Spatial calibration of an unstructured SWAN model forced with CFSR and ERA5 winds for the Black and Azov Seas

This study presents the first spatial calibration of SWAN model physical parameters based on satellite observations for the Black and Azov Seas. Currently, the wind-wave model performance is unknown in several coastal and offshore areas due to the gap in wave buoy measurements. The development and increase of altimetry satellites allow collecting a large number of observations in a reasonable time, and thus to calibrate spatially wave models. Therefore, the present study aims to determine the best physical parameterization of the unstructured SWAN model based on satellite data for coarse domain over Black and Azov Seas and for local domain implementations. For this purpose, the significant wave height (SWH) was simulated using 44 physical settings for both ERA5 and CFSR winds. The spatial calibration is based on observations from seven altimetry satellites (Jason-3, Sentinel-3A, Sentinel-3B, Cryosat-2, SARAL/AltiKa, CFOSAT, and Hai Yang-2B). The spatial calibration and evaluation of the wind-wave model performance show an informative variation in the wave model performance, which can depend on several factors such as coastal morphology and the wind source accuracy and the physical setting of the wind-wave model. It raises several findings concerning the spatial sensitivity of the SWAN model to the used wind field, the wind and whitecapping source terms, whitecapping coefficient ( Cds ), and the windscaling tunable parameter, and the wind growth formulates using the ST6 package. Thus, this study allows observing the spatial response of the SWAN model as a function of physical parameterizations. The SWAN model calibration improved the simulation accuracy considerably overall Black and Azov Sea areas, using both CFSR and ERA5. The SWAN model using the ERA5 winds has provided a higher correlation and better accuracy in large part of the seas. It is highly recommended for the whole Black and Azov Seas to define the wind input term and whitecapping term based on ( Komen et al., 1984 ), with a C ds coefficient of 0.8E-5. Considering the spatial error statistics, the same finding was obtained when the model results were compared with Utrish wave buoy measurements in the northeastern part of the Black Sea. Therefore, for local nested model implementation, it is also recommended to define the optimal configuration based on the error statistics mapped in the present paper.