Probability of detecting elevated concentrations of nitrate in ground water in a six-county area of south-central Idaho

A probability map constructed for this study
identified several areas in a six-county region of
south-central Idaho with high probabilities of
detecting elevated concentrations (greater than
2 milligrams per liter) of nitrate. An increasing
proportion of Idaho’s ground water being used
for drinking water and large increases in the inputs
of nitrogen to ground water in Cassia, Gooding,
Jerome, Lincoln, Minidoka, and Twin Falls Counties
have prompted concerns about the quality of
the resource. The probability map was constructed
to assist regulatory and resource agencies in managing
land use and protecting water resources.
To construct the probability map, hydrogeologic
and anthropogenic data were integrated with
ground-water quality data in a geographic information
system. The resulting data set contained
land use, geology, precipitation, soil characteristics,
depth to ground water, nitrogen input, and
ground-water velocity information for each of the
1,365 samples collected from 1991 to 2001. Logistic
regression analysis was used to determine the
most statistically significant variables related to
the detection of elevated nitrate concentrations.
The resulting multivariate probability model
showed that ground-water velocity, nitrogen input,
precipitation, soil drainage, land use, and depth to
ground water were significantly correlated with
elevated nitrate concentrations. A subset of the
water-quality data set was used to verify these
results. Linear regression of the percentage of predicted
probabilities of elevated nitrate concentrations
and the actual percentage of elevated nitrate
concentrations with the model data set and the verification
data set both showed good correlations:
r-squared values were 0.96 and 0.97, respectively.
Statistical comparisons of both data sets showed
that ground-water samples containing elevated
nitrate concentrations had significantly higher
probabilities of detection (p < 0.001) than samples
without elevated nitrate concentrations. On the
basis of these results, a map identifying the probability
of detecting elevated nitrate concentrations
was constructed. High-probability areas on the
map coincided with regions of agricultural land
use and high nitrogen input, except in southern
Gooding County and western Jerome County. In
these areas, high ground-water velocities representing
a predominance of regional ground water
resulted in a low probability of detecting elevated
nitrate concentrations. Areas of poor prediction
tended to be congregated along the transition zone
between high and low ground-water velocities in
Jerome and Gooding Counties, indicating a mix of
regional and recently recharged ground water.