assessment of statistical downscaling methods LARS-WG & SDSM in forecast of climate parameter variation

Now most reliable tool to produce climate scenarios is use of Atmosphere-Ocean General Circulation Model outputs which stands as AOGCM. One of the using major problems of AOGCM outputs is computational large cell size of their simulation in any region. So first must their outputs has been downscaled and then they used. Present several stochastically methods for downscaling AOGCM outputs to increase their accuracy in simulate. It should be noted that Deference in downscaling methods can cause deference in simulation results. So assess accuracy of downscaling methods is very necessary in any region. Many researchers around the world to check the accuracy of various downscaling methods have focused. Results of Research study around the world indicates that simulation of climate and hydrological parameters depending on output of AOGCM models and also quality and quantity of observation data are very deferent. The aim of this study is assessment of statistical downscaling methods for precipitation and temperature include LARS-WG and SDSM in Birjand synoptic station. Observation data of Birjand synoptic station include precipitation, maximum and minimum temperature and solar watch daily on 1960-2000 were taken of province Meteorological organization. The period 1960-1990 is used for models calibration (train) and 1991-2000 for validation (test) selected. Series of climate extremes indices evaluated for observed data of synoptic station and simulated by downscaling methods on validation period. Statistical tests are used for evaluation and analysis of downscaling methods performance. The sensitivity of the methods to large-scale anomalies (correlation between observed and simulated data) and their ability to replicate the observed data distribution in the validation period are separately tested for each index by Pearson correlation and Wilcoxon signed rank tests, respectively. By analysis of results defined that between of downscaling methods there isn’t significant superiority in person correlation test. While in correlation test in both model p-value of more 50% of observation and simulation indices is most of 0.05 and they acceptable. Results of performance models in Wilcoxon test showed that performance of weather generator technic is significantly better than linear regression method. Results of this test showed that more of 90% of indices have a suitable fit in LARS-WG. Also fit of temperature indices in SDSM-DC compared with LARS-WG were very weak. results of this study showed that LARS-WG method compared with SDSM-DC method is more accurate generally. This accuracy in forecast of distribution function was more tangible.