Retrieval of Ocean Surface Currents and Winds Using Satellite SAR backscatter and Doppler frequency shift

Sammanfattning: Ocean surface winds and currents play an important role for weather, climate, marine life, ship navigation, oil spill drift and search and rescue. In-situ observations of the ocean are sparse and costly. Satellites provide a useful complement to these observations. Synthetic aperture radar (SAR) is particularly attractive due to its high spatial resolution and its capability to extract both sea surface winds and currents day and night and almost independent of weather. The work in this thesis involves processing of along-track interferometric SAR (ATI-SAR) data, analysis of the backscatter and Doppler frequency shift, and development of wind and current retrieval algorithms. Analysis of the Doppler frequency shift showed a systematic bias. A calibration method was proposed and implemented to correct for this bias. Doppler analysis also showed that the wave contribution to the SAR Doppler centroid often dominates over the current contribution. This wave contribution is estimated using existing theoretical and empirical Doppler models. For wind and current retrieval, two methods were developed and implemented. The first method, called the direct method, consists of retrieval of the wind speed from SAR backscatter using an empirical backscatter model. In order to retrieve the radial current, the retrieved wind speed is used to correct for the wave contribution. The current retrieval was assessed using two different (theoretical and empirical) Doppler models and wind inputs (model and SAR-derived). It was found that the results obtained by combining the Doppler empirical model with the SAR-derived wind speed were more consistent with ocean models. The second method, called Bayesian method, consists of blending the SAR observables (backscatter and Doppler shift) with an atmospheric and an oceanic model to retrieve the total wind and current vector fields. It was shown that this method yields more accurate estimates, i.e. reduces the models biases against in-situ measurements. Moreover, the method introduces small scale features, e.g. fronts and meandering, which are weakly resolved by the models. The correlation between the surface wind vectors and the SAR Doppler shift was demonstrated empirically using the Doppler shift estimated from over 300 TanDEM-X interferograms and ECMWF reanalysis wind vectors. Analysis of polarimetric data showed that theoretical models such as Bragg and composite surface models over-estimate the backscatter polarization ratio and Doppler shift polarization difference. A combination of a theoretical Doppler model and an empirical modulation transfer function was proposed. It was found that this model is more consistent with the analyzed data than the pure theoretical models. The results of this thesis will be useful for integrating SAR retrievals in ocean current products and assimilating SAR observables in the atmospheric, oceanic or coupled models. The results are also relevant for preparation studies of future satellite missions.