| 18 Mar 2026
Automated Avalanche Hazard Indication for Southeast Alaska
Abstract. Snow avalanches represent a significant yet poorly characterized natural hazard in Alaska, where risk assessment is limited by extreme data scarcity, rapidly changing climate conditions, and land-use policies that often inadequately account for environmental uncertainty. We address these challenges in Southeast Alaska by developing the region’s first systematic, large-scale avalanche hazard indication maps to support public safety, infrastructure planning, and land-use decision-making. To overcome sparse observational records, we developed a hybrid modeling framework that integrates downscaled reanalysis for historical baselines (1981–2010) with dynamically downscaled climate projections for mid-century conditions (2031–2060). More than 3.5 million avalanche simulations were performed using RAMMS::LSHIM driven by downscaled snow inputs. Forest landcover masks were incorporated to represent both suppression of avalanche release and vegetation-induced braking during runout, recognizing that these simplified effects remain sensitive to landcover classification accuracy and assumed release-area configurations. The resulting maps reveal a heterogeneous response of avalanche hazards to climate change. At lower elevations, hazard extents generally decrease as warming temperatures shift precipitation from snow to rain. In contrast, select high-elevation areas of northern Southeast Alaska are projected to experience increased runout, where persistently low temperatures, combined with enhanced atmospheric moisture, lead to greater maximum snowfall. Collectively, these results provide the first region-wide, climate-informed assessment of avalanche susceptibility in Southeast Alaska, establishing a critical foundation for hazard adaptation, infrastructure resilience, and future mitigation strategies across Alaska’s sub-Arctic landscapes.
Competing interests: Author Marc Christen is the founder of RAMMS, AG. Author Yves Bühler is an Editor at NHESS.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
General Comments
In this manuscript, the authors present hazard indication maps for southeast Alaska, United States. They use historical (CFSR) and projected snow depth data (NCAR CCSM v4) combined with downscaled 4km WRF projected climate data for estimates of maximum 3-day snowfall for historical and the mid-century period. They then used these data to conduct millions of simulations using RAMMS with specific forest and land cover inputs over a large domain. They provide results in several populated areas and compare with previous runout assessments and local knowledge.
The manuscript is well organized, written, and has a logical order. I am, however, a bit confused by the Results section and associated reference to Results. The authors reference Wolken et al., (2026), but the doi does not work. I understand not making data available to the public prior to publication. However, the Data Availability section references this citation, and reviewers should have access to these results.
I think the Discussion (included in Results) could be bolstered by comparisons with other avalanche-climate studies throughout North American and the rest of the world. This would provide some context to the projected results.
Input and validation data are scarce or non-existent and, as the authors acknowledge, leads to a substantial amount of uncertainty. I think the authors treat this accordingly for the most part and do not overstate their results. However, I think a more quantitative approach to comparing this study’s output to the previous studies (i.e. runout distance or extent differences) would be beneficial. This would provide some context to the influence of uncertainty in this study especially given these previous studies are the main source of validation data. Additionally, the authors mention local avalanche knowledge (Lines 49 and 254 and Figure 12 caption) as methods of validation. What does this entail? Are actual observations used? More information regarding this type of validation and the period of record knowledge of these observations would be useful for the reader.
Overall, this is a useful approach for a data sparse area, but with substantial limitations, uncertainty, and assumptions. The manuscript is suitable for the journal, but I would strongly suggest that the authors provide a statement, at least in the Conclusions, about using these hazard indication maps with caution for operational or planning purposes given the uncertainty and lack of in-situ validation data.
Specific and Technical Comments
Line 59/Section 2: There are a lot of data and processing steps in this study. I think this manuscript would benefit greatly from a figure explaining the workflow that includes what datasets are used at each step and how each informs the other and perhaps the most important point of uncertainty with each dataset. This would also help with understanding uncertainty propagation through the workflow.
Line 82: How do you account for the scenario when only part of the PRA releases in a medium or large PRA? I assume that the major change in simulation is the release depth, but how do you account for partial propagation across the PRA? Or are you simply treating full PRA release in every simulation? This probably deserves some explanation.
Line 85 and onward: First abbreviated as NLCD, but then NCLD is often used.
Lines 167-168: This is a bit confusing as written. I assume the 30-year and 300-year return periods for frequent and rare are based on 3-day maximum snowfall occurrence and that the runout is simply a function of that rather than a more sophisticated 30-year and 300-year avalanche probability, correct? Consider clarifying.
Line 187: snow depths are given in m2. Should be m.
Line 210: Why did you choose 1000 m as a buffer?
Line 228 (Sec. 2.10): Given the coastal avalanche climate, wet snow avalanches (both winter and spring) are common in southeast Alaska. How did you account for a difference between wet and dry snow avalanches or mixed flow regimes (i.e. those that start dry but entrain wet avalanche debris as they run downslope) and how this would impact runout distance?
Line 361: Suggest including a sentence about what Sykes et al. (2022) did that would improve forest cover related runout and PRA identification in your study.
Line 385: “…inherent limitations of climate modeling…” and also forest cover inaccuracies?
Lines 545 and 559: This doi does not exist. See General Comment above. Wolken, G., Fischer, E., Hendricks, M., and Wikstrom, J.: Alaska Snow Avalanche Hazard Database, Tech. rep., Alaska Division of Geological & Geophysical Surveys, https://doi.org/10.14509/32076, 2026.
Figure 10: This figure is not referenced in the text. Additionally, please include a legend that identifies what the dark green and light green represent.
Figure 11: Please add location dots for the town centers and N arrows or grid tick marks on the exterior as in other figures.
Figure 13 right: It is difficult to see the blue outlines in the dark green areas.
Figure 15: ‘light greet’ to green.
Figure 19: caption “…show that expected precipitation [phase] changes are the...”
Figure 20: The resolution is quite poor in this figure.