Study On Drought Class Forecasting In Luanhe River Basin In The Context Of Climate Change

In the context of climate change,traditional drought forecasting model become less applicable.Meteorological droughts are teleconnected with large-scale oceanic-atmospheric patterns supported by previous literature.In this paper,a probabilistic forecasting model based on conditional distribution of multivariate normal distribution were evolved to involve two large-scale climatic indices at the same time and apply the forecasting model to 26 rain gauges of the Luanhe River basin in North China.In addition,the performance of conditional distribution model were evaluated and compared with traditional forecasting models.(1)Drought indicator calculation.First,Shapiro-Wilk normality test were used to test the normality of aggregated monthly precipitation.Then,Standardized Precipitation Index(SPI)with time scales 3,6,12 and 24 months for each gauge were calculated and analyzed and the results showed that SPI with shorter time scale responds to precipitation more quickly and,consequently,the conversion of wet phase and dry phase is more frequently.(2)Construction of drought forecasting model without climatic indices.Taking SPI as drought indicator,we constructed conditional distribution model,multivariate normal distribution model and Markov chain model and forecasted transition probabilities from a current SPI class to SPI classes 1,2 and 3 month later.Subsequently,we analyzed the impact of SPI time scale and lead time on transition probabilities calculated by conditional distribution model.In addition,a simple score method was used to evaluated and compared the forecasting accuracy of the forecasting models.(3)Selection of climatic indices.We used Pearson correlation and Shapiro-Wilk normality tests to test the correlation and multivariate normality between SPI time series and large-scale climatic indices including El Ni?o-Southern Oscillation(ENSO),(Pacific Decadal Oscillation(PDO),North Atlantic Oscillation(NAO),Arctic Oscillation(AO)and Atlantic Multidecadal Oscillation(AMO)and then took climatic indices with significant correlation and multivariate normality as climatic predictors.(4)Construction of drought forecasting model involving climatic indices.Based on climatic predictors and conditional distribution model,we constructed drought forecasting model to take the impact of climate change into account by involving large-scale climatic indices.Then,we calculated the transition probabilities between different SPI classes and analyzed the impact of climatic indices on them.A Monte Carlo approach is applied to give the significance of transition probabilities.Moreover,two climatic indices were involved into the conditional distribution model at the same time to forecast transition probabilities from a current SPI class to SPI classes 1,2 and 3 month later.The performance of the forecasting model involving climatic indices was evaluated and compared with those models without climatic indices.