Data-Driven Approaches toward Assessing Impacts of Industrial Development on Surface Water Quality in the U.S.A.

Data describing surface water quality in the continental United States are available from the Water Quality Portal (WQP; https://www.waterqualitydata.us) of the National Water Quality Monitoring Council. A compiled data set of all surface water data includes over 1000 million records and spans over 100 years from 1900 to 2018. We are currently focused on developing new data analysis tools for assessing the spatial and temporal trends for pH in surface water samples. Some problems in assessing these surface water data are that the spatial and temporal distribution of the data are not continuous. For example, the number of pH samples reported for a given location might be only one over the entire sampling history. The data also varies spatially: for example, the variance in pH in eastern U.S.A. is larger than that of western U.S.A.. In this study, we test the hypothesis that the observed temporal and spatial trends in pH are caused largely by trends in energy consumption and production, and specifically the switching in fuel used by the power generation industry. Our research design can be summarized into two steps: (1) we predict future pH based solely on the historic record of surface water pH in the U.S. with a spatio-temporal prediction model based on machine learning; (2) we measure the changes in global and local prediction accuracy by incorporating fuel types and fuel consumptions of power plants.