With the global climate change, drought disasters occur frequently. Drought prediction is an important content for drought disaster management, planning and management of water resource systems of a river basin. In this study, a short-term drought prediction model based on deep belief networks (DBNs) is proposed to predict the time series of different time-scale standardized precipitation index (SPI). The DBN model is applied to predict the drought time series in the Huaihe River Basin, China. Compared with BP neural network, the DBN-based drought prediction model has shown better predictive skills than the BP neural network for the different time-scale SPI. This research can improve drought prediction technology and be helpful for water resources managers and decision makers in managing drought disasters.

No abstract

Flood disaster management is an important part of flood risk assessment. A regional flood disaster risk assessment index system is established in this paper. Then principal component analysis (PCA) method and BP neural network are combined, and a regional flood disaster risk assessment of PCA-BP neural network model is established. PCA-BP neural network model analyze the loss of flood disaster about 30 China's provinces and cities in 2006 to assess the regional flood disaster risk, the results of the assessment are in line with actual situation, the flood disaster risk assessment model which is established in this paper is valid. Keywords-principal component analysist; BP neural network; flood disaster; risk assessmentI.