Spatial and temporal analysis of climate change in the future and Comparison of SDSM, LARS-WG and artificial neural network downscaling methods (Case Study: Khuzestan Province)

Drought is a complex danger. Therefore, the study and prediction of climate change could have a significant role in the management and planning. In this study, at first rainfall and temperature data were obtained daily from the period of 1985-2010 from eight selected stations in the region. The indices during the observation period with SPEI and SPI indexes was calculated using statistical methods and their zoning map was drawn. In this study, GCM data and HADCM3 model under two scenarios A2 and B1 were used for prediction of drought. Then, the GCM large-scale data were downscaled using three methods including SDSM, LARS-WG and artificial neural networks. The results of statistical measures of performance evaluation showed that the ability the ANN model to simulate rainfall is relatively more acceptable than other models. The results of the Man-Kendall statistics for drought indexes show that the predicted values by the LARS-WG model for the SPI and SPEI indexes are consistently more negatively correlated. This can be deduced by observing the zoning map of drought indicators in Khuzestan province that in the coming periods, the mean values of the two indices have always dropped but did not change significantly in terms of spatial.