Southern Caspian Sea wind speed retrieval from C-band Sentinel-1A SAR images

Abstract

ABSTRACT Synthetic aperture radar (SAR) imagery has shown the good capability for ocean wind speed retrieval. However, SAR wind speed retrieval above enclosed seas has been investigated rarely. In this study, the performance of geophysical model functions (GMFs) for the southern Caspian Sea wind speed retrieval from VV-polarized C-band Sentinel-1A SAR images is investigated. The used GMFs are recently developed empirical CMOD7 and C_SARMOD2, and two theoretical GMFs, which are derived in a simplified form based on the small-slope approximation (SSA) method with Hwang-Elfouhaily (H-E) and Hwang-McDaniel (H-M) composite sea spectra. Required wind directions for retrievals are obtained from wind streaks on SAR images by the local gradient (LG) method, a weather prediction model and in situ measurements. Estimations of all GMFs show good statistical agreements with predictions and in situ measurements. At 25-km resolution, all retrievals have mean biases better than ±0.7 m s−1, and root-mean-square deviations (RMSDs) and standard deviations below 2 m s−1 relative to predicted wind speeds. At 1-km resolution, using in situ measured directions, root-mean-square errors (RMSEs) of retrievals relative to in situ measured wind speeds are below 1.9 m s−1, and mean biases are better than ±0.7 m s−1. Of all the GMFs, CMOD7 has the best performance with the smallest RMSEs, and C_SARMOD2 has a good performance, except for very low winds. The theoretical GMF with H-M spectrum has a better performance than the model with H-E spectrum, and both of them can be used to retrieve Caspian Sea wind speeds from Sentinel-1A data, except for the crosswind condition with radar incident angle of about 45°.