Pesticide Occurrence 


A total of 135 environmental samples were collected among the 28 sites, including 18 replicates. All data are available from the USGS National Water Information System. Each sample was analyzed for as many as 175 compounds, and a total of 43 compounds were detected at least once: 2 isomeric forms of 1 fungicide, 26 herbicides and 3 of their metabolites, 9 insecticides and 2 degradates, and caffeine as an indicator of human waste (table 6). The median number of detections per sample was 4, although some samples had no detections; the maximum number of compounds detected in a single sample was 11 (in the 69th stormwater channel at Thurston Road in Springfield, site 11, during May 2004). To support appropriate comparison of detections for compounds that are associated with a wide range of detection levels (appendix A), detection frequency was determined in two ways—first, using the number of quantified values (both E-coded and unqualified) simply as a function of the number of analyses for the specific compound (frequency); and second, using the number of quantified values as a function of number of analyses for the specific compound based on a common detection level (common frequency) (Gilliom and others, 2006). The common detection level was defined as 0.1 µg/L, which represents the highest level at which most compounds (95 percent) were screened throughout the period of this study. Compounds are listed in table 6 in descending order of frequency within each compound class.


Most concentrations were low—the median value of maximum detected concentrations was 0.055 µg/L. The highest concentrations generally were detected for herbicide compounds, which comprised six of the seven compounds with maximum concentrations greater than 1 µg/L. The maximum detected concentration was 11.4 µg/L for caffeine, which was detected in a sample from downstream of a stormwater drain that was used for disposing gray water by a coffee kiosk in Springfield. This practice subsequently ceased, so this concentration should not be considered representative of current (2012) stormwater drain conditions. 


Twenty-one of these compounds may be considered to represent a potential threat to drinking- water quality as they are regulated by a drinking-water standard (n = 9) and/ or suspected to be endrocrine-disrupting compounds (n = 20). Caffeine was the most frequently detected compound, detected in nearly one-third of all samples. The seven compounds with the highest detection frequency accounted for 191 detections, or about 46 percent of all detections. These included caffeine plus four herbicides (hexazinone, 2,4- D, atrazine, and glyphosate), and one herbicide metabolite (aminomethylphosphonic acid or AMPA, the primary metabolite for glyphosate), and one insecticide (carbaryl). When considering detections based the common reporting level, the analysis is focused on fewer (19) compounds because those with only low-level detections are omitted. All detections for hexazinone and atrazine are excluded when screened at this level, although caffeine, 2,4-D, and glyphosate remain among the most frequently detected compounds. With the addition of diuron, these four compounds combined represent more than 60 percent of all detections when measured by the common frequency of detection. 


Physical-Chemical Characteristics and Reported Compound Use


Physical-chemical characteristics and reported or estimated use by urban, forestry, or agricultural land use were summarized for all pesticides detected and/or reported to be used in the basin (table 7). Of the 35 pesticide compounds that were detected (excluding caffeine and metabolites that are not applied, and considering the two isomers of propiconazole and 2,4-D plus the sum of 2,4-D and 2,4-D ME each as single compounds), 12 were not reported to be used by any of the sources consulted in this analysis. Although the largest proportion of the remaining 23 detected compounds were reported as associated with urban or residential land use, no single land use dominated: 15 were reported as used in urban and residential settings (3 insecticides exclusively urban); 17 were estimated as used in agricultural settings (3 herbicides and 1 insecticide exclusively agricultural); and 9 were reported as used in forestry settings (1 herbicide exclusively forestry). Detected compounds generally were fairly hydrophilic (log soil Koc < 3) with a moderate tendency to sorb to sediment particles (median log soil Koc=2.36) and were only moderately (T_1/2 < 100 days) persistent in soil (median T_1/2 = 47 days). In contrast, of the 25 compounds that were reported to be used in the basin but never detected, those associated with agricultural land use represented the largest proportion. Of the compounds never detected, a total of 22 compounds were estimated as used in agricultural settings (8 herbicides and 9 insecticides exclusively agricultural), although only 7 compounds were reported to be used in urban settings (1 fungicide, 1 herbicide, and 1 insecticide exclusively urban) and 2 were reported to be used in forestry (none exclusively forestry). These compounds were slightly more hydrophobic (median log soil Koc = 2.7), and less persistent in soil (median T_1/2= 30 days). 


These results suggest several factors to consider in assessing pesticide occurrence relative to reported or estimated use in the basin. First, it must be emphasized that pesticide use is difficult to evaluate based on use reports because they are not definitive. Compounds may be used by farmers without their use being reported, especially in emergency situations (Oregon Department of Agriculture, 2001). Alternatively, pesticides may be reported as used by forestry managers simply as placeholders for potential use, and may not subsequently be applied as reported. Urban surveys also may not accurately represent the real use of pesticides by the respondents for various reasons, including simple refusal of response to the survey or incomplete information about pesticides that are used. Some compounds (for example, diazinon, carbaryl, malathion, and chlorpyrifos) have been banned for residential use in recent years but will still be in use if homeowners retain supply in their basement or garage. Notwithstanding the uncertainty in the reported and estimated use data, however, a critical factor in the non-detection of so many presumed agricultural compounds in this dataset is certainly the small number of samples collected in agricultural streams, especially during significant storm-runoff periods. Because the agricultural streams have not been sampled with comparable effort to the other land-use categories, the present data are insufficient to adequately assess the presence of agricultural pesticides in the McKenzie River basin. 


Physical-chemical characteristics among detected compounds that are regulated by drinking- water standards or suspected of endocrine disruption (table 6) were compared to the remaining detected compounds to assess if potential differences in transport mechanisms might be inferred. No large differences were observed for regulated compounds compared to non-regulated compounds for log soil Koc (median = 2.1 and 2.5, respectively), or T_1/2 (median = 47 and 53 days, respectively). Larger differences were observed for suspected endocrine disruptors for soil sorption potential, however, which suggests a different mode of transport behavior. Suspected endocrine disruptors were relatively more hydrophobic (median log soil Koc = 2.8) compared to non-endocrine disruptors (median log soil Koc = 1.9), although they tended to be only slightly more persistent in soil (median T_1/2= 60 days) than others (median T_1/2= 46 days). These results suggest that detected compounds suspected of endocrine disruption may be more prone to be transported in association with sediment particles than the other compounds that have been observed in the basin. Because these compounds were detected in filtered samples, the implication for drinking water source protection is that contaminated sediments may be serving as a source for these compounds in the basin.


The physical-chemical characteristics of the eight pesticide compounds (excluding caffeine) that were detected at the treatment-plant intake (table 7) span a large range, suggesting a coincident range of potential source and transport conditions. The three herbicides (atrazine, hexazinone, and sulfometuron-methyl) are relatively hydrophilic (log soil Koc = 2.00, 1.73, and 1.89, respectively), while the three insecticides (carbaryl, cypermethrin, and diazinon) are more hydrophobic (log soil Koc = 2.36, 5.00, and 3.00, respectively) as are the two metabolites (AMPA and CIAT, log soil Koc = 3.96 and 3.78, respectively). A similarly large range of persistence in soil occurs, with sulfometuron-methyl being relatively non-persistent (T_1/2 = 20 days), cypermethrin, diazinon, and hexazinone being moderately persistent (T_1/2 = 30, 40, and 90 days, respectively), and atrazine being relatively persistent (T_1/2 = 146 days). 


Most compounds were reported to be used in multiple land-use applications, with only cypermethrin being associated with exclusively urban use. No single reported land use dominated among these compounds, suggesting that all categories of land use have an impact on the quality of water at the intake. Four compounds were either suspected of endocrine disruption (atrazine, carbaryl, cypermethrin, and diazinon), or additionally regulated by drinking-water criteria (atrazine). 


To provide some context for the results from the treatment-plant intake, they were compared to pesticide data from the water-treatment plant on the lower Clackamas River in Oregon during 2002–05 (Carpenter and others, 2008). A higher frequency of pesticide detections was observed in the Clackamas River, with a total of 14 compounds detected in 9 samples collected over 2 years. Four of these compounds also were detected in the McKenzie River (atrazine and its metabolite CIAT, hexazinone, and diazinon), although most were not widely detected in this study. Drinking-water intake concentrations generally were comparable, although the maximum concentrations differed by an order of magnitude (0.22 µg/L in the Clackamas River compared to 0.02 µg/L in the McKenzie River). 


Comparison to Water-Quality and Health‑Based Standards


Of the nine compounds that were detected at the EWEB treatment plant intake (table 8), most were frequently detected at other sites. The exception was cypermethrin, a pyrethroid insecticide that is strongly associated with urban land use, and that was never detected at any other site during this study. AMPA, carbaryl, and CIAT were detected twice at the intake, while diazinon, caffeine, hexazinone, atrazine, cypermethrin, and sulfometuron-methyl were each detected once, making a total of 12 detections overall over the 8-year course of this study. Of these 12 detections, 8 were E-coded values, indicating that the concentrations were less than the LRL. No concentrations exceeded the common reporting level of 0.1 µg/L. MCL was defined for one compound, and human-health benchmarks were available for six of the nine compounds, excluding caffeine and the degradation products. One of those with no benchmark available, AMPA, is a metabolite of glyphosate and has a similar toxicological profile; both are associated with low levels of toxicity (World Health Organization, 2004). Where defined, MCL and Benchmark Quotients were consistently one to several orders of magnitude less than 0.1, which is the suggested level for identification of compounds that warrant further monitoring to protect human health. Although these results indicate that adverse effects from pesticide concentrations at the drinking-water intake are expected to be negligible at the present time, they also document the occasional low-level presence of compounds regulated by drinking-water criteria or suspected of endocrine-disrupting activity. Additionally, several samples, notably in spring 2009 and 2010, contained detectable concentrations for multiple compounds. Even though the total concentrations of pesticide in these samples was uniformly very low (< 0.1 µg/L), the potential for additive or synergistic effects of pesticide mixtures on human health is not well understood, even at low concentrations.


Seasonality, Hydrology, and Land Use


The following sections evaluate the seasonal patterns of precipitation that were associated with the various land‑use sampling surveys in the McKenzie River basin, and the corresponding patterns of pesticide occurrence across the range of sampled seasons and land use. 


Seasonal Hydrologic Conditions


Almost all sample surveys were conducted during storms, either spring (n=4) or fall (n=7), with two additional storm surveys conducted during winter (table 2). Three non‑storm surveys also were conducted, two in spring and one in summer. The strategy for fall storm samples was simply to characterize the first major runoff-producing storm of the fall, while that for spring samples was to characterize the first major storm after spring pesticide application, generally considered to occur approximately in April. Spring sampling was not conducted during some years because of the frequently uneven nature of storms and runoff across the study area during that season, especially in non-urban environments where warmer temperatures and the presence of vegetation creates high evapotranspiration, which means more rain is needed to produce runoff. 


Most pesticide detections occurred during spring or fall storm surveys, reflecting the large number of these seasonal storm samples, while relatively smaller numbers of compounds were detected during winter storms or during non-storm surveys (table 9; fig. 5). Concentrations were roughly similar for spring and fall storm samples (median concentration = 0.03 µg/L for spring samples and 0.023 µg/L for fall samples). Nonetheless, despite the small number of spring storm surveys, spring samples showed the largest number of detections determined by the laboratory (202 for spring samples compared to 169 for fall storm samples), as well as detections determined by a common reporting level (53 for spring samples and 38 for fall samples). Fewer compounds were detected in winter storm samples (n = 23, or 4 based on a common reporting level), which were only collected at urban and mixed sites, reflecting that fewer pesticides are applied in the fall as well as suggesting that pesticide supply available for runoff from soil in urban areas may be depleted by the onset of the rainy season. 


Winter concentrations were low as well (median = 0.013 µg/L), presumably indicating the additional influence of dilution with greater streamflow. A small number of detections also were associated with non-storm surveys (n = 17, or zero when based on a common reporting level). These results are consistent with the general pattern that has been frequently observed (Anderson and others, 1996, 1997; Larson and others, 1997; Rinella and Janet, 1997; Wood, 2001; Gilliom and others, 2006), which is that the most significant pesticide transport occurs in non-irrigated areas during the spring and fall. Nonetheless, they also document that although pesticide transport was most pronounced in the spring, pesticides were detected in surface waters in the basin under a range of storm and non-storm conditions during the entire rainy season, especially at low concentrations. 


Focusing on the small number of compounds detected using a common reporting level of 0.1 µg/L, the three compounds most frequently detected during the spring storms were 2,4-D (n = 11), caffeine (n = 10), and diuron (n = 7). Combined, these three compounds represented 53 percent of all detections during the spring storm surveys; all these detections were associated with urban sites, except one detection of diuron. Two of these compounds (caffeine, n = 12, and 2,4-D, n = 7) also were among the top three compounds detected during the fall storm surveys, which including glyphosate (n = 7) represented 68 percent of all detections. Similarly, most of these detections were associated with urban sites (caffeine, n = 11; 2,4-D, n = 6; and glyphosate, n = 3), and the remainder were associated with mixed sites (caffeine, n = 1; 2,4-D, n = 1; and glyphosate, n = 4).


Precipitation conditions during storm surveys were evaluated by summing instantaneous precipitation data over the course of a day to provide daily totals for data from two precipitation gages (fig. 1). The gage at Springfield City Hall (site 2) was presumed to roughly represent precipitation conditions for sites located in the downstream region of the basin near the river mouth, including the treatment plant intake, the urban stormwater sites, the agriculture sites, and the sites at the mouths of Camp Creek and Cedar Creek (sites 5–18). The gage at Trout Creek (site 4) was assumed to better represent precipitation conditions for sites farther upstream—the forestry sites and mainstem sites at Leaburg and Finn Rock (sites 19–32). 


The four spring storm surveys were conducted in May or June, presumably following the period of spring pesticide application, and occurred in the midst of extended periods of precipitation approaching or exceeding 0.5 in. during the previous several months (fig. 6). The two winter storm surveys were conducted under similar conditions, following a series of large precipitation events exceeding 0.5 in. each. In contrast, the seven fall storm surveys were conducted in September or October during storms characterized by only one or two periods of antecedent precipitation approaching 0.5 in. during the previous week, with dry conditions prior to that for several months. 


Effect of Precipitation and Discharge


Precipitation data were summed over the 7 days prior to sampling to examine the potential relation between total precipitation volume and pesticide occurrence (evaluated as the number of detections and the sum of all concentrations in a single sample or total concentration). Data from the two precipitation gages (at Springfield City Hall and Trout Creek) were associated with individual sampling sites as described above so that pesticide occcurrence could be evaluated in the context of estimated local precipitation. Simple correlations were determined between precipitation volume prior to sampling and pesticide occurrence for sites grouped by dominant land use, and separately for the drinking water intake (table 10). No significant correlation was observed between precipitation volume and total concentration for any site group. The strongest association was observed for the number of detections at forestry sites (r = 0.68, p < 0.0001, n = 38), indicating that larger precipitation events are associated with increasingly more detections at these sites. Likewise, detections at the drinking water intake also were directly correlated with increased precipitation (r = 0.55, p = 0.02, n = 17); agricultural sites showed no significant correlations (p < 0.05), possibly due to the small sample size (n = 5). Urban sites showed a contrasting pattern, with a tendency toward fewer number of detections in response to increased precipitation (r = –0.33, p = 0.03, n = 41). These results suggest that the source of pesticides in urban setting may be depleted by greater volumes of precipitation, implying that urban pesticides may be rapidly discharged to the storm channels with the first flush of precipitation following application. 


Further analysis of conditions in Cedar Creek (site 9) focused on examining whether increased stormwater input as proportion of stream discharge was associated with a detectable impact on pesticide concentrations in the creek. The proportion of stormwater drain discharge was approximated by the ratio of a simple sum of mean discharge from the drains over the duration of each sample collection relative to the mean daily streamflow at Cedar Creek on the sample day. Despite the small sample size (n = 5), correlation analysis showed a positive, although statistically weak association between this proportion and total pesticide concentration over all seasons (r = 0.79, p = 0.11). These results are consistent with the hypothesis that pesticide transport in Cedar Creek is at least partially dependent on discharge from stormwater drains.


Effect of Sample Timing


Samples to contrast pre-storm and storm conditions were collected in stormwater drains during two fall storm surveys (2002 and 2004) and one spring survey (2004) (table 4). All sample pairs were associated with antecedent precipitation conditions at Springfield City Hall of approximately 0.5 in. within the previous 7 days (fig. 6). Presumably as a consequence of previous urban runoff, or possibly irrigation runoff from lawns and gardens, none of the pre-storm samples were devoid of pesticide compound, showing relatively high levels of detections although most were at low concentration (table 11). In fact, more than twice as many detections were observed in the pre-storm sample than in the storm sample during the fall 2004, although when screened at the common detection level of 0.1 µg/L, all pre-storm detections were excluded for the fall samples. Storm samples consistently contained relatively higher total pesticide concentrations, although some individual compounds showed either no change in concentration (that is, prometon in fall 2002), or were relatively diluted in the storm sample (for example, 2,4-D and caffeine in spring 2004). These results indicate that urban rainy season conditions may be associated with ongoing presence of pesticides in stormwater drains, although generally at lower concentrations than occur under storm conditions.


The influence of sampling over different regions of the hydrograph was evaluated by three sets of paired samples collected at stormwater sites during the fall 2002 and spring 2005. The first sample set, comparing data from the rise and fall of the storm hydrograph, shows both a higher number of detections and nearly twice the total concentration for the sample collected over the rise compared to the fall (1.2 versus 0.70 µg/L); this pattern was the same when concentrations were screened at the common reporting level (table 12). Similarly, the second two sets of samples collected early and later in the same storm while the hydrograph continued to rise, show a higher number of detections and larger total concentration for the samples collected nearer to the hydrograph peak (2.3 versus 5.0 µg/L and 0.49 versus 0.75 µg/L, respectively). Caffeine and 2,4-D were most consistently measured in both sets of these samples based on the common reporting level (table 12). These results demonstrate the variability inherent in sampling changing stormwater conditions, which is largely occurring at low concentration (< 0.1 µg/L). They reinforce the replicate analysis previously described, as well as the importance of documenting hydrograph conditions during sample collection. They also suggest that samples composited over the entire rise and fall of the storm hydrograph would tend to dampen this variability and thereby most comprehensively describe average conditions over the entire storm.


Effect of Land Use


The distribution of detected pesticide compounds across the range of land-use settings shows that the largest number of compounds was associated with urban land use (fig. 7). A total of 37 compounds were detected at urban sites at least once during this study, and 18 of these were uniquely detected at urban sites. In contrast, 14 compounds were detected at forestry sites and 8 compounds at agricultural sites, all of which were widely observed in a range of land-use settings, frequently associated with mixed land-use sites. Focusing on the smaller group of compounds detected using a common reporting level of 0.1 µg/L, the pattern is roughly the same: 16 compounds were detected at urban sites (9 unique), 3 compounds were detected at forestry sites (2 unique) and 3 compounds were detected at agricultural sites (none unique). Of these commonly screened compounds, those unique to urban sites included 2,4-DB (n = 1), carbaryl (n = 4), diazinon (n = 1), dicamba (n = 1), metsulfuron-methyl (n = 2), picloram (n = 3), simazine (n = 1), sulfometuron-methyl (n = 4), and tebuthiuron (n = 4); compounds unique to forestry sites included imazapyr (n = 1) and nicosulfuron (n = 1).


Urban sites also were associated with the largest pesticide concentrations (fig. 7). Of the 17 detected concentrations that exceeded 1 µg/L, all but 1 were detections from urban sites, specifically from stormwater drains. These urban compounds included 2,4-D (n = 1), caffeine (n = 5), carbaryl (n = 2), diuron (n = 2), picloram (n = 2), sulfometuron-methyl (n = 2), and tebuthiuron (n=2). The remaining concentration exceeding 1 µg/L was for triclopyr, and was detected at Ward Creek, a forestry site (site 26). This extreme value does not reflect forestry application exclusively, however, because it was subsequently associated with recent homeowner application near the stream.


Total concentrations were highest for urban sites (median = 0.70 µg/L, n = 37), and total concentrations generally were lowest for agricultural sites (median =  0.11 µg/L, n = 4) (fig. 8). Total concentrations generally were lower for forestry (median = 0.042, n = 20) and mixed (median = 0.072 µg/L, n = 33) sites than for urban sites. A strong seasonal pattern was observed overall, with fall (median = 0.14 µg/L, n = 42) and spring (median = 0.14 µg/L, n = 40) storm samples from all sites combined associated with the largest total concentrations compared to winter storm samples (median = 0.055, n = 8) and non-storm samples (median = 0.026, n = 5). The distribution of total concentrations across seasons and land use indicates that only urban sites showed a contrast among seasons that ranged across several orders of magnitude. Total concentrations were higher in spring storm samples (median = 1.4 µg/L, n = 13) than in fall storm samples (median = 0.70 µg/L, n = 18). Total concentrations were more similar across seasons for forestry sites (median for fall = 0.047 µg/L, n = 10; for spring = 0.04 µg/L, n = 9). Total concentrations for the mixed sites were higher in the autumn storm samples (median = 0.098 µg/L, n = 14) than the spring storm samples (median = 0.057 µg/L, n = 13). Total concentrations at agricultural sites, only sampled during spring storms, were relatively high but less than concentrations detected in urban sites (median = 0.11 µg/L, n = 5). Total concentrations for winter storm samples were uniformly low across all land-use sites (for urban sites, median = 0.065 µg/L, n = 5; for mixed sites, median = 0.084 µg/L, n = 2; and for forestry sites a single value of 0.014 µg/L). 


Land-Use Signatures


The potential to detect specific land-use signatures at sites characterized by a mixture of land-use activity was limited because the uncertainties in pesticide use reporting as well as the overlap in use among different land- use categories made it difficult to identify compounds that can serve to reliably identify specific land-use applications. Nonetheless, the data indicate that urban/rural residential and agricultural pesticides are important components of pesticide transport in tributary drainage basins with a mix of land use, despite the relatively small proportion of drainage basin area associated with these uses. In the lower basin, the Cedar Creek drainage is predominantly forested (about 62 percent), but with a significant component of urban and rural residential land use (about 24 percent) and a smaller proportion of agricultural land use (about 11 percent). A total of 17 compounds were detected in12 samples at the mouth of Cedar Creek (site 9) (total n for detections=44) (table 13). Most of these compounds also were associated with the range of reported agricultural, forestry, and urban applications, although two compounds (fipronil and its metabolite desulfinylfipronil) were exclusively associated with reported urban use. Detections based on a common reporting level were associated with four compounds—2,4-D, caffeine, glyphosate and its metabolite AMPA, which also represent a mix of reported agricultural, forestry, and urban applications.


Camp Creek (site 14) also is largely forested (about 85 percent), with less agricultural activity (about 8 percent) and a small component of rural residential land use (about 2 percent). Most of the 15 compounds that were detected in 10 samples from Camp Creek were reported to be widely used across all land uses (total n for detections=45 (table 13). Two compounds were reported as exclusively urban (imidacloprid) or agricultural (simazine). Detections based on the common reporting level were associated with four compounds—2,4-D, glyphosate, imidacloprid, and triclopyr. Imidacloprid is an insecticide associated with urban use, and the other three compounds are associated with a mix of reported uses. 


A comparable pattern of compounds from across the range of land-use applications was observed at the treatment-plant intake (McKenzie River above Hayden Bridge, site 5), as previously discussed (table 8). In 17 samples, a total of 9 compounds (total n for detections=12) were detected, reflecting reported land-use application that included all categories of land use even though urban and agricultural lands comprise a small component of the drainage basin (about 4 and 2 percent, respectively). None of the concentrations in samples from the intake exceeded the common reporting level of 0.1 µg/L. Farther upstream, effects of forestry or agricultural pesticide use were not observed in mainstem sites, possibly due to the small number of samples. Caffeine was the only compound observed at these upstream sites, with three detections in five samples from the McKenzie River at Hendricks Park Boat Ramp (site 16) and one detection in two samples from the McKenzie River at Holden Creek Road Bridge farther upstream (site 19). All these concentrations exceeded the common reporting level of 0.1 µg/L. Both of these sites are associated with forestry land use with varying degrees of limited residential or agricultural land use. However, even though the data are limited, these results indicate that effects of forestry pesticide use are negligible at these locations in the river system, and effects of agricultural pesticide use are similarly not detectable. 


Physical Characteristics and Land Use


Evaluating physical characteristics of all detections (excluding double counting of summed forms and degradates that are not applied to the landscape) across the range of land use, most were of compounds that were not highly hydrophobic (log soil Koc < 3), as measured by their tendency to sorb to soil particles (median log soil Koc =2.3, n=310) (fig. 9A). Those detections with the most hydrophobic character primarily were associated with urban and agricultural land use (fig. 9). Many of these also are suspected of endocrine disruption (table 6). Similarly, data for persistence indicates that most detections were not associated with highly persistent compounds (T_1/2 < 100 days) (median T_1/2 = 46 days, n = 310), although the most persistent compounds were associated with urban and agricultural sites (fig. 9B). Strongly hydrophobic and persistent compounds were also observed at mixed sites (especially Cedar Creek at Springfield (site 9), which is subject to significant urban influence. No strong seasonal pattern for physical characteristics was observed, although a slight increasing gradient in hydrophobic character was indicated from the non-storm samples through spring, winter, and fall (median log soil Koc = 1.9, 2.0, 2.2, and 2.4, respectively). 


First posted May 30, 2012