Click here to download Kelsea’s paper.

Abstract

Public avalanche forecasters in Canada increasingly rely on snowpack models to augment manual observations for avalanche hazard assessments, but several studies have recently demonstrated that precipitation inputs are one of the key sources of uncertainty in snowpack simulations. To address this challenge, this study evaluates the impact of precipitation data from four different sources—the Canadian High-Resolution Deterministic Prediction System (HRDPS), an operational adjustment to HRDPS by Avalanche Canada (AvCan), the operational HRDPS-Canadian Precipitation Analysis System (CaPA), and a new experimental version of CaPA (CaPA-Exp)—on snowpack simulations at 28 weather stations across western Canada over three winter seasons (2020-2022). For each station, we compare the snowpack structure of a reference simulation constrained by daily observed snowpack height with simulations driven by the four different precipitation data sources. To make these comparisons, we compute relative differences in snowpack height, differences in the prevalence of key grain types, and differences in the number of simulated weak layers to describe the snowpack structure. Linear mixed effects regression models are then used to explore the effect of precipitation data source, season, season period, and region on these performance measures. Our results show that simulations with all four precipitation data sources overestimate HS when compared to the reference simulations, but the CaPA-Exp analysis product is generally closest to the reference. However, regional differences exist, and CaPA-Exp is not always the best choice. The results of this study can help Canadian avalanche forecasters better understand the strengths and weaknesses of different precipitation products for snowpack simulations.