Searchable Index | Strategic Plan | 2000 Fact Book

NCEH Fact Sheets | NCEH Brochures

A Survey of the Quality of Water Drawn from Domestic Wells in Nine Midwest States

Introduction and Purpose

Domestic water wells supply water to 17.6% of the households in the upper midwestern states (Table 1). In the spring of 1993, flood waters covered some of the water wells in the Missouri and Mississippi river basins. River flooding can affect groundwater quality by raising the water table, altering hydraulic gradients, recharging from different areas, or flowing directly down the well casing. Many residents who tested water from their domestic well after the flood waters receded reported the presence of coliform bacteria or E. coli in these samples.

The coliform group of bacteria is recognized as a microbial indicator of drinking water quality because these bacteria are commonly found in the environment, are present in large numbers in feces, and are easily detected by simple laboratory methods. E. coli, a member of the coliform group, is found only in fecal material. The presence of coliform bacteria in a water system indicates vulnerability to contamination and ineffective disinfection whereas the presence of E. coli indicates fecal pollution. People drinking water with these bacteria are at increased risk of contracting a waterborne disease.

In addition to measuring bacteria, samples were collected for nitrate and atrazine analysis. The major sources of nitrate in groundwater include fertilizers, animal manure, seepage from septic systems, and atmospheric fallout from combustion of fossil fuel. Background levels of nitrate in ground water may reach 3 mg/L because of natural decomposition and soil bacteria. Higher nitrate levels are associated with anthropogenic activity (Mueller et al., 1995). The Environmental Protection Agency (EPA) established an MCL of 10 mg/L for nitrate-nitrogen in public water systems (EPA, 1994) because infants are particularly susceptible to nitrate and may develop methemoglobinemia (Coomley, 1945).

Triazines are organic herbicides introduced in the 1950s. These synthetic chemicals are among the most widely used and effective herbicides in the world. In the Midwest, atrazine is used seasonally to control grassy and broadleaf weeds in corn and wheat fields. The chemical is applied to the surface of the land and degrades quickly when exposed to light. However, the half-life of atrazine in soil or water is several months (EPA, 1984a). Atrazine is the most commonly found herbicide in ground and surface water because of its high use, persistence in the environment, and ability to dissolve in water. The chemical is mutagenic in bacteria and is considered a possible human carcinogen (IARC, 1991).

The purpose of the survey was to measure levels of coliform bacteria, E. coli, nitrate, and atrazine in water collected from households that are supplied water from a domestic well water system in nine midwestern states. This concern originated when many water samples from rural wells collected shortly after the 1993 midwest floods tested positive for coliform bacteria or E. coli. Public health officials from the affected states and from federal agencies met to discuss the contamination. They concluded that the available data was insufficient to characterize the nature and magnitude of the situation. They agreed to conduct a survey of the geographic distribution of chemical and bacteriological contamination of water from domestic well water systems in the affected states. The survey would collect information on the construction, maintenance, and condition of the well. To correlate health effects with contamination, participants in the survey would be asked whether they had a diarrheal episode in the 2 weeks before the water sample was collected from their house.

Methods

Any household in the nine upper midwestern states that used a domestic well to supply water for drinking, cooking, or bathing was eligible for the survey. The EPA defines a public water system as having at least 15 service connections or regularly serves an average of 25 people daily for 60 days out of the year (EPA, 1995). In this survey, a domestic well had fewer than 15 service connections and regularly served fewer than 25 people. Field personnel collected a water sample from the household closest to and within 3 miles of each intersection of a 10 mile grid overlaid on the 9 states. The grid was constructed by randomly choosing a starting point outside the 9-state region as the lower left corner (Gulf of Mexico). ArcInfo (Environmental Systems Research Inc., 1993) was the primary geographic information system (GIS) used to construct the sampling grid. This program also generated a list of the latitude and longitude of each grid intersection, a unique identification number for each intersection, and printed maps of each county showing the major rivers, roads, and railroads in the county, and the location and the unique identification number of each sampling unit (the area within a 3-mile radius of the intersect) in the county  ( Figure 1 ).

Figure 1. A county map used to locate households to be sampled in the 1994 Midwest Well Water Survey.  Households nearest to the intersection and within the circle and county sampled.

When a sampling unit included more than one county, field personnel did not enter the adjacent county to collect that sample. Most field personnel were familiar with the area in which they were assigned to collect samples. Real-estate plats, U.S. Geological Survey quadrangles, and municipal maps were also used to locate the households to be sampled. Field personnel were employed by the state agency that was conducting the survey.

A systematic geographical sampling approach was used because a list of domestic wells was not available and variables that affect water supply and quality (e.g. geology, soil type, topography, land use, etc.) are not randomly distributed. In addition, conducting a census of wells in each sampling unit would have been difficult and time-consuming.

Collection of water samples

Water samples were collected from May to November 1994. Field personnel located the household closest to the grid intersection and asked an adult resident for permission to collect a water sample. An eligible household received water from a domestic well, had at least one member who drank the water, and was within 3 miles of the intersection. In addition, the well must not have been chlorinated in the previous 4 days because chlorine that was used to disinfect the well may still be present. If the resident declined to participate or the well did not meet enlistment criteria, the field personnel proceeded to the next closest household. If no well was sampled in the designated sampling unit, field personnel proceeded to the next sampling unit. When no households with wells could be found in several sampling units within a county, the sampling unit within that county was extended to a 5-mile radius from the grid intersection.

When a household member granted permission, field personnel marked the approximate location of the sampled well on the survey map or recorded the latitude and longitude of the sampled well if geographical positioning system instruments were available. Water samples were collected from the faucet most commonly used to provide drinking water. When possible, aerators, strainers, hoses, water treatment devices, or other attachments were removed before the sample was collected. Taps were sanitized by wiping the inside and outside of the tap with a paper towel or cotton-tipped swab saturated with 100 mg/L sodium hypochlorite. The tap was opened fully for 3 to 5 minutes prior to sampling, and then the water flow was reduced during sample collection. The sample bottle cap was removed, and without rinsing, sufficient water was collected to fill four-fifths of the container. Water was collected in polyethylene bottles for bacteriologic analysis. Two milliliters of dilute sulfuric acid were added to the sample bottle for nitrate and atrazine analysis. The caps were immediately replaced without touching the interior of the cap or container. After collection, samples were placed on ice until they were delivered to the state laboratory. Microbiology testing begun within 30 hours of collection.

Duplicate samples were chosen in advance. In each state, the survey coordinator decided the rate at which duplicate samples were collected -- usually every eighth, ninth or tenth household -- and maintained this frequency throughout the state. Field surveyors collected the duplicate samples at the preselected rate. If no sample could be collected at the designated site, the sample was collected at the next available sample site.

Data Collection Form

In addition to collecting water samples, field personnel interviewed survey participants to obtain information on the construction, condition, and maintenance of the well; the potential sources of contamination; the number of people drinking water from the well; and the occurrence of diarrhea in the household (Appendix I). For most wells, a sanitary survey was performed to determine the condition of the well; the character of local geography; and the nature, distance, and location of potential pollution sources in the area.

Laboratory Analysis

Coliform Bacteria and E. coli. A 10-tube assay (Colilert, IDEXX Laboratories Inc., 1994) measured the concentration of coliform bacteria and E. coli in the water samples. In this procedure, an aliquot of the sample is placed in each of ten tubes containing nutrient broth and indicator chemicals. The broth turns yellow when coliform bacteria metabolize O-Nitrophenol-b-d-galactopyranoside and fluoresces under ultraviolet light when E. coli breaks down 4-methylumbellifery-b-d-glucuronide. The medium contains chemicals tha suppress the growth of noncoliform bacteria. The result, number of coliform bacteria or E. coli per 100 mL, is a statistical estimate of the mean density of bacteria in a water sample and is based on the number of samples testing positive. The assay had a quantitative range from 1.1 (95% confidence interval 0.0., 5.9) to 23 (95% confidence interval 8.1, 59.5) bacteria per 100 mL.

Nitrate. The colorimetric, automated, cadmium reduction method (APHA, 1992) measured nitrate concentrations as milligrams nitrate-nitrogen per liter (mg/L NO3-N). The preserved water sample was filtered and passed through a column containing granulated copper-cadmium. This step converts nitrate (NO3) tonitrite (NO2), which forms an azo dye when sulfanilamide couples with N-(1-naphthyl)-ethylenediamine dihydrochloride. The azo dye is measured colorimetrically and is proportional to the amount of nitrate in the sample. This assay had a limit of detection of 0.01 mg/L.

Atrazine. An enzyme-linked immunosorbent assay measured atrazine in the water samples (Ohmicron, 1995). This method used atrazine-selective antibodies linked to a peroxidase enzyme detector system. In the presence of atrazine, a colored product is formed that is inversely proportional to the concentration of triazines in solution. As with most immunoassays, structurally related chemicals may cross-react with the antibody. These include other triazines such as cyanazine, simazine, and terbutryn and the atrazine metabolites 6-hydroxy atrazine and, desisopropyl atrazine. This assay had a limit of detection of 0.05 ppb.

Quality Assurance

In an effort to produce data that is precise and comparable, standard protocols for sample collection and analysis were established by the laboratories conducting the water analysis. One quality control procedure involved collecting duplicate samples for every eighth to tenth well. The difference between the original and the duplicate samples for coliform bacteria, E. coli, or nitrate was not statistically significant (p = 0.14, student=s t-test). Other quality control measures used by the laboratories included standardized sample collection and transport procedures; standard solutions, reagents, and preservatives; and use of analytical reagents with the same lot number for the Colilert and the atrazine assays. Laboratories also performed routine internal quality control procedures.

Data Analysis

Data entry. State survey coordinators mailed completed data collection forms, county maps, lists of well identification numbers, and the latitude and longitude of each well, when available, to CDC. Forms were examined for completeness and logged into a program that monitored the progress of each form in the data-entry process. The latitude and longitude of each well were entered into an ArcInfo data base. The data were double-entered. Each state's well survey manager reviewed a data base of the information of the wells sampled in their state.

Contamination levels. The EPA established limits on the level of contaminants in drinking water to ensure that public water systems deliver water that is safe for human consumption. These limits are known as the maximum contaminant levels (MCLs) -- the highest allowable amount of a contaminant that a public water supply can deliver to a consumer. A violation occurs when an MCL is exceeded. The MCL is 10 mg/L for nitrate and 3 ppb for atrazine (EPA, 1994). For bacteriological monitoring, the EPA established the total coliform rule, which states that any water sample that tests positive for coliform bacteria must be analyzed for fecal coliform or E. coli. A positive test result is when coliform bacteria or E. coli concentration is at least one per 100 mL of sample. A repeat test is conducted for each positive sample and samples are collected within 24 hours of a positive test result. A violation occurs when coliform bacteria or E. coli are present in both the initial and repeat sample. While these standards pertain to public water systems, they served as guidelines for assessing the quality of water collected in this survey. Thus a water sample collected from a household served by a domestic well was considered to be contaminated if coliform bacteria or E. coli concentrations were detected, if nitrate concentrations exceeded 10 mg/L, or if atrazine levels were above 3 ppb.

Statistical analysis. Odds ratios were calculated to order to determine the strength of the association between a well feature (e.g., depth, presence of cracks in casing, pesticide use near the wellhead) and the presence of contaminants in the water samples (coliform bacteria, E. coli, nitrate, or atrazine). Results for atrazine are not reported because only 0.2% of the samples had levels that exceeded the MCL. An odds ratio less than one indicates that the well feature was associated with a lower contamination rate than the wells without that feature, an odds ratio greater than one implies that the well feature was associated with a higher contamination rate, and an odds ratio of one shows that the well feature had no association with the contamination rate. To examine the association between well construction and contamination, we chose drilled wells as the reference because they constituted the largest group and had samples with one of the lowest rates of contamination.

Epi Info version 6.0 was used for the descriptive analysis and calculation of odds ratios (Dean et al., 1994). SAS version 6.10 (SAS Institute Inc., 1991) was used to run the logistic regression to examine for associations between the analytes and well depth, age, and casing diameter. ArcInfo (Environmental Systems Research Inc., 1993) and MapInfo (MapInfo Inc., 1994) were used in the descriptive analysis of the spatial distribution of the analytes. ArcInfo was also used to examine for associations between the analytes and well location, political boundaries, bodies of water, soil type, household income, and the presence of multiple analytes in water samples.
 

Introduction and Methods  | Results  | Discussion | Acknowledgements
Summary | References  | Appendix I  | Appendix II