ANNUAL ESTIMATES OF EVAPOTRANSPIRATION, RUNOFF, AND RECHARGE FROM REMOTE SENSING, CLIMATE, AND STREAMFLOW DATA FOR THE CONTIGUOUS US

At present, there are no empirical high resolution, continental scale estimates of evapotranspiration, recharge and runoff for the contiguous US (CONUS). Having access to this data would allow for increased accuracy in large scale hydrologic and biogeochemical models and be a valuable resource to water resource policy makers. Here we synthesize several existing geospatial and in-situ datasets to create data-driven annual estimates for the CONUS at a ~1 km² spatial scale, averaged over the 2000-2013 timespan. These maps of estimates were co-developed and constrained by streamflow and precipitation data to ensure a closed water budget. The ET estimates are generated from an empirically-defined function of Landsat-derived land cover (USGS NLCD), precipitation, temperature, and daily average temperature range (PRISM), and calibrated to long-term streamflow data collected from 679 USGS-gaged watersheds across the CONUS. The runoff estimates are generated from a separate regression as a function of precipitation, surficial geology and soil saturated hydraulic conductivity, and are calibrated to runoff data generated by a hydrograph separation program (USGS-PART) run on data from 1509 watersheds. Recharge is then calculated as the remainder of streamflow minus runoff. The calibration of these equations to data from a variety of regional and climatic settings, and their subsequent broader application to datasets that are continuous across the country, allows us to produce estimates for these water budget components that extend across data sparse regions. The ET estimates are tested against independent measurements from AmeriFlux tower data from across the CONUS, and the recharge estimates are tested against recharge rates calculated from groundwater age dates from shallow wells, also across the CONUS. Both comparisons to field data show good agreement. These estimates of the closed water budget components of evapotranspiration, recharge, and runoff are expected to be reliable to first order, because they are derived from, calibrated to, and tested against measurements.