Correcting altimetry measurements on rivers for the satellite ground track shift bias – a case study of the Sentinel-3 altimetry on the Odra/Oder River

Satellite altimetry is a technique for measuring height. Originally designed to observe sea level dynamics, it is now also used to monitor inland waters. Using this technique, water levels (WL) are measured at so-called virtual stations (VS), which are defined as areas where satellite ground tracks intersect with the river. One of the assumptions of hydrological analyses based on altimetric data is that a satellite repeatedly flies over exactly the same place and measures WL. However, due to orbit perturbations, the ground track of a given satellite may be shifted by +/- 1 km, and thus the altimetric measurements on a given VS are carried out at different places on the river at different moments of observations. Since rivers are inclined water bodies, measurements taken upstream of the center of a VS have a positive bias, while measurements taken downstream reveal a negative bias.

To correct altimetric measurements for the error described above, it is necessary to calculate the distance of a given measurement from the central point of a VS and to calculate the slope of the studied river section. In this paper, we present two separate approaches to determine the slope: (1) using WL from two adjacent gauges referenced to a common vertical datum (Kronsztadt’86), which allows the determination of river slope at each satellite measurement time (gauge-based approach), as well as (2) using mean water levels from two adjacent VS, which results in one river slope value for the entire study period (VS-based approach). Both approaches resulted in similar river slopes, ranging from 24 cm/km to 30 cm/km. To verify the effectiveness of the proposed method, we consider WL from 16 VS of the Sentinel-3 satellites located on the middle Odra/Oder River (W Poland) and calculated using a modified DAHITI approach (https://dahiti.dgfi.tum.de/en/, last access: 29/12/2022). Finally, three datasets are obtained (WL without the river slope bias correction, WL corrected with the gauge-based slopes and WL corrected with the VS-based slopes), and each of them is compared to water level anomalies from neighboring gauges.

The uncorrected WL time series reveal mean root mean squared error (RMSE) of 22 cm. Both corrections lead to a similar statistically significant improvement by more than 25%, reducing the mean RMSE by 5.64 cm and 5.74 cm for the gauge-based approach and the VS-based approach, respectively. Only on one VS the correction slightly increases the RMSE (by less than 1 cm). In the remaining stations the improvement ranges from 0.7 cm to 13.4 cm, which is a percentage change from 4.99% to 53.23%. The proposed correction is especially recommended for altimetry-based WL of mountain rivers where the river slope bias is usually greater due to higher river slopes. It should also be mentioned that the VS-based approach utilizes only satellite data, therefore it can be applied globally, with no need for in situ observations. The research is supported by the National Science Centre, Poland, through the project no. 2020/38/E/ST10/00295. Our results have recently been published in Journal of Hydrology (https://doi.org/10.1016/j.jhydrol.2022.128761).