Comparison of SDSM and LARS-WG models for simulation of meteorological variables in northwest of Iran

In this study, the performance of two statistical weather generators the Statistical Downscaling Model (SDSM) and LARS-WG in simulating daily values of rainfall, maximum and minimum temperatures in northwest of Iran is compared. The study network was consisting 12 weather stations with minuim 40 years of daily temperature and precipitation data. The 1961-1990 period was used the baseline for models evaluation. In this study, two non-parametric tests of correlation and Mann Whitney were used in monthly basis for comparisons. Root mean square error (RMSE) was used to compare the accuracy of two models. The results showed that the skill of both model in simulating minimum and maximum temperature data is similar,but the number of month with higher correlation was more in case of using SDSM.For precipitation data, the mean RMSE values of SDSM and LARS-WG models simulations were 26.5 and 0.32 mm, respectively which indicates higher accuracy SDSM. No significant differences between the observed and simulated data were found using the Mann-Whitney nonparametric test.The number of month with significance correlation with slightly more in SDSM comparing the other model.