Climate-Resilient Snowpack Estimation Methods
Marianne Cowherd, University of California, Berkeley
A network of around 900 snow pillows — in-situ snow mass measurement instruments — located at snow telemetry (SNOTEL) stations critically inform operations by providing foundational information on Western United States (WUS) snow-water equivalent (SWE). In the 21st Century, warmer temperatures and changing atmospheric circulation will produce unprecedented changes in SWE, likely resulting in alterations in the timing, amount, and spatial patterns of snowpack across many catchments. In the future, the WUS SNOTEL network will be less sensitive to peak SWE. In this work, we explore a parameter space of data and model complexity to produce suggested frameworks for snowpack prediction in a warmer world. We explore the relative importance of model complexity and observational information for constraining SWE estimates in downscaled climate models over the WUS. We show that sufficiently complex data-driven models are likely to maintain SWE estimation skill even in future climate with 1) increased interannual variability and 2) ahistorical spatial and temporal correlations between predictors of peak SWE and the peak SWE field. In this framework, increased model complexity and additional observational datasets can compensate for loss of sensitivity to SWE that the SNOTEL network will experience. Nimble artificial intelligence-based models that incorporate partial, multi-modal SWE information can address these challenges and help ensure that peak SWE estimation in the WUS will be resilient to the future no-analog snowpack conditions.
Abstract Author(s): Marianne Cowherd, Colorado Reed, Stefan Rahimi, Utkarsh Mital, Shiheng Duan, Manuela Girotto, and Daniel Feldman