USGS Home
Contact USGS
Search USGS

U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2008–5025

Back to Table of Contents

Estimating the Probability of Elevated Nitrate Concentrations in Ground Water

Maps showing the probability of detecting elevated nitrate concentrations in ground water throughout Washington State were generated using the logistic regression models and a GIS (figs. 9-10). To generate the maps, a 500×500-meter grid was constructed that covered the entire State. A well was assumed to be in the center of each grid cell and the nitrate concentration in each well was assumed to represent the nitrate concentrations in ground water for the entire grid cell. For the model without hydrogeomorphic regions (fig. 9A), the amount of agricultural land use within a 4-km radius from the center of each cell was calculated; the annual average precipitation, population density, and soil drainage class were determined; a well depth of interest was selected; and probabilities of nitrate concentration exceedances were determined using equation (1) and the coefficients in table 2. The probabilities were then mapped using a GIS. The process for the model with hydrogeomorphic regions (fig. 9B) was similar, but included the hydrogeomorphic region of each cell.

Estimated probabilities of nitrate concentrations exceeding 2 mg/L in wells drilled to 145 ft below land surface (the median well depth for the calibration dataset) range from 0 to greater than 95 percent for both models with and without hydrogeomorphic regions (fig. 9). The high probabilities of elevated nitrate concentrations are most widespread in areas of eastern Washington that have a high density of agricultural land use. The high probabilities of elevated nitrate concentrations beneath the urban and agricultural areas in the Puget lowland are less widespread than in the agricultural areas of eastern Washington, but probabilities do exceed 50 percent in selected areas (fig. 10). Conversely, regions of the State with little agriculture or urban land use have less probability of exceeding a nitrate concentration of 2 mg/L. These types of maps can be generated for any depth of interest, with probabilities of elevated nitrate concentrations decreasing with increasing well depth.

In addition to maps displaying the probability of elevated nitrate concentrations, the logistic regression models can be used to generate maps showing the depth to which wells need to be drilled to have a specific probability of obtaining water with a nitrate concentration less than 2 mg/L. This can be done by selecting the probability of a specific elevated nitrate concentration, using the coefficients in table 2, and solving equation (1) for the well depth. For example, by fixing the probability of exceeding nitrate concentrations of 2 mg/L at 10 percent, wells would need to be drilled to a depth of more than 1,000 ft in some regions in order to have a 90-percent probability of obtaining ground water with nitrate concentrations less than 2 mg/L (fig. 11A). In general, wells installed in the agricultural areas of eastern Washington need to be drilled to a depth of at least 750 to 1,500 ft to have a 90-percent probability of obtaining water with nitrate concentrations less than 2 mg/L. Likewise, wells in the urban and agricultural areas of the Puget lowlands would need to be drilled to a depth of at least 400 ft (fig. 11B). Of course in some areas, it may not be possible to obtain drinkable water from a well drilled that deep so the probability of obtaining water with nitrate concentrations less than 2 mg/L would be less in a shallower well. Additionally, the inclusion of wells that are open to multiple hydrogeologic units could bias the estimated depths. In these cases, ground water with high nitrate concentrations enter the well from open intervals located at shallower depths. For example, many wells on the Columbia Plateau have wells that are only cased to the top of the basalts with very long uncased sections continuing down through the basalts. These shallow casings allow shallow ground water to enter the deeper wells. Casing depth has been shown to be preferable explanatory variable in these situations (Frans, 2000); however, casing depths were not available as part of the WDOH database.

Appropriate Uses of the Probability Maps

The probability maps developed for this study illustrate the estimated probability of nitrate concentrations exceeding 2 mg/L in ground water in Washington. The probability maps do not show actual nitrate contamination of ground water. In addition, inherent uncertainty is associated with these maps. Uncertainty in the estimates is the result of data errors in the well database and GIS-based explanatory variables and estimated error in developing the logistic regression coefficients. Higher resolution data, particularly for the soils and nitrate application data, would improve the models.

The maps in this report are intended for regional scale use only and have limitations for use at the field-scale. The maps are not appropriate at any scale larger than 1:250,000, as determined by the STATSGO soil data, which has the smallest scale of the explanatory variables used in the models. Many unaccounted for field-scale complexities affect the concentration of nitrate in ground water and in a given well. For example, the models do not account for point sources of nitrate or preferential pathways in the soil. Although a well may be installed in a region with an estimated high probability of elevated nitrate concentration, the well may in fact actually yield water with low nitrate concentrations due to complexities that cannot be represented in regional-scale models such as those developed in this study.

In addition to the significant explanatory variables of agricultural land use amounts, population density, precipitation, well depth, and soil drainage group, other variables also can have an important effect on nitrate concentrations. The absence of an explanatory variable from the model does not mean that the variable does not have an important effect on nitrate concentrations. In many cases, if two or more explanatory variables are closely correlated with each other, only one of the variables will be incorporated into the model to account for the effects of both variables. For example, soil drainage class and soil hydrologic group are well correlated, therefore, only one of the variables was included.

Back to Table of Contents