In Detail: Flood Mapping with Sentinel-1 and Sentinel-2 Imagery and Digital Terrain Models

Flood mapping in urban areas poses significant challenges for Synthetic Aperture Radar (SAR) sensors due to limitations in detecting water in regions characterized by dense vegetation, urban infrastructure, or complex surface conditions. These limitations include reduced sensitivity in vegetated or built-up areas and water-like backscatter effects on smooth, dry, or snow-covered surfaces. The Global Flood Monitoring (GFM) platform employs Sentinel-1 SAR backscatter data for automated flood delineation but recognizes the constraints posed by such conditions. To address these challenges, the GFM has integrated an exclusion mask that highlights regions prone to SAR-based misclassification.  

This recommended practice introduces a novel algorithm developed by the Joint Research Centre of the European Commission that combines SAR-derived flood layers with digital terrain models, the GFM exclusion mask and incomplete Sentinel-2 flood maps. By leveraging DTMs, water depth calculations and hydrodynamic propagation models are applied to infer flood conditions within exclusion mask areas, enhancing the reliability of flood extent delineations. 

Input Data:

  • GFM outputs for a flooded Area of Interest:
  • Flooded Area
  • Permanent and Seasonal Water Bodies
  • Exclusion Mask
  • Digital Terrain Model
  • Sentinel-2 flood map and cloud mask (calculated in GEE during this practice)

Sofware:

  • Global Flood Monitoring Database
  • Python
  • QGIS or other GIS software 

This practice can be used for any area with major flooding. The practice is especially relevant for large floodings, that extend into cities and vegetated areas, like forests and agricultural fields.

The contribution through the incomplete Sentinel-2 flood mask is especially important, if GFM underestimates the flooded area. In this context, the practice utilizes two incomplete and not perfect flood maps and combines them to get the best result possible. 

Strengths:

  • The approach is based on physics and the actual topography of the area. That means it overcomes many limitations associated with satellite imagery.
  • Utilizing two data sources strengthens the practice and makes it robust to several application  

Limitations:

  • The greatest limitations of the practice come from the spatial resolution and accuracy of the input flood delineation and of the Digital Terrain Model.
  • The quality of the output is dependent on the number of flooded pixels, provided with the input flood delineation. 

The workflow can be divided into two sections: 1) Preparedness before the flood and 2) Response after the flood.

Its implementation is done in GFM (Global Flood Monitoring Database), Python, and in the Google Earth Engine. 

A diagram of a flowchart

Description automatically generated
 

Hawker, Laurence; Uhe, Peter; Paulo, Luntadila; Sosa, Jeison; Savage, James; Sampson, Christopher; Neal, Jeffrey (2022): A 30 m global map of elevation with forests and buildings removed. In Environ. Res. Lett. 17 (2), p. 24016. DOI: 10.1088/1748-9326/ac4d4f.

Betterle, Andrea; Salamon, Peter (2024): Water depth estimate and flood extent enhancement for satellite-based inundation maps. In Nat. Hazards Earth Syst. Sci. 24 (8), pp. 2817–2836. DOI: 10.5194/nhess-24-2817-2024.

Expert Flood Monitoring Alliance, McCormick, N., Salamon, P., Global Flood Monitoring (GFM) – Product User Manual. European Commission. 2023. 

Attachment Size
Slides: RP_FLEXTH_S1S2 (2.02 MB) 2.02 MB