Spatial Probability Modeling of Nitrate Contamination in the Mid-Atlantic Region

• Activity Number: 399
• Type: Contributed
• Date/Time: Thursday, August 15, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics & the Environment*
• Author(s): Kerri-Ann Kelly*+
• Affiliation(s): University of Maryland, Baltimore County
• Keywords: Logistic Regression ; Ground water ; water quality ; model selection ; test statistic

Abstract

Nitrogen appearing in the form of fertilizer can help promote a healthy environment for plants and animals. This healthy environment becomes hazardous when the nitrates enter the ground water in elevated levels, causing the ground water to be unsuitable for consumption. Existing water-quality data obtained from the U.S. Geological Survey, State, and County records were combined with USGS geographic data to create the database from which this study is established. The goal of the study is to characterize the statistical relation between ground-water quality and geographic factors to generate susceptibility and ground-water vulnerability maps on a regional and local scale.

Based on Likelihood Ratio Test and the Wald Statistic, an optimal radius around each ground water well is identified as the area of influence. The probability of the presence of an elevated nitrate level in the shallow ground water is predicted using a logistic regression model. A spatial probability map for the region depicting the probability of the nitrate concentration exceeding a certain threshold and associated confidence is generated to identify and manage the areas deemed unsafe.