Joint Retrieval of Wind- and Total Surface Current Vectors from TanDEM-X Bidirectional Along-Track Interferometric Data

Abstract

Direct measurement of ocean surface velocity from space with a Synthetic Aperture Radar has shown to be a promising method to observe ocean surface currents. In this thesis report a method for Total Surface Current Vector (TSCV) retrieval using an experimental Bidirectional (BiDi) Along-Track Interferometric (ATI) acquisition mode with TanDEM-X is presented. Errors of retrieval results from simulated data and from real data are studied to assess the quality of the proposed method. The available data consists of a StripMap acquisition at the coast of Tromso and a data set acquired over Novaya Zemlya.
The measurement concept relies on the ATI phase, which provides an estimate of the first moment of the Doppler spectrum associated to total surface velocity. Observing with two beams squinted as far as 13.2 degree apart in azimuth on ground, allows the Doppler velocity to be observed in line of sight of the beams. Projection to the ocean surface gives a velocity field. This Doppler velocity field consists of a Normalized Radar Cross Section (NRCS) weighted average of velocities of sea-state dependent biases such as short wind generated waves, long swell waves and underlying currents.
Assuming the surface velocity is dominated by wind generated waves and underlying currents, the method attempts to solve for TSCV simultaneously with the surface wind vector by coupling geophysical model functions (GMF) for returned Doppler Centroid (DC) and NRCS from an ocean surface shaped by wind.
For NRCS the empirical GMF XMOD2 for X-band radar is used, based on the same regression algorithm as the widely used CMOD5 in scatterometry. For DC a GMF based on statistics of the sea surface and the Kirchhoff Approximation developed by IFREMER is used.
A cost function of the wind vector is defined as the squared difference between NRCS observations and values of the GMF in both beams. The wind speed magnitude and wind direction for which this cost function is minimal provide an estimate to the local wind vector and evaluating the GMF for DC with the estimated wind vector results in a component of surface motion caused by wind generated waves. Wind wave induced surface velocities and TSCV can then be separated.
Retrieval from simulated data shows that the wind retrieval algorithm gives an ambiguous result for the wind direction. To constrain the solution ECMWF ERA-5 reanalysis wind data is added to the cost function as an additional term with a low weight factor.
Error analyses on the propagation of data errors shows success of the method relies on calibration quality that itself depends on local conditions of the acquired data. Comparison of retrieved wind using different GMF's indicates there is a high uncertainty in the models. The average of retrieved wind vector field over the image is highly similar to the lower resolution ECMWF ERA-5 wind vector data. TSCV results appear good for data with small ATI phase errors, but are dependent on the accuracy of used GMF's.