Performance Test of Weather Research and Forecasting (WRF) Model for Central Anatolia and Black Sea Regions of Turkey

Creative Commons License

Özen C., Korkmaz E., Bağış S., Toros H.

8th Atmospheric Sciences Symposium, İstanbul, Turkey, 1 - 04 November 2017, pp.435-441

  • Publication Type: Conference Paper / Full Text
  • City: İstanbul
  • Country: Turkey
  • Page Numbers: pp.435-441
  • Istanbul Technical University Affiliated: Yes


Weather forecasting is quite challenging and significant issue since atmosphere is very hard to determine due to its dynamical and nonlinear structure. Despite the complexity of atmosphere, there are several methods to overcome and achieve success in weather forecasting like statistical and numerical weather prediction methods (NWP). Furthermore, NWP models give much more realistic and accurate results since it struggles with the atmospheric equations when it’s compare to the statistical methods. In this study, WRF – ARW model with 3.8.1 edition has been used as NWP model and its performance has been tested for Turkey. Two nested domains have been used for performance analysis and while the resolution of coarser domain has been set as 9 kilometers, inner domain has been set as 3 kilometers. On the other hand, Final Operational Global Analysis (FNL) data of National Centres for Environmental Prediction (NCEP) which has 1-degree by 1-degree resolution has been used as the initial and boundary condition for downscaling purposes. Besides, model has been runned for 4 months in 2016 and months have been chosen as the middle month of four different season in order to determine and investigate the seasonal dependence of model reliability. Furthermore, automatic weather observation stations which are located in the 13 provinces where especially in the northern and middle part of Turkey has been chosen for performance analysis of model. Consequently, two meter temperature variable has been chosen in order to make statistical comparisons which has been chosen as RMSE, Standart Deviation Error and Bias methods.