Integration of a Contaminant Source Land Use Regression Model in the Bayesian Maximum Entropy Spatiotemporal Geostatistical Estimation of Groundwater Tetrachloroethylene Across North Carolina

The assessment of groundwater tetrachloroethylene (PCE or PERC) exposure across North Carolina is currently hindered due to limited statewide spatiotemporal contaminant maps. In this study we incorporate data from multiple sources to create estimation maps of groundwater PCE. A land use regression (LUR) mean trend model was developed as a function of exponentially decaying contribution from contaminant sources in North Carolina. This mean trend model was integrated in a Bayesian Maximum Entropy (BME) framework to produce informative space/time (S/T) maps. We compare our method with standard geostatistical methods (i.e. kriging and BME with constant mean trends) and find a 25 % reduction in cross-validation mean square error. Our results suggest that dry cleaning and hazardous waste generator sites influence groundwater at distances of 1 km and 800 m respectively. This work introduces a novel integrated LUR and BME approach which produces accurate visual representations of PCE exposure across North Carolina.