerably less than those in agricultural or urban watersheds but greater
than those in forested watersheds because of the possible residual effects
of former land usage. The values presented in Figures 3 and 4 do, in fact,
support this thought. The mean annual nutrient concentrations for the range
categories reveal what might be expected from range land usage. The concen-
trations are generally less than those for streams draining mostly agricul-
tural watersheds but considerably greater than those for streams draining
forested areas. Oddly, the mean values for the >50% range; remainder largely
forest, cleared unproductive and/or wetland category are not significantly
lower than the other two range categories as one might expect, particularly
regarding nitrogen.
Interestingly, based on the mean concentration values for the nationwide
data set, phosphorus would generally be expected to be the limiting nutrient
for algal growth in surface water draining either forested or agricultural
areas. The total nitrogen to total phosphorus ratios are 33:1, 27:1, 25:1,
21:1, 19:1 and 33:1 for >90% forest, >75 forest, >50% forest, mixed, >50%
agriculture, >75% agriculture, and >90% agriculture watersheds, respective-
ly. Generally phosphorus is considered the limiting nutrient when the N:P
ratio is 15:1 or greater (Vollenweider, 1968).
Figures 3 and 4 also reveal some interesting relationships regarding
the inorganic forms (commonly regarded as the available forms) of both
nutrients. Figure 3 shows the relationships between orthophosphorus (P04
as P; technically, soluble reactive phosphorus) concentrations and overall
land use categories to be about the same as those between total phosphorus
(as P) and land use. Mean orthophosphorus concentrations comprised from 41
to 50% of the mean total phosphorus regardless of land use, except for land
use categories represented by fewer than 20 watersheds.
In comparing the mean inorganic nitrogen (NH3 + N02 + N03) values with
those of mean total nitrogen (total Kjeldahl--N + N02 + N03) in Figure 4, one
can see that mean inorganic nitrogen concentrations represent an increasing
percentage of the mean total nitrogen concentrations with increased amounts
of agricultural land use. The percentages of inorganic nitrogen are 18%,
20%, 35%, 45%, 52%, 56% and 79% for >90% forest, >75% forest, >50% forest,
mixed, >50% agriculture, 75% agriculture, and >90% agriculture categories,
respectively. This probably reflects the use of inorganic nitrogen fertil-
izers and the high water solubility of inorganic nitrogen compounds.
Mean inorganic nitrogen represented over 65% of the mean total nitrogen
in the 11, >40% urban, drainage areas. As mentioned earlier, no industrial
or municipal waste treatment facilities or outfalls were known within these
urban areas. However, the time and expense that would have been involved
did not allow field verification. The data probably reflect effects of
30
-------
runoff from streets and lawns and, to some extent, the effects of septic
tanks.
It should be noted at this point that the data set used for analyses
involving nutrient concentrations numbers 24 less (total = 904) than that
used for analyses involving nutrient exports. It was felt that omitted
watersheds had too few samples, due to low or no flow most of the year, to
reflect mean annual concentrations relative to those watersheds from which
samples were collected most or all of the sampling year. However, because
the dry periods that caused the insufficient flow were generally typical of
the particular areas involved and because flow is generally more significant
than concentration in determining export, the 24 watersheds were included in
the analysis of stream nutrient export—nonpoint source relationships.
The nutrient exports per unit area of watershed for both forms of
phosphorus and nitrogen are shown in Figures 5 and 6. These data indicate
that the differences in export from different land use categories are
considerably less pronounced than the differences in nutrient concentra-
tions. Total phosphorus export was only 2.9 times greater from predomi-
nantly agricultural (>90% agriculture) land than from predominantly forested
(>90% forest) lands, and total nitrogen was only 2.8 times greater. Regarding
the inorganic forms, orthophosphorus export from agricultural watersheds was
2.3 times greater than from forested watersheds, whereas the difference in
inorganic nitrogen export was over 13-fold. For sake of comparison, inorganic
nitrogen concentrations were, on the average, 39 times greater in streams
draining >90% agriculture watersheds than they were in streams draining >90%
forest areas. Partial explanation of the difference in the relationships of
concentrations to land use and export to land use apparently lies in the
differences in stream discharge (average flow per unit area per year) between
agricultural lands and forested, or previously forested, lands. When the
entire data set was considered, regression analysis of discharge to the
percent of land in forest plus the percent of land in cleared unproductive
revealed a relatively low correlation (r = 0.32). However, when the data
were grouped by region, the correlations were higher for two of the regions
which together comprise 77% of the study watersheds. For the east, the
correlation coefficient was 0.63; for the plains, 0.59; and for the west,
0-28 (for areal definition of these regions see the insert in Figure 23.)
The reason for the low correlation in the west as well as the entire data
set may be due to a masking caused by the several severely dissected and/or
mountainous, relatively arid, forested watersheds in the west. Possibly the
higher correlations between forested and previously forested land usage and
stream discharge were due to greater slopes and thinner soils in the forested
areas as opposed to the agricultural areas. The correlation coefficient
between mean slope and percent of drainage area in forest plus cleared
unproductive was 0.61. For the eastern region the correlation was 0.63; for
the plains, 0.49; and for the west, 0.54. One would expect direct surface
runoff to increase with increased slope but not average annual stream dis-
charge as was shown with these data; therefore, one or more covariants seem
probable.
Some additional explanation of the difference between relationships of
nutrient concentrations to land use and nutrient export to land use may be
31
-------
N
69
80
298
7
16
105
12
17
11
146
11
76
80
>90% Forest
s£75% Forest
>50% Forest
>75% Cleared
Unproductive
^50% Cleared
Unproductive
Mixe50% Range:
Remainder predominantly
forest
^75% Range
^50% Range:
Remainder predominantly
agriculture
>50% Agriculture
>40% Urban
>75% Agriculture
^90% Agriculture
total phosphorus export
Land Use
vs.
Mean Total Phosphorus and Mean
Orthophosphorus Stream Exports
Data from 928 "Nonpoint source-type" watersheds
distributed throughout the United States
orthophosphorus export
15.0 20.0 25.0 30.0 35.0
Kilograms per Square Kilometer per Year
40.0 45.0
50.0
Figure 5. Relationships between general land use and stream exports of total
phosphorus and orthophosphorus.
-------
&
69
80
298
7
16
105
12
17
11
146
11
76
80
^90% Forest
>75% Forest
^50% Forest
^75% Cleared
Unproductive
^50% Cleared
Unproductive
Mixed
>50% Range:
Remainder predominantly
forest
>75% Range
^50% Range:
Reaainder predominantly
agriculture
>50% Agriculture
>40% Urban
^75% Agriculture
^90% Agriculture
Land Use
vs.
Mean Total Nitrogen and Mean
Inorganic Nitrogen Stream Exports
Data from 928 "Nonpoint source-type" watersheds
distributed throughout the United States
inorganic nitrogen export
total nitrogen export
1 346.7
140.5
139.6
544.6
384.3
294.2
488.0
326.2
1 554.6
780.7
953.9
200
300
400
1000
Kilograms per Square Kilometer per Year
Figure 6. Relationships between general land use and stream exports of total
nitrogen and inorganic nitrogen.
-------
found in differences in mean annual precipitation patterns of study area
locations. From analysis of study area locations, it appeared that most of
the forested drainage areas were located in regions receiving somewhat
greater average annual precipitation than regions where most of the agricul-
tural watersheds were located. Greater annual precipitation would help
explain the greater stream flow in forested study areas. Although the
correlation between percent of subdrainage area in forest plus cleared
unproductive land and mean annual precipitation was found to be low (r =
0.29) for the entire data set, the coefficients for the eastern and plains
regions were substantially higher (0.56 and 0.73, respectively). Here, as
with the discharge—forest correlations, the several dissected, semi-arid,
forested watersheds in the west may have masked the relationship for the
nationwide data set.
The data also indicate that urban land usage seems to have a pronounced
effect on the amount of export. Again, the cause is probably increased flow
rates, but in this case it was more likely because of greater areas of
impervious surfaces.
In connection with the above explanation for understanding the differ-
ences in the relationships between concentrations and land use, and export
and land use, it is very important to recognize that the drainage areas
included in this study are not control plots. Rather, they are existing
drainage areas which represent typical land use- and geographically-related
characteristics in their respective areas. If one were studying control
plots, where slope, soil type, and climate conditions were similar from one
plot to another, one would expect a significantly higher nutrient export
rate (as well as a higher nutrient concentration) from plots in agricultural
land use than from forested plots. Runoff, as well, would probably be
somewhat greater from agricultural plots than from forested plots because of
less biomass in the agricultural areas and hence less water lost through
evapotranspiration. The drainage areas included in this study, on the other
hand, possess different topographic, soils, and climatic characteristics.
These characteristics are important in determining land use in the first
place. In general, flat to rolling terrain and rich soils, such as are
found in Iowa and southern Minnesota, lend themselves to agricultural land
use. Conversely, where the terrain is too mountainous or dissected or the
soil too poor for agriculture to be economically feasible, such as in much
of New England, forests are allowed to predominate.
The fact that study subdrainage area sizes vary, and that those cate-
gorized as forest were considerably smaller than any other category, sug-
gested that size might have a bearing on nutrient concentrations and/or
stream flow. If either factor was related to size, then export would also
be related. However, analysis of these data revealed no good correlations.
The frequency polygons (Figures 7 and 8) illustrate how the nutrient
concentration data were distributed for each of the land use types repre-
sented by large data sets (n >50). Comparison of these polygons with Figures
3 and 4 show the data for forested drainage areas were grouped much more
tightly around the mean values than were data for agricultural drainage
34
-------
5 90 % Forest (n.6B)
S 75 °/o Forest { n * 77)
t 50 °/o Forest (n* 295)
Mixed (n >103)
S 50 °/o Agriculture (n»l44)
t 75% Agriculture (n«72)
S 90% Agriculture (n«74)
•••••••••••
m 30
) .12 .15 .18 .21 .34 .27 .30 .33 .36 .3®
MEAN TOTAL PHOSPHORUS CONCENTRATION (mg/l)
*90% Forest (n»«B)
£ 76 % Forest (n- 77)
t 50% Forest (n- 295)
Mixed 74 )
eoeeeoes
MEAN ORTHOPHOSPHORUS CONCENTRATION
^ ^ l/U^T
.24 2ft .28 V .57
(mg/l)
^gure 7.
Frequency polygons of mean total phosphorus and mean orthophosphorus
concentrations fn streams by overall land use category.
35
-------
2 90% Forest (n-68) —
2 75 % Forest (n » 77)
2 50 % Forest (n. 295) —
Mixed (n »103)
2 50 % Agriculture (n»144) eeeeeeeeeee
£ 75 % Agriculture (n s 72)
2 90 % Agriculture (n ¦ 74)
V 40-
MEAN TOTAL NITROGEN CONCENTRATION (mg/l)
2 90 % Forest (n>68)
2 75 % Forest (n -77)
2 50 % Forest (n» 29S)
Mixed (n.|03)
2 50 % Agriculture (n > 144)
2 75 % Agriculture (n-72)
2 90 % Agriculture (n • 74)
eeeeeeeeeee
MEAN INOROANIC NITROGEN CONCENTRATION (mg/l)
Figure 8. Frequency polygons of mean total nitrogen and mean inorganic nitroqen
concentrations in streams by overall land use category.
36
-------
areas. It appears that the greater the percent of land in forest the lower
the mean nutrient concentration and the less the variability about the mean.
Regi onali ty
To refine the relationships of land use to nutrient runoff shown by the
bar graphs and frequency polygons, and particularly to explain some of the
variations about the mean for each land use category, regional analyses
employing both statistical and cartographic techniques were performed. The
analyses will be discussed later, but it is important to mention at this
point the part this method played in focusing on possible regionalities of
relationships. The statistical analyses indicated that, after the variation
in nutrient level related to general land use was explained with regard to
the entire data set, the importance of each of the other macro-watershed
characteristics being considered (animal unit density, mean slope, drainage
density, mean annual precipitation, geology-origin, surface soil pH, and
mean annual stream discharge rate) was either highly questionable or non-
existent. When the data were analyzed by region, land use remained very
important but the other characteristics varied in importance; i.e., a given
characteristic which did not correlate well with nutrient levels when the
set was taken as a whole or by region for some regions, might correlate well
in other regions.
Cartographic techniques, on the other hand, revealed more concerning
the variations about the mean. They illustrated regional patterns in the
relationships and interrelationships between all of the macro-watershed
factors and stream nutrient concentrations. By studying these patterns one
can gain an insight about which man-induced and natural characteristics may
be important in specific geographic areas.
A first attempt at understanding the general nature of these region-
al ities involved further defining, by region, the individual land use
categories shown in Figures 3 and 4. The regional delimitations used for
these analyses (Figures 9 and 10) and the regression analyses mentioned
earlier were based primarily on patterns evidenced by the maps commented on
in the "Areal Distributions of Data" portion of this section (pages 24-25).
Figure 9 indicates that both total phosphorus and orthophosphorus NPS
concentrations in streams generally are higher in the western region of the
United States than in the east. The graph also illustrates that for a given
land use category in the central region, phosphorus concentrations in streams
are generally lower than those in the west but are higher than those in the
east. This relationship appears to be the clearest with regard to forested
watersheds (in the >75% forest and >90% forest categories) where total P and
ortho-P concentrations are about twice as high in the west as in the east.
The relationship does not appear to be quite as well defined with regard to
agricultural watersheds (in the >75% agriculture and >90% agriculture cate-
gories) because of the small sample size in the western region.
The data presented in Figure 10 illustrate general regionalizes for
NPS nitrogen concentrations that are considerably different than those for
phosphorus. Some of the noteworthy regionalities these data suggest are:
37
-------
> 90%
Forest
> 75%
Forest
> 50%
Forest
> 50%
Agriculture
> 75%
Agriculture
> 90%
Agriculture
N
1 ,006
23 pKl .011
I 1.009
15MHE2Z3 .020
.022
30
1.007
mtm .015
2
3 VtV-i .025
.036
Land Use by Region
vs.
Mean Total Phosphorus and Mean
Orthophosphorus Stream Concentrations
Data from 730 'Nonpoint source - type' watersheds
distributed throughout the United States
^ orthophosphorus concentration
013
.036
99
43
2
45
19
8
60
11
total phosphorus concentration
Regions
,.u/v
^ I .070
.055
U.4-4-4- + 4-44- + »-f+ + + + + + 4- + 4-4» + 4-4»| i1T
l + + + + + + + .f + ++ Vr++ + + + + + + + + + ¦! .11/
.063
.123
AWMWAWA
&000000000000000000
134
.140
i uoy
»QCg^ .173
.154
^025 [050 !075 J00 J25 150
Milligrams per Liter
3 .214
.175 .200 .225
Figure 9. Regional relationships between general land use and total phosphorus
and orthophosphorus concentrations in streams.
-------
N
>90%
Forest
>75%
Forest
> 50%
Forest
> 50%
Agriculture
> 75%
Agriculture
> 90%
Agriculture
mm .658
* + * * TJ .501
077
glvVt%%vj .622
,070
O 1 .601
i .778
Land Use by Region
vs.
Mean Total Nitrogen and Mean
Inorganic Nitrogen Stream Concentrations
Data from 730 'Nonpoint source-type' watersheds
distributed throughout the United States
3 .907
99
43
2
1.858
l .647
+ + + + + + + + + + + + + + -I - -
S 4- 4- 4- 4. 4- +V+ 4- 4- + + 4- + +1 1./4I
1.832
1.615
19
8
60
11
- ' 7 ' 3.005
1.931
I i.oyy
xxxxxxxxxyyyyyxyxxssgror^x^^.S3 2.833
~~T~:
1.735
3 2.366
_L
1.0
2.0 3.0 4.0
Milligrams per Liter
n344
inorganic nitrogen concentration
~ .907
total nitrogen concentration
Regions
S.044
6.082
5.0
6.0
Figure 10. Regional relationships between general land use and total nitrogen
and inorganic nitrogen concentrations in streams.
-------
1. Total nitrogen concentrations in streams are higher in the eastern
region than in the central or western regions. The difference is
slight with respect to forested watersheds but fairly large for
agricultural areas. However, as was the case with the data shown
in Figure 9, the small set of agricultural watersheds in the west
qualify this analysis to some degree.
2. Inorganic nitrogen concentrations in streams are much higher in
the east than in the central or western regions particularly for
forested drainage areas. For the watersheds in categories with
the most forest (>75% forest and >90% forest), inorganic N concen-
trations are between 2.3 to 3.3 times higher in the east than in
the remainder of the country.
3. Associated with the above regionalities, inorganic nitrogen gener-
ally makes up a higher percent of total nitrogen concentrations in
streams in the east than elsewhere within the conterminous United
States. The differences are especially noticeable when comparing
data for forest land use categories. These percentages are given
in Table 4.
TABLE 4. PERCENT OF MEAN INORGANIC NITROGEN TO MEAN TOTAL
NITROGEN CONCENTRATIONS IN STREAMS BY REGION
Land Use Category
East
Regions
Central
West
>90% Forest
28
12
12
>75% Forest
32
16
12
>50% Forest
38
33
26
>50% Agriculture
58
37
*
>75% Agriculture
60
41
60
>90% Agriculture
82
32
*
^insufficient data points for adequate representation
Earlier, an interpretation was given based on mean values for the
nationwide data set (Figures 3 and 4) which indicated that phosphorus would
be expected to be limiting in surface waters draining either forested or
agricultural lands. The regional values shown in Figures 9 and 10 suggest
that nitrogen might be the limiting nutrient for algal growth in much of the
40
-------
agricultural parts of central and western United States. For the central
region (for regional definitions see Figure 23, page 64) total nitrogen to
total phosphorus ratios were 15:1, 14:1 and 11:1 for >50% agriculture, >75%
agriculture, and >90% agriculture. The ratios were 13:1, 16:1, and 11:1 for
>50% agriculture, >75% agriculture, and >90% agriculture watersheds in the
western region respectively. Again, the values for agricultural categories
in the western region are questionable because of the small sample size in
that region. The N:P ratios derived from Figures 9 and 10 greatly exceeded
15:1 for forested categories in all three regions and for agricultural
categories in the eastern region, which includes the "corn belt".
Another method for attempting to understand more about the varability
about the mean values shown in Figures 3 and 4 was to cartographically
Present the individual values for a given land use category or group of
categories for a given nutrient form that were within the highest third,
middle third, and lowest third of values for that particular data set (Fig-
ures 11 through 18). This method, in effect holds land use constant and
Graphically displays the possible regionalities relative to the geographical
locations of the study watersheds which the data represent. Maps were
compiled for both forested and agricultural categories for each form of each
nutrient. Data for agricultural watersheds included those from the >90%
agricultural and >75% agricultural categories. Maps showing data for for-
ested watersheds included values from watersheds in >90% forest and >75%
forest categories. The low, middle, and high ranges of concentrations for
seven of the eight maps were determined by studying frequency distributions
°f the concentrations and making divisions at logical, "even" points near
the one-third and two-thirds points. For one map (Figure 17), the range of
concentrations was divided into four parts because of a special interest in
these data and some peculiarities noticed in previous analysis.
The data presented in Figure 11 reveal that mean annual total phosphorus
concentrations in streams draining mostly forested watersheds were consis-
tently higher in the west and south central regions of the United States
than in the east. Generally, only in streams draining western high eleva-
tion- or National Park-watersheds were the phosphorus concentrations as low
as those observed in the forested eastern watersheds. Interestingly, the
characteristics the watersheds with low concentrations had in common that
the ones with high concentrations did not was a low density of domestic
animals (U. S. Department of Agriculture, Forest Service, 1972). Even for
watersheds with mixed land usage (but holding the mixture constant), there
were generally higher phosphorus concentrations in the west than in the
east.
Figure 12 reflects regionalities of total phosphorus concentrations in
sampled streams draining agricultural watersheds. The map reveals an
area of data points having relatively high mean annual values extending from
North Dakota through eastern South Dakota and eastern Nebraska as well as
"•nto central Iowa and parts of southern Minnesota. For the remainder of the
a9ricultural watersheds in states south of this region, the total phosphorus
concentrations are generally much lower. Another significant grouping
appears in the east central part of the "corn belt", from central Illinois
1nto central and northern Indiana, where the mean annual total P concentra-
41
-------
-p>
f>o
4
I
REGIONALIZES OF TOTAL PHOSPHORUS
CONCENTRATIONS' IN STREAMS
DRAINING FORESTED WATERSHEDS"
Pho$pHorv« Map
Concentrations (mg/li Symbol
< 0.015 0
0.016 to 0.025 ©
> 0.025 •
* representative of mean annual values
>75% Forest
Figure 11. Regionalities of total phosphorus concentrations in NES sampled
streams draining forested watersheds.
-------
c oo
Figure 12. Regionalities of total phosphorus concentrations in NES sampled
streams draining agricultural watersheds.
-------
tions in streams were generally lower than the mean for the agricultural
data set. It is difficult to see what aspect, or aspects, may be related to
these patterns that would help explain them. One possibility is that there
is a correlation with surface soil pH. Generally, soil pH is higher and
more near the level conducive to maximum availability of soil phosphorus in
the west central states where the stream concentrations are very high than
in the east-central part of the corn belt where the stream concentrations
are relatively low. The point of maximum phosphorus availability for soils
is pH 6.5; at higher pH values, calcium fixation limits the availability,
and at lower pH values aluminum fixation and iron fixation limit phosphorus
availability (Scarseth, 1962; Kling, 1977; and Moore, 1977). However, the
fact that relatively low phosphorus concentrations were obtained from streams
draining agricultural areas to the south where soil pH levels were similar
suggests that pH might not be the answer, although there are also north to
south differences in leaching and biological activity in the soil. Neverthe-
less, it appears that the patterns evidenced by the data on Figure 12, are
more closely associated with natural phenomena than anthropogenic activities.
Figures 13 and 14 illustrate patterns for orthophosphorus concentrations
in NES streams draining forested and agricultural watersheds which are very
similar to those for total phosphorus (Figures 11 and 12).
The maps illustrating regionalities of total and inorganic nitrogen
concentrations in NES streams draining forested watersheds (Figures 15 and
16), revealed markedly different regional patterns than those for phosphorus
and orthophosphorus (Figures 11 and 13). Figure 16 illustrates that, for
the NES data set, inorganic nitrogen concentrations were generally much
higher in streams draining forested areas in the east central and northeast-
ern states than elsewhere. There were a few exceptions to this, the most
notable was the group of watersheds in the Black Hills of South Dakota.
Inorganic nitrogen concentrations were particularly high in New York, Penn-
sylvania, and western Maryland. For the most part, the characteristic that
these high values have in common that the low ones did not, is proximity to
heavily industrialized and/or populated areas. There is a striking similar-
ity between value maps of these NES inorganic nitrogen stream concentrations
and isometric maps of observed acid precipitation (Likens, 1975). This
similarily tends to support Liken's (1975) conclusion that there is a rela-
tionship between annual atmospheric input of hydrogen ion and nitrate, and
that nitric acid is an important factor in explaining the recent increase in
acid precipitation. The total nitrogen patterns for forested watersheds
(Figure 15) were by contrast somewhat indistinct; for example, there were
mostly high concentrations in Pennsylvania; high, medium, and low concentra-
tions in New York; and medium and low concentrations for the remainder of
the forested watersheds in New England.
In general, mean annual total nitrogen concentrations in streams
draining predominantly agricultural watersheds were considerably higher in
the heart of the corn belt than elsewhere (Figure 17). Also, for agricul-
tural watersheds, there was a pattern of relatively low total nitrogen
concentrations in the plains states from South Dakota through Texas and
including much of Missouri and the southern tip of Illinois. The area with
high nitrogen concentrations in streams is also one of generally higher
44
-------
REGtONALITKS Of ORTHO PHOSPHORUS
CONCENTRATIONS' IN STREAMS
DRAINING FORESTED WATERSHEDS**
Ortho Phosphorus Mop
Concentrotiom (mg/l) Symbol
S 0.007 .... . . O
0.008 »o 0,015 w
> 0-015 •
* representative of mean annual values
**>75% Forest
Figure 13. Regionalities of orthophosphorus concentrations in NES sampled
streams draining forested watersheds.
-------
REGIONALIZES OF ORTHO PHOSPHORUS
CONCENTRATIONS * IN STREAMS
DRAINING AGRICULTURAL WATERSHEDS"
Ortho Phosphorus Mop
Corx entropions (mg/l) Symbol
S 0.030 O
0.031 to 0.089 0
> 0089 •
* representative of mean annual values
**2 75% Agriculture
Figure 14. Regionalities of orthophosphorus concentrations in NES sampled
streams draining agricultural watersheds.
-------
-fa.
REGION All TIES OF TOTAL NITROGEN
CONCENTRATIONS* IN STREAMS
DRAINING FORESTED WATERSHEDS"
Nitrogen Mop
Concentrations (mg/1) Symbol
S 0.500 O
0.501 to 0.900 ©
> 0.900 •
* representative of mean annuol values
"2 75* Forest
Figure 15. Regionalizes of total nitrogen concentrations in NES sampled
streams draining forested watersheds.
-------
REGIONALIZES OF INORGANIC NITROGEN
CONCENTRATIONS' IN STREAMS
DRAINING FORESTED WATERSHEDS"
Inorganic Nitrogen Map
Concentrations (wg/l) Symbol
< 0.050 O
0.051 to 0.100 ©
0.101 to 0.200 0
2 0.200 •
' representative of mean annual values
"> 75% Forest
Figure 16. Regionalities of inorganic nitrogen concentrations in NES sampled
stream draining forested watersheds.
-------
REGIONALIZES Of TOTAL NITROGEN
CONCENTRATIONS* IN STREAMS
DRAINING AGRICULTURAL WATERSHEDS
Nitrogen Map
Concentrations (mg/1) Symbol
S 2300 O
2.501 to S.000 ©
5.000 •
' raprtMntorivt of mean annual values
"i 75% Agriculture
Figure 17. Regionalities of total nitrogen concentrations in NES sampled
streams draining agricultural watersheds.
-------
fertilizer expenditures and higher domestic animal unit densities compared
to the areas where the concentrations are lower (U.S. Department of Commerce,
Bureau of Census 1972b). However, as was stressed in the previous section
on applications, this is not meant to imply that the high nitrogen concen-
trations in surface waters in the central part of the corn belt are a direct
result of fertilizer applications or any other agricultural practice.
Certainly the nutrient-rich soils of the area that played a major role in
determining the intensive agricultural land usage in the first place were
also an important factor in determining the high stream concentrations. The
proportion of the stream nutrient levels attributable to natural sources,
relative to the proportion attributable to anthropogenic sources, probably
cannot be determined at the present time. Nevertheless, one should not
disregard the possible relationship between NPS anthropogenic inputs of
nitrogen and high stream nitrogen levels. The regional patterns of inorganic
stream concentrations in NES agricultural study areas (Figure 18) were
similar to those of total nitrogen.
REGRESSION ANALYSES AND PREDICTIVE CAPABILITY
Early in this study, multiple regression analysis was selected as the
most likely method for developing predictive models from the data. In order
to accommodate the development of such predictive models, the data that were
collected for each study watershed were for those macro-watershed charac-
teristics thought to have a relationship to stream nutrient levels. However,
the analyses of these data suggested that only one of the macro-watershed
characteristics, land use, was related to stream nutrient concentrations and
exports on both the national and regional levels. Regarding the entire data
set, generally after the variations in stream nutrient concentrations related
to land use were explained, the other macro-watershed characteristics were
unimportant. When the data were analyzed by regions, and especially when
they were shown cartographically, it became apparent that certain charac-
teristics other than land use might be related to stream nutrient levels in
some regions but not others. The fact that makes identification of these
relationships unwieldy to seemingly impossible is that most of these charac-
teristics are interrelated with one another, and these interrelationships
vary spatially and temporally in such a way as to mask or otherwise affect
individual relationships with stream nutrient levels. Therefore, this
section primarily will be concerned with the relationships of land use to
nutrient levels in streams and to regionalities of these relationships.
Contributing Land Use Types
Figures 19 through 22 illustrate the relationships between land uses
generally considered nutrient contributing (% agriculture plus % urban) and
nutrients in streams. Several ways of examining the effects of land use on
nutrient concentrations or loads in streams were investigated. In general,
nutrient levels increased with increased percentages in agricultural land
usage and decreased percentages in forested land. Little to no correlation
was found between nutrient levels and percent of land in either cleared-
unproductive, range, urban or wetland. This was expected because of the
probable masking effects of agriculture and forest. Since increased or
decreased percentages of all general land use types appeared to have some
50
-------
oo
CJ1
REGtONAUTlES OF INORGANIC NfTROGCN
CONCENTRATIONS* IN STREAMS
DRAINING AGRICULTURAL WATERSHEDS"
Mop
Symbol
Inorganic Nitrogen
Concentrations (mg/t)
< 1.000 .
14>01 to 4.000
> 4.000 .
* npraianhitivi of m
**2 75% Agriculture
Figure 18. Regionalities of inorganic nitrogen concentrations in NES sampled
streams draining agricultural watersheds.
Inorganic Nitrogen Mop
Concentrations (mg/t) Symbol
< 1.000 O
1.001 to 4.000 9
> 4.000 •
-------
effect on nutrient levels, a land use ratio of contributing (agriculture +
urban) to non-contributing (forest + cleared-unproductive + wetland) types
was investigated for its utility as a single factor including all land use
types. Generally, relationships between these ratios and nutrient levels in
streams were found to correlate better than those considering only one land
use type. It was then determined that use of just the numerator (% agricul-
ture plus % urban) from the ratio provided more easily understood land use
values and eliminated the graphing problems encountered by working with
values to infinity.
Use of % agriculture plus % urban to relate effects of land use on
nutrient levels in streams appears to be appropriate where agriculture
and/or forest comprise the predominant type(s). These two land use cate-
gories comprise the bulk of the land use data gathered for this study but
also constitute by far the predominant land use in the eastern half of the
United States, as well as parts of the west where precipitation is adequate
to cause significant surface runoff most of the year. The use of % agricul-
ture plus % urban appears to compensate for minor amounts of the other
general land use types. However, its use is probably unsatisfactory for
predicting or estimating nutrient concentrations or loads for areas where
either urban, cleared-unproductive, or wetland land use types predominate
and particularly where urban use predominates. Insufficient data have been
collected for these types.
It should be mentioned for the relationships illustrated by Figures 19
through 22, that rather than construct one 1ine-of-best-fit for the entire
data set for each nutrient form, regional lines-of-fit were chosen. However,
where the intercepts are appreciably different between regions for one graph
(e.g. Figure 20), the scatter generally appears wider than it would if each
region were treated separately or if the intercepts are similar as in Figure
21.
Contributing Land Use--Nutrient Concentration Relationships
Figure 19 shows the relation between mean total phosphorus concen-
trations in streams and the contributing land use % agriculture plus %
urban. The equation for the relationship involving the entire data set
shown in Figure 19 is:
Log10 (PCONC) = -1.692 + .00816 (% agric. + % urban) (1)
The correlation coefficient (r) for this relationship is 0.65. The
utility of this equation for estimations can be illustrated by the use of a
multiplicative standard error. This factor, denoted by "f", when multiplied
by and divided into the value estimated by the equation, determines the
range which should contain roughly two-thirds of the observed concentrations
at a given level of contributing land use (% agric. + % urban). For equation
(1), the value of f is 2.05. The concentration ranges for different percent-
ages of contribution land use types are shown in Table 5.
52
-------
79-i
0-6-
0.5-
OJ 0.4 ¦
J.
— 0.3-
r = .52 (West)
r = .69 (Central)
r « 75 (Fast)
0.2
0.1
.07
.05
.04
.03
.02 -
.01-'
.007-
.005
.0045
-r-
70
~T-
90
10
—r~
20
r~
30
l
40
I
50
60
80
100%
% IN AGRICULTURE + % IN URBAN
Figure 19. Scattergram of "contributing" land use types related to total
phosphorus concentrations in streams.
-------
TABLE 5. ESTIMATED MEAN TOTAL PHOSPHORUS CONCENTRATIONS (mg/1)
% Ag + % Urb
Avg. PCONC
67% Limits
0
25
50
75
100
0.020
0.033
0.052
0.083
0. 133
0.010-0.042
0.016-0.067
0.025-0.106
0.041-0.170
0.065-0.272
For example, for streams draining areas with a combined agriculture
plus urban land use percentage of 25%, mean total phosphorus concentrations
average 0.033 mg/1. However, because of the variation around this estimation,
there is only 67% probability that the true value will fall in the range
0. 016 to 0.067 mg/1.
Another aid to the interpretation of this equation is given by a
factor denoted by "A" which is the approximate percentage increase in the
mean nutrient level corresponding to an increase of 10 percentage points in
contributing land use (% agric. plus % urban). For example, for equation
(1), W = 21% indicating that a watershed in which a total of 30% of the area
is in agricultural and urban land use can be expected to have NPS stream
phosphorus concentrations which are roughly 21% greater than those of a
stream draining a watershed in which only 20% of the area is in contributing
land use.
The equations for the regional regression lines shown in Figure 19
generally indicate higher correlations and better predictive capability than
the equation for the entire data set, especially for the eastern region.
The equations for regional models, along with the associated interpretive
aids, are as follows:
for the eastern region,
Log10 (PCONC) = -1.836 + .00971 (% agric. + % urban)
r = 0.74, f = 1.85, and A = 25%
(2)
for the central region,
Log10 (PCONC) = -1.744 + 0.01010 (% agric. + % urban) (3)
r = 0.69, f = 2.07, and A = 26%
and for the western region,
Log]0 (PCONC) = -1.548 + 0.00929 (% agric. + % urban) (4)
r = 0.52, f = 2.15, and A = 24%
54
-------
Differences in estimated values using the model for the nationwide data
set as compared to those for the regional models are shown in Table 6.
TABLE 6. ESTIMATED MEAN TOTAL PHOSPHORUS CONCENTRATIONS (mg/1)
FOR NATIONWIDE AND REGIONAL MODELS
Nationwide Regional Models
Model East Central West
% Ag + 67% 67% 67% 67%
% (Jrb Ave P Cone Limits Ave P Cone Limits Ave P Cone Limits Ave P Cone Limits
0
25
50
75
100
0.020
0.033
0.052
0.083
0. 133
(.010-.042)
(.016-.067)
(.025-.106)
(.041-.170)
(.065-.272)
0.015
0.026
0.045
0.078
0. 136
(.008-.026)
(.014-.048)
(.024-.083)
(.042-.144)
(.073-.251)
0.018
0.032
0.058
0. 103
0. 185
(.008-.037)
(.015-.066)
(.028-.120)
(.049-.213)
(.089-.382)
0.028
0.048
0.083
0.141
0.240
(.013-.060)
(.022-.103)
(.038-.178)
(.065-.303)
(.111-.516)
Figure 20 illustrates the relationship between mean orthophosphorus
concentrations in streams and "% agriculture plus % urban". The equation
for the nationwide set of data points is:
Log]0 (0PC0NC) = -2.033 + 0.00710 (% agric. + % urban) (5)
The correlation coefficient for the equation is 0.59; the multiplica-
tive standard error (f), 2.07; and A, 18%.
The equations for the regional data sets showed better correlations
and predictive capabilities, except for the western region. Equations,
together with the associated interpretive aids, for the regional regression
lines shown in Figure 20 are as follows:
for the eastern region,
Log10 (°PC0NC) = -2.222.+ 0.00934 (% agric. + % urban) (6)
r = 0.73, f = 1.86 and A = 24%
for the central region,
Log10 (0PC0NC) = -2.082 + 0.00868 (% agric. + % urban) (7)
r = 0.63, f = 2.09 and A = 22%
and for the western region,
Log]0 (0PC0NC) = -1.851 + 0.00863 (% agric. + % urban) (8)
r = 0.52, f = 2.05 and A = 22%
55
-------
.79 -i
0.5
_ 0.4-
cn
0.3-
Z
— 0.2
0.1
.07-
.05
.04
.03-
.02-
! .01 -+•'
.007
.005 -h —
.0045
r = .52 (West)
r = .63 (Central)
r = .73 (East)
I
20
~r
10
30
i i i n
40 50 60 70
% IN AGRICULTURE + % IN URBAN
80
—T"
90
100
Figure 20. Scattergram of "contributing" land use types related to orthophos-
phorus concentrations in streams.
-------
Figures 21 and 22 display somewhat higher correlations between stream
nitrogen concentrations and "contributing" land use than were evidenced for
phosphorus in Figures 19 and 20. Regarding the nationwide set of 904 water-
sheds, the equation and the interpretive factors for the regression shown in
Figure 21 are:
Log10 (NCONC) = -0.247 + 0.00814 (% agric. + % urban) (9)
r = 0.78, f = 1.62 and A = 21%
It should be noted that the factor A appears to be a relatively poor
interpretive aid for the regressions shown on Figures 21 and 22 because of
the possible curvilinear relationships involved. As was mentioned earlier
in the discussion of data shown in Figure 4 (page 29), increases in stream
nitrogen concentrations proportionate to increases in agricultural land
usage were greater for watersheds with >75% agricultural land usage than
those with less agricultural land usage. However, analysis of Figure 4 and
identification of the individual data points in the upper right hand corner
of Figures 21 and 22, suggest the curvilinear relationship may only be
characteristic of the eastern data set.
A rough idea of the utility of equation (9), using the multiplicative
standard error of 1.62, is given in Table 7.
TABLE 7. ESTIMATED MEAN TOTAL NITROGEN CONCENTRATIONS (mg/1)
% agric. + % urban
Avg. NCONC
67% Limits
0
0.57
0.35-0.92
25
0.90
0.56-1.46
50
1.45
0.89-2.34
75
2.31
1.43-3.74
100
3.69
2.28-5.97
Similar to the findings for total phosphorus, two of the three equa-
tions for the regional regression lines shown on Figure 21 indicated better
predictive capability than the equation for the entire nationwide data set.
Only for the western region, which includes less than one-fourth of the
study watersheds, was the correlation coefficient lower. Although the
correlation coefficient is slightly less for the central regional model than
that for the nationwide set, the multiplicative standard error is lower
indicating less variation about the mean and greater predictive capacity.
Equations for the regional regression lines shown on Figure 21, together
with the related interpretive factors are:
57
-------
15.85 T-
10.0-
rr 7.0-
0>
^ 5.0-
~ 4.0-1
o
< 3.0-
0£
~—
z
ui 2.0-
r * .82 (East)
r = .50 (West)
x - .71 (Central)
"# -
> . \
m
¦
W£S7
CENTRAL
#•*#" " ¦ «
» m
0.3
0.2
0.1
"T"
10
I
20
T
30
I
40
I
50
I
60
r~
70
-1—
80
I
90
100%
% IN AGRICULTURE + % IN URBAN
Figure 21. Scattergram of "contributing" land use types related to total
nitrogen concentrations in streams.
-------
for the eastern region,
Log1Q (NCONC) = -0.292 + 0.00932 (% agric. + % urban) (10)
r = 0.85, f = 1.52 and A = 24%
for the central region,
Log]0 (NCONC) = -0.246 + 0.00662 (% agric. + % urban) (11)
r = 0.72, f = 1.54 and A = 16%
and for the western region
Log10 (NCONC) = - 0.200 + 0.00704 (% agric. + % urban) (12)
r = 0.50, f = 1.84 and A = 18%
Differences in estimated concentrations using the regional and nationwide
models are illustrated in Table 8.
TABLE 8. ESTIMATED MEAN TOTAL NITROGEN CONCENTRATIONS (mg/1)
FOR NATIONWIDE AND REGIONAL MODELS
Nationwide
Regional Models
Model
East
Central
West
% Ag + 67%
% Urb Ave N Cone Limits
67%
Ave N Cone Limits
67%
Ave N Cone Limits
67%
Ave N Cone Limits
0
0.57
(0.35-0.92)
0.51
(0.33-0.77)
0.57
(0.37-0.88)
0.63
(0.34-1.16)
25
0.90
(0.56-1.46)
0.87
(0.57-1.32)
0.83
(0.54-1.28)
0.95
(0.52-1.75)
50
1.45
(0.89-2.34)
1.49
(0.98-2.26)
1.22
(0.79-1.88)
1.42
(0.77-2.61)
75
2.31
(1.43-3.74)
2.55
(1.68-3.88)
1.78
(1.15-2.74)
2.13
(1.16-3.92)
100
3.69
(2.28-5.97)
4.37
(2.87-6.64)
2.61
(1.69-4.02)
3.19
(1.73-5.87)
Figure 22 portrays the relationship of inorganic nitrogen concentra-
tions in streams to contributing land usage. Equations for the relationship
involving the complete data set and the regional breakdowns are:
for the nationwide data set,
Log-|Q (INC0NC) = -0.978 + 0.01404 (% agric. + % urban) (13)
r = 0.79, f = 2.28 and A = 38%
59
-------
15.85-r
r = .83 (East
r = .54 (West
r = .65 (Central
wesr
0.7-
0.5 J
.03-
central
100 %
% IN AGRICULTURE + % IN URBAN
Figure 22. Scattergram of "contributing" land use types related to inorganic
nitrogen concentrations in streams.
-------
for the eastern region,
Log-j0 (INCONC) = -0.890 + 0.0146 (% agric. + % urban) (14)
r = 0.83, f = 1,95 and A = 39%
for the central region,
Log10 (INCONC) = -0.953 + 0.00975 (% agric. + % urban) (15)
r = 0.65, f = 2.18 and A = 25%
and for the western region,
log1Q (INCONC) = -1.048 + 0.01278 (% agric. + % urban) (16)
r = 0.54, f = 2.70 and A = 34%
Comparison of Figure 22 with Figure 21 reveals greater variation about
the mean for the inorganic form of nitrogen than for total nitrogen. This
is also evidenced by the substantially higher "f" values for the inorganic
nitrogen equations as compared to those for total nitrogen.
Contributing and Forest Land Use~-Nutrient Concentration
Relationships
The preceeding equations reflect simple relationships between a com-
bination of land use categories that had a positive correlation with NPS
nutrient concentrations in streams. They afford a fairly good predictive
capability for the eastern region of the United States and, for the total
form of both nutrients, appear quite useful for the central region as well.
For the west, however, their utility appears limited. This is probably due
in large part to the greater percentages of range and cleared-unproductive
land usage in the west. In the eastern and central regions, most of the
area that was not in agricultural and urban land usage was largely forested.
Since this is often not the case in the west, a slightly more complicated
model was constructed which considered the negative correlation between
percent of watershed in forest and stream nutrient concentrations in addition
to the positive relationship between "contributing" land use percentage and
stream nutrient concentrations. Equations for these models are given in
Table 9 and the relationships are illustrated in Figures 23 and 24.
61
-------
TABLE 9. REGIONAL STREAM NUTRIENT CONCENTRATION PREDICTIVE MODELS
Nutrient Form Model, Correlation Coefficient and Multiplicative Standard
Region Error
Total phosphorus
East Log10 (PCONC) = -1.8364 + 0.00971 (% agric + % urb)
r = 0.74, f = 1.85
Central ^og^ (PCONC) =-1.5697 + 0.00811 (% agric + % urb) -0.002312 (% for)
r = 0.70, f = 2.05
West Log10 (PCONC) =-1.1504 + 0.00460 (%agric + %urb) -0.00632 (% for)
r = 0.70, f = 1.91
Qrthophosphorus
East Log10 (OPCONC) = -2.2219 + 0.00934 (% agric + % urb)
r = 0.73, f = 1.86
Central '-°9]o (OPCONC) = -2.0815 + 0.00868 (% agric + % urb)
r = 0.63, f = 2.05
West Log10 (0PC^NO = -1.5513 + 0.00510 (% agric + % urb) -0.00476 (% for)
r = 0.64, f = 1.91
Total nitrogen
East Log10 (NC0NC) = -0.08557 + 0.00716 (% agric + % urb) -0.00227 (% for)
r = 0.85, f = 1.51
Central Log1Q (NCONC) = -0.01609 + 0.00399 (% agric + % urb) -0.00306 (% for)
r = 0.77, f = 1.50
West Log]0 (NCONC) = -0.03665 + 0.00425 (% agric + % urb) -0.00376 (% for)
r = 0.61, f = 1.75
(Continued)
62
-------
Table 9. (Continued)
Nutrient Form Model, Correlation Coefficient and Multiplicative Standard
Region Error
Inorganic nitrogen
East Log10 (INCONC) =-0.3479 + 0.00858 (% agric + % urb) -0.00584 (% for)
r = 0.84, f = 1.93
Central Log1Q (INCONC) = -0.5219 + 0.00482 (% agric + % urb) -0.00572 (% for)
r = 0.71,f = 2.06
West Log10 (INCONC) = -0.6339 + 0.00789 (% agric + % urb) -0.00657 (% for)
r = 0.65, f = 2.45
Comparison of "r" and "f" values for these equations to those involving
only "contributing" land use indicates that an improvement in predictive
capability for the western region was gained through addition of percent of
watershed in forest to the models. For the eastern and central regions,
the addition of % in forest to the total P and ortho-P models generally
resulted in little or no increase in the predictive capability. Where this
occurred, new models incorporating the addition were not constructed for
inclusion in Table 9 and Figures 23 and 24; instead, the regional models
including only "contributing" land use are shown.
Figures 25 through 28 offer some qualitative refinement of the models
shown in Table 8 by showing geographical areas where nutrient concentrations
can be expected to be greater, much the same as, or less than those predicted
by the models.
The data in Figure 25 illustrate some fairly obvious regional patterns.
They indicate that total phosphorus concentrations in streams are generally
higher than those predicted by the regional models in a very large region
extending from Wisconsin, southern Minnesota, and Iowa west into Nebraska,
South and North Dakota, and even northwestern Wyoming and southeastern
Montana. Another area where total-P concentrations are, for the most part,
higher than those predicted by the models comprises much of eastern Texas,
Louisiana, and western Mississippi. This area appears to join a region of
similarly high concentrations which extends from northwestern Tennessee
through western Kentucky into southern and eastern Illinois and western
Indiana. A very large region where phosphorus concentrations are nearly the
same as, or lower than, those predicted by the model includes most of the
Northeast from West Virginia, Maryland, and Pennsylvania through Maine, with
63
-------
REGIONAL' RELATIONSHIPS BETWEEN LAND USE AND PHOSPHORUS CONCENTRATIONS IN STREAMS
EAST
02 r =.75
025 %
\ *
\.03
\
\
K-
> in Agric + % in Urban
EAST
•UVO
\o, %
A
\
015 V
CENTRAL
r = .70
% in Agric + % in Urban
CENTRAL
rs .63
%
%
«y015 /O
\
v02 "V
..025 «
T JT I *1
WEST
west
r = .70
in Agric + % In Urban
WEST
in Agric * % in Urban
in Agric * % in Urban
in Agric ~ v in Urban
Figure 23. Differences in predictive characteristics between regional land
use-phosphorus concentration models.
-------
REGIONAL* RELATIONSHIPS BETWEEN LAND USE AND NITROGEN CONCENTRATIONS IN STREAMS
Regions
EAST
CENTRAL
in Agric + % in Urban
in Agric + % in Urban
WEST
west
Cen-
% in Agric + % in Urban
EAST
CENTRAL
WEST
% in Agric + % in Urban
in Agric + % in Urban
in Agric + in Urban
Figure 24. Differences in predictive characteristics between regional land
use-nitrogen concentration models.
-------
s
£
*
mu imhimh ttsnuu tmtsn m amm
KHMM (a) Ml* FW IK IK KU KHMMU V
IK HUMM niMU* HKU
U|_ (TP)- -UK* * JM71(X I* If. ~ X k Mm)
m (ir}'-uii;*.niii(xhi|. ~ * )• mm)-mxa(*i«F«**t
ta»JW)—UK* ~ JMN(X hi H. ~ X h Mn) - JKU (X la Ml
IniHi
It*
1.1
U ta U
II MU
-UkU
1J ti-IJ
U It-IJ
-1.1
Figure 25. Areal distribution of residuals from regional predictive models for
total phosphorus concentrations in streams.
-------
the exception of the area centered on southeastern New York and Connecticut.
Other smaller areas where phosphorus concentrations are mostly lower than
those predicted are south-central Texas, central Ohio, and a small region
extending from southern Missouri through northern Arkansas.
The regional patterns for orthophosphorus concentrations (Figure 26)
are generally similar to those for total phosphorus. There are some notable
exceptions, however. The large area in the Northeast, where total-P concen-
trations are generally the same as or lower than those predicted by the
model, is more extensive on the orthophosphorus map. The same is true for
the smaller area centered on northern Arkansas and southern Missouri which
had concentrations consistently lower than those predicted by the total
phosphorus model.
Regional patterns were also evident for the total nitrogen predictive
models (Figure 27). The regional patterns illustrated by residuals of the
nitrogen estimative model show little or no resemblance to those of the
total phosphorus model. It is also noteworthy that neither map reveals
patterns that appear to have any clear correlation with map units of the
natural macro-aspects (such as physiographic regions, climatic characteris-
tics, soil types, and geology) that one might expect to affect the regional
patterns of the residuals. One exception to these general notations may be
seen in the Pacific Northwest where the data indicate both nitrogen and
phosphorus concentrations in streams draining the well-watered forested
watersheds in western Oregon and western Washington to be generally well
below those predicted by the models.
Figure 27 indicates three major areas where observed total nitrogen
concentrations were above those predicted by the regional models. The first
area is centered on the corn belt and includes most of Iowa, northern and
central Illinois and western and central Indiana. The second area covers
much of northern and central California. The third high area may exist in a
region extending from the Black Hills of South Dakota through most of Wyoming
and into the southern edge of Montana and southeastern tip of Idaho, but
large gaps in data points make definition of this area difficult to support.
This map also reveals a large area centered on eastern Pennsylvania, New
Jersey, and southeastern New York (but including much of Delaware, east-
central Maryland, central Pennsylvania, and northeastern New York and Connect-
icut) where total nitrogen concentrations in streams are near or above those
predicted by the regional models. Figure 27 also suggests that throughout
most of the Appalachian highlands and adjacent parts of the Piedmont, southern
Illinois, southern Indiana, and southern Ohio, total nitrogen concentrations
in streams can be expected to be near or below the levels the models predict.
Another area of lower total-N concentrations is apparent in the western
parts of Oregon and Washington, with the exception of the Puget Sound lowland.
Figure 28, depicting the regional patterns of residuals from the
inorganic nitrogen regional models, illustrates both similarities and
dissimilarities with the-map of total-N residuals. Similar to the total-N
map, Figure 28 reveals patterns of high inorganic-N concentrations in the
corn belt and central part of the northeast. On the other hand, the high
concentrations of total-N indicated for Wyoming and adjacent areas and
67
-------
<8
&
«?
1
TK IN HI HSMIU
la«|(|lf)«-znit * JMMIX ti *«.~ X It Mm
(X (¦*).«% la Mm
U|r»T)-IJII»*
Wwli
tai (W)>-MSU * .«*»(% k l|.*X \m Mm)
a Mitt
U
U H U
UNM
1.1 ft U
-IJ ia-U
-M M-IJ
-IJ
Figure 26. Areal distribution of residuals from regional predictive
models for orthophosphorus concentrations in streams.
-------
d?
%
?*
i
w
IK nUMIM
J*7W(X ta l|.*X li Mm) -JK2J(% h Farnt
(%kl|.*Xklita)-JaiN(Xkhrnt
* JMS(X It H. * % k Ma) - JHJ« (* k hmt
a Mas
U
1J to 1.S
U to IJ
-UtoU.
-IJ to-U
U to-U
« - U
Figure 27. Areal distribution of residuals from regional predictive
models for total nitrogen concentrations in streams.
-------
>3°
*
<*»
w»
%
%
uwtua m ntMtn
KfUTM (v) Mm HM IK LM KM B8MU *
IK nUtWIM HMMU' MHI:
farnl
UijM)* - JM •
(Xtat|.*X ii Ma) - JHM(X h
(Xta»|.*X kMa)-JBIIl«k
{X la l|. * X ii Ma) - jm(X li Nntt)
¦ U •
u a u •
UkU •
-U to U «
-I.lt.-U... O
-u a-u . . . . o
«-!.! o
0
e® °oe
«e I
V
o
o
'*A
&
to#
*%*>
QO
o %
o •
-&¦
Q
ese
'o |