Â鶹ÊÓƵ

Across much of the western U.S., the prediction of seasonal water supply (defined here as April-July total streamflow volume) depends on annual snow accumulation, with snowmelt accounting for up to 70% of total runoff. Although many operational forecasts rely heavily on sparseÌýin situÌýsnow measurements, previous work by the author team found that in many cases, satellite-derived snow timing variables (e.g., the day of snow disappearance) are as, or even more, effective thanÌýin situÌýdata in predicting seasonal water supply when applied to data-driven regression models. This work expands previous spatial and temporal ranges to include approximately 100 study basins and satellite-derived data from the water years 1985-2021. This larger set of basins includes both ‘monitored’ basins — equipped withÌýin situÌýsnow monitoring stations — and ‘unmonitored’ basins — allowing the team to assess the ability of remotely sensed snow data to replace in situ measurements in data-scarce locations. This analysis considers the effect of varied discretization methodologies (i.e., dividing each basin into subdomains by aspect, clustering algorithms, etc.) on model skill. By quantifying the importance of these subdomains and their skill in forecasting seasonal water supply across the western U.S., this analysis provides insights that will guide future data-driven analyses.
Ìý