Sensitivity analysis of regional climate model COSMO-CLM over Turkey

The aim of this thesis is to evaluate the performance of the Regional Climate Model COSMO-CLM in simulating spatial characteristics of surface temperature and precipitation over Turkey. Sensitivity of COSMO-CLM is tested by using nine different configurations based on the physical parameterizations, data exchanges and nudging technique. The results are compared with the temperature and precipitation observations taken from Turkish State Meteorological Service (TSMS). The CCLM-Starter Package cclm-sp_2.4 version was used in this study in order to exchange data sets of aerosol optical depth and surface albedo. The ERA-Interim reanalysis data were driven by climate mode of COSMO (COSMO-CLM) with 0.11° (~10-12 km) spatial resolution over 200x100 grid points domain during the period between August 1st, 2007 and December 31th, 2009. The first five months of model runs were assumed as enough time for the spin-up of the model since the model is initiated beginning of the autumn season when the soil moisture is near minimum in the study domain. For both temperature and precipitation variables, 212 common observation stations were used in this study to be used to investigate the model performance. Two extreme climate years in the last decade according to the annual total precipitation data of TSMS, 2008 and 2009, represent dry and wet years during two years simulation period, respectively. Although the temperature averages of dry and wet years are similar, the annual total precipitation averages of all 212 stations over Turkey are 430 mm and 715 mm for 2008 and 2009, respectively. Taking this into consideration, 2008 and 2009 were selected as the dry and wet years. In total, seven COSMO-CLM runs were performed with different options in terms of albedo, aerosol, and vertical variation of critical humidity. Moreover, two runs with spectral nudging method were completed to improve model performance. For analysis of results, the domain was divided in seven regions according to climate zones of Turkey that are redefined by Ünal et al. (2003). Because of the complex topography of Turkey, the elevations of observation stations and model are different at common usage points. Also, this difference depends on the resolution of the simulations. The elevation of the model is higher than the stations' on complex topography over major part of Turkey. Increasing resolution ameliorates the altitudinal representation of the model especially over complex topographical regions. This sensitivity analysis project focuses on 2m temperature and total precipitation in order to present the main climate features over selected region. The sensitivity simulations were compared with reference configuration, observation data and MSWEP precipitation dataset for both dry year and wet year, separately. Model biases and differences between sensitivity configurations and reference run were calculated and presented. The eight sensitivity configurations produce 1-2°C higher results with respect to reference configuration over major part of Turkey during the dry and the wet years, except B1 and C1. Model simulations have cold biases over Turkey domain against observation data for both extreme years. COSMO-CLM simulated higher precipitation amount in the wet year with respect to the dry year. On the other hand, the negative biases of precipitation were observed over major part of Turkey against observation data.