Ven Te Chow Faculty Scholar in Water Resources University of Illinois, Urbana-Champaign
Several studies have used USGS national water usage estimates as the primary data source for municipal and industrial water withdrawals in the contiguous United States (CONUS). However, the county-level spatial scale and 5-year frequency make the estimates less appropriate for analysis at a greater resolution and frequency. This study attempts to develop a dataset for municipal and industrial water uses at the sub-county level in the CONUS. We have collected water use records for the period 2011-2021 from more than 130 small and mid-sized cities (population less than 250000) within CONUS. We use a Bayesian Neural Network framework to map the relationship between annual water use and socioeconomic factors such as income, housing units, and number of employees in specific sectors (obtained from the American Community Survey), as well as climatic variables such as precipitation, temperature, and humidity (from the PRISM dataset). Using sensitivity analysis and partial derivatives obtained from the model, we investigate the variation in important factors influencing water usage across space and time. These results are used to understand spatiotemporal trends and the fundamental drivers of water use across the CONUS. Such inferences, along with the understanding of the local water use drivers, will be used to interpolate water use for any small and mid-sized cities in the CONUS.
