U.S. Geological Survey investigations reveal widespread contamination of the nation's water resources.

ROBERT J. GILLIOM, JACK E. BARBASH, DANA W. KOLPIN, AND STEVEN J. LARSON

     One of the most striking findings was that one or more pesticides were found in almost every stream sample collected. More than 95 percent of the samples collected from streams and almost 50 percent of samples collected from wells contained at least one pesticide. Seventy-four of the 83 pesticide compounds analyzed were detected at least once in streams or groundwater. Major rivers, as well as agricultural and urban streams, had relatively similar high frequencies of detection.

     Analysis of patterns in pesticide use revealed that concentrations of herbicides and insecticides in agricultural streams, and in most rivers in agricultural regions, were highest in those areas of the nation with the greatest agricultural use. Herbicide concentrations were greatest in central U.S. streams, where use is most extensive. Insecticide concentrations were highest in urban streams.

     For drinking water, NAWQA results are generally good news regarding individual pesticides in relation to current regulations and criteria. However, important questions remain about risk to humans because the criteria cover a limited number of pesticides and a limited range of potential effects. Drinking water criteria--EPA Maximum Contaminant Levels (MCL), Health Advisory Levels (HAL), or Risk-Specific Doses (RSD)--have been established for 43 of the 76 pesticides analyzed but none of their transformation products (1). Peak levels of several herbicides frequently occurred above drinking water criteria in some agricultural streams. But annual average concentrations, upon which the criteria are based, were only rarely observed to exceed the criteria. Pesticide concentrations in wells also seldom exceeded drinking water criteria.

     NAWQA results show a high potential for pesticide impacts on aquatic life in some streams, particularly those in which concentrations of more than one pesticide approach or exceed aquatic-life criteria for an extended period of time. Matthiessen (2) recently (ES&T 1998, 32(19), 460A-461A) illustrated that mixtures may have toxic effects on aquatic organisms that are not anticipated from our limited data on individual compounds. The NAWQA study reveals that low-level mixtures are the most common form of pesticide occurrence in streams and groundwater. Long-term exposure to low-level mixtures of pesticide compounds, punctuated with seasonal pulses of high concentrations, is an exposure pattern that may not be adequately accounted for in present criteria.

Program scope and structure
Congress directed the U.S. Geological Survey in 1991 to begin the NAWQA Program to assess the quality of the nation's streams and groundwater. Pesticides were one of NAWQA's first priorities. NAWQA's evaluation of the geographic and temporal distribution of pesticide occurrence in relation to land use is painting a picture of the nature and distribution of pesticides in streams and groundwater. Further, the NAWQA results for pesticides frame key questions--which cannot yet be fully answered--about the significance of pesticide exposure to human health and aquatic ecosystems.

     The building blocks of the national assessment are water quality investigations in major watersheds, referred to as study units. The study units cover about 40% of the conterminous United States, encompassing 60-70% of national water use and population served by public water supplies. The study units are divided into three groups; each group is studied on a rotational schedule of three-year periods of intensive data collection. The first 20 NAWQA study units, the focus of this article, are widely distributed throughout the United States (see figure below).

Study locations and total herbicide distributions

     The first phase of the NAWQA investigations, performed during 1992-1996, included analyses of 76 pesticides and 7 pesticide degradates in more than 8000 water samples from streams and wells in 20 of the nation's major watersheds. The 76 pesticides studied account for about 75% of national agricultural use (mass basis) and a substantial portion of urban and suburban use.

     Study Design and Data Selection: This summary focuses on results from the most standardized components of the NAWQA study design (3), a brief outline of which is provided here. Data for streams were restricted to the single most complete year, and for groundwater, to one sample per well.

     The national study design for streams uses a network of "indicator" basins with relatively small drainages dominated by a single agricultural or urban land use, and "integrator" basins with large drainages and mixed land use influences. Water samples were collected from 40 agricultural streams, 11 urban streams, and 14 larger rivers with mixed land uses; typically, 20-40 samples were gathered over a 1-year period at each site. Samples were collected on a regular fixed-frequency schedule, such as weekly or monthly, supplemented by a smaller number of samples during selected high-flow conditions. The fixed-frequency sampling at most sites was increased during seasonal periods when pesticide concentrations were expected to be elevated.

     The national study design for groundwater focuses on recently recharged shallow groundwater associated with agricultural and urban land use in specific hydrogeologic settings and on major aquifers that are presently used for water supply. Most major aquifers are affected by a wide range of different land use activities and are considered as mixed land use. Wells were sampled in 36 agricultural land use studies, 13 urban land use studies, and 32 aquifer surveys. Each land use study and aquifer survey was a one-time sampling of 20-30 randomly selected wells within the geographic area and aquifer zone targeted for study.

     Chemical Analysis: Most NAWQA pesticide samples were analyzed by two different analytical methods for a total of 76 pesticides and 7 transformation products. Compounds were analyzed by gas chromatography/mass spectrometry (GC/MS) (4) and by high-performance liquid chromatography (HPLC) (5). Detection limits for the HPLC method are 5-50 times higher than detection limits for the GC/MS method. Details on the compounds analyzed and analytical performance are available via the World Wide Web at http://water.wr.usgs.gov/pnsp/anstrat.

Pesticide detections in streams and groundwater

Occurrence in natural waters
The widespread presence of pesticides in streams and groundwater is a principal finding of the study (see figure above). Frequencies of concentrations 0.01 g/L were lower for groundwater than for streams in all land-use categories and decreased from 45% of samples for shallow groundwater in agricultural areas to 37% for shallow groundwater in urban areas and to 20% for major aquifers with mixed land uses. Compared with streams, groundwater had fewer concentrations above 0.05 g/L in all land use and hydrogeologic settings, reflecting the overall lower concentrations in groundwater.

Pesticide detections in streams and groundwater

     The geographic distribution of pesticide concentrations generally follows regional patterns in agricultural use and the influence of urban areas, although this relation is stronger for streams than for groundwater. Compared with streams, the occurrence of pesticides in groundwater is more strongly governed by compound properties and hydrogeologic factors that affect transport from land surface to a well. The geographic distribution of herbicides in streams and groundwater exemplifies results from the first 20 NAWQA study units (see figure on previous page). The annual 75th percentile of monthly median concentrations of total herbicides (sum of all herbicides) is shown for (A) streams and (B) the overall detection frequency for groundwater. Values for each site or study were ranked by national quartiles for color coding on the maps.

     Herbicide use is extensive in the central United States, and stream concentrations are correspondingly high. All five agricultural streams and two major rivers in the White River Basin (Ind.) and Central Nebraska Basin (Nebr.) study units had herbicide concentrations in the highest national quartile. Urban streams had the highest insecticide concentrations; 7 of 11 had total insecticide concentrations in the upper 25% of all streams sampled, although some agricultural streams in irrigated agricultural areas of the western United States also had high levels. Results for groundwater show that herbicides were highest in shallow groundwater within agricultural areas and lowest in major aquifers, but the locations of areas with the highest detection frequencies did not follow use patterns as clearly as did streams. Insecticides were seldom detected in groundwater although about 8% of samples of shallow groundwater from agricultural and urban areas had insecticide detections at concentrations less than 0.01 g/L.

     In most agricultural areas, the highest levels of pesticides occur as seasonal pulses--usually during spring and summer--lasting from a few weeks to several months during and following high-use periods. Total pesticide concentrations in streams draining urban areas are generally lower than in agricultural areas, but seasonal pulses last longer and the concentrations are more dominated by insecticides.

Use-detection relationships
A relatively small number of heavily used compounds accounts for most detections. The most frequently detected pesticide compounds in agricultural areas (see figure on next page) were the major herbicides atrazine (and its transformation product, deethylatrazine (DEA)), metolachlor, cyanazine, and alachlor, ranked first, second, fourth, and fifth in national herbicide use for agriculture. Results from other studies suggest that transformation products of metolachlor, alachlor, and cyanazine would also have been frequently detected if they had been analyzed (6, 7). The most heavily used herbicides also account for most of the detections in rivers and major aquifers and many of the detections in urban streams and shallow groundwater.

Detection frequencies of pesticide compounds

     The herbicides found more often in urban than agricultural areas (see figure on next page) are simazine, prometon, 2,4-D, diuron, and tebuthiuron. Simazine and prometon account for most detections in streams and shallow groundwater in urban settings. Note that 2,4-D and diuron could not be evaluated at the 0.01 g/L level because of analytical limitations and could only be approximately compared with other compounds at the 0.05 g/L detection level. The use of 2,4-D and prometon rank 1st and 14th among herbicides in frequency of home and garden applications. In addition, 2,4-D, simazine, and diuron rank 3rd, 18th, and 23rd, respectively, in national herbicide use for agriculture. Prometon and tebuthiuron have no reported agricultural use.

     Insecticides were detected more frequently in urban streams than in most agricultural streams and were seldom detected in groundwater in either land use. Most detections were accounted for by diazinon, carbaryl, malathion, and chlorpyrifos, which nationally rank 1st, 8th, 13th, and 4th among insecticides in frequency of home and garden use.

     Several pesticides that are used extensively in agriculture were infrequently detected, even in streams. These include the herbicides pendimethalin, linuron, propachlor and propanil, and the insecticides methyl parathion, terbufos, and disulfoton. Of these compounds, linuron, propachlor, and propanil were not heavily used in any of the basins studied. In addition, the physical and chemical properties of most of these compounds or their methods of application create a low potential for transport to streams by runoff (8).

Environmental significance
Pesticides in streams are a potential concern for human health if they affect a drinking water source or occur where there is recreational use. They also are a potential concern for aquatic life in all streams. The primary issue for groundwater is drinking water quality, although groundwater may also play a role as a source of pesticides to surface water. For protection of drinking water and aquatic life, water quality criteria have been established for some pesticides. These criteria provide widely used benchmarks that serve as starting points for evaluating the potential effects of exposure.

Detection frequencies of pesticide compounds

     Drinking Water: Although NAWQA was designed as a broad water resource assessment and did not specifically target drinking water supplies, its findings are relevant to drinking water quality. However, some limitations must be kept in mind. Most of the major aquifers and about half of the shallow groundwater zones sampled are drinking water sources. Thus, NAWQA results for groundwater are directly relevant to potential drinking water concerns even though many of the wells sampled are not used for domestic supply. Most of the streams sampled are not directly used as sources of drinking water. The agricultural and urban indicator sites can be viewed as extreme examples of what drinking water sources would be like in highly developed watersheds within the region. Some sites sampled on major rivers are close enough to water-supply withdrawals that they reasonably represent the source water. Others, however, are in regions where groundwater or remote surface water sources are used for drinking water.

     Annual average pesticide concentrations in streams--upon which drinking water criteria (DWC) are based--only exceeded the MCL for atrazine in one agricultural stream and the HAL for cyanazine in this same stream and one other agricultural stream. Neither of the streams-- Kessinger Ditch, (Ind.) or Prairie Creek, (Nebr.)--are used directly for drinking water, nor do they flow into larger streams that are used until much farther downstream.

     Pesticide concentrations seldom exceeded DWC in wells. For shallow groundwater in agricultural areas, 4 of 925 wells (0.4%) had concentrations greater than a criterion: 1 for atrazine, 1 for cyanazine, 2 for dieldrin (RSD), and 1 for dinoseb (MCL). For shallow groundwater in urban areas, 10 of 301 wells (3%) had concentrations greater than a criterion: 1 for atrazine and 9 for dieldrin. In major aquifers, only 1 of 933 wells (0.1%) exceeded a criterion, with a concentration of dieldrin above the RSD.

     Although NAWQA results indicate few problems for drinking water based on current criteria, conclusions must be tempered by several concerns:

      Criteria are not established for many pesticides.

      Cumulative exposure from drinking water plus food and other avenues is not considered.

Frequency of mixtures in streams and groundwater

      Mixtures and transformation products (see box below and figure above) are not considered.

Sidebar: Pesticide mixtures

      The effects of seasonal exposure to high concentrations have not been evaluated.

      Some types of potential effects, such as endocrine disruption, reproductive or nervous system disorders, and unique responses of sensitive individuals, have not yet been assessed.

Stream concentrations can exceed aquatic-lifelife criteria

     For these reasons, together with the pervasive uncertainty of extrapolating results from laboratory animals to humans, estimating the risk associated with long-term consumption of drinking water that contains pesticides, even at levels below current regulatory standards, is speculative. The 1996 Food Quality Protection Act will expand the numbers of pesticides considered in drinking water assessments and consider all major exposure routes. However, it will be hampered by uncertainty in the degree of risk associated with common exposure patterns, such as the sustained presence of low-level mixtures and seasonal pulses of high concentrations. Improved risk assessment will require an iterative approach that relies on the simultaneous improvement of exposure characterization and effects assessment, with frequent feedback between the two.

     Aquatic Life: Assessment of the pesticide risk to aquatic life is hampered by many of the same problems discussed in relation to human health, but existing water quality criteria are also more often exceeded (see figure above, right). Many of the streams studied had only one or two samples with concentrations above a criterion, but concentrations of atrazine and diazinon above or near criteria values for sustained periods were common in some streams.

Stream concentrations can exceed aquatic-lifelife criteria

     For aquatic life, NAWQA results indicate a high potential for effects in some streams, particularly in urban areas. In addition to high concentrations of individual compounds, some streams have seasonal periods during which mixtures of several compounds approach or exceed criteria.

Implications for NAWQA
Results from the first phase of NAWQA have clarified our understanding of the nature and degree of pesticide exposure in streams and groundwater. However, the assessment remains incomplete in several ways: the range of environments sampled, the range of chemicals assessed, our ability to extrapolate exposure characterization to unsampled areas, our understanding of the factors that govern pesticide behavior and define potential management options, and our understanding of the risk level to humans and aquatic life.

     The second phase of NAWQA investigations, to be completed this year, and the third phase, to be completed in 2002, will more than double the number of pesticide measurements in streams and groundwater, expand the range of environments sampled, and selectively expand analytical capabilities to keep pace with new compounds and degradation products. Coupled with this expansion of data collection will be a focus on the use of empirical and deterministic models to improve the estimation of pesticide exposure patterns for unsampled water resources.

     Finally, the relationship between the pesticide exposure patterns observed in the ambient environment and potential effects on humans and aquatic life must be better understood. Modifications of the NAWQA design are being considered to improve the assessment of effects on aquatic life in future studies. For evaluating potential effects on humans, however, NAWQA will focus on collaboration with human-health researchers in other organizations.

References

(1) Gilliom, R. J.; Mueller, D. K.; Nowell, L. H. Methods for Comparing Water-Quality Conditions Among National Water-Quality Assessment Study Units, 1992-1995; U.S. Geological Survey Open File Report 97-589; U.S. Geological Survey: Reston, VA, 1998.

(2)  Matthiessen, P. Aquatic risk assessment of chemicals: Is it working? Environ. Sci. Technol. 1998, 31(19), 460A-461A.

(3)  Gilliom, R. J.; Alley, W. M.; Gurtz, M. E. Design of the National Water Quality Assessment Program--Occurrence and Distribution of Water Quality Conditions; U.S. Geological Survey Circular 1112; U.S. Geological Survey: Reston, VA, 1995.

(4)  Zaugg, S. D.; Sandstrom, M. W.; Smith, S. G.; Fehlberg, K. M. Methods of Analysis by the U.S. Geological Survey National Water Quality Laboratory--Determination of Pesticides in Water by C-18 Solid-Phase Extraction and Capillary-Column Gas Chromatography/Mass Spectrometry With Selected-Ion Monitoring; U.S. Geological Survey Open File Report 95-181; U.S. Geological Survey: Reston, VA, 1995.

(5)  Werner, S. L.; Burkhardt, M. R.; DeRusseau, S. N. Methods of Analysis by the U.S. Geological Survey National Water Quality Laboratory--Determination of Pesticides in Water by Carbopak-b Solid-Phase Extraction and High-Performance Liquid Chromatography; U.S. Geological Survey Open File Report 96-216; U.S. Geological Survey: Reston, VA, 1996.

(6)  Herbicide Metabolites in Surface Water and Groundwater; Meyer, M. T., Thurman, E. M., Eds.; ACS Symposium Series 630; American Chemical Society: Washington, DC, 1996.

(7)  Kolpin, D. W.; Thurman, E. M.; Linhart, S. M. The environmental occurrence of herbicides: The importance of degradates in groundwater. Arch. Environ. Contam. Toxicol. 1998, 35, 385-390.

(8)  Goss, D. W.; Wauchope, R. D. The SCS/ARS/CES pesticide properties database: II Using it with soils data in a screening procedure. In Pesticides in the Next Decade: The Challenges Ahead; Weigman, D. L., Ed.; Virginia Water Resources Research Center: Blacksburg, VA, 1990; pp. 471-493.

At the U.S. Geological Survey, Robert J. Gilliom is chief of the Pesticide National Synthesis Project. Jack E. Barbash and Steven J. Larson are chemists, and Dana W. Kolpin is a hydrologist on the project team. All are with the U.S. Geological Survey's National Water Quality Assessment Program.