Dynamic risk prediction based on discriminant analysis for maize drought disaster

This study presents a discriminant analysis-based method for prediction of agriculture drought disaster risk. We selected the Chaoyang city in the Northeast China as the study area. We employed multi-scale standard precipitation index (SPI) to reflect drought hazard. We used the yield losses to indicate the drought disaster risk, which was divided into no, low, or high drought risk. We used the multi-scale SPI and drought disaster risk as the input factors for the discriminant analysis-based risk prediction model. The results showed that the model's prediction accuracy varied between 40 and 82.4 %. The accuracy of high drought disaster risk category was higher than low and no drought disaster risk category. The prediction accuracy of the milky maturity stage was highest. We use leave-one-out cross-validation method to validate the model's accuracy. And the results showed that the model validation accuracy of high drought group could reach 70.6 % in milky maturity stage. This study showed discriminant analysis is an effective and operable method for disaster risk prediction. This model can provide timely information for decision makers to make effective measures for drought disaster management and to reduce the drought effects to yields at the minimum level.