The impact of the satellite ground track shift on the accuracy of altimetric measurements on rivers: A case study of the Sentinel-3 altimetry on the Odra/Oder River

This study is the first attempt to investigate the bias of the Sentinel-3 altimetry measurements over rivers resulting from the satellite ground track shift and the associated river slope. Altimetry-based water levels are measured at the so-called virtual stations (VS), which are defined as the area where the satellite pass intersects with a river channel. Since the ground tracks of the Sentinel-3 satellites can shift up to 1 km away from the nominal track, the calculated heights over VS correspond to different locations on the river. However, all measurements at a given VS are combined into one time series of water levels, assigned to a single reference position. Because rivers are inclined water bodies, the upstream measurements are characterized by a positive bias, while the downstream measurements reveal a negative bias. In this study, we investigate water levels measured at 16 VS of the Sentinel-3 satellites, located on the middle Odra/Oder River. To correct the measurements for the bias, we calculate the river slope by employing two separate approaches: (1) using in situ water levels, referenced to a common vertical datum (Kronsztadt'86), calculating the slope for each satellite measurement time, as well as (2) using the means of water levels from VS, calculating the slope once for the entire study period. The uncorrected water level anomalies, compared to the anomalies from the neighbouring gauges, are characterized by a mean root mean square error (RMSE) of 22 cm. The correction of water levels utilizing both approaches led to similar outputs, and resulted in a statistically significant improvement in mean accuracy by 5.64 cm and 5.74 cm for the gauge-based and VS-based approaches, respectively (i.e. over 25% improvement in mean RMSE). The percentage improvement varied from 4.99% to 53.23%, depending on VS. This study confirms the importance of the bias caused by the satellite ground track shift and the associated river slope, determines its contribution to the overall altimetry measurement error budget, and provides a fully automated approach to correct the time series for the slope effect.