Evaluation of SDSM, LARS-WG, and ANN Methods in Down scaling of Temperature and Precipitation for Two Different Climates

General Circulation Models are one of the important and practical models in regional scale climate change studies. Since the simulation of climate parameters in these models is done on a large spatial and temporal scale, their output is scaled down using different methods. In this study, the results of three downscaling models SDSM, LARS-WG and ANN has been compared in the climatic parameters of daily rainfall and minimum and maximum temperature in two stations of Taybad (hot and dry) and Qouchan (cold and moderate) in Khorasan Razavi province. For this purpose, the 20-year period of rainfall and minimum and maximum temperature data of the stations between 1986-2005 was used. To compare the accuracy of the models, the indicators of coefficient R2, RMSE and MAE were used. The results showed that LARS-WG and SDSM models have the least amount of error in simulating minimum and maximum daily temperature. The average MAE of the SDSM model error for the minimum temperature parameter in the two stations of Taybad and Qouchan is 0.61 and 0.71. The error value of the mentioned parameter in the LARS-WG model is 0.65 and 0.43 in the two stations of Taybad and Qouchan. On the other hand, in the ANN model, the absolute mean of the minimum temperature error is 1.78 in Taybad station and 1.73 in Qouchan station. Finally, SDSM and LARS-WG models have higher accuracy in simulating daily temperature climatic parameters and especially the minimum temperature parameter in Taybad and Qouchan stations compared to ANN model.