Search

• Traditional solar radiation modeling and time-decay snow albedo functions lead to errors in snowmelt rate and snow depletion timing for process based models. • Combining numerical weather prediction solar radiation and remotely sensed snow albedo data products improved modeled snowmelt rate and snow depletion timing relative to in situ measurements. • Snowmelt accuracy enhancements increase the potential for process based models to be adapted in local water supply forecast operations. Snow that accumulates seasonally in mountain headwaters is traditionally a vast and consistent natural reservoir, providing water as the snow melts in the spring and summer. This resource is at risk due to declining and more variable snow cover, increasing the need to accurately forecast snowmelt. The timing and magnitude of snowmelt, first order controls on downstream water resources, are primarily driven by the amount of absorbed (net) solar radiation controlled by the snow albedo. However, solar radiation and snow albedo are not commonly measured at mountain instrumentation sites despite their high degree of spatial variability. With the sparsity of observations, physically based snow models often use simplified solar radiation modeling and time-decay albedo functions, leading to errors in snowmelt rate and snow depletion timing. Here, this limitation has been addressed by combining two independent gridded solar radiation data products; 1) incoming solar radiation output from the High-Resolution Rapid Refresh (HRRR; U.S. National Weather Service) numerical weather prediction model and 2) remotely sensed snow albedo derived from the Moderate-Resolution Imaging Spectroradiometer (MODIS). The hourly HRRR and snow albedo products were used to update net solar radiation in a spatially distributed snow energy balance model over two water years (2021, 2022) in the East River Watershed, Colorado, USA. Results were assessed through time against two observation sites within watershed boundaries and spatially against snow extent from two airborne lidar flights in 2022. Updating net solar radiation improved modeling of melt rates and reduced errors in snow depletion timing from 15 – 33 days later (baseline runs) to 1 – 6 days later relative to the observation sites. The updates additionally improved spatial agreement of where snow had already been depleted from 87% to 97% during the melt season relative to lidar. These enhancements using open-access gridded products available over the continental US increase the potential for adaptation of process-based models into local water supply forecast operations to ultimately improve runoff predictions in snow dominated watersheds. [ABSTRACT FROM AUTHOR]