Advanced Statistical Models for Large Datasets of Pesticide and Emerging Contaminants in Groundwater

Background/Need
Nitrate and atrazine are one of the most widespread groundwater contaminants in Wisconsin, and emerging contaminants such as per- and polyfluoroalkyl substances (PFAS) are of increasing regulatory concern. More than 60% of Wisconsin residents rely on groundwater for drinking water, yet decisions about where to sample, how deep to drill or remediate existing wells, and whether prohibition or best-management-practice policies are working have historically relied on simple trend analyses or two-dimensional spatial summaries. Large, publicly available groundwater monitoring datasets from the Wisconsin DNR and DATCP create an opportunity to apply modern spatial statistics and machine learning to give regulators and well owners more accurate and cost-effective guidance.

Objectives

• Develop a semiparametric point process model to characterize the spatial relationship between pesticides in groundwater and environmental covariates hypothesized to drive contamination
• Use the model from Objective 1 to characterize temporal trends of pesticides in Wisconsin groundwater.
• Extend the model in Objective 1 to account for biased, preferential spatial sampling of groundwater contaminant data through a thinned point process formulation.