Colin Doyle

Flood mapping in the lower Mekong River Basin using daily MODIS observations

ABSTRACT In flat homogenous terrain such as in Cambodia and Vietnam, the monsoon season brings significant and consistent flooding between May and November. To monitor flooding in the Lower Mekong region, the near real-time NASA Flood Extent Product (NASA-FEP) was developed using seasonal normalized difference vegetation index (NDVI) differences from the 250 m resolution Moderate Resolution Imaging Spectroradiometer (MODIS) sensor compared to daily observations. The use of a percentage change interval classification relating to various stages of flooding reduces might be confusing to viewers or potential users, and therefore reducing the product usage. To increase the product usability through simplification, the classification intervals were compared with other commonly used change detection schemes to identify the change classification scheme that best delineates flooded areas. The percentage change method used in the NASA-FEP proved to be helpful in delineating flood boundaries compared to other change detection methods. The results of the accuracy assessments indicate that the −75% NDVI change interval can be reclassified to a descriptive ‘flood’ classification. A binary system was used to simplify the interpretation of the NASA-FEP by removing extraneous information from lower interval change classes.

Near Real-Time Flood Monitoring and Impact Assessment Systems

Flood detection through satellite remote sensing is a well-established practice (Brakenridge and Anderson 2006; Nigro et al. 2014) and has the potential to significantly enhance flood early warning and response systems (Brakenridge and Anderson 2003). Flood extent can be determined from optical sensors like Landsat and the Moderate Resolution Imaging Spectroradiometer (MODIS; Brakenridge and Anderson 2006), radar sensors like Sentinel, and land surface hydrologic routing models. Optical imagery can be obtained frequently at high to very high resolution. Radar sensors are sensitive to water and can penetrate through cloud cover at high resolution. Land surface models are used for hydrologic routing simulations of microwave radar precipitation data to determine flood depth and inundation extent (Brakenridge and Anderson 2006).