A verification of numerical weather forecasts for avalanche prediction

• Published in: Cold regions science and technology 2001-12, Vol.33 (2), p.189-205
• Main Authors: Roeger, Claudia, McClung, David, Stull, Roland, Hacker, Joshua, Modzelewski, Henryk
• Format: Article
• Language: English
• Subjects: Applied meteorology; Avalanche forecasting; Earth, ocean, space; Exact sciences and technology; External geophysics; High-resolution weather models; Mountain weather in British Columbia, Canada; Numerical weather prediction; Snow. Ice. Glaciers; Weather forecast verification
• ISSN: 0165-232X; 1872-7441
• DOI: 10.1016/S0165-232X(01)00059-3

Past, current, and future meteorological conditions are key parameters for snowpack instability and hence, the risk of avalanches. For regional, computer-assisted avalanche forecasts, highly accurate weather predictions are needed. The objective of this research is to help weather and avalanche forecasters in their decision-making process based on meteorological predictions by (1) quantifying some of the many uncertainties of meteorological forecasts and (2) combining numerical weather and avalanche prediction. Case studies were used to verify and quantify output from high-resolution numerical weather prediction (NWP) models as input for avalanche forecasting models. At the University of British Columbia, two high-resolution, real-time, numerical weather forecast models are currently run every day. Their output of the fine grid spacing of 3.3 km for the Whistler/Blackcomb ski area in the British Columbia Coast Mountains, and 2 km for Kootenay Pass in the Columbia Mountains, are verified here. The forecasts are compared with surface observations of manual and automatic weather stations using standard statistical methods. The results are very good, especially regarding the mountainous terrain: all forecasted surface parameters are within range to their observed value. Precipitation rate has results in the same order of magnitude, which is very good for this variable that is very difficult to forecast. Here, the MC2 2-km grid has a better bias ratio than the MC2 10-km grid. In general, the NMS model produces comparable results even though the resolution is lower. For temperature, an error reduction as much as 50% was achieved using the post-processing Kalman-predictor correction method. With such small errors (around 0.7 K), it looks quite promising that the forecast can be used for avalanche forecast models such as at Kootenay Pass where air temperature is a primary variable for wet avalanche prediction.