MINIMIZING "FALSE-STABLE" STABILITY TEST RESULTS: WHY DIGGING MORE SNOWPITS IS A GOOD IDEA

The worst nightmare for an avalanche worker is to assess an unstable slope as stable since the consequence of such an assessment is that you, your clients or the public could be caught in an avalanche. Thus, a primary goal in avalanche forecasting is to minimize such "false-stable" errors. In this paper we analyze the first season of data from the SnowPilot database. Starting with nearly 1,000 snowpits and 3,500 stability tests, we use stability test scores, shear quality, and weak layer depth to identify what we term the "critical weak layer" in each pit. We also divide the pits into "stable" and "unstable" categories based on the assessed snow stability and observations of obvious signs of instability (collapsing, cracking and recent avalanche activity). This filtering leaves us with 289 compression, rutschblock and stuffblock stability tests that fractured on the critical weak layer on unstable slopes. Of those 289 tests, 38 of them (13%) presented "false-stable" results, which we define as CT21 or greater, RB5 or greater, or SB drop heights 40 cm or greater. If we include shear quality and consider strong test results with a Q1 shear to be unstable, we decrease our false-stable cases to around 9% of the total. This implies that - if we use only stability test results - around 1 in 10 times we assess unstable slopes we will conclude that it is stable, which is unacceptably high. Recently spatial variability research has led some to argue that digging snowpits is unnecessary or futile, but we believe our data reinforce the idea that the key to analyzing snow stability lies in digging more rather than fewer pits, and using a holistic approach that considers much more than simple stability test results. Though our dataset is limited, it suggests that digging multiple pits might be an effective strategy for minimizing false-stable situations. In fact, having stability tests and associated shear quality from two different, but representative locations on the slope might decrease the chance of a false-stable error from around 10% to closer to 1%.