Abstract

Accurate prediction of snowpack evolution and ablation is critical to supporting weather and hydrological applications. Convection-permitting modeling has been shown to well capture observed snowpack evolution over many western United States (U.S.) mountain ranges, but some significant ablation b... Show moreAccurate prediction of snowpack evolution and ablation is critical to supporting weather and hydrological applications. Convection-permitting modeling has been shown to well capture observed snowpack evolution over many western United States (U.S.) mountain ranges, but some significant ablation biases still remain. In this study, we conduct process-level snowpack analyses of a widely used convection-permitting (4-km) weather research and forecasting (WRF) modeling product (WRF4km) for the contiguous U.S. to understand the mechanisms causing its unobserved early-spring snow ablation over Utah mountains. Analyses across Utah Snowpack Telemetry (SNOTEL) sites show that the unobserved snowpack ablation during mid-February to late-March in WRF4km is driven by multiple strong melting events. The melting results from the enhanced downward sensible heat flux to snowpack and enhanced ground solar radiation absorption, with generally larger contributions from the former before early March and from the latter after early March. The enhanced downward sensible heat flux to snowpack is mainly due to the enhanced surface heat exchange coefficient induced by high surface wind speeds. The enhanced ground solar radiation absorption is driven by both enhanced surface downward solar radiation and strong melting-induced snow cover reduction that is caused by deficiencies in Noah-MP snow-related parameterizations used in WRF4km. The substantial snow cover reduction during melting decreases surface albedo and hence triggers a positive albedo feedback that further accelerates melting. Our analyses reveal possible deficiencies in WRF and Noah-MP (e.g., canopy processes and snow albedo) and shed light on future directions for model improvements. Show less