Estimating a floodwater from MODIS time series and SRTM DEM data

Extreme climate event, such as heavy rainfall and Typhoon, is anticipated to escalate extreme floods. In fact, many flood plains in the Asian-Pacific region have already experienced a rising number of flood disasters. In this circumstance, real-time flood mapping with automatic detection technique is increasingly important in emergency response efforts. However, current mapping technology is still limited in accurately expressing information in flood areas such as inundation depth and extent. For this reason, the authors attempt to improve a floodwater detection method with a simple algorithm for a better discrimination capacity to discern flood areas from turbid floodwater, mixed vegetation areas, snow, and cloud. In this research, pixel classification was performed on the Moderate Resolution Imaging Spectroradiometer (MODIS) time series images (8-day composites, MOD09A1, 500-m resolution) for floodwater detection. The purpose of this image classification was to estimate a flood area based on the spatial distribution of a nation-wide flood from near real-time MODIS images coupled with a digital elevation model (DEM). Moreover, the authors improved the accuracy of the water extent boundary using a 8-direction tracking algorithm to find the same level between flood-prone area and non-flood area. The results showed the superiority of the developed method in providing instant and accurate flood mapping by using three algorithms, which indicates decision tree, modified land surface water index (MLSWI) and 8-direction tracking based on DEM data.