Water Supply Prediction in Unmonitored Basins: Integrating Statistical Models and Remotely Sensed Snow Data
Accurate predictions of seasonal water supply are vital to all communities – regardless of their size, population, or location – as they are the basis for informed water resource decisions. Throughout the western U.S., predictions of total annual streamflow often rely upon spatially limited in situ snow measurements, which may not be available in all watersheds. However, previous work by the author team showed that these in situ measurements can be supplemented (or even replaced) by remotely sensed snow timing data. Initial findings for fifteen snow-dominated basins during the years 2001-2019 indicate the existence of a significant (p ≤ 0.05) predictive linear relationship between remotely sensed day of snow disappearance (DSD) and seasonal water supply, with mean DSD explaining roughly half of the variance in AMJJ total flow volume. This work expands on the spatial and temporal extents of previous research, describing the skill of these remotely sensed variables as predictors of water supply in over one hundred basins with varied watershed characteristics (elevation, SWE/P ratio, etc.) Further, we are particularly interested in the utility of remotely sensed snow disappearance in basins that lack in situ monitoring. By comparing the skill of watershed scale Monte Carlo linear regression models across monitored and unmonitored basins, this analysis provides new insight into the potential for remotely sensed data-driven models across the western U.S.