Influence of environmental factors on the accuracy of the Sentinel-3A altimetry over Polish rivers

Satellite altimetry is a technique of measuring height. Originally developed to observe sea level dynamics, altimetry has proven its usefulness in monitoring inland waters. Over the recent years these observations became an important supplement to the classical river gauge records. Due to the improvement of the accuracy of altimetric measurements, river water levels are being used in numerous hydrological projects, aiming to calculate water storage or to predict water levels and river discharges. Despite the improving quality of altimetric data, the accuracy of river stage measurements is still in the decimetre range, an order of magnitude lower than altimetry-based sea level observations. This is due to several factors that can lead to the deterioration of altimeter readings.

Our study is the first attempt to assess the accuracy of water levels measured by the Sentinel-3A altimetry at virtual stations (intersections of a satellite ground tracks and a river channel, hereinafter abbreviated as VS) located along Polish rivers. Further, this study aims to investigate the influence of the environmental factors on the data accuracy. The study is conducted on six biggest Polish rivers (Vistula, Odra, Warta, Bug, Narew, San) which drain predominantly lowlands, and – based on width – can be classified as small and medium rivers (40–610 m in width).

In order to assess the accuracy of measurements at virtual sites, we compare water level anomalies of these readings with stages from two adjacent gauges: one downstream and one upstream a VS. In this study we used Sentinel-3A water levels from the Hydroweb database (http://hydroweb.theia-land.fr/ – last access 09.01.2022). The time span of gauge and altimetry data ranges from April 2016 to August 2019. Since the virtual sites are located up to 73 km away from the adjacent gauges (with mean distance of 20.12 km), we decided to calculate the time shift occurring between the analysed stations. Such a unification of times is based on a two-gauge relationship, calculated for each of the satellite measurements.

We found that the root mean square error ranges from 0.12 to 0.44 m, with mean of 0.22 m. The Nash–Sutcliffe efficiency (NSE) varies between 0.40 and 0.98 (with mean of 0.84) for 67 pairs of time series, out of 68 considered. We found no correlation between the accuracy of Sentinel-3A water levels and the river width, neither for the small nor medium river sections. Likewise, land cover (determined using the Corine Land Cover 2018 data) has not been identified as an environmental factor to constrain the data accuracy. However, we found that complex river channel morphology (i.e. the occurrence of sandbars) and the unfavourable geographical setting of the VS (river channel parallel to satellite ground track or its multiple crossing) occur more often at VS with lower NSE (⩽0.8).

This study confirms the usability of the Sentinel-3A altimetry over Polish rivers and identifies factors to constrain its accuracy. The research is supported by the National Science Centre, Poland, through the project no. 2020/38/E/ST10/00295. Our results were recently published in Journal of Hydrology (https://doi.org/10.1016/j.jhydrol.2021.127355).