In late August 2017, the gulf coast of Texas found itself in the path of Hurricane Harvey. In a 4-day period, Harvey brought immense rainfall (more than 40 inches 1,000 mm) to a number of communities along the coast, including Houston – the 4th largest city in the United States. Needless to say, this hurricane caused catastrophic flooding throughout the area.
Remote sensing offers tremendous possibilities to generate important information on water levels and flooded areas for people on the ground in a timely manner. The new Sentinel-2A and Sentinel-2B satellites provide a high spatial resolution of 10 m, a high spectral resolution with 13 bands ranging from blue to SWIR and a fairly high temporal resolution of 5 days. This combination makes the system highly suitable for mapping and monitoring widespread disaster events.
eCognition serves as powerful tool to create robust, transferable analysis strategies capable of combining multiple data sets to to quickly extract valuable information from images. In eCognition a classification of flooded areas is a fairly simple task.
My colleague, Matthias Stängel, developed a fast prototype rule set based on two different Sentinel-2 acquisition dates (T1 – pre-Harvey: January 07, 2017 & T2 – post-Harvey: August 30,2017) to detect flooded areas south-west of Houston. For a more sophisticated analysis, elevation models, LiDAR data and shapefiles (eg. of houses) could be integrated to provide more detailed information for the crisis management teams on the ground (e.g. number of houses within flooded area or flood water depth).
Within this Sentinel-2 scene, a total of 735 square kilometers of land was classified as flooded. Below are several close ups showing the classification results in more detail:
If you have any questions, please feel free to contact us.