Within the last decade it has been recognised that the Doppler shift as measured by SAR satellites (“Doppler Centroid Anomaly”) contains useful information about the velocity of the ocean surface in the line of sight of the radar, though at a coarser resolution than the traditional SAR roughness measurements, of the order of 5 km compared to 100 m. The retrieved surface Doppler velocity includes a contribution from the ocean surface currents, and a contribution from wind waves. This happens through a complex interplay of mechanisms such as orbital motion caused by longer waves, phase velocity of the shorter “Bragg” waves, and motion of irregular ocean surface elements originating from wave breaking. The Doppler velocity is therefore a new resource in addition to the roughness for retrieval of both ocean wind and surface current from SAR.

Recognising that retrieval of wind, waves and current from SAR is a highly coupled problem, the objectives of the INCUSAR projects are two-fold: the first part of the study focuses on a method to improve the SAR wind retrieval by taking benefit from the Doppler information. This will benefit the second part, which concerns a consistent inversion of wind, waves and the surface current.

A prerequisite for using the Doppler shift information from Envisat ASAR is to correct the given Doppler Centroid Anomaly for non-geophysical contributions from relative Earth rotation velocity and antenna mispointing errors. This is done by respectively subtracting a modeled Doppler shift from satellite attitude parameters, and by determining the instrument bias as a function of incidence angle (primarily by averaging the Doppler shift over land, which is expected to be zero).

Wind speed over the ocean is traditionally calculated from SAR imagery with empirical algorithms (e.g. CMOD) by combining the NRCS with wind direction from a numerical forecast model. Often, however, the direction from the model is not correct at smaller scales, leading to erroneous wind speed estimates. The SAR Doppler shift is however an additional source of information, which is in particular helpful in situations with sharp changes in wind direction, such as in relation to cyclones and close to the coast. A novel algorithm, based on Bayesian statistics, is implemented to take advantage of the Doppler information in the SAR wind retrieval scheme. The basic concept is to utilize all available information sources, weighted by the inverse of their uncertainty (variance). The new algorithm is seen to produce more realistic wind fields in many cases. Due to challenges with the calibration, the uncertainty of the Doppler information is quite high, about 5 Hz corresponding to some 30 cm/s. For cases with sharp wind fronts, the uncertainty of the forecast model wind is however also quite high, and in these cases the Doppler shift is proven to be an invaluable source to correct positioning of fronts.

In most cases, the Doppler velocity is dominated by contributions from wind-induced waves. However, in the general case, a contribution from surface currents is also present. By subtracting the wind contribution from the total Doppler velocity, the anomaly should reveal the surface current. For this purpose, the accuracy is however of even greater concern, as most surface currents are of the order of the accuracy of the retrieved Doppler velocity, except for very strong current systems like Gulf Stream and the Agulhas, where current signatures are frequently visible in single SAR Doppler velocity images. The accuracy is however greatly increased by averaging the Doppler velocities over time, as has been done for the North Atlantic. Climatological signatures of the persisting topography-steered currents are retrieved in fine details, in good agreement with altimetry, and in some cases with more details.

The planned inversion of consistent wind, waves and surface current will combine both SAR Doppler and roughness measurements, and will utilise a complex radar imaging model (forward model) to simulate the radar signatures for given wind, wave and current fields. An iterative approach will be attempted to converge to a consistent solution. A major challenge is however that the Doppler velocity is provided on a rather coarse resolution (~10 km) with high uncertainty (~20-30 cm/s) whereas NRCS is provided with fine resolution (~100 m) with a good accuracy (~0.5 dB). For the future, more accurate Doppler estimates are to be expected from Sentinel-1 due to dedicated design specifications for the Doppler. For Envisat ASAR, more work is still needed on the calibration issue, and in parallel more knowledge about Doppler for retrieval of surface currents can be obtained by exercises with the radar imaging model and by validating the averaged Doppler velocities.