The avalanche forecast regions in Canada range from 100 to 50,000, far larger than the 10 km2 covered in a typical backcountry day. This difference in scale could cause the local avalanche danger to differ from the regional bulletin. This study assessed the relationship of field observations and snowpack tests with the local avalanche danger. Data were collected over 6 winters during 425 field days. Univariate and multivariate cross-validated classification trees were created using the observations to predict the local danger. The univariate trees show the critical value of an observation that indicates elevated danger. The multivariate trees show how
the observations can be combined. These trees provide objective data that could form the framework of a decision support tool to help recreationists localize the danger. Contrary to popular belief, field observations were more valuable than stability tests for localizing the danger.