ENSEMBLE PREDICTION OF AVALANCHE HAZARD

The prediction of avalanche hazard results from an analysis of current snow conditions, the upcoming meteorological conditions and their combined impact on the future state of the snowpack. The SAFRAN – SURFEX/ISBA-Crocus – MEPRA (S2M) chain of numerical models is used by French avalanche forecasters to estimate present and future avalanche hazard over areas assumed to be meteorologically homogeneous (massifs), primarily as a function of altitude. The input provided until now to S2M in terms of meteorological forecast comes from the deterministic numerical weather prediction model ARPEGE with a lead-time of 2 days. Here we introduce an application to avalanche hazard forecast of ensemble meteorological prediction using the output of an ensemble of 35 ARPEGE forecasts, which are used to feed S2M and thus provide an ensemble of 35 different predicted snowpack conditions. A posteriori ensemble forecasts were generated and evaluated in the French Alps for the winter 2013-2014 with 4 days lead time, initialized each day at 6 UTC. Forecasts over the Pyrenees during the exceptional winter and spring 2012-2013 are also described. Results indicate that accounting for the uncertainty in meteorological forecast improves significantly the skill and the usefulness of the modeling chain, regardless of the prediction lead time. The predictability of snowpack conditions using the ensemble forecast technique remains good at a 4 day lead time. This will allow, in the future, building probabilistic estimates of simulated avalanche hazard level in support of operational avalanche hazard forecasting activities.