Eric F. Vowinkel, Hydrologist

U.S. Geological Survey, 810 Bear Tavern Road, Suite 206, West Trenton, NJ 08628

Phone: (609) 771-3931; E-mail: vowinkel@usgs.gov

A numerical rating model was developed to assess the vulnerability of water from community water-
supply wells in New Jersey to contamination by pesticides. The model was used to rank the community water-supply wells into groups of low, medium, and high vulnerability to pesticides on the basis of hydrogeologic-sensitivity and pesticide-use-intensity variables. Sensitivity variables used in the model were distance of the well from the outcrop area of the aquifer from which water is withdrawn, soil organic-matter content at the wellhead, and depth to the top of the open interval of the well. Pesticide-use-intensity variables used in the model were predominant land use surrounding the wellhead, distance of the well from agricultural land, and distance of the well from a golf course.

Of the 1,945 wells for which sufficient hydrogeologic and land-use data were available, the vulnerability index calculated with the model was low for 26 percent, medium for 70 percent, and high for 4 percent of the wells. To test the reliability of the model, 90 wells were randomly selected for sampling, and water samples were analyzed for 140 pesticides and nutrients. Pesticides were detected in 6 of the 90 wells. None of the five pesticides that were detected was present at a concentration that exceeded a U.S. Environmental Protection Agency or New Jersey Department of Environmental Protection (NJDEP) maximum contaminant level for drinking water. The frequency of pesticide detection in water from wells by vulnerability index was 0 percent in the low-vulnerability group, 5 percent in the medium-vulnerability group, and 19 percent in the high-vulnerability group. The median nitrate concentration in water from wells in the low-vulnerability group was less than the detection level of 0.1 milligrams per liter.

The NJDEP used the results of the vulnerability study to evaluate strategies for monitoring requirements for pesticides. Monitoring requirements can be waived or reduced for wells that are not vulnerable to contamination. The NJDEP estimated that monitoring waivers for community water-supply wells will save water purveyors about $5 million in analytical costs annually.

About 50 percent of the 8 million people in New Jersey obtain their drinking water from ground-water sources—about 39 percent from community supply wells and 11 percent from privately owned domestic supply wells. In 1995, about 400 Mgal/d (million gallons per day) of water was withdrawn from more than 2,300 community water-supply wells. Regulations enacted by the U.S. Environmental Protection Agency (USEPA) and the New Jersey Department of Environmental Protection (NJDEP) require water companies to monitor well water for pesticides routinely (Louis et al 1994). NJDEP determined that 10 of the 23 regulated pesticides are used in New Jersey and have been detected in water supplies; therefore, some monitoring is required (Louis et al 1994). Monitoring requirements for pesticides in water samples can be waived if (1) the part of the aquifer from which the water is withdrawn is insensitive to contamination by pesticides or (2) the aquifer is sensitive to contamination but pesticides are not used in the area near the wellhead. A screening tool was needed for evaluating the vulnerability of wells to pesticides at a statewide scale.

The U.S. Geological Survey (USGS), in cooperation with the NJDEP, developed a geographic information system (GIS) data base and a numerical rating model to provide a measure of the vulnerability of water from community water-supply wells to contamination by pesticides (Vowinkel et al 1994; Vowinkel et al 1996). The GIS data base and the rating model were used to rank community water-supply wells (fig. 1) into groups of low, medium, and high vulnerability. To test the results of the model, water samples were collected from 90 wells and analyzed for pesticides, nitrate, and other nutrients. Results of a previous investigation indicate that concentrations of nitrate in water from domestic wells were larger in those wells in which pesticides were detected than in those wells in which no pesticide was detected (Vowinkel and Tapper, 1995). The frequencies of detection and concentrations of the pesticides and nitrate in water from wells were compared among the vulnerability groups. This paper describes the method and results of the application of the numerical rating model used to predict vulnerability of water from wells to contamination by pesticides.

New Jersey’s principal aquifers can be classified into three groups: (1) unconsolidated sediments of the Coastal Plain, (2) unconsolidated glacial sediments, and (3) bedrock containing fractures and (or) solution cavities. The aquifers of the Coastal Plain differ in areal extent and thickness, and generally are permeable units of unconsolidated sand and gravel that are separated from each other by less permeable units of silt and clay. Aquifers in the Coastal Plain generally are confined except where they crop out. The glacial aquifers are mostly valley-fill deposits consisting of narrow, belt-like deposits scattered throughout New Jersey, most commonly north of the terminal moraine of the Wisconsinan glaciation. The bedrock aquifers include fractured shale and sandstone units of the Newark Supergroup in the Piedmont physiographic province, weathered and fractured crystalline rocks in the New England province, and sedimentary rocks in the Valley and Ridge province (fig. 1).

Types of land use and land cover in New Jersey are diverse. In the early 1970’s, the land was about 24 percent urban (16.5 percent residential and 7.5 percent nonresidential), 25 percent agricultural, and 51 percent undeveloped forests and wetlands. Some areas that were agricultural and undeveloped during the 1970’s have since been developed. Because ground water moves slowly in most aquifers in New Jersey, however, the effects of land use on ground-water quality may be more closely related to previous land use than to current land use at wells where land use has changed. Digital land-use data from the early 1970’s were used to test hypotheses concerning relations of land use to ground-water-quality data that were collected mainly during the 1980’s and 1990’s.

Pesticide use in agricultural, golf-course, and residential areas was estimated from available data collected as part of the NJDEP Pesticide Control Program (R.L. Myers, New Jersey Department of Environmental Protection, written commun., 1997). Most of the agricultural land in New Jersey is cropland, with smaller amounts of pasture and orchards. The amount of permitted active pesticidal ingredients applied in 1985 in agricultural areas was reported to be about 3.8 ´ 10⁵ kg (kilograms), at a rate of about 420 kg/km² (kilograms per square kilometer). In 1990, reported pesticide use at 202 golf courses in New Jersey was about 2.0 ´ 10⁴ kg, at a rate of about 442 kg/km². Pesticide use in residential areas is difficult to estimate. Permitted applications of pesticides by registered lawn-care companies were about 1.7 ´ 10⁵ kg in 1995; application rates are unknown. The amount of pesticides applied by homeowners is unknown.

The vulnerability of a public supply well to contamination is defined by the equation

Vulnerability = Sensitivity + Intensity.

A numerical rating model was developed to rank wells by using variables selected to describe (1) the sensitivity of the aquifer to contamination from the land surface and (2) the intensity of pesticide use in areas where the aquifer is sensitive to contamination. The sensitivity and intensity variables were coded such that larger scores for the sensitivity of the aquifer and intensity of pesticide use indicate greater vulnerability to contamination by pesticides. Individual sensitivity and intensity variables were coded from a low of 0 to a high of 5 (table 1). Possible sensitivity and intensity scores for a given well ranged from 0 to 15.

The sensitivity of an aquifer to contamination was defined by properties describing position within the ground-water flow system that control the introduction and transport of contaminants into the system. Ground water is most sensitive to contamination by pesticides in shallow parts of the outcrop area and in soils containing little organic matter, and least sensitive in deep parts of an aquifer downdip from the outcrop area. Hydrogeologic sensitivity of a well was based on three variables—distance from the outcrop area, soil organic-matter content, and depth to the top of the open interval. A numerical score of 0 was classified as low sensitivity, whereas a score of 10 or greater was classified as high sensitivity.

Because the intensity of pesticide use is not uniform among land-use types, the potential for contamination of ground water by pesticides is greater in recharge areas and areas in which the land use is predominantly agricultural or urban-residential than in discharge areas and areas in which the land is undeveloped. Pesticide-use-intensity scores for wells that are in sensitive parts of an aquifer were based on three variables-surrounding land use, distance from the nearest agricultural land, and distance from the nearest golf course. Agricultural land use was weighted higher than residential land use because most of the regulated pesticides are used for agricultural purposes. A well with a numerical score of 0 was classified in the low-intensity group, whereas a well with a score of 10 or greater was classified in the high-intensity group.

The wells were grouped into a three-by-three matrix of low, medium, and high sensitivity and low, medium, and high pesticide-use intensity. To simplify the grouping of wells, wells in the low-sensitivity and low-intensity groups were rated as low vulnerability, whereas wells in the high-sensitivity and high-intensity groups were rated as high vulnerability (fig. 1). All other wells were rated as medium vulnerability (Vowinkel et al 1996).

This model differs from other vulnerability models in that wells in the low-sensitivity group were not assigned an intensity rating because the water in the wells was not believed to be hydraulically connected to water at land surface that could be affected by the land use at the wellhead. The confining units above the open intervals of the wells are sufficiently thick to prevent the movement of contaminants from the land surface to the well or the times of travel of water from the outcrop area to the wells most likely are greater than 50 years and pesticides were not used prior to the time the water recharged the aquifer.

Water samples were collected from a stratified set of 90 randomly selected wells to test the accuracy of the results of the numerical rating model (Clawges et al in press). Samples were analyzed for 140 pesticides (Mogadati et al 1994) at Rutgers University laboratory and for nutrients at the USGS National Water Quality Laboratory (NWQL). Historical nitrate water-quality data for samples from all 640 community water-supply wells previously collected and analyzed by the USGS also were used to test the results of the model.

Of the 1,945 community water-supply wells for which data were available, the aquifer sensitivity was low for 26.3 percent, medium for 15.8 percent, and high for 57.9 percent (fig. 2A). All of the wells in the low-sensitivity and low-vulnerability groups are screened in confined aquifers in the Coastal Plain. Most wells in the bedrock or glacial aquifers were classified in the high-sensitivity group because they are in the outcrop area of the aquifer in which the well is screened and the depth to the top of the open interval of the well is less than 20 meters from the land surface. About 4.5 percent of the wells were grouped in the high-pesticide-use-intensity group (fig. 2B). The 77 wells (4 percent) classified in the high-vulnerability group (fig. 2C) typically are in outcrop areas with shallow depths to the top of the open interval and are in predominantly agricultural areas.

Pesticides were detected in water samples from 6 of 90 wells. Five pesticides—atrazine, dinoseb, simazine, metolachlor, and metalaxyl—were detected. Concentrations ranged from 0.01 to 2.2 g /L (micrograms per liter). No pesticide concentration exceeded any U.S. Environmental Protection Agency (USEPA) maximum contaminant level (MCL). Concentrations of nitrate were significantly larger (p = 0.005) in water from those wells in which a pesticide was detected than in those wells in which no pesticide was detected (fig. 3).

Pesticides were not detected (fig. 4A) and concentrations of nitrate were smallest (fig. 4B) in water from wells in the low-vulnerability group. Pesticides were detected in about 5 percent of the wells in the medium-vulnerability group and about 19 percent of the wells in the high-vulnerability group (fig. 4A). Similarly, nitrate was not detected in water from any of the wells in the low-vulnerability group (fig. 4B). The distributions of nitrate concentrations are significantly different (p < 0.001) among the three vulnerability groups. The 75^th-percentile concentration of nitrate (as N) in water from wells in the medium-vulnerability group is 2.1 mg/L (milligrams per liter) compared to 4.3 mg/L in water from wells in the high-vulnerability group.

The distributions of historical concentrations of nitrate in water from community water-supply wells were compared by vulnerability group (fig. 5). The median concentration of nitrate (as N) in water samples from wells in the low-vulnerability group was less than 0.1 mg/L and the nitrate concentration was larger than 0.5 mg/L in water from only 2 of 157 wells (1.3 percent). The 75^th-percentile concentration of nitrate was less than 2 mg/L in water from wells in the medium-vulnerability group. The median concentration of nitrate (as N) was greatest in water from wells in the high-vulnerability group (2.0 mg/L).

On the basis of the current and historical data analyses, the likelihood is small of detecting a pesticide or nitrate in community water-supply wells classified in the low-vulnerability group. The results indicate that nitrate concentration in water from a well may be a crude predictor of possible contamination of the water by pesticides.

The numerical rating model is a means for conducting a preliminary assessment of the potential vulnerability of water from community water-supply wells to contamination by pesticides. The numerical rating model has several limitations that result from the necessary simplifying assumptions made in the analysis. These model assumptions are that (1) the solubilities of all pesticides and nitrate in water are similar; (2) pesticides are applied uniformly in agricultural, residential, and golf-course areas; and (3) differences in the scales of the GIS hydrogeologic and land-use data do not affect the results of the statistical analyses.

The NJDEP used the results of the vulnerability study to evaluate strategies for monitoring requirements for pesticides. Monitoring requirements can be waived or reduced for wells that are not vulnerable to contamination. The NJDEP estimated that monitoring waivers for community water-supply wells will save water purveyors about $5 million in analytical costs annually (Vowinkel et al 1996).

Modified from Vowinkel and others, 1994, p. 505.

[>, greater than; ³ , equal to or greater than; - -, this numerical score not assigned to this variable]