Validations of T639 model forecast skill for severe snowfall events in the warm sector in northern Xinjiang

Using the precipitation products from T639 model, we have conducted an analysis and validation on the forecast skill for 15 severe snowfall events in the warm sector in the Altai region of northern Xinjiang between November the current year and March next year from 2009 to 2014. The results indicate that the prediction performance of T639 model to snowstorm events in Altai region is better than that to heavy snow. The forecast accuracy is the highest within forecast time length of 12-24 hours. For 60h forecast, the forecast accuracy reduce, the missing alarm does not present an obvious change, and the false alarm increases obviously with the extension of time length. Furthermore, the false alarm is obviously higher than the missing alarm. The forecast accuracy of heavy snow (snowstorm) at the automatic weather stations (AWSs) with large frequency of heavy snowfall in the warm sector is higher than that of at the AWSs with low frequency of heavy snowfall. The forecast of T639 model to severe snowfall in the warm sector presents significantly a large systematic bias. The precipitation forecast products from T639 model have a certain reference value on the severe snowfall in the warm sector in the Altai region within the time length of 48 hours. The precipitation products from T639 model can seldom forecast the shortwave low trough type severe snowfall in the warm sector, while they can often forecast the low vortex center type. Owing to a forecast error in water vapor flux divergence and vertical velocity, the precipitation products from T639 model present a different forecast skill in the Altai region.