Item: AUTOATES V3.0: AUTOMATED ATES MAPPING AT SCALE ACROSS WESTERN CANADA
-
-
Title: AUTOATES V3.0: AUTOMATED ATES MAPPING AT SCALE ACROSS WESTERN CANADA
Proceedings: Proceedings, International Snow Science Workshop, Whistler, BC, Canada, 2026
Authors:
- John Sykes [ Simon Fraser University Avalanche Research Program, Burnaby, BC, Canada ]
- Declan Knies [ Simon Fraser University Avalanche Research Program, Burnaby, BC, Canada ]
- Pascal Haegeli [ Simon Fraser University Avalanche Research Program, Burnaby, BC, Canada ] [ WSL Institute for Snow and Avalanche Research SLF, Davos, Switzerland ]
- Kristin Anthony-Malone [ Simon Fraser University Avalanche Research Program, Burnaby, BC, Canada ]
- Grant Statham [ Parks Canada, Banff, AB, Canada ]
Date: 2026-09-28
Abstract: Avalanche Terrain Exposure Scale (ATES) maps help backcountry travelers understand how exposed a piece of terrain is to avalanche hazard. Making those maps by hand is slow and expensive, so they exist for only a few high-use areas. Producing consistent maps for every avalanche forecast region in western Canada, 400,000 km², is a different problem: it removes the option of local tuning. This paper describes the working autoATES v3.0 model for that production run, and the changes from autoATES v2.0 that the domain forced. A domain-wide Sentinel-2 forest product, built from paired summer and winter scenes at 10 m spatial resolution, supplies a consistent canopy layer from the Coast Range to the Rockies. The classifier uses the area of openings in that canopy (gap area, in hectares), not average canopy percentage, as the forest variable. Treeline is estimated per forecast subregion rather than as a single elevation. Elevation comes from a global 30 m surface model, chosen for consistency over a lidar patchwork. Potential release areas and runout are each produced as a typical scenario and a more inclusive infrequent scenario, so low-angle persistent-weak-layer terrain can be represented without compromising the typical layer. The ATES classifier follows the published ATES v.2 technical model: floors for the bold defaults, then an averaging step for the remaining evidence. Typical and infrequent PRA and runout are the regional calibration surface. Canadian production maps use both scenarios and classes 0–4; the same method can rate a typical scenario alone as classes 1–4, with Simple as the default. Observed avalanches and expert maps answer different questions; neither is ground truth. Connaught Creek, mapped independently of autoATES, is the transfer test: the classifier agrees with the expert consensus at the same level that three mappers agree with each other. Maps edited from an autoATES draft are how the reference set can grow. Finished maps will be hosted by Avalanche Canada and made freely available. A best-practices document is in development. The aim is a defensible, consistent product for western Canada, not a closer fit to any one expert map.
Object ID: ISSW2026_P4.48.pdf
DOI: https://doi.org/10.15788/1790099379
Language of Article: English
Presenter(s): John Sykes
Keywords: Automated avalanche terrain mapping; Avalanche Terrain Exposure Scale (ATES); runout simulation; model validation; trip planning
Page Number(s): 2139 - 2148
-