Onderzoeksproject S0/00/022 (Onderzoeksactie S0)
Context and objectives
Floods in Flanders are a regularly recurring event. Floods can often cause tremendous economic damage. To reduce the damage from floods, people (read the Flemish government) need to be well informed.
For its flooding management policy, the Flemish water administration (Afdeling Waterbouwkundig Laboratorium en Hydrologisch Onderzoek (WLH) and Afdeling Water) has already developed computer models of the most important streams under its authority. With these models the Flemish water administration tries to imitate the floodings and to predict their geographical extent. For a good validation of these model, it is essential that a correct delineation of the flooded areas is available.
The Flemish water administration (Afdeling Water) has also developed an inundation database of the natural flooding areas (NOG database) and the recentrly flooded areas (ROG database) in Flanders from 1988 to 2003. This database is an important instrument for the policy of regional planning and the operational water system management. This database has been built up with information from local authorities, Flemish administrations and consultancy agencies. Now, the problem arises to keep this inundation database up to date.
To ensure the production of validation material for existing flood models and to ensure the actualisation of the inundation database, the FLOODMAP-project developed an operational processing chain to extract flooded areas from RADAR-imagery. The merits and limitations of various extraction techniques for flood mapping have been investigated.
Once the flooded areas are extracted, the delineations of the flooded areas can be put on the geo-portal website ‘Geo-Vlaanderen’ on the internet so that local water managers can manually add some additional descriptive terrain information to the inundation database.
Methodology
Since RADAR-sensors are designed to overcome the most common problem associated with monitoring floods, i.e. getting information regardless of weather conditions, it was decided to develop a flood monitoring system based on radar imagery.
The flood monitoring system has been made operational for only ENVISAT and RADARSAT imagery and is developed in such a way that it requires both a flood and a reference image. Two different methods were used for extracting flooded areas out of RADAR-imagery: 1) the active contour technique and 2) the object-oriented classification technique, based on the commercial software eCognition. The results of both techniques are fused together and the minimum and maximum area extent of flooded areas determined. The end result can be put on the geo-portal website ‘Geo-Vlaanderen’ that can be freely accessed via internet.
Results
The project objectives i.e. 1) the developing of an operational processing chain to extract flooded areas from RADAR-imagery and 2) the visualisation of the extracted flooded areas on a geo-portal were acheived by the development of three end products.
End product 1. Development of an operational processing chain for mapping flooded areas based on RADAR-imagery.
The developed image processing chain is divided into two main parts: the pre-processing chain (image registration and speckle reduction) and the flood extraction processing chain (end product 2)
End product 2. Semi-automatically delineation of flooded areas
Two classification algorithms were developed to transform the pre-processed RADAR-imagery into a flood map product i.e.1) the active contour technique and 2) the object oriented classification technique based on eCognition.
End product 3. Implementation of a web-based geo-portal ‘delinieation of flooded areas’
The geo-portal ‘flooded areas’ has been extended with extra tools so that local water managers can 1) visualize the delineation of flooded areas and 2) possibly add some additional descriptive terrain information (height of the water level, source of flooding, date of flooding,…) for a certain flooded location.
Products and services
- flood monitoring system
- geo-portal ‘flooded areas’ + extra tools
