Data sources and methodology Watersheds were selected from among those investigated in the U.S. Geological Survey's (USGS) National Water Quality Assess ment program. These program study units were first imple mented in 1991. A modified version of the scheme employed by Smith and others (S) was used to classify 114 watersheds based on the percent of various land use classes (agriculture, forest, urban, mixed). This provided an indication of the land use in the water shed that would be most likely to have the greatest impact on water quality (Table 1). Watersheds were also classified accord ing to region (Western, Central, Southeastern, Northeastern), providing a means of grouping watersheds along broad hydrogeologic, climatic, and vegetation zones. Estimates of nitrogen and phosphorus in commercial fertilizer and in animal manure were derived from national databases of fertilizer sales for 1985-1991 (3) and from Census of Agriculture animal population figures for 1982 and 1987 (R. B. Alexander, 1993, written communication). County-level estimates of nitrogen and phosphorus in commercial fertilizer were apportioned to watersheds based on the proportion of agricultural land in coun ties located within the watershed. Atmospheric deposition inputs of nitrate and ammonium were spatially averaged using an inverse-weighted average of the distance from the National Atmospheric Deposition Program/ National Trends Network stations ( 10) to a central point in each
sheds. For example, animal manure accounted for more than half of the nitrogen inputs in Virginia's Shenandoah River watershed whereas, across the Po tomac River in Maryland, fertilizer accounted for more than half of the nitrogen in the Monocacy River watershed. One interesting finding was that fertilizer nitro gen inputs per unit area of agricultural land were re lated to the dominant land use in the watershed. In general, agricultural land in those watersheds clas sified as predominantly agricultural received larger nitrogen inputs from fertilizer than agricultural land in predominantly urban watersheds, probably as a result of differences in the intensity of agricultural practices. These findings underscore that nutrient in puts to watersheds are dependent on variations in land use practices as well as the mixture of land use in the watershed. It is not enough to know what the land is used for. It is equally important to know the type and intensity of that use. Considerable uncertainty remains over the rela tive importance of the major nonpoint sources with respect to their contributions to stream loads. Mass balance studies of individual watersheds have sug gested that although animal manure may represent a potentially large nutrient input to the watershed, it may contribute only a small portion of stream load (15-17). However, statistical studies of water qual ity trends have suggested that increases in total ni trate and phosphorus are associated with increases in livestock population densities and fertilized acre age (8, 18). There is mounting evidence that atmospheric ni trogen inputs may be a more important nitrogen source than has been previously suggested (19). Fisher and Oppenheimer (15) and Jaworski and his col-
watershed and then multiplied by the area of the watershed to estimate total inputs. Adjustments were made for droplet deposi tion, urban effects, and dry deposition (6). Point-source estimates were developed from data collected by EPA's Permit Compliance System. Effluent volumes were mul tiplied by the reported concentrations of nitrogen and phospho rus for each facility located in the watershed to yield estimates of the total input to a given stream ( 11). Mass transport of total nitrogen and total phosphorus in streams was calculated for typical-flow years (using the mini mum variance unbiased estimator method [12, 13\) with concen tration data available from USGS and the EPA databases. Retention was calculated as the fraction of the total nutri ent inputs to the watershed from manure, fertilizer, air, and point sources that could not be accounted for in the stream load. Although these four sources are not the only sources of nitrogen and phosphorus, they are believed to represent the majority that contributes to stream loads in most of the studied watersheds. Statistical tests (Kruskal-Wallis and Tukey's Honestly Signifi cant Difference) were used to look for differences among groups (14). All inputs and loads were compared on the basis of unit areas (metric tons per square kilometer). For each water shed, fertilizer and manure totals were apportioned to agricul tural land areas, atmospheric deposition and stream loads were apportioned to the total area, and point sources were appor tioned to the urban areas.
FIGURE 5
Percent of stream load attributed to point-source discharges The x-axis is the percent of either the nitrogen or phosphorus stream load that is attributed to point-source discharges; the y-axis counts the number of streams that fell within that percent range. L 50 ι *
VOL. 29, NO. 9, 1995/ENVIRONMENTAL SCIENCE & TECHNOLOGY • 4 1 3 A