CoAuthors

Pierre Rieu (CLS, France); Jérémie Aublanc (CLS, France); Matthias Raynal (CLS, France); Ngan Tran (CLS, France); Pierre Thibaut (CLS, France); François Boy (CNES, France); Nicolas Picot (CNES, France); Franck Borde (ESA/ESTEC, The Netherlands); Constantin Mavrocordatos (ESA, The Netherlands)

By exploiting the coherence of continuously recorded pulses, the new generation of radar altimeters, such as the ones carried on board Cryosat-2 and Sentinel-3 satellites (and for the upcoming Sentinel-6), can achieve much higher spatial resolution of the ocean surface than conventional real aperture radars. The choice of the coherent processing can be focused, unfocused or in between, depending on the level of resolution that is requested by users. For unfocused SAR altimetry, the along-track resolution is typically of the order of a few hundreds of meters, whereas the fully focused SAR processing allows theoretically to produce resolution down to half meter. If lowering the spatial resolution is an asset in measuring non-ocean topography from space (ice and land regions), this is much more doubtful over ocean surfaces since it was stressed that long ocean waves affect unfocused SAR-processed data [Aouf and Phalippou, 2015]. This impact has been most noticeably evidenced for the longest waves and swell fields that are propagating in a direction parallel to the satellite track. In such swell conditions, the assumption of a Gaussian distribution of the sea surface elevations used within the ocean retracker model is no longer met, resulting in non-negligible impacts on parameter retrievals [Moreau et al., 2016]. To this, one should add possible inhomogeneity between looks within the stack caused by the surface movement during the multi-looking process (2.5s).

Given the limitations in the actual SAR-altimeter processing to cope with measurements of long ocean waves, ESTEC and CNES decided to jointly run studies with the support of CLS to test different processing approaches for tackling this issue. The LR-RMC method, originally designed by Thales Alenia Space for the analysis of its Cryosat-2 Acceptance Test Campaign [Phalippou and Demeester, 2011] and currently defined as an experimental mode for Sentinel-6, has rapidly established itself as an innovative and promising method, demonstrating its capability to limit swell impact on retrieval performances, and also to improve the measurement precision over ocean [Boy et al., 2017].

Similarly to unfocused SAR processing, the LR-RMC method coherently combines pulses from each burst to obtain decorrelated beams. The stacking approach however differs from that used in SAR operational processing: the processed-beams are range migrated with respect to the nadir beam then all the beams contained in a radar cycle (4 bursts of 64 beams) are incoherently averaged. The effective footprint is thus as large as the illuminated area in conventional altimetry, allowing to filter out small sub-mesoscale structures over ocean surfaces (< 1km). On the other hand, the number of looks is as high as in SAR mode processing, providing significant noise reduction. The LR-RMC method has the added benefit to reduce the integration time (50ms compared to 2.5s) limiting possible surface movement effects.

Preliminary analyses of this method had evidenced high sea-surface slope sensitivity affecting the range/SWH measurement accuracy [Boy et al., 2017]. Since then, this issue has been handled. The LR-RMC method has been improved by adding an along-track sea surface slope information (from S3A mean profiles or MSS model) to the slant range correction to better align beams within the stack. A thorough assessment of this method has been carried out on a large amount of data from Sentinel-3A SAR mode and compared to operational products. Results confirm the insensitivity of the LR-RMC method to long swells, and the global significant noise level reduction achieved over ocean, making it particularly adapted for the observation of small ocean scale signals (as small as the size of the real-antenna footprint ~10km).

This paper reviews briefly the method and provides a qualitative and quantitative evaluation of its performance.