FloodDAM-DT FloodML mapping

FloodML is a machine learning algorithm that maps flood extents from remote sensing satellite data. It produces three different outputs.

From either Sentinel-1, Sentinel-2, Landsat 8, Landsat 9 and TerraSAR-X images, it generates in a matter of minutes a pixel by pixel classification in a raster format, with a majority filter applied to improve data consistency.

The algorithm also provides an automatically generated report that shows the flooded areas along with the permanent water bodies extracted from the Global Surface Water dataset, with the characteristics of the input image.

The third product is a raster that takes into account the built-up and forest areas from the ESA world cover dataset of 2021 along with the previously flooded areas classification. This allows to put in context areas that might not be observed as flooded but that can be indeed flooded, mainly due to surface hiding from trees and buildings.This product has 8 levels of flooded surface classification :

• 0: non classified
• 1: water/flood
• 2: forest
• 3: flooded forest
• 4: built
• 5: flooded built
• 6: clouds
• 7: shadow

This dataset is generated on demand using the OGC Process API standard after a flood detection and alert.

License available here.