An Agricultural Aquifer Sensitivity Regression Model for the Prediction of Nitrate in the Glacial Drift of Southwest Van Buren County, Michigan

Richard N. Passero, Michael R. Stoline, Terri M. Smith, Julie A. Scott, Mark S. Swartz, Steve A. Miller

Nitrate application rates from commercial fertilizers in southern Van Buren County averaged 35.7 pounds/acre/year and 22 percent of the 535 well analyses compiled by the United States Geological Survey exceeded 10 milligrams/liter nitrate-nitrogen. After evaluating the hydrogeology and aquifer sensitivity of the County using the DRASTIC and AQUIPRO models, a four-township area which experienced the greatest fertilizer loading and highest nitrate levels was selected to develop an agricultural aquifer sensitivity model for use in preparation of a statewide pesticide management plan.

The study area is transected by the Valparaiso Moraine which is bordered on the southeast by an outwash plain and the northwest by outwash and ground moraine. The glacial drift ranges in thickness from less than 100 feet to greater than 500 feet. Clay and clayey sediments attain maximum thicknesses of about 80 and 60 feet respectively beneath the moraine and till plain. Correlation, analysis of variance and regression methods were chosen to screen 68 variables to conceptualize a prediction model using the natural logarithm of nitrate-nitrogen concentrations from 174 well water samples which ranged from non-detect to 78.5 milligrams/liter. Pesticides and herbicides were generally below detection limits. The step-wise regression model includes 6 independent variables (irrigated row crop , depth of well submergence, soil, drainage, recharge potential, till thickness, depth to water) accounting for 38.2 percent of the variance.

AAPG Search and Discovery Article #91020©1995 AAPG Annual Convention, Houston, Texas, May 5-8, 1995