Innovative validation of Sentinel-3 SAR altimetry measurements over rivers for the evaluation of hydrology-dedicated retrackers

The measurements from radar altimetry missions providing accurate Water Surface Height (WSH) estimates over continental areas at global scale are of key interest to monitor the fresh water stocks. In comparison to previous data versions, the recent generation of the ESA Sentinel-3 Level-2 hydrology thematic products has improved the performance of the WSH estimates over lakes and rivers. Over rivers, this performance is usually evaluated by comparing altimetric and in situ measurements with the standard nadir validation at the virtual station, where the satellite ground track intersects the river. However, the combined impact of the satellite orbit excursion and the river slope makes the WSH comparison difficult between the virtual station and the in situ station. In this study, we propose an innovative validation method of satellite altimetry measurements with a reduction of such impact, thus enabling a better estimate of the retracking algorithms bias over rivers. The so-called ”off-nadir” validation takes advantage of the SAR processing to select measurements in the near cross-track direction, where the SAR footprint intersects the river section monitored by the in situ station. This validation technique is applied over French rivers that are monitored by hundreds of in situ stations near Sentinel-3 ground tracks, to statistically evaluate the performance of the Offset Centre Of Gravity (OCOG) and sinc 2 retrackers over isolated narrow rivers. After applying an a priori editing on the dataset of off-nadir measurements, our results indicate that the WSH bias of the sinc 2 retracker, that is statistically computed as the median value of the WSH difference between Sentinel-3 measurements and in situ measurements, is around 4 cm. In contrast, the OCOG retracker causes a bias of around 22 cm for the specularity associated with the rivers analyzed in this study. We then statistically illustrate that defining off-nadir virtual stations rather than nadir virtual stations increases the ratio of Sentinel-3 WSH values with a small deviation from the in situ WSH values by a factor of 2. We finally show that, when considering the uncertainties of the off-nadir validation due to the along-track footprint of the Sentinel-3 measurements, the residual bias of the sinc 2 retracker reaches a value close to zero.