A damped Newton variational inversion method for SAR wind retrieval

Abstract

AbstractThe variational inversion for synthetic aperture radar (SAR) wind retrieval can take all sources' errors into account, but its iterative computation is very time consuming. For the wind vectors, (u, v) components are commonly used for variational inversion, but they are not intuitive for practical applications. In this paper, we modify the decomposition of wind vectors in the cost function into wind speed and wind direction and adopt the damped Newton method (DNVAR) to solve the cost function. Experimental results on simulated data show that DNVAR can effectively reduce background wind vector errors. Additionally, the average number of iterations is reduced drastically compared to prior arts. Furthermore, a detailed comparison between direct SAR wind retrieval (DIRECT) and DNVAR is performed. Simulations reveal that the DNVAR errors are smaller than the background wind vector errors in all considered cases. Thus, DNVAR could be employed to retrieve SAR sea surface wind. For practical applications, when the background wind speed is within moderate and high wind speed range, DNVAR has higher accuracy and is thus preferred. Otherwise, both DNVAR and DIRECT are feasible, considering the unknown actual errors of both background wind vectors and geophysical model function. Experimental results on Envisat/advanced synthetic aperture radar data show that the wind speed accuracy of DIRECT is largely affected by the background wind direction errors, but DNVAR can reduce the wind direction errors with minor effect on the wind speed errors in comparison with the background wind errors.