Flood and waterlogging detection

Organizations involved:   FÖMI (Hungary)

The workflow provides a highly automated pre-processing and classification of satellite images for detection of flooded and waterlogged areas. This complex workflow contains multiple algorithms. Compared to the current („as-is“) solution implemented using commercial software components at FÖMI, it provides a higher level of automation via smarter algorithms; it improves overall processing time and implies a better use of human resources.

It consists of the following steps: pre-processing of satellite imagery; Calculation of spectral indices (NDxI); Raster segmentation; Clustering of raster data; Thematic classification, tailored to different water affection levels of land surfaces. The outcome is a thematic raster containing predefined classes related to different amounts of water presence on land surfaces. See Figure 6, for intermediate results and Figure 7. The final result is a thematic raster containing the following categories:

  • Natural waters
  • Flood/waterlogging
  • Seriously affected soil
  • Moderately affected soil
  • Weakly affected soil
  • Vegetation standing in water
  • Dry areas
  • Clouds
  • Areas out of interest / not supported areas

On this figure can be observed the intermediate results for Spectral Indices (NDXI) Left: Vegetation index (NDVI), Middle: Soul Index (NDSI); Right: Water Index (NDWI)

On this figure the final result is represented with thematic classification for the distinguished classes.