EGU23-3415, updated on 22 Feb 2023
EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Mapping of large-scale low water situations using satellite-based water-land boundaries

Bastian Gessler1, Silke Mechernich2, Robert Weiß3, and Björn Baschek4
Bastian Gessler et al.
  • 1The German Federal Institute of Hydrology (BfG), Koblenz, Germany (
  • 2The German Federal Institute of Hydrology (BfG), Koblenz, Germany (
  • 3The German Federal Institute of Hydrology (BfG), Koblenz, Germany (
  • 4The German Federal Institute of Hydrology (BfG), Koblenz, Germany (

In the summers of 2018 and 2022, low water levels of German waterways massively restricted the transport performance of freight ships. Furthermore, oxygen and temperatures were critically high for the ecosystem. In such hydrological extreme situations, information on the location and shifting of the boundaries between water and terrain (water-land boundary) is relevant, e.g. for improved forecasting and monitoring of sediment displacements.

Satellite-based methods are an effective way to monitor such situations and can be used to observe large areas in a short time. Due to their independence from solar illumination and weather conditions, radar data offer considerable advantages compared to optical data. Particularly the radar satellite Sentinel-1 (ESA, Copernicus) is of great relevance, since the data are available free of charge and a continuous future supply is assured. For this reason, we use Sentinel-1 data as basic information in the project "Sat-Land-Fluss".

Here, we will present an example of S-1 water-land-boundary detection for the low water event in 2018 at the Middle Rhine. Comprehensive validation data are available, as an imagery flight was assigned by BfG on behalf of the Freiburg Waterways and Shipping Authority (WSA) at the lowest water level in November 2018. The water-land boundaries were derived from the 10-cm-resolution aerial photographs by the Federal Institute of Hydrology.

The water surfaces from S-1 data is obtained by a thresholding method of backscatter intensity. Various ancillary data were integrated and their potential for improving the result was analyzed, e.g.:

  • the location of the shipping channel (©WSA Rhein) led to a significant reduction of misclassifications, since e. g. overlay effects from ships or bridges can be removed.
  • The land cover information (©ESA World Cover 2020) allowed the correct classification of areas with low backscatter effects (e.g. agriculture) as non-water.
  • The HAND (Height Above Nearest Drainage) index from the high-resolution terrain information (DTM-5 of the Federal Agency for Cartography and Geodesy) helped to exclude areas that could be classified as not covered by water due to their topographic location.

The algorithm based only on S1-data yields about 85-92 % of correct water-classification, and together with the additional data in a)-c) we gain up to approximately 94-98 % of correct classification depending on the S-1 scene. We highlight that particularly the usage of landcover data and high resolution DTMs highly improves the reliability of the water-land boundary from S-1 data. The main remaining weaknesses are located near the water-land-boundary within approximately 50 m. Since the spatial resolution of S-1 data is rather low with about 5 x 20 m, the resulting spatial accuracy of the water-land-boundary is less than about 10 m. To improve this, the integration of a 1-m-digital terrain model of the water course (DGM-W) together with measured or predicted water level information is ongoing. This will provide water level information in areas where Sentinel-1 is not able to record information (e.g. areas of radar shadow due to vegetation, buildings, bridges or topography).

How to cite: Gessler, B., Mechernich, S., Weiß, R., and Baschek, B.: Mapping of large-scale low water situations using satellite-based water-land boundaries, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-3415,, 2023.

Supplementary materials

Supplementary material file