Abstract: In this study, the spatiotemporal distribution, intensity and duration of low temperature in Wuwei in the east Hexi Corridor during the period of 1960-2009 were systematically analyzed using low temperature data (the lowest temperature ≤-20 ℃) observed by 5 meteorological stations. The results show that low temperature occurrence was regional, and low temperature days in high mountains and the marginal zone of north desert were obviously more than those in oases. Days and intensity of low temperature were significantly decreased with climate warming. Low temperature occurred mainly in January, February and December. Proportion of low temperature duration more than 2 days varied in a range of 41.4%-64.4% of the total, and severe low temperature occurred in low temperature process. The factors were initially elected with Press criterion and chosen with stepwise regression forecast method by selecting the ECMWF grid field data in January, February, November and December during the period of 2003-2007, the equations were developed with the optimal subset regression for forecasting the minimum temperature in January, February, November and December. The optimal significance equations for forecasting the minimum temperature were determined with the CSC double grading criterion, and the forecast equations passed through a significance test at level of  α=0.01. Back substitution fitting rate of the prognostic equations varied in a range of 70.7%-76.5%, and the forecast accuracy rate was in a range of 63.5%-68.9%. An objective and effective instruction product was provided for forecasting low temperature. Forecast marginal value and forecast rank of low temperature were determined with the highest approach principle.

Key words: low temperature, climate characteristic, ECMWF, numerical forecast, Hexi Corridor