Checking billy barr’s snow log…
Persistent slab avalanches are a leading cause of avalanche fatalities, in part because persistent weak layers (PWLs) are difficult to detect.A key risk factor is spatial variability in snow depth: shallow areas promote stronger temperature gradients, favor weak-layer formation, and increase the likelihood of human triggering, yet are often indistinguishable from the surface by eye.We present a feasibility study developing basin-wide "shallow snow" maps derived solely from historical, high-resolution Airborne Snow Observatories (ASO) lidar snow depth records for three case-study watersheds spanning contrasting snow climates: Truckee and Tuolumne, California, and East River, Colorado.We evaluated five depth-only statistics across three general approaches: sliding-window comparison to the local neighborhood (raw anomaly, normalized anomaly, and local percentile), local spatial autocorrelation (Local Moran's I), and residual analysis relative to dominant spatiotemporal patterns via low-rank singular value decomposition (SVD).We assessed the sliding-window statistics via leave-one-year-out cross-validation against matched-truth intersection-over-union (IoU) scores, Normalized anomaly proved the most promising, reproducing held-out years with IoU of roughly 0.5-0.6,well above the chance baseline (IoU ≈ 0.03-0.10),and reaching most of its achievable skill within just a few years of training flights.Local Moran's I and the SVD residual showed moderate success in qualitative comparison, but each carries a distinct limitation, an elevation bias tied to flight timing, and untested sensitivity to reconstruction rank, respectively.Both require further refinement before either is ready for the same leave-one-year-out evaluation as the sliding-window methods.These results demonstrate that historical lidar alone, without real-time data or terrain covariates, can identify terrain that is persistently prone to shallow snow, supporting further development of a shallow-snow layer for integration into backcountry mapping platforms to strengthen terrain-based decision support for persistent slab avalanches.
Items connected by shared entities, co-authorship, citations, or semantic similarity.