Comparing the Performance of the Dynamic-Statistical Hybrid Method with the Dynamic Method for Downscaling of CMIP5 Precipitation Data

One of the most important challenges of hydrological studies is the projection of precipitation variability in a basin in future periods. Atmospheric Ocean General Circulation Models (AOGCMs) can project this variability but in large-scale regions. Various versions of these models have been released, one of which is the CMIP5 series with about 40 dynamic models. In this study, the CCSM4 climate model has been used. The data output of this model is in 1 × 1 (Latitude× longitude) degree. To downscale the precipitation data of this model, the hybrid of two dynamic (WRF1) and statistical methods has been proposed. The study area is the Poldokhtar subbasin in the ​​Karkheh basin and rainfall data between 1996-2005 have been used. First, the precipitation data were downscaled by the WRF model from the range of 1 × 1 degree to 9 × 9 km, and then these data were converted to the range of 3 × 3 km separately by two methods a) hybrid and b) WRF. The results showed that the dynamic downscaling by the WRF model in the range of 9 × 9 km quite shows the precipitation fluctuations during the statistical period. In the 3 × 3 km range, the hybrid method performed better than the WRF method. Finally the results showed that downscaling of precipitation in this basin using both methods is associated with underestimation and should be accompanied by an bias correction method. As a result, by reducing the time and cost of dynamical downscaling of climate models, the findings of this paper can be of great help in making decisions about water resources management in the wet and dry years of the coming decades of a basin.