Item: SNOWPACK MODEL SENSITIVITY TO OBSERVATIONAL AND WEATHER MODEL INPUTS
-
-
Title: SNOWPACK MODEL SENSITIVITY TO OBSERVATIONAL AND WEATHER MODEL INPUTS
Proceedings: Proceedings, International Snow Science Workshop, Whistler, BC, Canada, 2026
Authors:
- Ryan Szczerbinski [ Utah Avalanche Center, Salt Lake City, UT, USA ]
- Travis J. Morrison [ Utah Avalanche Center, Salt Lake City, UT, USA ]
- McKinley Talty [ Utah Avalanche Center, Salt Lake City, UT, USA ]
- Jim Steenburgh [ University of Utah, Salt Lake City, UT, USA ]
Date: 2026-09-28
Abstract: The SNOWPACK model has shown skill in predicting the formation and evolution of various layers within the seasonal snowpack, including instances of avalanche-conducive properties which have supported human forecasters. While ground truth atmospheric and snowpack observations are ideal for SNOWPACK inputs, weather stations typically experience some intermittent failure, leading to gaps in the observational records. Additionally, they are expensive to install and maintain, and only provide point information, which has limited relevance to broad mountainous regions. In response, some agencies are turning to Numerical Weather Prediction (NWP) model data to entirely drive SNOWPACK over broader spatial scales. While this method improves spatial coverage, it comes at the cost of accuracy. Little Cottonwood Canyon's Atwater (ATH20) snow study site provides a continuous observational record, which enables an opportunity to study the sensitivity of SNOWPACK results to observed versus NWP inputs. ATH20 site observations and raw and downscaled values from NOAA's 3-km resolution High-Resolution Rapid Refresh (HRRR) model were each used as inputs to SNOWPACK over the 2025 snow year, as well as various observation-HRRR combinations. Results were validated against manual snow profile observations by quantifying profile similarities with dynamic time warping (DTW). HRRR downscaling improved the rate of SNOWPACK convergence and had a better DTW result than the raw HRRR data compared to ATH20-driven SNOWPACK. All SNOWPACK runs including the fully ATH20-driven performed similarly against the manual snow profiles, indicating a need for manual profile data collection tailored to numerical model validation. We conclude by discussing the organization's future model validation and operational forecasting intentions.
Object ID: ISSW2026_P4.12.pdf
DOI: https://doi.org/10.15788/1790099265
Language of Article: English
Presenter(s): Travis Morrison
Keywords: Snowpack and Weather Modeling, Observation-Model Comparisons, Snowpack Properties, Weather Model Downscaling
Page Number(s): 556 - 562
-