National Water-Quality Assessment (NAWQA) Project

 Go to:      NAWQA Home

Pesticide National Synthesis Project

Home Publications National Statistics Data Pesticide Use Water-Quality Benchmarks PNSP Internal

Pesticides in Streams of the United States--Initial Results from the National Water-Quality Assessment Program

By Steven J. Larson, Robert J. Gilliom, and Paul D. Capel

U.S. GEOLOGICAL SURVEY
Water-Resources Investigations Report 98-4222
Sacramento, California, 1999


EFFECTS OF LAND USE ON OCCURRENCE OF PESTICIDES IN STREAMS

The use of most pesticides is directly related to land use. The largest amounts of pesticides are used on agricultural land, with different combinations of chemicals used on different crops and in different climates. The total amount of pesticides applied in urban areas is less than the total amount applied in agricultural areas, but urban pesticide use is often more intensive than agricultural use in terms of the amount applied per unit area. Little or no pesticides are applied to undeveloped land. Thus, understanding the relation between land use and pesticide use is key to evaluating the causes of pesticide occurrence in streams.

Overview

A mixture of the target compounds was present in most samples from the three types of sites. The number of different compounds detected at most of the agricultural and urban indicator sites and at most of the integrator sites was similar (fig. 12). Between 17 and 26 of the target compounds were detected at most of the sites, including 13 to 17 herbicides and 3 to 8 insecticides. A number of different compounds were detected at all sites, but the number varied considerably among the sites of a particular type. Variability was greatest among the agricultural sites, with the number of detected compounds ranging from 7 to 37 (U.S. Geological Survey, 1999). This wide range is to be expected because of the variety of different crops represented by the agricultural sites. In addition,pesticides not included in the target compounds are used extensively in some of the agricultural basins (table 3). At some of these sites, the number of compounds detected in the streams may have been relatively low because the pesticides that had the highest use in the drainage basin were not targeted in this study. The number of detected compounds at the 11 urban sites ranged from 11 to 29, which was smaller than the range for the agricultural sites, and the number of detected compounds at the 10 integrator sites ranged from 10 to 33 (U.S. Geological Survey, 1999).

Because of the seasonal nature of both pesticide use and the occurrence of pesticides in streams, the number of compounds detected in individual samples from a site usually was much lower than the total number of compounds detected at that site. The average number of different compounds detected in each sample from each site is shown in figure 13. The values shown in figure 13 were adjusted to account for differences in the number of samples collected at each site and for the uneven sampling frequency at most sites. Thus, these values are estimates of the average number of compounds that would be detected in samples collected at even intervals throughout the year. At most of the agricultural sites, an average of 3 to 8 of the target compounds were detected in each sample, including 3 to 7 herbicides and 0 to 1 insecticide. At most urban sites, an average of 5 to 8 compounds were detected in each sample, including 3 to 6 herbicides and 1 to 2 insecticides. At most integrator sites, an average of 5 to 8 compounds were detected in each sample, including 4 to 6 herbicides and 0 to 1 insecticide. Similar to the number of compounds detected at each site, the variability in the number of compounds detected in each sample was greatest at the agricultural sites. The average number of compounds detected in each sample from the agricultural sites ranged from less than 1 to more than 13 compared with an average of 3 to 9 compounds at the urban sites and 4 to 10 compounds at the integrator sites (U.S. Geological Survey, 1999).

Detection Frequency

Overall detection frequencies of all target compounds are shown for each of the three types of sites in figure 14. The range of detection frequencies for the most commonly detected herbicides and insecticides among sites in each group is shown in figure 15. As in figure 6, the detection frequencies in both figures 14 and 15 are unbiased with respect to sampling frequency and are based on a common reporting level of 0.01 µg/L.

In general, the same compounds were detected at all three types of sites, but at different frequencies. The agricultural and the integrator sites were very similar in terms of which pesticides were detected (fig. 14A,C), but the detection frequencies at the integrator sites were slightly higher. This was expected because the integrator basins generally are larger and have a wider variety of crops and other land uses than most of the agricultural basins and because high-use pesticides are applied in nearly all of the integrator basins. Use of some of the high-use compounds was low (or zero) in some agricultural basins, which were chosen to represent specific crops or groups of crops.

Many compounds, including some with relatively high national agricultural use, such as trifluralin, butylate, molinate, phorate, and terbufos, had low detection frequencies at all three types of sites. In terms of the aggregated data for each type of site (fig. 14), annual average detection frequencies were less than 5 percent at all three types of sites for 13 herbicides and 14 insecticides. Most of the compounds with low average detection frequencies, however, were detected frequently at one or a few sites (fig. 7).

Several compounds, including the herbicides prometon, simazine, and tebuthiuron and the insecticides carbaryl, chlorpyrifos, diazinon, and malathion, were detected most frequently at urban sites. Atrazine and metolachlor, herbicides used almost exclusively in agricultural applications throughout most of the United States, also were detected frequently at most urban sites. The frequent detections of these compounds are due, in part, to the presence of some agricultural land in most of the urban basins. These compounds also have been detected frequently in precipitation in the Midwest (Capel, 1991; Wotzka and others, 1994; Goolsby and others, 1997), which is another potential source of these compounds to urban basins.

The boxplots in figure 15 indicate that many of the detection frequencies shown in figure 14 for the combined sites are actually the mean of widely varying values among the sites within each type of site. For example, detection frequencies for atrazine, simazine, metolachlor, and DEA ranged from near zero at some agricultural sites to 100 percent at other sites (fig. 15). For the agricultural and the integrator basins, much of this variability probably is due to large differences in the use of specific pesticides among the basins. The variability in detection frequencies for many of the compounds at the urban sites also may primarily be due to differences in use, but data on nonagricultural pesticide use are not available. Among urban sites, variability of detection frequencies for the compounds most characteristic of urban streams (simazine, prometon, diazinon, carbaryl, and chlorpyrifos) generally is lower than the variability for major herbicides at agricultural sites (fig. 15). This implies that use of these five compounds is relatively consistent among the 11 urban areas compared with the use of major herbicides which is much more variable in the 37 agricultural basins. The agricultural basins represent several agricultural crops and related pest-management situations.

A few compounds were frequently detected at nearly all the sites (fig. 15). Detection frequencies for atrazine were greater than 80 percent at more than one-half of the agricultural and the urban indicator sites and at more than three-fourths of the integrator sites. Detection frequencies for metolachlor were greater than 50 percent at more than one-half of the agricultural and the integrator sites. Detection frequencies for simazine and prometon were greater than 90 percent at more than one-half of the urban sites and greater than 40 percent at more than one-half of the integrator sites. These high detection frequencies indicate that these compounds are present for much of the year at many of the sampling sites.

Concentrations

The distributions of monthly median total herbicide and total insecticide concentrations are shown in figure 16 for the three types of sites for a 1-year period. The distribution of total herbicide concentrations was similar for all three types of sites (fig. 16A). Approximately 60 percent of monthly median concentrations were greater than 0.1 µg/L and less than 12 percent were greater than1 µg/L. Monthly median concentrations greater than 5 µg/L occurred only at the agricultural indicator sites and the integrator sites, reflecting the relatively high concentrations resulting from the spring flush that occurs at many of these sites.

The distribution of total insecticide concentrations for the urban indicator sites was distinctly different from the distributions for the agricultural and the integrator sites (fig. 16B). At the urban sites, nearly 100 percent of monthly median concentrations were greater than 0.01 µg/L, indicating that one or more insecticides were detected for most of the year at nearly all the urban sites. In contrast, only about 30 percent of monthly median insecticide concentrations were greater than 0.01 µg/L at the agricultural and the integrator sites. Higher insecticide concentrations also were much more common at urban sites than at agricultural or integrator sites. At the urban sites, approximately 80 percent of monthly median total insecticide concentrations were greater than 0.1 µg/L and about 10 percent were greater than 1 µg/L. At the agricultural and the integrator sites, only about 10 percent of monthly median concentrations were greater than 0.1 µg/L, and virtually none were greater than 1 µg/L.

The concentration distribution plots in figure 16 show the general concentration patterns observed for the three types of sites. It is important to note that the median concentrations shown in these plots minimize the influence of extreme values in the distribution of actual concentrations. Maximum total herbicide concentrations at some agricultural sites were much higher than is indicated in figure 16A, with concentrations greater than 10 µg/L common in individual samples at some sites during May and June. In addition, the apparent similarity of herbicide concentrations at the three types of sites shown in figure 16 is somewhat misleading because of the use of monthly median concentrations. Peak concentrations of herbicides at urban sites and at most integrator sites generally were lower than at agricultural sites, but concentrations usually remained elevated for longer periods of time. The concentration distribution plots in figure 16, which show monthly median concentrations and give equal weight to concentrations for each month, tend to smooth out the differences between the three types of sites. The variability in concentrations among agricultural, urban, and integrator sites is discussed in separate sections of this report.

Seasonal Patterns

Pesticides concentrations varied seasonally at nearly all the sampling sites, but temporal patterns were most apparent at the agricultural indicator sites and the integrator sites. A critical period was defined for each site (fig. 4) during which both the number of pesticides present and the pesticide concentrations were highest. The average number of compounds detected in individual samples during the critical period and during the rest of the year is shown in figure 17. At most agricultural sites, an average of 4 to 10 compounds were detected in each sample during the 5-month critical period compared with 2 to 6 compounds detected in each sample during the rest of the year. At the integrator sites, the difference was slightly less evident, with 6 to 10 pesticides detected in each sample during the critical period compared with 3 to 7 during the rest of the year. At most urban sites, however, there was little difference in the number of compounds detected in each sample during the critical period and during the rest of the year.

The difference in seasonal patterns among the three types of sites also is evident in terms of concentrations. In figure 18, total pesticide concentrations are plotted for each month for each of the three types of sites. At most agricultural and integrator sites (fig. 18A,C) there was a distinct seasonal peak in herbicide concentrations during May, June, and July, with monthly median concentrations greater than 10 µg/L at some sites. During autumn and winter, monthly median concentrations were less than 0.5 µg/L at nearly all sites. This pattern was much less obvious at most urban sites (fig. 18B). Although concentrations at some urban sites were elevated from May through August, monthly median concentrations at most sites remained less than 1 µg/L throughout the year. In general, total pesticide concentrations were less variable during the year at the urban sites than at the agricultural and the integrator sites. Three urban indicator sites that have significant portions of agricultural land in their drainage basins--Cedar Run in Pennsylvania (lsus-cedar), Rush Creek in Texas (trin-rush), and Little Buck Creek in Indiana (whit-little) (see fig. 3); these sites were are not included in figure 18B so that the concentrations shown are primarily the result of urban influences. This restriction is especially important for assessing seasonal patterns for urban basins because the concentrations during the spring flush from even a relatively small amount of agricultural land would tend to overwhelm concentrations resulting from urban influences.

Annual Pesticide Yields

The effects of land use on pesticide occurrence in streams also can be assessed by comparing the yields of pesticides in streams influenced by different land uses. Pesticide yield is defined as the mass of a pesticide (or a selected group of pesticides) transported in a stream for a specific period (load) per unit area of land. Comparisons that are based on yields minimize the influence of differences in stream discharge, which can affect both detection frequencies and concentrations.

Annual total herbicide and total insecticide yields at individual sampling sites are shown in figure 19. The yields varied widely, especially among the agricultural indicator sites. Yields at the agricultural sites ranged more than three orders of magnitude for both total herbicides and total insecticides. Total herbicide yields at most agricultural sites were between 0.01 and 4.1 kg/km2/yr, and total insecticide yields generally were about 10 times lower, ranging from about 0.0001 to 0.4 kg/km2/yr (fig. 19). To a large extent, the variability among the agricultural sites can be attributed to differences in the pesticides used on the various crops grown in these basins. This variability is examined more closely in the next section. Variability in weather, soils, and topography among the agricultural sites may also be important factors (Leonard, 1990). In addition, load estimates (and therefore yield estimates) for streams with small drainage basins have greater uncertainty than load estimates for streams with larger basins, a consequence of the relatively short duration of elevated pesticide concentrations and the greater short-term variability in discharge in most of these small streams. The probability of sampling small streams during peak concentrations is lower than the probability of sampling peak concentrations in larger streams because elevated pesticide concentrations in larger streams usually are spread out over longer periods. The range in yields was somewhat smaller for the integrator sites, with values ranging from about 0.03 to 1.9 kg/km2/yr for total herbicides and 0.0002 to 0.15 kg/km2/yr for total insecticides (fig. 19). The integrator sites are considerably larger than most of the agricultural sites, and the estimates of yield for these sites are probably more accurate.

Total herbicide yields at the eight urban sites with little or no agricultural use were between 0.008 and 0.45 kg/km2/yr (fig. 19). This range is at the lower end of the range observed for the agricultural sites. Total insecticide yields at urban sites ranged from 0.015 to 0.31 kg/km2/yr, which is at the upper end of the range for the agricultural sites, meaning that total insecticide yields were higher at many of the urban sites than at most of the agricultural sites. This is consistent with the detection frequency and concentration data discussed earlier, which indicated that insecticides are detected more frequently and at higher concentrations in streams draining urban areas than in streams draining many agricultural areas. These data show that, in terms of mass per unit area, the amount of insecticides transported to streams from urban basins exceeded the amount transported from most of the agricultural basins. Compared with agricultural and integrator sites, yields among urban sites were less variable, reflecting a general similarity in pesticide use among different urban areas.

Agricultural Indicator Sites

The agricultural indicator sites in the NAWQA Program were chosen to represent streams affected by selected agricultural settings; the agricultural sites are defined primarily by specific crops or groups of crops. Comparison of the pesticide levels measured in streams draining basins with different types of agricultural settings can provide information on the relative contribution of different types of agricultural activities to the occurrence of pesticides in streams. An important consideration in making this type of comparison, however, is how much of the pesticide use in a particular basin is accounted for by the target pesticides. Table 3 shows that coverage of total pesticide use by the target compounds varies widely among the agricultural basins. Because the coverage of pesticide use for certain crops, such as corn and soybeans, is more complete than coverage is for other crops, such as wheat or peanuts, certain comparisons are not appropriate. For example, comparisons of total pesticide yields from corn-growing areas and wheat-growing areas would be biased because the coverage of pesticide use in the corn-growing areas is more extensive. On the other hand, comparisons of the loads of individual pesticides as a percentage of their use in the basin are not biased and therefore can provide useful information on the behavior of pesticides in different agricultural settings. In addition, comparisons of detection frequencies and concentrations of pesticides measured at sites representing different crop groups can be made as long as the differences in coverage of pesticide use are considered.

The agricultural sites have been classified according to the major crops grown within the drainage basin (table 4). The classifications for these basins are derived from a national-scale classification of agricultural land in the United States (Gilliom and Thelin, 1997), which uses percentage criteria to define combinations of 1 to 3 crops that account for 50 percent or more of harvested cropland within a county. Crop data used for the classification of the basins in this study are derived from the 1992 Census of Agriculture (U.S. Department of Commerce, 1995) and from the U.S. Geological Survey's Land Use and Land Cover database (U.S. Geological Survey, 1990), as described in the methods section. For most of the basins in this study, the classifications are based on the percentage of harvested cropland planted in specific row crops, such as corn, soybeans, alfalfa, cotton, and wheat and small grains. For two of the agricultural basins in the San Joaquin study unit--Orestimba Creek and the Merced River--orchards and vineyards account for a substantial portion of the agricultural land use in the basin; thus, the classification of these two basins is based on the amounts of specific orchard and vineyard crops grown. No classification was assigned to the two basins in the Ozark Plateaus study unit because the estimates of harvested cropland in these basins account for less than 10 percent of the total basin area and because there is no reported orchard or vineyard acreage. The classifications in Gilliom and Thelin (1997) were developed to account for major crop groups at a national scale. In some cases, the combinations of crops grown in the small agricultural basins included in this study do not fit exactly into any of the categories established in the national classification scheme (for example, Salt Slough in California). In such cases, the site was assigned the category that most closely matched the distribution of crops grown in the basin. In other cases, an important crop in a particular basin was not included in the assigned category (for example, cotton in the Chambers Creek Basin in Texas or peanuts in the Tucsawhatchee Creek Basin in Georgia). With these limitations in mind, the crop-group classifications provide a useful way of organizing the agricultural sites for comparisons of pesticide occurrence in various agricultural settings.

Detection Frequencies

Herbicides were detected much more frequently than insecticides at most of the agricultural indicator sites (fig. 14A). The most commonly detected herbicides include atrazine, metolachlor, alachlor, cyanazine, and EPTC--herbicides used heavily in corn and soybean growing areas. DEA, a transformation product of atrazine and the herbicides simazine and prometon, which have significant nonagricultural use, also were detected frequently at many of the agricultural sites. The remaining 19 herbicides and all the target insecticides were detected in less than 10 percent of samples at most of the agricultural sites.

Detection frequencies for some compounds varied widely among the agricultural sites (fig. 15). Much of this variability is caused by differences in the amounts of these compounds used in the various agriculture basins. This variability in use is illustrated in figure 20, which shows the amounts of specific compounds used for agriculture in these basins. In figure 20, pesticide use is normalized to basin area to account for differences in basin size. For some compounds, use varies widely. For example, atrazine use exceeded 25 kg/km2 in 6 basins, but was less than 1 kg/km2 in 14 basins. Use of metolachlor, alachlor, cyanazine, pendimethalin, and chlorpyrifos also varied widely. Use of terbacil, diazinon, azinphos-methyl, and propargite was relatively high in a few basins but low (or zero) in all others. A few compounds, including the herbicides prometon, tebuthiuron, thiobencarb, and propanil, and the insecticide lindane, had very low (or zero) reported agricultural use in all the agricultural basins.

In general, the relation between detection frequency at the agricultural sites and pesticide use in the basins was significant, but weak. Figure 21 shows the relation between annual detection frequencies and the amount used for agriculture in the basins for all target herbicides and insecticides with reported agricultural use. A general positive relation is evident for both herbicides and insecticides, as most compounds with higher use were detected more frequently in most basins. However, there is considerable variability. For example, detection frequencies for herbicides used at rates of 1 to 10 kg/km2 ranged from 0 to 100 percent. Many compounds with relatively high use were detected infrequently or not at all. Linear regression of the data used in figure 21 shows that only about one-fourth of the variability in detection frequency is accounted for by differences in pesticide use (r2 = 0.30 for herbicides and 0.23 for insecticides). For some individual compounds, the correlation between use and detection frequency was somewhat higher, but only two compounds (azinphos-methyl and disulfoton) had values of r2 higher than 0.5. Clearly, factors other than the amount used, such as pesticide properties, weather, and agricultural practices (Leonard, 1990), can have a major influence on the occurrence of pesticides in streams. Perhaps most importantly, the weakness of correlations between pesticide occurrence and use may result largely from inaccuracies in the pesticide-use estimates for the agricultural basins for the year that samples were collected.

Detection frequencies for specific compounds at the agricultural sites are shown in table 5. The sites in this table are arranged by crop-group classification. Detection frequencies greater than 50 percent are indicated with dark shading and those greater than 10 percent but less than 50 percent are indicated with lighter shading. In general, when a wide variety of crops were grown in a basin, a wider variety of pesticides were detected in the streams. The presence of certain crops, such as vegetables, orchard and vineyard crops, and nursery stock, commonly was associated with detection of a higher number of compounds. The patterns evident in table 5 help illustrate the similarities and differences among sites within a given crop-group classification and among sites with different classifications.

In terms of compounds that were detected most frequently, a generally consistent pattern is evident for sites where corn is a major crop (table 5). The herbicides atrazine, metolachlor, alachlor, and cyanazine were detected frequently at nearly all of these sites. Simazine also was detected frequently even though it was not one of the major compounds used on corn in these basins (Gianessi and Anderson, 1996). Several other herbicides, including metribuzin, pendimethalin, and trifluralin, were detected frequently at corn and soybeans sites; these compounds commonly are applied to soybeans (Gianessi and Anderson, 1996). Most of the sites where corn is a major crop had fewer than three insecticides with detection frequencies greater than 10 percent. Carbaryl, chlorpyrifos, and carbofuran were the most commonly detected insecticides at these sites. The pattern for Lonetree Creek in Colorado (splt-lone) is an exception to the generally consistent pattern among the corn sites. At this site, several other compounds also were detected relatively often, including the herbicides EPTC, DCPA, butylate, ethalfluralin, and linuron and the insecticides propargite and methyl-parathion. Agricultural land in this basin generally is irrigated, and stream discharge during much of the growing season is primarily due to irrigation return flows, which may contain elevated levels of pesticides (Kimbrough and Litke, 1996).

Among the sites where wheat (but not corn) is a major crop, fewer compounds had detection frequencies greater than 10 percent. At most of these sites, detection frequencies for all insecticides were less than 10 percent. Only a few sites had any herbicides with detection frequencies greater than 50 percent. Atrazine and simazine were detected relatively often at several of the wheat sites. The lower number of herbicides detected at these sites was due, in part, to poor coverage of herbicide use in these basins by the target compounds (table 3). Several major herbicides used on wheat, including 2, 4-D, bromoxynil, dicamba, diuron, and MCPA, are not included in the target compounds; however, most of the insecticides used on wheat, including chlorpyrifos, disulfoton, methyl parathion, permethrin, and phorate, are included in the target compounds. These compounds were detected infrequently or not at all at most of the wheat sites. Two sites in the Central Columbia Plateau study unit--Crab Creek Lateral (ccpt-crab.rl) and EL68 Wasteway (ccpt-el68)--are exceptions to the general pattern among the wheat sites (table 5). At both of these sites, a number of herbicides and insecticides had detection frequencies greater than 10 percent. The higher number of compounds detected at these two sites may be due to the presence of a relatively small amount of land with orchards (mainly apples) and greenhouse and nursery crops in these basins (table 4). The results for these two sites suggest that minor crops with relatively high pesticide use may have a strong influence on pesticide occurrence in streams.

Peanuts are the major crop in two basins--Little River (gafl-little) and Aycocks Creek (acfb-aycocks) in Georgia. Substantial portions of three other basins are also planted in peanuts--Pete Mitchell Swamp in North Carolina (albe-pete) and Lime Creek (acfb-lime) and Tucsawhatchee Creek (gafl-tucsa) in Georgia (table 4). No clear pattern in detection frequencies is evident in table 5 for sites in these basins. In general, few compounds had detection frequencies greater than 10 percent in the streams draining these basins. Several of the major pesticides used on peanuts, including the herbicides 2,4-DB, paraquat, bentazon, and acifluorfen and the insecticide aldicarb, are not included in the target compounds. Metolachlor, which is used extensively on peanuts and other crops grown in these basins (Gianessi and Anderson, 1996), was detected frequently at most of the sites in these basins. Several other compounds used on peanuts, including the herbicides pendimethalin, trifluralin, and ethalfluralin and the insecticides carbaryl, chlorpyrifos, and phorate, were detected infrequently or not at all in streams draining these basins.

Cotton accounts for more than 10 percent of harvested cropland in seven of the basins (table 4). Similar to sites with peanuts, no clear pattern is evident in the detection frequencies shown in table 5 for the sites with cotton. The target compounds cyanazine, pendimethalin, trifluralin, malathion, azinphos-methyl, methyl parathion, and parathion are used extensively on cotton; use of these compounds, however, varies considerably in these seven basins (Gianessi and Anderson, 1996). None of these compounds were detected consistently among the sites where cotton is a major crop. Several pesticides were detected frequently in Salt Slough (sanj-salt) and the Merced River (sanj-merced); a substantial amount of cotton is grown in these basins. Many other crops also are grown in these basins, including a variety of vegetables and several orchard and vineyard crops, making it difficult to determine the source of a particular pesticide detected in streams draining these basins. The detections of molinate and thiobencarb in these streams probably are the result of applications to rice, which is the only crop on which these compounds are used in the United States (Gianessi and Anderson, 1996).

In the two agricultural basins in the Willamette River valley in Oregon--Zollner Creek Basin and Pudding River Basin--the major crops are grass seed and a wide variety of vegetables. In addition, there are plant nurseries in both basins, and several orchard crops are grown in the Pudding River Basin (E&S Environmental Chemistry, Inc., and Tetra Tech, Inc., 1995). A number of pesticides were frequently detected at the will-zollner and will-pudding sampling sites in these basins which reflects the diverse agricultural activities in these basins. The pattern of detections at these two sites is very similar, with many of the same compounds detected at both sites (table 5). This was not unexpected because the Zollner Creek Basin is within the larger Pudding River Basin. Detection frequencies for all agricultural pesticides are higher for Zollner Creek than for Pudding River, except for DCPA, and several compounds were detected in Zollner Creek that were not detected in the Pudding River. This also was expected as the Zollner Creek Basin is much smaller and more intensively farmed than the Pudding River Basin as a whole. Fifteen different pesticides had detection frequencies greater than 10 percent in Zollner Creek (table 5). For several compounds, the detection frequencies for Zollner Creek and the Pudding River were higher than at the other 56 sites. These compounds include napropamide, an herbicide used primarily on various berry crops; ethoprop, an insecticide used primarily on green beans and sweet corn; and carbofuran, an insecticide used primarily on strawberries in this area (Gianessi and Anderson, 1996). Diazinon, which frequently was detected in Zollner Creek, but not in the Pudding River, is applied to a variety of crops in this area, including hops, vegetables, and nursery crops. Atrazine, detected in all samples at both sites, is used mainly on sweet corn in this area (Gianessi and Anderson, 1996).

Orchards and vineyards are the dominant uses of agricultural land in the Merced River and Orestimba Creek basins in California (table 5). A variety of vegetables and other row crops also are grown in both basins. Similar to the sites in the Willamette River valley, a number of pesticides were frequently detected in both basins. Simazine, which is used on several orchard crops and in nonagricultural applications in these basins, was the herbicide detected most frequently at both sites. Azinphos-methyl, chlorpyrifos, and diazinon also were detected relatively frequently at both sites, primarily as a result of application to orchard crops. Seventeen pesticides had detection frequencies greater than 10 percent in Orestimba Creek. Several pesticides, including alachlor, ethalfluralin, fonofos, metolachlor, pebulate, and propargite, were detected more frequently in Orestimba Creek than in the Merced River. These compounds are applied primarily to beans and other vegetables in the Orestimba Creek Basin (Panshin and others, 1998).

Concentrations

The concentrations of specific compounds varied considerably among the agricultural indicator sites, which is consistent with the widely varying agricultural use among the basins (fig. 20). Monthly median concentrations of the compounds detected most frequently at agricultural sites are shown in figure 22 (herbicides) and figure 23 (insecticides). These figures show concentrations for a 1-year period for the 37 agricultural sites, so that there are (at most) 12 values for each site in each of the plots. The sites are arranged by crop-group classifications and are presented in the same order as the sites in tables 4 and 5. Monthly median concentrations less than the detection limit are shown as points on the x-axis. The median of the monthly values for each site, shown in red, represents the midpoint of the monthly values. The plots in figure 22 and 23 provide a relatively unbiased representation of the concentrations measured at each of the agricultural sites because each month is given equal weight. In addition, the use of medians, rather than means, minimizes the influence of extreme values in the distribution of concentrations. Corresponding plots for the urban indicator sites and the integrator sites are discussed in the sections on these types of sites that follow. Figures 22 and are used in the next two sections to illustrate the concentration ranges of specific herbicides and insecticides measured at the agricultural sites.

Herbicides

Monthly median total herbicide concentrations at the 37 agricultural sites are shown in figure 22A. At the 17 sites in corn-growing areas, median concentrations generally ranged from 0.1 to 1 µg/L for most months. Monthly median concentrations exceeded 10 µg/L during seasonal peaks at several sites in the corn crop group. Concentrations were lower at two sites in the corn group--Albemarle Canal Basin in North Carolina (albe-albe) and Lime Creek Basin in Georgia (acfb-lime)--which is consistent with the generally lower detection frequencies for herbicides at these two sites (table 5). The percentage of cropland planted in corn was somewhat lower in Albemarle Canal and Lime Creek basins than in most of the other basins in the corn group (table 4), and the amounts of the target herbicides used in these two basins also was lower than in most of the other basins in corn-growing areas (table 6). At the 11 sites where wheat (but not corn) is a major crop, total herbicide concentrations generally were an order of magnitude lower than at sites in corn-growing areas, with monthly median concentrations of 0.01 to 0.1 µg/L during most months. At the two sites where peanuts are the major crop (gafl-little and acfb-aycocks), concentrations also were low, with monthly median total herbicide concentrations less than 0.1 µg/L for much of the year. Concentrations at the sites in the San Joaquin and Willamette River basins were similar to concentrations at most of the corn sites, with monthly median total herbicide concentrations between 0.1 to 1 µg/L for much of the year. Total herbicide concentrations at the two sites in the Ozark Plateau study units (ozrk-dous and ozrk-yocum) were very low. Harvested cropland accounts for less than 10 percent of the basin area at these two sites (table 3), and use of the target herbicides is very low (table 6). The concentrations of the six herbicides detected most frequently at the agricultural sites are discussed below.

Monthly median concentrations of atrazine, metolachlor, and cyanazine generally ranged from 0.01 to 1 µg/L at most sites in corn-growing areas (fig. 22B,C,D). Atrazine concentrations also were in this range in Zollner Creek (will-zollner) and the Pudding River (will-pudding), probably as a result of application to sweet corn. Concentrations of these three herbicides generally were lower at sites representing other crop groups, which is consistent with the much lower use of these compounds in these basins.

Concentrations of the atrazine transformation product DEA (fig. 22E) followed a spatial pattern quite similar to that of atrazine concentrations, but concentrations were substantially lower at most sites. DEA was detected throughout the year at most sites in corn-growing areas of the Midwest and West, with median concentrations between 0.01 and 0.1 µg/L for most months. Concentrations of DEA were very similar to atrazine concentrations at several of the sites in corn-growing areas of the East (lsus-mill, hdsn-canaj, poto-muddy, albe-pete, and lsus-eastm). The relatively narrow range of monthly median concentrations at these sites also shows that DEA concentrations were fairly constant throughout the year. Similar to atrazine, concentrations of DEA were considerably lower at sites representing crop groups other than corn even though it was present for much of the year at several of these sites. It should be noted that the concentrations of DEA shown in figure 22 probably are biased low because of the low analytical recovery of this compound (table 2). On the basis of mean analytical recoveries of atrazine (98 percent) and DEA (16 percent), it is possible that the actual DEA concentrations may have been higher than atrazine concentrations during some months at several sites.

Among the agricultural sites, concentrations of the herbicide prometon (fig. 22F) were highest at three sites with few apparent similarities--Mill Creek in Pennsylvania (lsus-mill), in a corn-growing area with relatively high population density; Lonetree Creek in Colorado (splt-lone), with major crops of corn, wheat, and grains and very low population density; and Devils Cradle Creek in North Carolina (albe-devils), with major crops of wheat, soybeans, and tobacco and a medium population density (table 1). Prometon concentrations at these sites were between 0.01 and 0.3 µg/L for much of the year (fig. 22F) The relatively narrow range of concentrations at these sites indicates that prometon concentrations were fairly uniform throughout the year. Detections of prometon were rare and concentrations very low at sites in the San Joaquin and Willamette basins and in most wheat-growing areas. Prometon is used almost exclusively in nonagricultural settings, such as transportation rights-of-way, and little quantitative information is available on the amounts used (Capel and others, 1999).

Monthly median concentrations of simazine (fig. 22G) generally ranged from 0.01 to 0.1 µg/L, with highest concentrations at sites in the San Joaquin Basin in California (sanj-salt, sanj-merced, and sanj-orest) and the Willamette Basin in Oregon (will-pudding and will-zollner) and at two sites in corn-growing areas of the East (lsus-mill and poto-mono). This herbicide is used on a variety of agricultural crops, including orchard crops, and has substantial nonagricultural use as well. Several orchard crops are grown in the San Joaquin and Willamette basins. In addition, the two eastern sites with the highest simazine concentrations, Mill Creek in Pennsylvania (lsus-mill) and the Monocacy River in Maryland (poto-mono), and the two sites in the Willamette Basin (will-pudding and will-zollner) have the highest population densities of any of the agricultural sites (table 1). The simazine concentrations measured at these sites are likely the result of a combination of agricultural and nonagricultural use.

Insecticides

Monthly median total insecticide concentrations at the 37 agricultural sites are shown in figure 23A. Concentrations were less than 0.1 µg/L during most months at sites in corn-growing areas. Concentrations were lower at most sites in wheat-growing areas, except for the Crab Creek Lateral (ccpt-crab.rl) and El68 Wasteway (ccpt-el68) sites in Washington. Higher insecticide concentrations at these sites are consistent with the higher detection frequencies at these sites discussed earlier and may be due to the presence of orchard and nursery crops in these two basins. Insecticide concentrations were very low at the two peanuts sites (gafl-little and acfb-aycocks) and at the sites in the Ozark Plateau (ozrk-dous and ozrk-yocum) study unit, with most or all monthly median concentrations less than the detection limit. Insecticide concentrations generally were much higher at the sites in the San Joaquin (sanj-merced, sanj-orest, and sanj-salt) and the Willamette River basins (will-pudding, and will-zollner). Monthly median total insecticide concentrations at these sites ranged from 0.01 to 1
µg/L during most months (fig. 23F). These higher insecticide concentrations are consistent with the higher detection frequencies and the greater number of insecticides measured at these sites where a wide variety of crops are grown.

The insecticides most frequently detected at agricultural sites were chlorpyrifos, diazinon, carbofuran, and carbaryl. These insecticides were detected less frequently at agricultural sites than the most frequently detected herbicides, and concentrations generally were lower. Monthly median concentrations of these four insecticides are shown in figure 23 B-E. Note that the maximum value on the concentration scale in these plots is 1 µg/L compared with 10 µg/L in the plots for individual herbicides. At most sites, monthly median concentrations of these compounds were less than the detection limit for much of the year. During the remainder of the year, monthly median concentrations were less than 0.1 µg/L at most sites. Concentrations of these four insecticides were substantially higher at the five agricultural sites in the San Joaquin (sanj-merced, sanj-orest, and sanj-salt) and Willamette (will-pudding and will-zollner) River basins (fig. 23). Monthly median concentrations of chlorpyrifos and diazinon were between 0.01 and 0.1 µg/L for much of the year at several of these sites. Concentrations of all four of the most frequently detected insecticides were elevated for much of the year at the Zollner Creek site (will-zollner) in the Willamette River Basin (fig. 23). The Zollner Creek site had the highest carbofuran concentrations of any of the 58 sites discussed in this report.

Seasonal Patterns

Seasonal patterns in pesticide occurrence were apparent at most of the agricultural indicator sites. Both the number of pesticides detected (fig. 17) and their concentrations (fig. 18) were highest during a relatively well-defined period at most sites. The period of elevated concentrations extended from April or May through July at most of the agricultural sites. At some of the agricultural sites, elevated concentrations occurred during autumn or winter.

Seasonal concentration patterns for eight agricultural sites are shown in figure 24. The plots in this figure illustrate the general temporal patterns of total herbicide and insecticide concentrations for a 1- to 2-year period. The plots in figure 24 (A-C) show concentrations measured at sites in corn-growing areas of the eastern, midwestern, and western United States. The temporal patterns of total herbicide concentrations in Mill Creek (fig. 24A), Kessinger Ditch (fig. 24B), and Lonetree Creek (fig. 24C) were similar, with elevated concentrations occurring primarily during the months of May and June and very low concentrations occurring during the remainder of the year. This pattern is typical of most sites representing corn-growing areas, where herbicides are applied shortly before or after planting in the spring. The duration and the levels of elevated herbicide concentrations at these three sites varied widely, however, depending on basin size, amount and intensity of rainfall or irrigation, and soil properties. The temporal pattern is not as well-defined for insecticide concentrations at the sites in figure 24 (A-C), with spikes in concentration occurring periodically from spring through autumn.

Figures 24 (D-F) show temporal patterns in pesticide concentrations observed at the three agricultural sites in the Central Valley of California. Seasonal patterns at these sites were different from those at sites in corn-growing areas and varied considerably among the three sites. Concentrations of herbicides and insecticides were elevated in samples collected in January and February from these three sites; the elevated concentrations were due to applications to orchards (primarily almonds and walnuts) during the dormant season. In Salt Slough and Orestimba Creek, concentrations of insecticides remained at detectable levels for much of the year. This is consistent with the variable periods of pesticide application to the wide variety of crops grown in the Salt Slough and Orestimba basins (Panshin and others, 1998). Seasonal patterns in these basins also are complicated by water-management policies for a system of canals, wasteways, and reservoirs and by regulation of irrigation return flows. The large insecticide peak in Orestimba Creek in August of 1993 is due to a high concentration (26 µg/L) of propargite in one of the samples (U.S. Geological Survey, 1999).

Figures 24 (G,H) show the temporal patterns in concentrations for the two agricultural sites in the Willamette River Basin in Oregon. The concentration patterns are somewhat similar for these two sites, with detectable levels of herbicides and insecticides throughout much of the year and elevated concentrations in early summer and autumn. Concentrations in Zollner Creek, however, generally were much higher than concentrations in the Pudding River. Similar to the San Joaquin River valley, a wide variety of crops is grown in Willamette River Basin and use of pesticides is high.

The plots of total herbicide and insecticide concentrations in figure 24 show some generally consistent patterns among these sites. Each site, however, had its own unique temporal concentration pattern, and therefore, it is difficult to generalize, even among sites representing similar crop groups. For example, the peaks in insecticide concentrations shown in figure 24 (A-C) were caused by different insecticides at each site. In Mill Creek in Pennsylvania (lsus-mill), the main contributors to the peaks in total insecticide concentration were diazinon, carbaryl, and chlorpyrifos. These peaks could be the result of nonagricultural pesticide use. In Kessinger Ditch in Indiana (whit-kess), the peaks in total insecticide concentration primarily were due to carbofuran and carbaryl. In Lonetree Creek in Colorado (splt-lone), each peak in total insecticide concentration was due to a different insecticide or group of insecticides. The first peak, in late May of 1993, represents elevated concentrations of carbofuran, terbufos, phorate, and chlorpyrifos. The second peak, in late August of 1993, represents concentrations of carbaryl and diazinon. The third peak, in July of 1994, was almost entirely due to a high concentration of propargite. Thus, despite the generally consistent temporal patterns in pesticide concentrations for many of these sites (fig. 24), the timing and the magnitude of elevated concentrations of specific pesticides can be influenced by unique situations that can occur within individual drainage basins.

Urban Indicator Sites

Detection Frequencies

Several pesticides were detected much more frequently in streams draining urban basins than in streams draining agricultural basins (figs. 14 and 15). The herbicides simazine and prometon and the insecticides diazinon, carbaryl, chlorpyrifos, and malathion were detected much more frequently at most urban sites than at most agricultural sites. The annual mean detection frequencies of these compounds are compared in figure 25 for urban and agricultural sites. Together, these six compounds serve as a signature of urban influences in a basin because their concentrations were elevated at nearly all urban indicator sites for much of the year. The herbicide tebuthiuron also was detected more frequently at most urban sites than at most agricultural indicator or integrator sites (fig. 15). All seven of these pesticides have significant nonagricultural use (Meister, 1996; Larson and others, 1997), but data on the actual amounts of these pesticides used in urban areas are not currently available.

Pesticides that are used primarily for agriculture, such as atrazine and metolachlor, also were detected frequently at many of the urban indicator sites (fig.14B). The frequent detection of these pesticides probably is due to the presence of some cropland in most of the urban basins (fig. 3). Atrazine is registered for use on turf grass in several southeastern states; the actual amounts used are unknown. Atrazine was detected frequently in some urban basins with very little cropland; these basins are in parts of the country where atrazine is not registered for home use. For example, in Cherry Creek in Colorado (splt-cherry) and Fanno Creek in Oregon (will-fanno), each of which have about 2 percent cropland in their drainage basins, the detection frequencies of atrazine in 1993 were 88 and 100 percent, respectively (U.S. Geological Survey, 1999). Atrazine and several other commonly used agricultural herbicides have been detected frequently in precipitation and air samples in several regions of the United States (Capel, 1991; Wotzka and others, 1994; Goolsby and others, 1997). The atmospheric deposition of these compounds may partly explain their frequent detection in streams draining urban basins. The presence of agricultural pesticides in urban streams helps explain why the number of pesticides detected in streams draining urban and agricultural basins is similar in most cases (fig. 12). Most of the urban streams discussed in this report contain the urban signature compounds, as well as low-level concentrations of agricultural pesticides used in the region.

Concentrations

The distribution of total herbicide and insecticide concentrations measured at the urban indicator sites is shown in figure 16. The general characteristics of these distributions were discussed previously. The most obvious difference between concentrations at the urban indicator sites and concentrations at the agricultural indicator and the integrator sites is the higher incidence of elevated levels of insecticides at the urban sites. The distributions of total herbicide and insecticide concentrations at the urban sites are nearly identical if the aggregated data at the urban sites are used. This distribution is in sharp contrast to the distribution for the agricultural and the integrator sites where elevated herbicide concentrations were much more common than elevated insecticide concentrations. Nearly 100 percent of the monthly median concentrations of total herbicides and insecticides were greater than 0.01 µg/L at the urban sites (fig. 16), which indicates that both types of pesticides were at detectable levels in streams draining urban basins for much of the year.

The range of concentrations of individual pesticides detected at the 11 urban indicator sites is shown in figure 26. These plots are analogous to those in figures 22 and 23 and show monthly median concentrations of the pesticides that were detected most frequently at the urban sites. The 3 herbicides and 4 insecticides shown in figure 26 were detected more frequently at urban sites than at agricultural or integrator sites. The plots in figure 26 are used in the next two sections to illustrate the concentration ranges of specific herbicides and insecticides detected at the urban sites. Herbicides

Monthly median concentrations of simazine ranged from 0.01 to 1.0 µg/L at most urban indicator sites (fig. 26A). Concentrations at these sites were similar to, or higher than, concentrations of simazine at most of the agricultural indicator sites and integrator sites. Monthly median concentrations greater than 0.1 µg/L were common at four sites--Accotink Creek in Virginia (poto-acco), Sope Creek in Georgia (acfb-sope), Rush Creek in Texas (trin-rush), and Las Vegas Wash in Nevada (nvbr-lasvegas). Only one site--Accotink Creek (poto-acco)--had a monthly median concentration greater than 1 µg/L. Simazine concentrations in individual samples from this site were greater than 1 µg/L for 4 consecutive months during 1994. There was no reported agricultural use of simazine in Accotink Creek Basin during the sampling period (Gianessi and Anderson, 1996) which suggests that these relatively high concentrations were the result of nonagricultural use. Two sites--Norwalk River in Connecticut (conn-norwalk) and Lisha Kill in New York (hdsn-lisha)--had much lower simazine concentrations than the other urban sites for most of the year. At most of the urban sites, concentrations of simazine varied substantially throughout the year, as shown by the large range in the monthly median concentrations (fig. 26A). Monthly median concentrations were higher than the detection limit for the entire year at 6 of the 11 sites.

Monthly median concentrations of prometon (fig. 26B) ranged from 0.01 to 0.1 µg/L at most urban sites. This relatively narrow range indicates that prometon concentrations were relatively consistent for much of the year at most of the urban sites. Prometon concentrations at the urban sites were higher than concentrations at most of the agricultural and the integrator sites. Prometon is used almost exclusively in nonagricultural settings, although there is little quantitative information available on the amounts used.

Monthly median concentrations of tebuthiuron (fig. 26C) were between 0.01 and 0.1 µg/L at 3 of the urban sites (lsus-cedar, acfb-sope and will-fanno) throughout much of the year, but concentrations were considerably lower at the other 8 sites. Monthly median concentrations were less than the detection limit for the entire year at 4 sites. Tebuthiuron has virtually no reported agricultural use in the United States (Gianessi and Anderson, 1996), and little information is available on the quantities used in nonagricultural settings.

Insecticides

Carbaryl concentrations (fig. 26D) varied considerably among the urban indicator sites. At several sites, monthly median concentrations ranged from 0.01 to about 0.2 µg/L during most months. Carbaryl concentrations at these sites were higher than concentrations at nearly all the agricultural sites (fig. 23). The highest carbaryl concentrations were detected in Cherry Creek in Denver, Colorado (splt-cherry) where monthly median concentrations were greater than 0.1 µg/L for 6 of the 12 months. Concentrations were substantially lower at seven urban sites, with most monthly median concentrations less than 0.01 µg/L.

Chlorpyrifos concentrations (fig. 26E) also varied among the urban sites. At five sites, most monthly median concentrations ranged from about 0.002 to 0.05 µg/L. Chlorpyrifos concentrations at these sites were comparable with concentrations detected at agricultural sites in the San Joaquin and Willamette River basins but higher than concentrations detected at all other agricultural sites (fig. 23). Concentrations of chlorpyrifos were lower at the other six urban sites, with monthly median concentrations less than the detection limit during most months. Monthly median concentrations were less than the detection limit for the entire year at two urban sites, Norwalk River (conn-norwalk) and Lisha Kill (hdsn-lisha).

Diazinon concentrations (fig. 26F) ranged from 0.01 to 0.1 µg/L for much of the year at nine of the urban sites. The highest concentrations were detected in Rush Creek in Arlington, Texas (trin-rush), where monthly median concentrations were greater than 0.1 µg/L for most of the year and greater than 1 µg/L during several months. Diazinon concentrations at most urban sites were comparable with concentrations at agricultural sites in the San Joaquin and Willamette River basins and higher than concentrations at all other agricultural sites (fig. 23). Concentrations were much lower at the other two urban sites--Norwalk River in Connecticut (conn-norwalk) and Cedar Run in Pennsylvania (lsus-cedar) --where concentrations were less than the detection limit during most months.

Although malathion was detected more frequently at the urban sites than at the agricultural or the integrator sites, concentrations were quite low for much of the year (fig. 26G). For months in which malathion was detected, monthly median concentrations generally were less than 0.1 µg/L. Only at Rush Creek (trin-rush) were more than half of the monthly median concentrations of malathion greater than the detection limit.

Seasonal Patterns

Seasonal patterns in the occurrence of pesticides were less obvious for the urban indicator sites than for the agricultural indicator sites and the integrator sites. In terms of both the number of compounds detected in each sample (fig. 17) and the monthly median concentrations (fig. 18), there was little difference between the critical period and the rest of the year at most urban sites for both herbicides and insecticides.

The temporal patterns for pesticide concentrations at urban sites can be examined more closely by looking at results from individual sites (fig. 27). Although concentrations varied throughout the year at several urban sites, no clear seasonal pattern is evident among the eight sites shown in figure 27. At several sites, including Norwalk River (conn-norwalk), Lisha Kill (hdsn-lisha), Lafayette Creek (gafl-lafayette), and Fanno Creek (will-fanno), total herbicide concentrations were relatively low in all samples. At other sites, including Sope Creek (acfb-sope), Las Vegas Wash (nvbr-lasvegas), and Accotink Creek (poto-acco), herbicide concentrations were elevated for long periods at various times of the year. Elevated total herbicide concentrations at most of the urban sites were almost entirely due to concentrations of simazine, prometon, and atrazine. Insecticide concentration patterns also were variable at the urban sites. Distinct spikes in insecticide concentrations were observed at several sites, including Norwalk River (conn-norwalk), Sope Creek (acfb-sope), and Las Vegas Wash (nvbr-lasvegas). Insecticide concentrations were elevated for much of the year at other sites, including Cherry Creek (splt-cherry), Lafayette Creek (gafl-lafayette), and Accotink Creek (poto-acco). At most sites, elevated total insecticide concentrations were due to concentrations of diazinon, carbaryl, and malathion.

Urban Integrator Sites

Detection Frequencies

At the integrator sites, the overall detection frequencies of several agricultural herbicides were similar to or somewhat higher than the detection frequencies at the agricultural indicator sites (figs. 14 and 15). This was true for all the most commonly detected agricultural herbicides, including atrazine, metolachlor, alachlor, and cyanazine and the atrazine transformation product DEA. These somewhat higher detection frequencies are not surprising because a wider variety of crops generally are grown in the integrator basins and because all the integrator basins likely received applications of these widely used herbicides. The agricultural indicator sites, on the other hand, represent specific crops or groups of crops, some of which received little or no application of these compounds. At the integrator sites, detection frequencies of pesticides with substantial nonagricultural use, including several insecticides and the herbicides prometon and simazine, generally were between the detection frequencies at the agricultural and the urban sites. Between 3 and 7 percent of the land within most integrator basins is urban land (fig. 3), and therefore, detection of compounds with substantial urban use is likely. In most of the urban basins, however, more than 60 percent of the land is urban land, resulting in a much stronger urban signal and higher detection frequencies for these pesticides.

Concentrations

Aggregated data from each group of sites indicate that total herbicide and insecticide concentrations measured at the integrator sites were very similar to concentrations measured at the agricultural indicator sites (fig. 16). The general characteristics of the distribution of concentrations for the integrator sites were discussed previously. Monthly median total herbicide concentrations between 0.01 and 0.1 µg/L were slightly more common at the integrator sites than at the agricultural indicator sites, but distributions at these two types of sites were nearly identical for concentrations greater than 0.1 µg/L (fig. 16A). The more common occurrence of herbicides at moderate concentrations at the integrator sites is consistent with the general pattern of a longer period of elevated concentrations in streams with larger drainage basins (Richard and Baker, 1993; Larson and others, 1995). The distributions of total insecticide concentrations were very similar at the integrator sites and the agricultural indicator sites (fig. 16B).

The range of concentrations of individual pesticides measured at the 10 integrator sites are shown in figure 28. These plots are analogous to those in figures 22, 23, and 26 and show monthly median concentrations of pesticides detected frequently at the integrator sites. The plots in figure 28 are used in the next two sections to illustrate the range in concentrations of specific herbicides and insecticides measured at the integrator sites. Herbicides

Monthly median concentrations of atrazine were less variable among the integrator sites (fig. 28A) than among the agricultural indicator sites (fig. 22B), with concentrations between 0.01 and 0.1 µg/L for most of the year at most sites. The two integrator sites with the highest atrazine concentrations--the White River in Indiana (whit-white) and the Platte River in Nebraska (cnbr-platte)--had relatively high atrazine use compared with the other integrator sites (table 6). A strong seasonal peak in atrazine concentrations at these two sites is evident in figure 28A, with monthly median concentrations greater than 1 µg/L for several months during the growing season. Atrazine concentrations were lowest in the San Joaquin River in California (sanj-sanj) where atrazine use in the basin is relatively low. Monthly median concentrations were greater than 0.01 µg/L for the entire year at seven of the integrator sites (fig. 28A).

Concentrations of DEA (fig. 28B) generally followed the same spatial pattern at the integrator sites as concentrations of atrazine, but the monthly median concentrations were 5 to 10 times lower at most of the sites. Detectable levels of DEA were present for most of the year at seven of the integrator sites. The concentrations of DEA shown in figure 28 are probably biased low because of the low analytical recovery of DEA (table 2).

At most of the integrator sites, monthly median concentrations of metolachlor were similar to the monthly median concentrations of atrazine, ranging from 0.01 to 0.1 µg/L for much of the year (fig. 28C). Concentrations of metolachlor were more consistent among the integrator sites than were concentrations of atrazine, which is consistent with the more uniform use of metolachlor in these basins (table 6). The highest metolachlor concentrations were in the White River (whit-white). Use of metolachlor was the highest (per unit area) in the White River Basin than in any other of the integrator basins.

Concentrations of cyanazine were variable among the integrator sites (fig. 28D). Monthly median concentrations were less than the detection limit for most of the year at the four eastern sites where use of cyanazine was relatively low. Concentrations ranged from 0.01 to 0.1 µg/L for much of the year at the other six sites. The White River site (whit-white) had the highest cyanazine concentrations (fig. 28 D) and the highest reported use of cyanazine (table 6) compared with the other integrator sites. Concentrations of cyanazine at the other integrator sites did not correlate well with agricultural-use data, as concentrations were similar at sites with widely differing values for cyanazine use (table 6).

Concentrations of simazine (fig. 28E) also were variable among the integrator sites, with monthly median concentrations between 0.01 and 0.1 µg/L for much of the year at several sites but much lower at other sites. Simazine concentrations generally were highest in the San Joaquin River (sanj-sanj), which is the only integrator basin with substantial agricultural use of simazine (table 6). Simazine concentrations were low, but detectable, for much of the year at sites in areas with low population density--the Platte (cnbr-platte) and Red River basins (redn-rr.fargo and redn-rr.em) (table 1). Concentrations also were low in the Mohawk River in New York (hdsn-moh), however, which drains an area of relatively high population density.

Prometon concentrations (fig. 28F) were between 0.01 and 0.1 µg/L for much of the year at five of the integrator sites but much lower at the other sites. Similar to simazine, concentrations were low at sites in the Platte and Red rivers (cnbr-platte, redn-rr.fargo, and redn-rr.em) where population density is low, but concentrations also were low in the Mohawk River (hdsn-moh) where population density is relatively high. Monthly median concentrations of prometon were less than the detection limit for the entire year in the San Joaquin River (sanj-sanj).

Insecticides

With the exception of diazinon, concentrations of insecticides generally were low at the integrator sites (figs. 28 G-K). Median concentrations of chlorpyrifos, malathion, carbofuran, and carbaryl were less than 0.01 µg/L during most months at all 10 integrator sites. These concentrations are similar to those measured at most of the agricultural sites for these insecticides, with monthly median concentrations less than the detection limit for much of the year. Concentrations of chlorpyrifos, diazinon, and carbaryl were higher in the San Joaquin River (sanj-sanj). Monthly median concentrations of diazinon were greater than the detection limit for much of the year at three sites--the San Joaquin River (sanj-sanj) where diazinon has extensive agricultural use, the Tar River in North Carolina (albe-tar) where diazinon has moderate agricultural use and population density is relatively high, and the White River in Indiana (whit-white) where there is little reported agricultural use, but population density is high. Previous studies of diazinon occurrence in large midwestern rivers have reported a positive correlation between population density in a drainage basin and the amount of diazinon transported in the river (Larson and others, 1995).

Accessibility FOIA Privacy Policies and Notices

Take Pride in America home page. FirstGov button U.S. Department of the Interior | U.S. Geological Survey
URL: http://water.usgs.gov/nawqa/pnsp/pubs/wrir984222/effects.html
Page Contact Information: gs-w_nawqa_whq@usgs.gov
Page Last Modified: Tuesday, 04-Mar-2014 14:44:34 EST