Item: IMPLEMENTATION OF SNOWPACK AND MACHINE LEARNING MODELS IN AN OPERATIONAL FORECASTING WORKFLOW
-
-
Title: IMPLEMENTATION OF SNOWPACK AND MACHINE LEARNING MODELS IN AN OPERATIONAL FORECASTING WORKFLOW
Proceedings: Proceedings, International Snow Science Workshop, Whistler, BC, Canada, 2026
Authors:
- Chad Brackelsberg [ Utah Avalanche Center, Utah, USA ]
- Paige Pagnucco [ Utah Avalanche Center, Utah, USA ]
- Travis J. Morrison [ Utah Avalanche Center, Utah, USA ]
Date: 2026-09-28
Abstract: Snowpack modeling and machine-learning–based decision-support tools offer increasing potential for operational avalanche forecasting, yet their complexity can limit practical adoption in time-constrained forecast environments. The Utah Avalanche Center (UAC) has integrated SNOWPACK, instability metrics, and danger-level models into its daily avalanche forecasting workflow using an operations-first development framework focused on usability and forecaster needs. The primary objective of this effort is to incorporate advanced model output into routine forecasting without overly increasing cognitive load or disrupting established workflows. System design was guided by four operational goals: improving forecasting efficiency, supporting forecast accuracy, providing clear and interpretable visualizations, and delivering information not readily available from field observations alone. To guide implementation, we identified more than 35 questions commonly considered by avalanche forecasters during daily forecast development, including assessment of snowpack structure, tracking temporal changes in instability, and identifying spatial trends across elevation and aspect. Model outputs and visualizations were explicitly designed to address these questions, allowing forecasters to quickly contextualize model information within their existing decision-making process. This question-driven approach supported consistent interpretation and facilitated routine use during daily operations. At the UAC, SNOWPACK and machine-learning–derived models are not intended to replace field observations or professional judgment. Instead, they function as an additional decision-support layer to help forecasters synthesize complex information, recognize emerging patterns, and prioritize limited time and field resources. Particular emphasis has been placed on transparency, consistency, and building forecasters' trust through iterative refinement and evaluation in operational settings.
Object ID: ISSW2026_P2.2.pdf
DOI: https://doi.org/10.15788/1790098929
Language of Article: English
Presenter(s): Chad Brackelsberg
Keywords: snowpack, machine learning, operational forecasting, workflow
Page Number(s): 55 - 62
-