HYDROGEOLOGIC AND WATER-QUALITY VARIABLES ASSOCIATED WITH ELEVATED ARSENIC CONCENTRATIONS
A total of 54 hydrogeologic and water-quality variables were evaluated for a statewide population of 5,033 wells sampled for arsenic. Univariate logistic regression was used to test the significance of the individual factors as indicators of elevated arsenic concentrations. These models were judged on the basis of the Likelihood Ratio and its associated probability (p-value) and the Hosmer and Lemeshow Goodness-of-Fit Test (HL). Three factors were retained in the final model that exhibited a significant correlation (95 percent confidence interval) and an HL greater than 0.05: dissolved oxygen (n = 640, HL = 0.5510), chloride (n = 3660, HL = 0.1701), and zinc (n = 3292, HL = 0.0623). The significance of these variables in the logistic regression model is consistent with previously described associations of arsenic with zinc in metallic-sulfide mineral deposits in Pennsylvania and the control of arsenic mobility by adsorption to hydrous ferric oxide (HFO). Arsenic that had been sorbed by HFO may be mobilized by anion exchange (elevated chloride) and/or by reductive dissolution of the HFO under reducing conditions (low dissolved oxygen). Thus, sampling to obtain arsenic concentration data may be warranted for groundwater that has low concentration of dissolved oxygen, elevated concentrations of chloride, and/or elevated concentrations of zinc.