Temporal Trends of Selected Agricultural Chemicals in Iowa's Groundwater, 1982-95: Are Things Getting Better?

Dana W. Kolpin, Debra Sneck-Fahrer, George R. Hallberg, and Robert D. Libra

---

METHODS

Field

All groundwater samples were collected by USGS personnel using protocol designed to obtain a water sample representative of the aquifer (Detroy, 1985). Thus, all samples were collected as close to the wellhead as possible and prior to any water treatment. All wells were pumped at least 20 min prior to sampling to allow for complete removal of stagnant water from the casing and for the sampled water to reach chemical stability (based on successive field measurements of water temperature, specific conductance, and pH). Water samples collected for the analysis of nitrate were filtered through a 0.45-µm cellulose acetate filter into opaque polyethylene bottles, preserved with sulfuric acid (1 mL), and immediately chilled. Water samples collected for pesticide analysis were unfiltered, stored in 1-L borosilicate glass bottles having polypropylene lids with teflon liners (specifically prepared for collecting samples for pesticide analysis), and immediately chilled. The chilled water samples for nutrient and pesticide analysis were sent to UHL (typically within 24 hours), where all the analytical work was performed.

Laboratory

The laboratory methods used to obtain concentrations of the selected agricultural chemicals examined for this study have not changed significantly through the duration of the IGWM. Refrigerated water samples for nitrate were analyzed according to U.S. Environmental Protection Agency USEPA Method 353.3 (U.S. Environmental Protection Agency, 1983) employing copper-cadmium reduction and colorimetric quantitation with an automated analyzing system. The analytical reporting limit for nitrate using this method was 0.10 mg/L.

Refrigerated water samples for pesticide analysis were usually extracted within two to three days from sample collection but always within a 7-day holding period. These extracts were refrigerated until analyzed. Analysis of pesticides was performed using a USEPA multi-residue, gas chromatography method (U.S. Environmental Protection Agency, 1980, 1988; i.e., USEPA Method 507, Determination of Nitrogen- and Phosphorus- Containing Pesticides in Water by Gas Chromatography with a Nitrogen-Phosphorus Detector). Laboratory established analytical reporting limits determined by USEPA Method 507 for alachlor, atrazine, cyanazine, and metolachlor are all 0.10 µg/L. Residues were extracted using methylene chloride and exchanged into hexane (adjusted to 10% acetone concentration to improve recovery of cyanazine). Varian 3400 gas chromatographs with capillary columns and nitrogen and phosphorus detectors were used for the analysis. Over the duration of the IGWM, minor improvements in methods and equipment have been made (e.g., changes from packed columns to capillary columns), but these have not substantively changed the precision or accuracy of the routine pesticide analyses. For confirmation, the laboratory used two dissimilar chromatographic columns for separation and quantitation of each compound. Instrument calibration was performed routinely using appropriate standards for these pesticides. For the pesticides included in this study, percentage recovery from laboratory spiked samples generally ranged from 84 to 110% during the period of study. Reported concentrations were not adjusted for percentage recovery.

As part of its quality-assurance effort, UHL participates in numerous inter-agency and inter-laboratory proficiency and performance evaluation programs including: USEPA Water Supply Series, Water Pollution Series, Laboratory Proficiency Testing Program, and Solid Waste Series; the USGS Standard Reference Sample Program; and the American Industrial Hygiene Association programs.

Project Design

Because there could be some seasonal fluctuations in concentrations of agricultural chemicals in unconfined aquifers (Barbash and Resek, 1996; Kolpin et al., 1994), only IGWM samples collected during the months of April to September of each year were used for study. To eliminate potential bias in the data set, if an individual well had multiple water samples collected between April-September in a particular year, the water sample closest to the middle of this time period was selected. Even though problems of seasonality may not have been completely eliminated using this criterion (includes both spring and summer seasons), it was felt that further restrictions on the months of sample collection would have eliminated too many samples -- decreasing the temporal nature of this dataset. A total of 1,936 water samples from 1,019 wells were obtained using these criteria (Fig. 1). The wells are distributed throughout the state except in portions of southern Iowa where surface-water supply systems are dominant. The data were divided into three roughly equal time periods (1982-86, 1987-91, and 1992-95) to evaluate long-term patterns in chemical concentrations. The selected wells have a median well depth of 39.3 m. The wells are grouped into the following aquifer types: alluvial (34.2%), glacial-drift (19.6%), and bedrock (46.2%).

To ensure that the sampling design of the IGWM did not cause misidentifications of temporal trends in the data, a subset of all sampled municipal wells was selected for a more in-depth investigation of temporal patterns. Possible misidentifications may be caused by the changing population of wells sampled for a given year (not following the same set of wells through time). Furthermore, the selection of a portion of wells to be sampled during a particular year was not always completely random -- during early years of the IGWM, specific aquifers were targeted because of suspected vulnerability, and some individual wells were sampled because of known water-quality problems. The criteria for selection were that a well have at least a water sample collected before 1988 and after 1991 for the analysis of agricultural chemicals. These criteria yielded 89 wells with the longest record of water-chemistry data available in the IGWM (Fig. 2) and assured that the same set of wells were examined through time. As with the entire IGWM data set, only the April - September samples for each well were used for this study. A median chemical concentration from samples for these wells was calculated for each time period under investigation. For these 89 wells, 69 had samples collected during 1982-86; 76 had samples collected during 1987-91; 89 had samples collected during 1992-95; and 56 were sampled during all three time periods. These 89 wells have a median well depth of 17.7 m. The wells are grouped into the following aquifer types: alluvial (68.0%), glacial-drift (16.0%), and bedrock (16.0%). Thus, the subset of wells for in-depth study tended to be shallower and had a greater proportion of alluvial aquifers than the total IGWM data set.

Ancillary data compiled for this study included statewide estimates of pesticide use for the years 1979, 1985, 1990, 1991, 1992, 1993, 1994, and 1995 (Hartzler and Wintersteen, 1991; U.S. Department of Agriculture, 1995) and annual statewide estimates of nitrogen-fertilizer use for the period 1982-1995 (Vroomen, 1989; U.S. Department of Agriculture, 1995) (Table 1). To obtain average pesticide application rate (intensity of use, mass per unit area), the estimated total mass applied was divided by total crop hectares planted. For atrazine and cyanazine, the divisor used was total corn (Zea mays L.) hectares planted; for alachlor and metolachlor, the divisor was total corn plus total soybean [Glycine max (L.) Merr.] hectares planted.

Nonparametric statistical techniques were used for this study. These methods were appropriate because the data were not normally distributed, and a large percentage of the data were censored; that is, many of the determinations were less than the analytical reporting limit. The Kruskal-Wallis test (Helsel and Hirsh, 1992) was used to test for differences in the medians of two or more groups. The Wilcoxon signed-ranked test (Helsel and Hirsh, 1992) was used to determine whether the median difference between paired observations equaled zero. The long-term chemical use data (Table 1) were used to determine whether a one- or two-tailed Wilcoxon sign-ranked test was appropriate. If a continuous increasing or decreasing long-term trend in use was noted for an individual compound in Table 1, then a one-tailed Wilcoxon test was used. Otherwise a two-tailed Wilcoxon test was used. A significance level of 0.05 was used for all statistical tests in this study.

---

Next: Results