A snow cover numerical model has been tested with respect to operational avalanche forecasting. The site of the experiments is characterized by deep snow-packs with usually moderate temperature gradients. Particular attention was given to parameters which are critical in stability assessment. CROCUS proved itself being very efficient in modeling snow depth, density and temperature profiles. Results for the grain simulation show that satisfactory accuracy is difficult to achieve in the case of moderate gradients with an order of magnitude of -5deg/m. Improved grain simulation is obtained for gradients with greater magnitudes. Promising results are shown when using heat exchanges simulation to track surface hoar occurences. / Un modèle numéri...