Water Resources and Power

In order to explore a reliable drought prediction method, large scale precedent climatic indices closely related to droughts were initially selected based on the identification of historical drought events in Jiangsu Province, and different predictors with 0-12 month lagged were obtained. Subsequently, these predictors were classified into three groups, which were namely the atmospheric circulation pattern group, SST-based group and the atmospheric hybrid group. These three groups of different forecast period predictors were used to build partial least squares regression(PLSR) models with 0-12 monthly forecast period. The verified model was used to quantificationally analyze the predictability of seasonal drought considering the climatic indices of different pre-drought forecast periods. The results show that the predictability of these three groups models with 0-12 monthly forecast periods had various performance. The PLSR model based on 0-month-lagged circulation pattern predictors could predict drought grades during the 1994 summer-autumn droughts well. Furthermore, the PLSR models based on SST-based predictors with 0-9 months pre-drought could indicate development of the whole 2010-2011 winter-spring drought process, and the predicted processes were steady. The thermal conditions of the Western Pacific warm pool and the Western Pacific Subtropical High pre-drought 0-9 months took the most weights among these three groups of models with different forecast periods, and the maximum weight of those indices was of PLSR model with 3 months forecast periods, which can be used as the predictable indices for seasonal drought prediction in Jiangsu Province. The research results can provide scientific basis for seasonal drought prediction and comprehensive prevention of drought disasters.