Scientific Investigations Report 2007–5186
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5186
Back to Table of Contents
Yields are useful when making comparisons between catchments because the yield calculation normalizes loads for catchment area (for example, pounds per square mile of catchment). This provides a measure of the “intensity” of pollutant sources, which is useful for watershed management. Although drainage areas were available for 91 of the 100 catchments, WY 2000 yield estimates were made for only 70, 81, and 65 catchments for TN, TP, and SS, respectively, because of either a lack of adequate data to build a model or poor model performance. Basic land-use characteristics for the 91 delineated catchments are shown in figure 23. The catchments tended to have large percentages of forested and undeveloped land and small percentages of agricultural and urban land. The amount of agricultural and urban land in one-half of the catchments was less than 5 percent and 2 percent of the total land area, respectively, whereas the amount of forested or undeveloped land in one-half of the catchments was greater than 86 percent of the total land area. However, some catchments included relatively large areas of agricultural and urban land: in eight catchments, agricultural areas made up more than 50 percent of the total land area; and in five catchments, urban areas made up more than 50 percent of the total land area. Appendix A contains the predicted yields for WY 1997, 2000, and 2001 for TN, TP, and SS, and table 13 summarizes the annual yields for the 3 water years. For TN, TP, and SS, respectively, the median yields were 2.2, 2.7, and 3.8 times greater in 1997 (the high-flow year) compared to 2000 and 2.2, 2.4, and 5.0 times greater in 2000 compared to 2001 (the low-flow year).
The values for yields and annual catchment nutrient loads were not normally distributed. For these variables, the log-transformed (base 10) values were used when relations between catchment yields and catchment characteristics were evaluated. This transformation resulted in approximately normally-distributed values. The two-sided Pearson’s correlation coefficient (r) was calculated for all comparisons. The Pearson’s correlation coefficient expresses the similarity between two sets of data. The significance of each correlation coefficient (p-value) also was calculated. A 5 percent significance level was selected.
Figure 24 shows the TN, TP, and SS yields estimated for the 91 catchments. In this figure, catchments are represented by the monitoring station location (catchment outlet). Table 14 shows the ranges for the WY 2000 yield quartiles used in figure 24. TP yields were highly correlated with SS yields (r=0.84, p≤0.05), whereas TN yields were less correlated with SS yields (r=0.65, p≤0.05). This reflects the fact that phosphorus can adsorb onto and be transported by soil particles, whereas the predominant species of nitrogen do not exhibit this property as strongly. TN yields were highly correlated with TP yields (r=0.81, p≤0.05), which probably is explained by the similarities in the sources of nitrogen and phosphorus in the catchments. The yields for TN, TP and SS generally were greater on the western side of the Cascade Mountains (the Willamette River and the Puget Sound basins) compared to the eastern side of the Cascade Mountains (t-test p-value = 0.00 for all three constituents). The median yield (untransformed) for the western-side catchments was 5.6 times greater for TN, 4.6 times greater for TP, and 5.8 times greater for SS than the median yield for the eastern-side catchments. The western side of the Cascades experiences much more precipitation and surface runoff than the eastern side. When the annual mean streamflows for WY 2000 were normalized for catchment area, the median value for the western-side catchments was 4.7 times greater than the value for the eastern-side catchments, and the two groups were significantly different. This probably explains most of the difference in the regional pattern for yields, because stream loads (and yields) are dependent on flow and, therefore, should be highly correlated with spatial patterns in precipitation.
Notable exceptions to the regional pattern in yields were Granger Drain in the lower Yakima Basin, Granger, Wash. (fig. 1, site 50), which was in the 4th quartile for TN and SS yields and in the 3rd quartile for TP yields, and the Snake River at Flagg Ranch, Wyo., WY 2000 (fig. 1, site 100), which was in the 4th quartile for SS yields. The Granger Drain catchment contains intensive agricultural activity that relies almost exclusively on irrigation during the growing season. In addition, the agricultural activity is near the drain. With 60 percent agricultural land, it was the 6th most agricultural catchment of the 91 catchments with land-use data. This probably explains the relatively high yields for this catchment. The catchment characteristics available for the Snake River at Flagg Ranch, Wyo., could not explain the high SS yield for WY 2000. Of the remaining 14 catchments in the 4th quartile for SS yield, 13 were in the Puget Sound watershed, and 1 was in the upper Willamette Basin.
Thirty-one of the 70 delineated catchments with annual in-stream TN loads predicted for WY 2000 had non-zero estimates for annual WY 2000 point-source TN load, and 31 of the 81 delineated catchments with annual in-stream TP loads predicted for WY 2000 had non-zero estimates for annual WY 2000 point-source TP load. The catchments with point-source nutrient loads of zero could have contained nutrient point sources, but these point sources either were not identified in the available data bases or did not meet the criteria for inclusion in the analysis. Table 22 (at back of report) lists the total WY 2000 estimated point-source TN and TP loads for the catchments. Estimated annual point-source loads as a percentage of the annual in-stream loads are also listed in table 22. For each catchment, the annual loads for all point sources were expressed as a percentage of the annual in-stream WY 2000 nutrient load at the catchment outlet. Because nutrients can undergo many transformations in a stream (for example, settling or uptake by plants), a direct comparison between point-source and annual in-stream load in a catchment is not possible. However, a determination of the relative importance of point-sources in a catchment is possible. Using this approach allowed for the following statement—During WY 2000, the estimated annual point-source load for TN in catchment A was at least X percent of the annual in-stream TN load at the outlet of catchment A.
The annual point-source nutrient loads in the catchments ranged from zero to 85 percent of the predicted annual in-stream TN loads and 98 percent of the predicted annual in-stream TP loads. The annual point-source loads from industrial sources (not WWTPs) generally were a small percentage of the annual in-stream loads (less than 5 percent). The one exception was the Spokane River (fig. 1, site 73), where the annual TP loads from paper mills were 17 percent of the annual in-stream load. Figure 25 shows the point-source nutrient loads for each catchment as a percentage of the annual in-stream load. In this figure, the catchments are represented by the location of the monitoring station. The percentage of the annual in-stream load is represented by four symbol size classes: less than 25 percent, 25–50 percent, 50–75 percent, and greater than 75 percent (table 15 shows the number of sites with these percentages for TN and TP).
The one catchment with point-source annual TN loads greater than 75 percent of the annual in-stream load, the Walla Walla River near Touchet, Wash. (fig. 1, site 46), contained a WWTP that discharged relatively large annual loads of TN (ranking 13th of 140 TN point sources). Five of the six catchments with annual point-source TN loads between 25 and 50 percent of the annual in-stream loads were in the more developed areas of the basin (the Willamette basin, Spokane, and Boise). Although no sites with point-source contributions were in the Seattle-Tacoma area, this does not mean that the streams in that area are necessarily free from point-source effects. Rather, the results likely reflect the lack of water-quality sampling sites downstream of major point sources in the area. The other catchment in this group contained the WWTP for Page, Idaho, which ranked 61st of 140 TN point sources.
As with TN, the one catchment where annual point-source TP loads were greater than 75 percent of the annual in-stream load, the Spokane River (fig. 1, site 73), contained a WWTP that discharged relatively large annual loads (ranking 5th of 130 sources for TP). One of the two catchments with point-source annual TP loads between 50 and 75 percent of the annual in-stream loads, another Spokane River site (fig. 1, site 99), contained WWTPs that discharged relatively large annual loads of TP (if they were aggregated into one point-source, they would rank 14th of 130 TP point sources). The other catchment in this group contained the WWTP for Page, Idaho (fig.1, site 97), which ranked 63rd of 130 TP point sources. All 15 catchments with annual point-source TP loads between 25 and 50 percent of the annual in-stream loads were in some of the more developed areas (Willamette River/lower Columbia River basins, Spokane River, Snake River Plain, and lower Yakima River/lower Snake River/Colombia River confluence). No sites with annual TP point-source loads between 25 and 50 percent were in the Seattle-Tacoma area.
The results from the point-source analysis showed that streams away from major population centers and large point sources can still be significantly affected by point sources, if the annual in-stream load is relatively low. The results also showed that streams downstream from many large point sources are not necessarily largely affected by those upstream point sources. An example of this was the Columbia River at Beaver Army Terminal (fig. 1, site 44). Although this water-quality station is the integrator site for the Columbia River basin (the site is downstream of a large and complex drainage basin and contains multiple environmental settings and point sources), the annual WY 2000 TN load for all estimated point sources in the Columbia River basin was 11.8 percent of the annual in-stream TN load, and the estimated annual TP point-source load was 22.5 percent of the annual in-stream TP load. These results indicate that, in large catchments with a well-mixed land-use profile, nonpoint sources, such as fertilizer, manure, and atmospheric deposition, are responsible for most in-stream load. The 12 largest catchments generally had a diverse mix of urban and agricultural land (1–6.5 percent urban [mean of 3.1 percent] and 1.8–22.7 percent agricultural [mean of 13.7 percent]). The annual point-source loads for these catchments, as a percentage of annual in-stream loads, ranged from 0 to 22.6 percent for TN (mean of 8.9 percent) and from 0 to 42.1 percent for TP (mean of 26.2 percent). For these catchments, all annual point-source loads were less than the mean value of 27 percent for the catchments with an estimated point-source TN load, and all but one annual point-source load were less than the mean value of 39 percent for the catchments with an estimated point-source TP load.
Table 22 shows the predicted annual in-stream TN and TP loads and the estimated annual TN and TP loads from point sources for WY 2000, and the estimated annual fertilizer and manure nitrogen and phosphorus loads and atmospheric nitrogen deposition for calendar year (CY) 2002. The data for CY 2002 were used because manure data were available for that year but not CY 2000. Table 22 also includes (1) annual point-source loads expressed as a percentage of the annual in-stream load, (2) annual fertilizer and manure loads expressed as a percentage of the annual in-stream load, (3) annual atmospheric nitrogen deposition expressed as a percentage of the annual in-stream load, (4) annual point-source loads expressed as a percentage of the annual fertilizer and manure loads, (5) annual point-source TN load expressed as a percentage of the annual atmospheric deposition of nitrogen, and (6) annual point-source TN loads expressed as a percentage of the land-applied nitrogen loads.
Relative allocations of annual TN and TP loads to the catchments from fertilizer and manure, point sources, and atmospheric deposition (nitrogen only) are shown in figure 26. For TN, values ranged from near 0 to nearly 100 percent for both fertilizer and manure (median of 86 percent) and atmospheric deposition (median of 12 percent), and from 0 to close to 50 percent for point sources (median of 0 percent). For TP, values ranged from near 0 to close to 100 percent for fertilizer and manure, and from 0 to close to 80 percent for point sources (median of 0 percent). However, 60 of the 91 catchments had no estimated point-source TN or TP loads. The sites are generally located at the left sides of each plot, indicating that the estimated point source loads tended to be a small percentage of total TN and TP loads in the catchments compared to fertilizer and manure application and atmospheric deposition.
One catchment, the South Fork Coeur d’Alene River (fig. 1, site 97), had annual point-source TP loads greater than the total land-applied phosphorus loads, and 10 catchments had annual point-source TP loads greater than 10 percent of the land-applied phosphorus loads. Annual point-source loads for TP in the South Fork Coeur d’Alene River were equal to 60 percent of the annual in-stream load (the median value was 10.5 percent for all 31 catchments with point-source TP loads). Higher values of annual point-source nutrient loads as a percentage of annual land-applied loads were positively related to higher values of annual point-source nutrient loads as a percentage of annual in-stream loads. In smaller, urbanized catchments, point sources tended to contribute more to the total nutrient load than in larger, less urbanized catchments.
A multiple linear regression of catchment yields (the response variable) on catchment characteristics (the explanatory variables) was performed to determine which characteristics best explained the variability in TN, TP, and SS yields. Seven explanatory variables were used in the regressions for TN and TP: (1) percentage of the catchment land area categorized as urban; (2) percentage of the catchment land area categorized as agricultural; (3) mean slope for the catchment in percent; (4) mean annual precipitation for the catchment for 1980–1997; (5) annual load of nitrogen or phosphorus to the catchment from fertilizer and manure application (CY 2002), normalized for catchment area and log-transformed; (6) annual load of nitrogen to the catchment from atmospheric deposition (CY 2002), normalized for catchment area and log transformed; and (7) annual load of nitrogen or phosphorus to the catchment from point sources (WY 2000), normalized for catchment area and log-transformed. The linear model selected provided the best estimate of the response variable using some linear combination of the explanatory variables, based on the model standard error (a measure of the differences between the measured and the predicted values). The linear model took the form: Y = aX1 + bX2 + cX3 + dX4 + eX5 + fX6 + gX7, where Y was the predicted yield, X1- X7 were values for the catchment characteristics, and a–g were the coefficients. The statistically significant predictive capability of each explanatory variable in the presence of the other variables was provided by the p-values. The magnitude of each regression coefficient indicated the response of the yields to a particular catchment characteristic. The results from the multiple linear regressions are summarized in table 16. Also, included in the table are standardized regression coefficients. To obtain these coefficients, all variables (response and explanatory) were standardized before fitting the multiple regression equation by subtracting the mean for that variable and dividing by the standard deviation. Each standardized regression coefficient represents the change in the response of yield in standardized units to a unit change in the standardized explanatory variable (for example, one standard deviation unit change in precipitation corresponds to some standard deviation unit change in yield). Because standardized regression coefficients are independent of scale units, they are useful in interpreting the results from multiple linear regressions.
Although average annual precipitation was the major explanatory factor for TN, TP, and SS yields, the relation was slightly stronger for TP (on the basis of the standardized regression coefficients). However, precipitation was only one factor that influenced catchment yields; for TN and TP yields another important factor was the amount of nutrients applied to the catchments from point and nonpoint sources. The results from this analysis showed that, although a unit change in point-source loads of phosphorus tended to have about twice the effect on TP yields as a unit change in nonpoint-source loads (fertilizer and manure), a unit change in point-source loads of nitrogen tended to have only about two-thirds the effect on TN yields as a unit change in nonpoint-source loads (fertilizer and manure or atmospheric deposition). The decrease in TN and TP yields with an increase in mean slope may have been an indication of the importance of topography and subsurface flow paths on nutrient delivery to streams. These results might indicate, for example, that greater overland flow and nutrient transport and delivery occurred in catchments dominated by relatively flat topography. However, slope also could have been acting as a surrogate for the percentage of forested land in the catchments because mean slope was strongly correlated with the percentage of forested land. Land use in the catchments also was a factor that influenced nutrient yields—the percentage of agricultural land influenced TN, TP, and SS yields (about as much as point-source loads for TN yields), whereas the percentage of urban and agricultural land influenced TP yields (with the percentage of urban land having a stronger influence). The analysis described above provided an estimation of the factors affecting nutrient and sediment transport in some of the catchments of the Pacific Northwest. A more complete analysis would include many more water-quality monitoring sites, all the catchments, and a more rigorous geospatial statistical approach. A study is planned that will incorporate all of those components. This study will use the SPARROW model to estimate nutrient and sediment sources and transport in the surface waters of the Pacific Northwest that have been mapped on the RF1 hydrologic network (U.S. Environmental Protection Agency, 1996). The SPARROW model makes these estimates by correlating water-quality data with data on pollutant sources (for example, atmospheric deposition, fertilizers, and human and animal wastes) and climatic and hydrogeologic properties (for example, precipitation, topography, vegetation, soils, and water routing) that affect contaminant transport (Schwarz and others, 2006). Much of the data collected for the study described in this report will be used in the upcoming SPARROW study.
Back to Table of Contents