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

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

NATIONAL OVERVIEW OF PESTICIDE OCCURRENCE

A broad overview of national findings provides the context for a more detailed analysis of pesticides in streams in relation to land use, pesticide use, and environmental significance. The national overview addresses which pesticides were detected in streams, how often they were detected, their concentrations, and the seasonal patterns in pesticide occurrence.

Detection Frequencies

The annual mean detection frequencies for all the target compounds at 50 of the 58 sampling sites are shown in figure 6. The 50 sites include 33 agricultural indicator sites, 10 urban indicator sites, and 7 integrator sites. For the other eight sites, sampling was not sufficient during some periods of the year for calculation of a comparable annual mean detection frequency. Each bar in figure 6 represents an estimate of the mean of monthly detection frequencies for a 1-year period for each of the 50 sites. This is a relatively unbiased estimate of detection frequency for a year because each period of the year is equally represented regardless of the number of samples analyzed during that period. The range of detection frequencies for each pesticide among the 50 sites is shown in figure 7.

All the target compounds were detected in at least one sample from at least 1 of the 58 streams. Herbicides generally were detected more frequently than insecticides (fig. 6). Among the most commonly detected herbicides were atrazine, metolachlor, alachlor, and cyanazine. These compounds are used extensively in agriculture primarily in the Midwest. The herbicides simazine and prometon also were detected frequently in most basins. Simazine is used in both agricultural and nonagricultural settings throughout the United States. Prometon is used almost entirely in nonagricultural settings (Capel and others, 1999). The insecticides detected most frequently include diazinon, carbaryl, chlorpyrifos, carbofuran, and malathion. With the exception of carbofuran, which is used primarily in agriculture, these compounds are used extensively in both agricultural and nonagricultural settings.

Several pesticides that are used extensively for agriculture on a nationwide basis (see table 2) were detected infrequently at nearly all sites (fig. 7). These pesticides include the herbicides pendimethalin, linuron, propachlor, and propanil and the insecticides disulfoton, terbufos, and methyl parathion. Of these compounds, linuron, propachlor, and propanil had low use in all basins in this study. In addition, the physical and chemical properties of most of these compounds and the methods used to apply these compounds for agricultural applications result in a low potential for removal from agricultural fields in runoff (Goss and Wauchope, 1990).

A wide range in detection frequency among sites is evident for many of the pesticides (fig. 7). Detection frequencies for several of the most commonly detected compounds, including atrazine, simazine, metolachlor, prometon, DEA, cyanazine, and diazinon, ranged from zero percent at some sites to 100 percent at others. Some compounds, such as napropamide, terbacil, butylate, and ethoprop, were detected frequently at a few sites but rarely or not at all at other sites. Much of this variability can be attributed to differences in the amounts of these pesticides used in the different basins.

Concentration Ranges

Distributions of total herbicide and insecticide concentrations at 50 sites are shown in figure 8. Monthly median concentrations are used in this plot to minimize the effects of the uneven sampling frequency described earlier. Each line in figure 8 represents values for 600 monthly median concentrations--one value for each month for each of the 50 sites. The plots in figure 8 show the overall distribution of concentrations measured at all three types of sites, although the distributions are most strongly influenced by concentrations at the agricultural indicator sites (33 of the 50 sites).

Herbicides generally were detected at higher concentrations than insecticides. Approximately 90 percent of monthly median herbicide concentrations were greater than 0.01 µg/L compared with about 40 percent for insecticide concentrations. Monthly median concentrations were greater than 1 µg/L about 10 percent of the time for herbicides compared with about 2 percent of the time for insecticides. To help put the concentration levels shown in figure 8 into perspective, aquatic-life criteria values range from 1 to 10 µg/L for most herbicides and from 0.01 to 0.1 µg/L for most insecticides (Canadian Council of Resource and Environment Ministers, 1991; Nowell and Resek, 1994). Because equal weight was given to the samples collected during each month, figure 8 indicates that detectable levels of herbicides were present in samples for most of the year at nearly all sites.

In general, the pesticides detected most frequently (fig. 6) also had the highest concentrations. Monthly median concentrations of the most frequently detected herbicides and insecticides at the 50 sites are shown in figure 9. Again, each of the lines in these plots represents 600 values--12 monthly median concentrations from each site. For the seven most frequently detected herbicides, less than 25 percent of monthly median concentrations were greater than 0.1 µg/L and less than 6 percent were greater than 1 µg/L (fig. 9A). Atrazine was detected more frequently than any of the other herbicides over the entire concentration range. For the five most frequently detected insecticides, less than 5 percent of monthly median concentrations were greater than 0.1 µg/L and less than 0.1 percent were greater than 1 µg/L (fig. 9B). For most of the compounds shown in figure 9, the distribution of concentrations is highly skewed, with low concentrations occurring most of the year and elevated concentrations occurring only as seasonal pulses.

Seasonal Patterns

Seasonal patterns in pesticide occurrence were evident at nearly all sites. More compounds were detected during May, June, and July than during the rest of the year at most sites (fig. 10). Two to 8 of the target compounds were detected in most of the samples collected during August through April. During May, June, and July, however, most of the samples contained 5 to 12 target compounds, and approximately 25 percent of the samples collected during these 3 months contained more than 11 of the target compounds. Most of the increase in the number of compounds detected in each sample during these 3 months was due to an increase in the number of herbicides detected. In many of the agricultural indicator basins and integrator basins, herbicides are applied during May and early June, and some fraction of the amount applied is transported to surface water in runoff resulting from spring rains or irrigation. This phenomenon is called the "spring flush"; it has been described in several studies, particularly in studies of the Midwest (Thurman and others, 1991; Goolsby and Battaglin, 1993; Schottler and others, 1994).

A similar seasonal pattern also was evident in pesticide concentrations at most sites. Figure 11 shows the distribution of total pesticide concentrations in samples collected at the 58 sites each month of 1993, 1994, and 1995. Each of the boxplots in figure 11 represents 74 to 274 samples, depending on the month. A higher proportion of samples collected during May, June, and July, and to a lesser extent during August, had elevated total pesticide concentrations compared with the samples collected during the rest of the year. During these 4 months, total pesticide concentrations were near or greater than 1 µg/L in approximately 25 percent of samples. During May and June, total pesticide concentrations were greater than 5 µg/L in more than 10 percent of samples. During the remainder of the year, total pesticide concentrations were less than 1 µg/L in more than 90 percent of samples.

The use of aggregated data from the 58 sites (fig. 11) conceals some differences in seasonal patterns among the sites. At some sites, the highest pesticide concentrations were measured during autumn or winter. At many of the urban indicator sites, the seasonal pattern was less obvious, with elevated pesticide concentrations occurring for a longer period. These different seasonal patterns are obscured in figure 11 because of the large number of sites that exhibit the spring flush phenomenon. Seasonal patterns observed at the different types of sites will be discussed in more detail in the next section.