Abstract: Winter wheat is one of the major grain crops in Gansu province, so it is important to forecast its yield dynamically for agricultural production and food security.Based on daily meteorological observation data at 16 weather stations and winter wheat yield data of Gansu province from 1985 to 2013, the main meteorological factors and the key periods influencing winter wheat yield in Longzhong, Longdong and Longnan regions in Gansu province were analyzed according to principles of integral regression.Dynamic forecast models for winter wheat yield in the last ten days of March, April and May were established.The results show that the precipitation and temperature have great influence on winter wheat yield.Increased precipitation has clearly positive effect on the winter wheat yield, especially at the fallow period and reviving jointing stage of winter wheat.Warmer temperature has an obviously negative effect on winter wheat in seedling-before winter stage and jointing-booting stage, whereas warming has a positive effect on winter wheat in late winter-reviving stage in Longzhong region and Longdong region.Forecast verification test in the recent five years suggests that average accuracy rate of the dynamic yield forecasting model reaches above 96%, which can meet the demand by operational service.

Key words: Winter wheat, Integral regression, Meteorological element, Key period, Dynamic forecast