The present paper applies Synthetic Aperture Radar (SAR) based on Local Gradient-Modified (LG-Mod) algorithm to retrieve wind directions from Sentinel-1 data in the Camargue and the Wadden Sea protected coastal areas. Wind speeds are estimated through the inversion of the C-band MODel 5.N (CMOD5.N) backscattering model. Both Interferometric Wide Swath (IW) and Extra Wide Swath (EW) Level 1 products were evaluated for wind fields retrieval at high (5 km) and medium (12.5 km) output spatial resolutions. SSW fields from Sentinel-1 were compared with Numerical Weather Prediction (NWP) models and in situ data. Exploitation of the LG-Mod provided wind direction with a related marginal error parameter (i.e., ME αROI ) which proved useful for selecting the optimal input pixel size of SAR data processing. When compared to in situ data, the selection of the optimal pixel size reduced the Root Mean Squared Error (RMSE) values of LG-Mod wind directions up to 7° and about 45° for Wadden Sea and the Camargue site, respectively. In turn, such reduction provided a decrease of the wind speed RMSE values up to 0.7 m/s and 2.1 m/s, for Wadden Sea and the Camargue site, respectively. In addition, the LG-Mod gave better performance than the global NWP model European Centre for Medium-Range Weather Forecasts (ECMWF) in estimation of wind direction, at 12.5 km output spatial resolution, for both sites. The ME αROI exploitation in the directional analysis of IW and EW products evidenced that at high resolution (5 km) the percentage of reliable wind directions from IW images (84.5%) resulted much larger than that obtained from EW images (30.1%). At medium resolution (12.5 km) instead, the percentage values resulted quite close to each other (99.2% and 86.3%, respectively). IW images proved optimal for high resolution SSW retrieval, whereas EW images suitable for medium resolution. With respect to NWP models, the spectral analysis confirmed the suitability of Sentinel-1 to represent the local wind fields spatial variability in coastal areas, at both high and medium output resolution. Our findings suggest that the combination of the LG-Mod algorithm with NWP models could better resolve spatially wind patterns in complex coastal areas.
Rana, F. M., Adamo, M., Lucas, R., & Blonda, P. (2019). Sea surface wind retrieval in coastal areas by means of Sentinel-1 and numerical weather prediction model data. Remote Sensing of Environment, 225, 379–391. https://doi.org/10.1016/j.rse.2019.03.019