Conclusions


In 135 environmental samples collected at 28 sites in the McKenzie River basin during 2002–10 that were analyzed for as many as 175 compounds, a total of 43 compounds were detected at least once. Concentrations tended to be low (< 0.1 µg/L) (median for maximum concentration for each compound = 0.055 µg/L). Most samples contained compound mixtures (median number of detections per sample = 4), although some samples had no detections. Caffeine was the most frequently detected compound, and with hexazinone, 2,4-D, atrazine, glyphosate and its metabolite AMPA, and carbaryl, accounted for approximately 46 percent of all detections. When detections were measured using a common LRL (0.1 µg/L), the number of compounds was reduced to 19, reflecting the occurrence of many compounds only at low concentration. When screened at the common LRL, no detections were observed for two of the most frequently detected compounds (hexazinone and atrazine), although caffeine, 2,4-D, and glyphosate (with the addition of diuron) remained among the most frequently detected compounds. Twenty-one compounds (nine when based on the common LRL) are either regulated by a drinking-water standard (n = 9 versus n = 4 for common LRL) and/or suspected to be endocrine-disrupting compounds (n = 20 versus n = 6 for common LRL). 


Nine compounds were detected at the treatment-plant intake (none when screened at the common LRL), most of which were frequently detected at other sites. Concentrations were uniformly quite low, most of them E-coded as “estimates” or less than the LRL. Human-health benchmarks were available for six of these compounds and were several orders of magnitude higher than measured concentrations, indicating that pesticide concentrations at the drinking‑water intake present a negligible threat to human health. Nonetheless, multiple compounds were occasionally detected in a single sample, and the potential for synergistic effects of occasional low-level presence (< 0.1 µg/L total concentration) of compound mixtures is not well understood.


The largest number of pesticide detections occurred during spring storm surveys, primarily associated with urban stormwater drains. Urban sites also were associated with the highest concentrations, occasionally exceeding 1 µg/L. Many of the compounds detected at urban sites were relatively hydrophobic (log soil Koc > 3), persistent (T_1/2 > 100), and suspected of endocrine disruption; these patterns held for both total detections and for detections screened at the common LRL. When screened by the common LRL, caffeine and 2,4-D were most frequently detected in spring and fall storms; additionally, diuron was an important compound detected during the spring and glyphosate was among the most frequently detected during the fall. 


Even though pesticides were detected across the range of sites under both storm and non-storm conditions, not all categories of sites responded in the same way to increased precipitation. No simple relation was observed for any category of sites between total (summed) pesticide concentration and precipitation volume, although larger precipitation events were significantly correlated (p < 0.05) with more detections at forestry sites and the drinking-water intake. In contrast, increasing precipitation volume was associated with fewer detections at urban sites. These results suggest that pesticides show a different pattern of runoff and transport in the urban environment compared to less impervious environments associated with forested lands. 


The data show a tendency for increasing total pesticide concentration in Cedar Creek with increasing proportion of flow from stormwater drains, suggesting that stormwater is an important influence on pesticide transport in that tributary. Additional data from stormwater drains show that pesticides can be detected even under non-storm conditions, although total concentrations were relatively reduced compared to storm concentrations. Considerable variability in total pesticide concentration was observed in samples collected from stormwater drains over the range of a storm hydrograph.


Definitive land-use signatures impacting conditions at the drinking-water intake were difficult to establish because of the potential for many observed chemicals to be used across a range of land uses, as well as the uncertainty in pesticide use estimates. Nonetheless, the occurrence of pesticide detections across all categories of sites indicates that all land-use applications contribute to pesticide runoff. This pattern was observed for all detections as well as the subset of detections screened at the common LRL. No significant detections based on the common LRL of any pesticide compounds were observed at the drinking-water intake or any mainstem river site, indicating that concentrations in the McKenzie River itself were consistently low. Although forest land use is predominant in the basin, and forestry pesticide use can be detected in small tributaries draining forested lands following application, these compounds rarely were detectable in the McKenzie River. Forestry pesticide use, therefore, probably is not a potential threat to drinking-water quality at the present time. Agricultural pesticide runoff is not well characterized by the limited data available. Nonetheless, agricultural pesticide use is likely to pose a greater potential threat because of the large number of relatively hydrophobic agricultural compounds reported to be used in the basin (most of which were never detected, presumably because of limited sampling of agricultural streams) and the moderately high concentrations that were observed in small tributaries draining agricultural lands. More complete understanding of agricultural chemicals in runoff in the McKenzie River basin requires further investigation. In contrast, results of this analysis are sufficient to strongly suggest that urban pesticide use is an important source for pesticides of concern for drinking water in runoff, not limited exclusively to storm conditions. A large number of compounds and high concentrations (> 0.1 µg/L) were observed in stormwater drains, many of them relatively hydrophobic (log Koc > 3) and persistent (T_1/2 > 100 days). 


Conceptual Model


Because this was a reconnaissance study, the data were not intended to provide a consistent framework for comprehensive and rigorous analysis. Nonetheless, the results from this study and others in Oregon represent a useful foundation for generating a hypothetical conceptual model describing pesticide contamination and transport in the McKenzie River basin. This conceptual model depends on and is consistent with current scientific thinking about pesticides in surface waters (Larson and others, 1997). Nonetheless, because limited data currently exist to fully characterize pesticide occurrence and transport in the McKenzie River basin, this model can be understood essentially as a set of hypotheses that are proposed to serve as the foundation for future monitoring in the basin. 


The McKenzie River basin is dominated by forested land in the High and Western Cascade physiographic provinces with a large groundwater component, and consequently by mainstem streamflow that is clean and relatively stable during base flow. As such, it serves as a valuable source for drinking water of superior quality for the City of Eugene. The relatively small number of pesticide detections and low concentrations detected at the treatment plant intake indicate that this high quality of water is not seriously compromised at this time. Nonetheless, pesticides can be detected in stormwater channels and streams that drain a range of land use in the basin including urban/residential settings, agricultural applications, and forestry management. 


A large number of compounds are reported or estimated to be used within the context of these land-use activities. Some fraction of the many pesticides applied across the mixture of land-use settings in the basin is transported from the site of application to surface waters in the basin, presumably primarily through surface runoff, so that all land-use activities in the basin generate measurable pesticide runoff. Because most pesticide applications occur in the middle and lower regions of the basin, where precipitation generally occurs as rain, pesticide transport is pronounced during storm conditions in the spring and fall. Pesticide compounds also may be present at low levels in surface waters throughout the year, suggesting ongoing supply from some sources and/or some degree of groundwater input in addition to surface-water runoff.


A large number of pesticides are used in urban settings, some of them estimated to be unique to urban use. Pesticides are applied in urban areas according to a fairly continuous pattern throughout the spring, summer, and fall (Larson and others, 1997). Once precipitation begins to fall, the larger proportion of impervious area in urban environments means that runoff is generated quickly. Virtually all pesticides that reach impervious areas are transported to stormwater drains, where concentrations can become quite high. As a result, pesticides may be detected in urban stormwater drains whenever significant runoff occurs. The summed or total pesticide concentrations in storm drains vary considerably over the course of individual storm events, often being highest near the peak of the hydrograph; at the same time, urban pesticide sources also may be depleted or diluted relatively quickly by increasing volume of precipitation. Urban runoff via stormwater drains during storms is an important source of discharge to Cedar Creek, and a lesser source to Keizer Slough, both of which flow into the McKenzie River close to the treatment-plant intake.


The occurrence of agricultural pesticides in streams in the McKenzie River basin is not well documented, although it is assumed to accord with patterns observed in other Oregon streams (Anderson and others, 1996, 1997; Rinella and Janet, 1997; Wood, 2001). These patterns include runoff of agricultural pesticides during large spring storms following application, with reduction in runoff occurring during the low-flow period in the summer. Another peak in concentration is anticipated in agricultural streams during the first major precipitation event in the fall. 


Most compounds that are used for forestry applications in the McKenzie River basin are widely used for other applications in the basin, although a small number are unique to forestry use. Because forestry applications are relatively limited in both time and space, forestry pesticide use is less of a concern than urban or agricultural use. Pesticides, predominantly herbicides, are utilized in forestry management primarily for site preparation before planting and reduction of competition from non-target vegetation. This pattern of application means herbicides are applied only once or twice in the period of two to five decades between planting and timber harvest (Larson and others, 1997). As a result, in any given year pesticides are applied to a small number of forested watersheds, which are also relatively small in size. Furthermore, depending on the compound and mode of application, pesticide runoff from forestry use occurs fairly quickly during the first few storms following application (Neary and others, 1993). Forestry pesticides are therefore transported in surface waters only briefly, occurring in short-lived pulses that are quickly reduced by dilution downstream (Larson and others, 1997). As a result, the seasonal pattern in forestry pesticide transport is relatively ephemeral and site specific, with pesticides showing a strong response to precipitation and primarily mobilized within the first few months after they are applied (Neary and others, 1993).


Most compounds that are detected in water by conventional laboratory analysis tend to be relatively hydrophilic and are therefore transported primarily in the dissolved phase. Additional compounds have been reported or estimated to be used in the basin, especially for agricultural applications, but have not yet been observed in surface water. Laboratory analysis has been conducted exclusively on filtered samples, so chemicals that are more hydrophobic in character and occurring primarily in association with sediment particles are not likely to be detected. Data from a study of sediment released from Cougar Reservoir in the South Fork McKenzie River during a construction project in 2002–05 document that metabolites of DDT were detected in fine sediments deposited downstream, including below the confluence with the mainstem McKenzie River (Anderson, 2007). DDT was widely used in a forestry application in the upper McKenzie Basin during the 1950s. No species of DDT or its metabolites were detected in water during that study, although they have been detected in very low concentrations, well below reporting limits for conventional analysis, in extracts from passive samplers deployed in the lower basin (USGS, unpublished data). These data suggest that contaminated sediment may be serving as a persistent and low-level source for these types of compounds in the basin. 


Potential threats to drinking water quality are specifically identified in this report by the occurrence of pesticide compounds that are regulated for drinking water or suspected of endocrine disruption. Observed concentrations of pesticides at the drinking water intake, while uniformly low (< 0.1 µg/L), include one regulated compound and four suspected endocrine disruptors, making the reduction of sources for these compounds a high priority for EWEB. These compounds were estimated to be either associated with exclusive urban (cypermethrin) or predominantly agricultural (atrazine and diazinon) land use, or both urban and agricultural applications (carbaryl). Furthermore, while little data exist to characterize agricultural pesticide runoff, a large proportion (40 percent) of compounds detected in agricultural streams is currently regulated by drinking water criteria. Similarly, a large number (50 percent) of compounds detected in urban runoff are known or suspected endocrine disrupting compounds. In contrast, fewer compounds detected in forestry streams are associated with either category (< 30 percent). Accordingly, while pesticide sources include all land use activities occurring in the basin, those compounds presenting the greatest recognized potential threat to drinking water quality in the McKenzie River are largely related to urban and agricultural pesticide applications. 


Implications for Monitoring


A scientifically-based monitoring program is one with clear objectives and a sampling strategy that builds on current data to refine understanding and provide new insights. In terms of drinking water source protection, appropriate objectives are directly related to identifying and reducing perceived potential threats to drinking water quality. The data presented here, and the proposed conceptual model describing pesticide transport in the McKenzie River Basin, indicate that pesticide runoff occurs across the range of land use activities in the basin. Nonetheless, the majority of compounds that present a documented threat to drinking water quality, in terms of water‑quality regulations or suspected endocrine disruption, are associated with agricultural and urban land use applications rather than forestry. These data suggest that agricultural and urban land use areas are the most important to target for future monitoring efforts, and eventually for developing management strategies to reduce pesticide runoff.


Conventional monitoring approaches for drinking water source protection are focused on identifying conditions that result in mobilization of contaminants, especially those with documented adverse health consequences. These programs are typically based on discrete samples that characterize storm runoff across a range of land use sources, similar to the approach taken in this reconnaissance study. In terms of identifying potential threats to drinking water quality, it is clearly important to quantify acute concentrations of toxic chemicals during peak periods of runoff and to identify possible sources for those chemicals. An important issue for further consideration, however, is that pesticides and other relatively hydrophobic chemicals may be present at very low concentrations (below the detection level for conventional analytical techniques), especially in compound mixtures with unknown synergistic effects. This is especially true for many of the compounds associated with urban and agricultural use in the basin. Since health-based criteria are based on long-term exposure, an important component of the potential threat from the most hydrophobic of these chemicals may be not be acute or short-term concentrations, which are not likely to be detected with discrete samples in any case. Passive sampling methods that sequester relatively insoluble compounds over a period of weeks may be more suitable for documenting the low-level presence of many of these compounds, and are increasingly being used for current use pesticides and other organic compounds of interest (Alvarez and others, 2008). Furthermore, utilization of the passive sampling approach provides an alternative perspective that expands the temporal scale of observation to incorporate more chronic conditions, thereby expanding our understanding of pesticide occurrence.


A separate concern relative to assessing threats to drinking water is that MCL criteria and human-health benchmarks that identify the potential for toxicity are defined for compounds in isolation, without accounting for exposure to compound mixtures. These criteria and guidelines do not represent the actual exposure risk of chemical mixtures that are typically observed in the stream environment. The use of metabolic assays that target specific biological responses to chemical mixtures provides a means to quantify specific toxic behavior resulting from the synergistic effect of multiple chemicals, some of which may not be measured or even detectable with conventional analysis (Routledge and Sumpter, 1996).


In summary, there are several components to consider relative to monitoring a complex system like the McKenzie River: the elements of the system, which include the individual chemicals and their sources; the interconnections within the system, which can be conceived as the land use and climate factors that interact to mobilize and transport chemicals; and the function or behavior of the system, which includes the detrimental effect of chemical mixtures on human health. The conventional approach to water quality monitoring generally is to document the first two components, at least at the scale of discrete point-in-time conditions. An important next step in developing our understanding of pesticide occurrence is expanding the temporal scale of pesticide measurement beyond the specific conditions that occur during storm runoff to characterize more persistent exposure, especially for relatively hydrophobic and toxic chemicals that occur in very low concentrations in water (Alvarez and others, 2008). Furthermore, including some measure of system behavior in the form of synergistic effects of compound mixtures creates a more effective monitoring program capable of providing new understanding about existing threats to drinking water.


First posted May 30, 2012