The application of Synthetic Aperture Radar (SAR) techniques to nadir-pointing radar altimetry offers the potential to significantly enhance Earth surface mapping. The Sentinel–3 Surface Topography Mission altimeter will be able to operate in SAR mode over the ocean, and will aim to achieve high–resolution high–accuracy altimetric mapping over the ocean in regions of high mesoscale variability and in coastal areas, and also over inland waters.

The SAMOSA project was initiated in 2007 to investigate the improvements that SAR mode altimetry can offer in measurements over ocean, coastal and inland water surfaces, developing practical implementation of new theoretical models for the SAR echo waveform as part of this process. The team, led by Satellite Oceanographic Consultants (SatOC,UK), brings in expertise from the National Oceanographic Centre (NOC) and De Montfort University (DMU) in the UK, STARLAB in Spain and DTU-Space in Denmark. The team was also strongly supported by Dr Keith Raney of Johns Hopkins University (USA) – a highly regarded international expert in the field of Delay Doppler altimetry

The project team succeeded in defining novel retracking techniques for SAR Mode (SARM) altimeter echoes over water surfaces and in evaluating the performance of SARM altimetry compared to conventional pulse-limited altimetry. The performance of SARM in terms of range retrieval accuracy was analysed by retracking simulated Cryosat data, airborne data and CryoSat-2 data, and with estimates of achievable precision of SARM through the Cramér-Rao Lower Bound (CRLB) method. In addition, the “Berry Expert System” (BEST) was applied to simulated data over complex inland water scenarios to assess SARM performance over lakes, estuarine and wetlands.

The SAMOSA project led to the definition of two new theoretical models for SAR waveforms over water. The first model (“SAMOSA1”) assumes Gaussian ocean wave statistics and a circular antenna pattern, and includes the effect of Earth curvature and antenna mispointing in the along track direction only. The second model, “SAMOSA2” is a more complex formulation that includes non-Gaussian ocean wave statistics, Earth curvature and a better representation of mispointing effects both along- and across-track. The SAMOSA2 model also comprises radial velocity effects and an elliptical antenna pattern. All SAMOSA theoretical models were implemented as SAR ocean retrackers and applied successfully to both simulated and real CryoSat-2 SAR waveforms.

The SAMOSA1 SAR ocean retracker was documented in a Detailed Processing Model (DPM) in support of the Sentinel-3 Surface Topography Mission. The DPM was based on the original SAMOSA1 formulation and did not include enhancements of the SAMOSA1 model for low sea states.

A technique was developed for the reduction of SARM data to emulate LRM and implemented in the “RDSAR” software. Waveform retracking applied to simulated Cryosat data over ocean surfaces allowed for quantitative comparison of “Low Rate Mode” (LRM - conventional altimeter approach) and “SAR mode” (SARM) over identical sea state conditions. Results showed that the RDSAR data offer the same retrieval accuracy as LRM, and an almost two-fold improvement in range retrieval with SARM compared to LRM and RDSAR, thus confirming earlier results from in the literature. However, no improvement was found in the retrieval of SWH from SARM data (when using simulated data).

The SAMOSA1 ocean retracker performance was evaluated against airborne SAR altimeter data acquired with ASIRAS during the Cryovex’2006 campaign. Even though the ASIRAS doesn't exactly replicate the geometrical parameters at the surface, over 96% of the waveforms were successfully fitted by the SAMOSA1 model when the ASIRAS data was processed to have 64 pulses per burst and a maximum look angle of 1.4 degrees.

The SAMOSA1 Enhanced model was used to successfully retrack real Cryosat-2 SAR waveform data from different oceanic regions. Results confirmed a marked, almost two-fold, improvement in range retrieval accuracy with Cryosat-2 SAR compared to Jason-2 LRM. Results also indicated that retrieval of significant wave height is at least as good for SARM as for LRM, although SARM overestimated SWH slightly compared to LRM, particularly in low sea states.

The SAMOSA2 waveform model was also implemented as a SAR ocean retracker and applied to simulated data. The SAMOSA2 waveform model being more complex, it required longer computation time than SAMOSA1. Results confirmed the findings with SAMOSA1 of an approximately two-fold improvement in range retrieval accuracy with SAR compared to LRM. Analytical solutions have been identified to speed-up the computation of SAMOSA2 and could be incorporated in future implementations.

The performance of both SAMOSA1 and SAMOSA2 models were evaluated numerically in terms of precision with Cramér-Rao Lower Bound techniques. The SAMOSA2 model was found to be more robust than SAMOSA1. The impact of the various model improvements was investigated and quantified separately in terms of their effect on the precision of range retrieval. The modification of the model to include non-Gaussian ocean statistics had the greatest effect on precision. However, the change in precision resulting from these improvements was found to be small in terms of the overall precision error budget.

Simulated LRM and SARM data were obtained also for scenarios representing inland waters, including a lake scenario, an estuarine scenario and a wetland scenario. These were processed with BEST and successful retracking of the SAR waveforms (more than 62% for the wetland, and up to 85% for the lake scenario) and recovery of small-scale topographic features was demonstrated.

In summary, the SAMOSA project successfully demonstrated the potential improvements offered by SAR mode altimetry over water surfaces. Through the development of new theoretical models for SAR waveforms over water and their application as SAR ocean altimeter retrackers to simulated and real Cryosat-2 L1B SAR data, SAMOSA confirmed earlier expectations of improvement in range retrieval accuracy and finer along-track spatial resolution.

Further work is recommended, including: a more efficient software implementation of the SAMOSA2 retracker; validation and testing of the SAMOSA SAR retrackers with a wider range of Cryosat-2 data co-located with ground truth or other reference data; updating the DPM to include (at least) the SAMOSA1 Enhancement to resolve the numerical singularities at low wave heights and allow retracking over the full range of waveform gates. It is also recommended that the SAMOSA results with simulated Cryosat data should be cross-validated against output from the Sentinel-3 mission simulator for the same scenarios, and that the SAMOSA SAR ocean retracking method are tested against other retrieval approaches in a benchmark exercise. Finally, the RDSAR technique to reduce SAR mode to Low Rate mode data should be validated using real data, to prepare for its application to Sentinel-3 data and provide continuity over the altimeter SAR/LRM mode transitions.