SEMI-AUTOMATIC FLOOD DETECTION USING HISTORIC SATELLITE IMAGERY.

Worldwide, flooding causes more damage than any other natural hazard. The growing extensive flood risk has a large impact on the present and future economy, especially in developing countries, which have limited funds for prevention and mitigation measures. Moreover, the allocation of these funds is often done without sufficient and adequate information on high-risk and flood-prone areas, since the acquisition of flood data is very expensive as well. Therefore, this research focuses on a low-cost methodology to create the necessary, accurate flood maps in data-sparse regions. In this methodology, historic Sentinel-2A satellite imagery was used to detect water bodies, which in later research will be the basis for detecting flooded areas and creating flood maps. Multiple indexes that combine several bands of the satellite image to enhance the difference between water bodies and land use, were tested to determine which combination gives the best results. Two study areas, one in an urban and one in a rural environment, were chosen and compared since these environments have completely different characteristics, which result in different spectral properties and reflectance values in the satellite images. The indexes that were tested in both study areas are the NDWI (Normalized Difference Water Index), the MNDWI (Modified Normalized Difference Water Index) and the AWEI (Automated Water Extraction Index) for images with shade and without shade. The last index gives the most promising results, for urban areas as well as rural areas. Not only different indexes were tested, but also the threshold value of water per index was determined. The complete methodology was automated in a Python script, so the end user can easily create 2D flood maps without the need of extensive background knowledge. The first results of this tool are very promising in determining water bodies and flood areas. Future research will focus on combining multiple flood maps, generated with open source satellite images, into a flood prediction map that can help developing countries in understanding future flood risk. [ABSTRACT FROM AUTHOR]