PROFILING AREAS OF GROUND-WATER CONTAMINATION BY PESTICIDES IN CALIFORNIA - PHASE-II - EVALUATION AND MODIFICATION OF A STATISTICAL-MODEL

A well sampling study was conducted to evaluate an empirical approach to classifying areas of land in California as vulnerable to ground wat er contamination by pesticides (Troiano et al., 1994). Wells were samp led from sections of land that had no previous detections of pesticide residues. The sections had been classified into vulnerable soil clust ers or into a not-classified group using a procedure based on Principa l Components Analysis (PCA). Grape, citrus, and olive growing areas of Fresno and Tulare Counties were targeted, areas where pre-emergence h erbicide residues had been detected in well water. Overall, herbicide residues were detected in 75 of 176 sampled wells, a high frequency of detection in relation to results from previous targeted well sampling studies. Since residues were also detected in the not-classified grou p, the classification procedure was modified using an approach based o n Canonical Variates Analysis (CVA). More sections were classified int o vulnerable soil clusters with the CVA approach than with the PCA met hod. Data from two other explanatory variables, depth to ground water and amount of pesticide used per section, were included to illustrate how additional information can be incorporated into this approach of i dentifying vulnerable areas.