This study determines the qualities of atmospheric wind field data in comparison with wind measurements at five locations along the Black Sea coast. For this purpose, four different wind fields were obtained from three different weather centers (NCEP, NASA, and ECMWF). Three of these are reanalyzed winds (Climate Forecast System Reanalysis CFSR, Modern-Era Retrospective analysis for Research and Applications MERRA, ECMWF reanalyses ERA-Interim), and one is an operational dataset (ECMWF operational). Their performances were determined using the wind measurements from 2000 to 2014 at five coastal locations along the southern coastline of the Black Sea (Kumköy, Amasra, Sinop, Giresun, Hopa) and from 2006 to 2009 at the offshore location (Gloria) off the coast of Romania. The performances of these wind fields were determined based on statistical characteristics (mean, standard deviation and variation coefficient, etc.), the statistical error analysis for all data and for different wind speed intervals, the wind roses and the probability distributions. Additionally, long-term variations of the yearly error values (SI and bias) of wind speeds from wind data sources during 2000 - 2014 were discussed. Finally, it was concluded that the CFSR wids give the best performance at most stations. The ECMWF datasets yield better results along the western side but the CFSR wind fields have shown better performances along the eastern side of the Black Sea coast and at the Gloria offshore location.

A high-resolution SWAN wind wave hindcast model was implemented for the Sea of Marmara. For this, we focused firstly on the quality of two data sources for the wind forcing, viz., the ERA-Interim winds from the ECMWF and CFSR winds from the NOAA/NCEP. These were compared against wind measurements for 2013 collected at the Silivri offshore buoy in the north of the Sea of Marmara. A sensitivity analysis was performed to find the optimal numerical settings and wind source. This analysis showed that the CFSR winds are most suited for wave modeling in the Sea of Marmara. As the Sea of Marmara can practically be considered as deep water, we calibrated the SWAN model for different combinations of wind input and whitecapping source terms. The calibration was performed by varying the whitecapping coefficient for different combinations. The model setting giving the lowest errors and highest correlation via sensitivity analysis was determined as the calibrated model. Thirdly, the calibrated model was validated against measurements at the Silivri buoy for the years 2014, 2015 and 2016. This validation confirmed that the calibrated SWAN model with CFSR wind forcing performed better than the default settings. Lastly, the performance of the calibrated SWAN model was assessed for different wave height ranges, wind sources, annually and per season, their directional properties using wind and wave roses, their distribution function and Quantile – Quantile plots along with extreme waves. The calibrated model offers almost the same extreme waves with different recurrent periods as the measurements.