|TITLE||BIAS CORRECTION OF PRECIPITATION SIMULATED BY REGIONAL CLIMATE MODEL OVER TURKEY|
|AUTHORS||Ceren Balli (1)|Yurdanur S. Unal (1)|Stefanie Vogl (2,3)|Harald Kunstmann (2,3)|
- Istanbul Technical University, Faculty of Aeronautics and Astronautics, Department of Meteorology , Istanbul, Turkey
- University of Augsburg,Institute for Geography, Regional Climate and Hydrology, Augsburg, Germany
- Karlsruhe Institute of Technology(KIT), Institute for Meteorology and Climate Research (IMK-IFU), Garmisch-Partenkirchen, Germany
|ABSTRACT||In this study, ECMWF ERA40 reanalysis data are downscaled to first 50 km and then 10 km resolutions over Turkey by regional climate model, RegCM4.3 for 1971-2000. The RegCM is coupled with two different land surface models: BATS and CLM. Coarse resolution simulations are carried out for BATS and CLM while high-resolution simulations only for BATS.
The simulated precipitation climatology of two different land surface models is compared to the CRU data set. Also, 245 station-observations of Turkish Meteorological Service are used to estimate model biases in the precipitation field. Mean Value (MV) and Quantile Mapping (QM) bias correction methods are utilized for monthly, seasonal and yearly bases. The QM method is also applied with two fitted distributions, Gamma (Gamma QM) and the best-fitted cumulative distribution functions (best-fitted QM) and the results are compared with each other. The results of these methods are tested by three quantitative validation measures such as Spearman rank Correlation, Root Mean Square Error (RMSE) and Nash-Sutcliffe Efficiency (NSE).
Generally, RegCM is good at modeling the general precipitation patterns. Mostly, the positive biases have been occurred over Turkey and the mountainous regions when compared with CRU observations during winter and spring seasons. However, the fall season has been simulated drier than the CRU climatology. The highest positive bias is estimated in the spring season and the dryness of the summer season is very well captured by two configurations.
The bias correction factors are calculated for coarse and high resolutions by considering the first 20 years of the study period, and then the constructed models are validated for the last 10 years. The results reveal that the precipitation is underpredicted over the Black Sea and Mediterranean Sea coasts and overpredicted over the Aegean Sea and inland of Turkey. Generally, CLM produces more precipitation throughout inland, whereas BATS generates more precipitation over high topography.
Although the Gamma QM is mostly preferred in the literature, the results of the best-fitted QM corrections are better than Gamma QM. The MV correction methods generate smaller RMSEs than the QM correction methods while the NSE results of the month-based MV methods show a perfect match over Turkey except Mediterranean shorelines. According to the validation measures, the month-based MV bias correction method has the best performance especially for the high resolution.|