Scientific Investigations Report 2007–5186
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5186
Back to Table of Contents
For this study, trend sites were selected based on the availability of data collected from 1993 to 2003, meaning that data for each site generally were available for most seasons in each year during this period. Consequently, data for the sites were not necessarily statistically representative of water-quality trends in each of the Pacific Northwest states. The results represent conditions at the sites used in this study and should not be extrapolated to other areas in each of the states. Table 17 shows the breakdown of the trend sites by collecting/analyzing agency. There were 11 USGS sites, 14 ODEQ sites, 23 WDOE sites, 2 CWS sites, and, depending on the constituent, 18–20 sites in Oregon, 20–25 sites in Washington, and 4–5 sites in Idaho. Figure 6 shows the distribution of drainage areas for the 45 trend sites with delineated catchments (indicated by blue bars). Delineated drainage areas ranged from 25 to 658,000 km2, although most catchments had drainage areas between 1,000 and 5,000 square km2. Only two catchments had drainage areas less than 100 km2, and only one had a drainage area greater than 100,000 km2. Consequently, the trend results were not necessarily representative of conditions occurring in the smallest and largest basins.
The distribution of land use for the trend sites with delineated catchments is shown in figure 23 (red circles). Total agricultural land use ranged from 0 to 95 percent, with most sites less than 20 percent; total urban land use ranged from 0.5 to 60 percent, with most sites less than 10 percent; and total forest land use ranged from 4 to 100 percent, with most sites greater than 70 percent. The many forested watersheds in the study area, especially in the mountainous parts of the basin where rainfall is high and nutrient concentrations are relatively low, likely provided dilution of nutrient concentrations from point sources and agricultural runoff downstream. Depending on forest activities, the forested areas also may have contributed significantly to the suspended-sediment loads, especially during winter storms and high-flow, snow-melt periods.
For the statistical trend analysis used in this study, the null hypothesis being tested is “There is not a trend in the constituent concentrations for either flow-adjusted trend in concentrations (FATC) or non-flow-adjusted trend in concentration (NFATC) at a site.” A type 1 error in the analyses occurs if the null hypothesis is rejected when it is true. For this study, the tolerable probability level—level of significance—for committing a type 1 error was set at p=0.05; that is, there is a 5-percent chance of rejecting the null hypothesis when it is true (Ott, 1993). As p becomes smaller, the data show more evidence for rejecting the null hypothesis. For example, at p=0.20, there is a 20-percent chance of rejecting the null hypothesis when it is true, and at a p≤0.051, there is only a 0.1-percent chance of rejecting the null hypothesis when it is true. The time span from 1993–2003 for assessing trends in this report is relatively short, considering that the data used in this analysis could be far fewer than 12 samples per year (for example, as few as 56 samples over an 11 year period). Given the paucity of the data, significant trends could have actually existed in the streams even when the reported p-level in this report was computed to be greater than 0.05 because the amount of data analyzed may have been insufficient to reveal the significant trend. In an attempt to compensate for the relatively small data sets, the direction of the trends in the basin that were not significant (p>0.05) also are mentioned in the text to further paint a picture of basin-wide trends. Although not significant, the majority of these trends often were consistent with the direction of the significant trends; thereby indicating that the patterns in trends could be wider spread throughout the basin, and might have been apparent if more data had been available for analysis.
Only streamflow at 2 of the 50 sites: Crab Creek near Beverly, Wash. (fig. 1, site 54) and Rock Creek above Hwy 30/93 crossing at Twin Falls, Idaho, (site 89), had significant trends (decreasing). Unlike the data from most sites, which extended from October 1992 through September 2003, the streamflow data for these sites began in October 1993 (WY 1994). Three of the other four sites with streamflow data starting in WY 1994 rather than WY 1993 also showed a decreasing (but not significant) trend in streamflow. A re-analysis of all trends in streamflow, using WYs 1994 through 2003, showed only one additional site, the Okanogan River at Malott, Wash. (fig. 1, site 79), with a significant trend in streamflow (decreasing). In addition, changing the analytical time period did not substantially change the results for flow-adjusted trend in concentration, nonflow-adjusted trend in concentration, or trend in load. Except at the two sites where there was a trend in streamflow, any site showing a FATC could only show a significant NFATC that was in the same direction (increasing or decreasing).
Because significant NFATC occurred at about 90 percent of the sites with significant FATC for all three constituents, and the values for the NFATC and FATC were generally close in magnitude, only the results for FATC are presented in this section. Overall distributions of FATC for TN, TP, and SS, expressed in percentage per year and milligrams per liter per year, are shown in figures 27, 28, and 29. The figures show distributions for the sites where the FATC was significant (p < 0.05) and the sites where it was not significant. Although many sites showed relatively large FATC for TN based on percentage change, more than 90 percent of the sites had relatively small FATC for TN based on absolute change in concentration (≤ 0.05 mg/L/yr). Most sites showed relatively small FATC for TP based on both percentage change and absolute change in concentration (≤ 0.003 mg/L/yr). Although many sites showed relatively large FATC for SS based on percentage change, more than 90 percent of the sites had relatively small FATC for SS based on absolute change in concentration (≤ 2 mg/L/yr).
The numbers of sites with significant FATC for TN, TP, and SS are shown in table 17. Unless indicated otherwise, all the FATC results presented in the remainder of this section were significant (some FATC results that were not significant also are described to provide a more complete picture of the patterns in trends). For most catchments, the net change in nonhydrologic characteristics (land use and other human activities) was not great enough to cause any FATC for TN (52 percent), TP (68 percent), or SS (60 percent) during this relatively short time period (WY’s 1993–2003). However, some nonhydrologic changes might still have occurred in these catchments, but either they cancelled out (for example, decreasing nonpoint source loads and increasing point-source loads) or were too far upstream to have much effect on water quality. Although a few sites had increasing FATC for TN, TP, and (or) SS (5, 14, and 8 percent of the sites, respectively), more sites showed decreasing FATC (43, 18, and 31 percent of the sites for TN, TP, and SS, respectively).
Some differences in trend results were observed among the agencies (table 17), which might have been due to each agency having sampled sites with different land uses and hydrologic characteristics, using different sampling protocols and (or) laboratory methods, and targeted sampling of problem areas by the State regulatory agencies. For example, USGS data showed decreasing FATC for SS at 73 percent of sites (8 of the 11 sampled), whereas ODEQ and WDOE data showed decreasing FATC for SS at 7 and 26 percent of sites, respectively. In addition, decreasing FATC for TP were measured at only one USGS site and no WDOE sites, compared to 43 percent of the ODEQ sites (table 17). In all three states, the number of decreasing FATC frequently exceeded the number of increasing FATC, except for TP in Idaho and Washington (table 18).
Complete results from the trend analysis for TN are shown in table 19, and figure 30 shows the FATC for TN at each available trend site in the study area. Figure 30 indicates whether there was a significant FATC at each site and, if there was, shows the direction of the trend and indicates its magnitude in percentage per year and milligrams per liter per year (some sites with relatively large percentage change in FATC had small changes in concentration). Twenty-one (48 percent) of the 44 sites had FATC for TN (2 increasing, 19 decreasing, and 6 sites had insufficient water-quality data to determined trends). Trends in load for TN were observed at 10 sites (all decreasing), and 77 percent of the sites had decreasing (but not necessarily significant) FATC and trends in load for TN. The decreasing trends in load for TN generally corresponded to decreasing FATC but not decreasing trends in streamflow, indicating that the decreasing loads resulted solely from decreasing FATC. However, at Rock Creek (fig. 1, site 89) the decreasing trend in load for TN may have been caused by the decreasing FATC and (or) the decreasing trend in streamflow.
Seven sites in the Willamette River Basin had decreasing FATC for TN: the Willamette River main-stem sites at Harrisburg (fig. 1, site 7), Albany (fig. 1, site 11), Salem (fig. 1, site 14), and Portland (fig. 1, site 31), and tributary sites on the Long Tom River (fig. 1, site 8), Johnson Creek (fig. 1, site 23) and the Tualatin River (fig. 1, site 25). Although not significant, the FATC at Little Abiqua Creek (fig. 1, site 15), the Willamette River at Portland (fig. 1, site 33), and the Willamette River at SP&S Railroad Bridge (fig. 1, site 37) also were decreasing.
Other sites in Oregon, Washington, and Idaho with decreasing FATC for TN were in catchments with a wide range of agricultural activity, covering as little as 0.8 percent to as much as 56 percent. The two sites with increasing FATC for TN had different land-use profiles. Agricultural land covered 95 percent of the catchment for Zollner Creek near Mt. Angel, Oreg. (fig. 1, site 16), whereas agricultural land covered only 0.5 percent of the catchment for the Boise River at Glenwood, Idaho (fig. 1, site 93). However, the Boise River at this location received a large point-source load of nitrogen (estimated to be about 40 percent of the in-stream TN load), which indicates that changes in point-source load could have been responsible for the increasing FATC for TN.
Complete results from the trend analysis for TP are shown in table 20, and figure 31 shows the FATC for TP at each available trend site in the study area. Sixteen (32 percent) of the 50 sites had significant FATC for TP (7 increasing and 9 decreasing). Trend in load for TP was observed at six sites (3 increasing and 3 decreasing). The number of sites with decreasing FATC for TP was about one-half of those for TN. Much of this difference can be attributed to the WDOE data, which showed eight sites with decreasing FATC for TN but no sites with decreasing FATC for TP (table 17). About 50 percent of the sites had decreasing (but not necessarily significant) FATC and trend in load for TP.
Six sites in the Willamette River Basin, Oregon: Willamette main-stem sites at Albany (fig. 1, site 11), Salem (fig. 1, site 14), and Portland (fig. 1, site 31), and tributary sites on the Long Tom River (fig. 1, site 8) and the Tualatin River (fig. 1, sites 19 and 25) had decreasing FATC for TP. The four other sites in the Willamette River Basin had decreasing (but not significant) FATC for TP. Increasing FATC for TP was measured in the Puget Sound at four sites: Deschutes River (fig. 1, site 56), Green River (fig. 1, site 63), Puyallup River (fig. 1, site 57), and Skokomish River (fig. 1, site 61). In these catchments, more than 14 percent of the land cover was urban, and less than 9 percent was agriculture. In the eastern side of the Columbia Basin, increasing FATCs for TP were observed in the Methow River (fig. 1, sites 78 and 80) and South Fork Coeur d’Alene River (fig. 1, site 97). These catchments were predominately forested with low agricultural and urban use. Twenty-two of the 24 sites with increasing (but not necessarily significant) FATC for TP were in catchments with less than 15 percent agricultural land, whereas the FATC for TP at 14 sites in catchments with more than about 15 percent agricultural land generally decreased (one-half were significant). Increasing FATC for TP was measured (but not significant) at two larger integrator sites (the Columbia River at Vernita Bridge [fig. 1, site 51] and the Willamette River at Portland [fig. 1, site 33], located in catchments with 18.4 and 22.7 percent agricultural land, respectively).
Complete results from the trend analysis for SS are shown in table 21, and figure 32 shows the FATC for SS at each available trend site in the study area. Nineteen (40 percent) of the 48 sites had significant FATC for SS (4 increasing, 15 decreasing, and 2 sites had insufficient water-quality data to determine trends). Trend in load for SS was observed at seven sites (one increasing, six decreasing). More than 65 percent of the sites had decreasing (but not necessarily significant) FATC and trend in load for SS. All six sites with decreasing trends in load for SS also had decreasing FATC for SS, and one site with an increasing trend in load for SS also had an increasing FATC for SS. Five of the six sites with decreasing trend in load for SS had no trend in streamflow, indicating that the decreasing FATC for SS at these sites was the cause of the decreasing trend in load. At the sixth site, Rock Creek [fig. 1, site 89]), both FATC and trend in streamflow were decreasing, meaning that either could have contributed to the decreasing trend in load at this site.
Significant increasing FATC for SS were measured at four sites: Long Tom River (fig. 1, site 8, with an average increase of 0.6 mg/L/yr from 1993 to 2003) and Hood River (fig. 1, site 42, 1.1 mg/L/yr), which are agricultural catchments; Deschutes River (fig. 1, site 10, 0.1 mg/L/year), a predominately forested and undeveloped catchment; and Willamette River at Salem (fig. 1, site 14, 0.2 mg/L/yr), which integrates runoff from a mix of forested, agricultural, and urban lands. The increasing FATC for SS in the Hood River (fig. 33) was most likely the result of high elevation rain on Mt. Hood in late September, early October 2000. The rain caused a glacial outwash of sediment that filled the Hood River channel and took several years to wash out. This increase in trend probably will not continue at this site, unless another glacial outwash occurs.
Decreasing FATC for SS were measured at four of the five sites in Idaho: Boise River (fig. 1, site 93), Snake River (fig. 1, site 91), Rock Creek (fig. 1, site 89), and South Fork Coeur d’Alene River (fig. 1, site 97). The fifth site, the Spokane River (fig. 1, site 99), also had a decreasing (but not significant) FATC for SS. Three sites in the Willamette River Basin had significant decreasing FATC for SS: Willamette River at Portland (fig. 1, site 33), Zollner Creek (fig. 1, site 16), and Little Abiqua Creek (fig. 1, site 15). Two sites had significant increasing FATC for SS: Long Tom River (fig. 1, site 8) and Willamette River at Salem (fig. 1, site 14). Six sites in the Willamette River Basin had increasing (but not significant) FATC for SS: Middle Fork Willamette River (fig. 1, site 4), Willamette River at Harrisburg (fig. 1, site 7), Albany (fig. 1, site 11), Hawthorne Bridge (fig. 1, site 31), and SP&S Railroad Bridge (fig. 1, site 37), and Johnson Creek (fig. 1, site 23).
Decreasing FATC for SS were measured at three of the four sites with the largest percentages of agricultural land, ranging from about 42 to 95 percent: Zollner Creek (fig. 1, site 16, 95.2 percent agricultural land); Palouse River (fig. 1, sites 52 and 53, 72.5 percent); and Walla Walla River (fig. 1, site 46, 56.1 percent). One site (Crab Creek [fig. 1, site 54, 42.2 percent]), showed decreasing FATC for SS that was not significant. The Walla Walla River, Palouse River, and Crab Creek are in southeastern Washington, and Zollner Creek is in the Willamette River Basin, Oregon. Other than at these four sites, the percentage of agricultural land in a basin did not appear to be related to the increasing or decreasing FATC for SS.
Ancillary data on (1) nitrogen and phosphorus loads for fertilizer and manure use for farm and on-farm applications, (2) nitrogen loads from atmospheric deposition, and (3) nitrogen and phosphorus loads from point sources (table 22) were acquired for 45 trend catchments. For this study, point-source nutrient loads were estimated only for CY 2000, and total fertilizer and manure applications were estimated only for CYs 1992, 1997, and 2002, because manure estimates were available only for those 3 years. Because of the scarcity of nutrient data from point sources for the 1993–2003 period, temporal changes in point-source input trends generally were not assessed in this report; however, the relative importance of point-source discharges, on the basis of CY 2000 data, is addressed in relation to their potential effect on the trend results.
Estimates of point- and nonpoint-source contributions of nitrogen and phosphorus are shown in table 22 and figure 26. Point-source loads could have an immediate effect on water quality and trends, because the effluent generally is piped directly into the stream, whereas the transport of nutrients from nonpoint sources (for example, land applications of inorganic fertilizers and manure) generally is delayed by the timing of erosional processes or by way of seepage into the ground water, which eventually discharges to the streams. Nutrients from nonpoint sources first contact the land surface (for example, impervious surfaces, soil particles, and vegetation) and their transport to the stream is delayed by sorption, biological uptake, and (or) degassing to the atmosphere (for example, denitrification). Eventually, part of the nonpoint-source nutrients will flow into the streams, but the process could take hours to decades depending on the characteristics of the flow path. Soluble nitrate fertilizer not taken up by plants likely will move through the soils into the ground water and gradually seep back to the stream. In contrast, phosphorus fertilizer is more likely to adsorb to soils and erode to streams during periods of intense irrigation or precipitation events, especially during high-flow years. Therefore, land-applied nutrients will likely show a temporal delay in being transported to streams. To some extent, the transport to streams of atmospheric nitrogen also may be delayed; however, during the cold, wet winters in the Pacific Northwest when the soils are saturated and biological uptake is minimal, transport of atmospheric nitrogen is likely to be greater during the large rainfall/snowmelt events.
Of the 43 trend sites with in-stream and point-source TN load estimates, point-source loads accounted for less than 25 percent of the in-stream nitrogen load at 37 sites (86 percent). Of the 44 sites with in-stream and point-source TP load estimates, point-source loads accounted for less than 25 percent of the in-stream phosphorus load at 31 (70 percent) of the sites (table 22). Assuming that (1) a point-source load accounted for 25 percent (or less) of the in-stream load for a catchment, and (2) a decreasing trend in the in-stream nutrient load was 3–4 percent per year (which was typical), the decrease in the point-source load would have to be 12–16 percent or more per year to account for the decreasing trend observed in the stream. Over an 11 year period, this would equal 132–176 percent. Because point-source nutrient loads in most catchments during this period probably increased due to increases in population, a decrease in point-source loads of this magnitude does not seem likely. Consequently, in catchments with point-source loads equal to less than 25 percent of the in-stream load, the trends observed were not likely caused solely by reduced point-source nutrient loads. If this study had focused on seasonal rather than annual trends, the effect of point-source nutrient discharges might have been apparent during periods of low flows, because this is when point-source loads can often contribute a larger part to the in-stream nutrient load. Additionally, seasonal trends would not necessarily be consistent with the annual trend results shown in this report. For example, FATC for TN during summer low flows might have been increasing at a number of sites because of increasing populations, rather than generally decreasing (as was shown in the annual data).
Results from this study indicate that nonpoint-source loads of nutrients probably have decreased over time in many of the catchments. Some data for the Willamette River Basin support this hypothesis. Between 1991 and 2000 the sum of the annual average flows for the largest nutrient point sources in the basin (those contributing close to 90 percent of the total measured point-source flow in 2000) increased from 199 Mgal/d (Tetra Tech, Inc., 1992) to 233 Mgal/d, an increase of 17.3 percent. This increase in total annual average flow corresponded to an increase in population in the basin during this period. In addition, the estimated total annual TP load from a subset of these sources increased by 130 percent from the early 1990’s (Tetra Tech, Inc., 1992) to 2000. Although this subset represented only 37 percent of the total TP load in 2000, the trend indicates that total TP load from point sources in the Willamette River Basin increased overall since the early 1990’s.
Point sources generally accounted for less than 20 percent of the total load (from point and nonpoint sources) of nitrogen and phosphorus to the trend-site catchments (fig. 26). Except for a few sites, point-source loads were equivalent to less than 10 percent of the land-applied nutrients in fertilizers and manure. In all the trend-site catchments, nonpoint sources of nutrients from land applications of fertilizers and manure exceeded point-source loads and, generally, even nitrogen loads due solely to atmospheric deposition exceeded point-source loads. The few exceptions are shown in table 22.
In spite of the generally small contribution of point-source nutrient loads, they still may have been partially responsible for the significant trends at several sites assessed in this study, especially those sites where the point-source nutrient loads exceeded 25 percent of the in-stream load. These sites were the Willamette River at Highway 99E, Oreg. (Harrisburg) (fig. 1, site 7), the Walla Walla River near Touchet, Wash. (fig. 1, site 46), and the Boise River at Glenwood, Idaho (fig. 1, site 93) for TN, and the Willamette River at Albany (fig. 1, site 11), at Marion Street (Salem) (fig. 1, site 14), and at Hawthorne Bridge (Portland) (fig. 1, site 31), and the South Fork Coeur d’Alene River near Pinehurst, Idaho (fig. 1, site 97) for TP. The catchment for the Boise River site had the following characteristics: (1) the point-source TN load was 40 percent of the in-stream load, (2) the population increased by about 20 percent from 1990 to 2000 (about 2 percent per year), and (3) the FATC for TN was significant at 2.1 percent per year. These observations indicate that a possible increase in nitrogen load from point sources could have accounted for the increasing FATC for TN. Although the FATC for TP at this site was not significant, it did increase by 3.3 percent per year, indicating that a possible increase in phosphorus load from point sources could have accounted for the increasing FATC for TP. Except for the limited data for the Willamette River basin, temporal point-source data were not available from 1993 through 2003 to accurately assess water-quality trends throughout the region in relation to changes in point-source loads.
Sixteen of the 19 sites with decreasing FATC for TN also had decreasing annual loads of atmospheric nitrogen in their catchments between CY’s 1992 and 2002 (ancillary data were not available for three of the catchments). However, no consistent temporal pattern in the load of nitrogen from fertilizers and manure was observed across the 16 catchments during this time period. Zollner Creek near Mt Angel, OR (fig. 1, site 16) and the Boise River at Glenwood, ID (fig. 1, site 93) were the two sites that showed increasing FATC for TN. More than 99 percent of the nitrogen in the Zollner Creek catchment came from land-applied fertilizers and manure, which also had an upward trend in nitrogen application from 1992 to 2002. The increasing FATC for TN in the Boise River catchment could not be explained with the available fertilizer and manure data. However, as discussed above, the increasing FATC for TN and TP at this site might have been due to the influence of point sources. Although data on annual phosphorus loads from fertilizers and manure were available for 13 of the 15 sites having FATC for TP, the trends could not be explained by the annual patterns in the application of phosphorus from fertilizers and manure. However, the limited data on point sources showing large increases in TP loads to the Willamette River Basin between 1991 and 2000 strongly indicate that an overall decrease in nonpoint-source phosphorus load was the reason for the decreasing significant FATC for TP at the three lower Willamette River sites.
Back to Table of Contents