METAMODELING OF GROUNDWATER WELL VULNERABILITY TO CONTAMINATION FOR ELUCIDATING POTENTIAL IMPACTS OF SHALE GAS DEVELOPMENT ON WATER QUALITY

The expansion of shale gas development (SGD) in rural communities that heavily rely on shallow aquifers for drinking water is accompanied by public concern over contamination risks and possible health impacts. We previously proposed a physically based groundwater vulnerability assessment approach utilizing the concept of capture probability to better characterize the risks from SGD and circumvent the paucity of detailed data on actual contamination events. We demonstrated the method in a 190 sq km topographically controlled groundwater system in northeast Pennsylvania. This physically based approach required intensive computational resources to adequately resolve the groundwater velocity field. Thus, its application at very large geographic scales remains practically prohibitive. In the current study, we use metamodeling to reproduce the most pertinent insights from the physically based vulnerability model using spatial descriptors such as terrain and hydrological attributes, as well as measures of proximity to SGD locations. The trained metamodels exhibit high accuracy and are used to classify the vulnerability at 94 household wells distributed across ~2900 sq km that we sampled in 2018. The predicted vulnerability, alongside the observed inorganic and organic chemistry in the samples, as well as violations reports and historical groundwater quality data, provide a coherent framework for identifying sites potentially impacted by SGD. Metamodeling enables efficient physics-informed risk assessment at large geographic scales, generating critical insights for the design of adequate groundwater protection measures.