ABSTRACT
Geostatistical prediction modeling required three steps: i) aggregation of measured As concentrations to reduce spatial variability with the dependent variables, and binary-coding using the WHO threshold value of 10 μg/L; ii) logistic regression to obtain the weighting coefficients for the environmental proxies; and iii) calculation of the probability of As exceeding the WHO threshold. The geospatial datasets used as independent variables in e.g. the Southeast Asia model (Winkel et al. 2008a) were sedimentary depositional environments as a proxy for aquifer conditions and soil variables as a proxy for the drainage and chemical maturity of sediments.
