Wind Field Retrieval from Satellite Radar Systems

Wind observations are essential for determining the atmospheric flow. In particular, sea-surface wind observations are very useful for many meteorological and oceanographic applications. In this respect, most of the satellite remote-sensing radar systems can provide sea-surface wind information. This thesis reviews the current wind retrieval procedures for such systems, identifies the most significant unresolved problems, and proposes new methods to overcome such problems. In order to invert the geophysical model function (GMF), which relates the radar backscatter measurement with the wind speed and direction (unknowns), two independent measurements over the same scene (wind cell) are at least needed. The degree of independence of such measurements is given by the azimuth (view) angle separation among them. This thesis is focused on improving the wind retrieval for determined systems (two or more measurements) with poor azimuth diversity and for underdetermined systems (one single measurement). For such purpose, observations from two different radar systems, i.e., SeaWinds and SAR (Synthetic Aperture Radar), are used. The wind retrieval methods proposed in this book for determined (Multiple Solution Scheme, denoted MSS) and underdetermined (SAR Wind Retrieval Algorithm, denoted SWRA) systems are based on Bayesian methodology, that is, on maximizing the probability of obtaining the true wind given the radar measurements and the a priori wind information (often provided by numerical weather prediction models), assuming that all wind information sources contain errors. In contrast with the standard procedure for determined systems, the MSS fully uses the information obtained from inversion, which turns out to positively impact the wind retrieval when poor azimuth diversity. On the other hand, in contrast with the various algorithms used nowadays to resolve the wind vector for underdetermined systems, the SWRA assumes not only that the system can not be solved without additional information (underdetermination assumption) but also that both the algorithms and the additional information (which are combined to retrieved the wind vector) contain errors and these should be well characterized. The MSS and the SWRA give promising results, improving the wind retrieval quality as compared to the methods used up to now. Finally, a generic quality control is proposed for determined systems. In general, high-quality retrieved wind fields can be obtained from scatterometer (determined systems) measurements. However, geophysical conditions other than wind (e.g., rain, confused sea state or sea ice) can distort the radar signal and, in turn, substantially decrease the wind retrieval quality. The quality control method uses the inversion residual (which is sensitive to inconsistencies between observations and the geophysical model function that are mainly produced when conditions other than wind dominate the radar backscatter signal) to detect and reject the poor-quality retrievals. The method gives good results, minimizing the rejection of good-quality data and maximizing the rejection of poor-quality data, including rain contamination.