Validation of Sentinel-6MF based lake levels – An assessment with in situ data and other satellite altimetry data

Abstract

The importance of satellite altimetry for monitoring inland water bodies is increasing, with the increasing need for fresh water. With the addition of the Sentinel-6 Michael Freilich (MF) in 2020 to the satellite altimetry record, we have parallel high- and low-resolution inland water observations. In this study, we validate Sentinel-6MF low- and high-resolution lake water levels, as determined with a state-space model, with data from 124 gauges in 85 lakes (of size >20 km2) from the United States, Canada, Sweden, and Australia. Furthermore, we estimate the scatter of the observations at 1330 globally distributed lakes. We address the difference in performance when using a physical (ocean) and empirical (OCOG) retracker. Using the 124 gauges as ground truth, we determine a best-case unbiased root mean squared error (u-RMSE) of 6.4 cm for Sentinel-6MF, when using the high-resolution mode and the OCOG retracker. The u-RMSE is between 6.4 and 12.3 cm when correcting for ice-covered lakes, and 15.6 to 31.3 cm for all data, depending on retracker and resolution mode. The corresponding median standard deviation of water surface elevations per transect from the 1330 lakes is 3.5 cm and 8.1 cm for high and low-resolution OCOG-based water levels, respectively. The corresponding median standard deviation of Sentinel-3A/B OCOG (enhanced measurement, 20 Hz mode, baseline-3/4) is 5.9 cm and 6.0 cm. We find a lower u-RMSE and standard deviation for high-resolution data, and the ability to observe more lakes, compared to the low-resolution mode. From this, we see that Sentinel-6MF yields valid estimates of the water surface elevation, including smaller lakes, and would provide additional value to future water level monitoring via remote sensing.