High-Resolution Daily Flood Extent Depiction from Microwave Remote Sensing: Global Results

Abstract:

The need for frequent, accurate, high-resolution characterization of the temporal and spatial progression of flood hazards is evident, but has been beyond the capabilities of remote sensing methods. The surface is too often obscured by cloud cover for visual and infrared sensors and observations from radar sensors are too infrequent to create consistent historical databases or for monitoring current conditions. Passive microwave sensors, such as SSM/I, AMSR-E, and AMSR-2, acquire useful data during clear and cloudy conditions, have revisit periods of up to twice daily, and provide a continuous record of data from 1987 to the present. In this presentation, we will describe results from a flood mapping system capable of producing high-resolution (100-m) flood extent depictions from lower resolution (10-40-km) microwave data. The system uses the strong sensitivity of microwave data to surface water extent combined with land surface and atmospheric data to derive daily flooded fraction estimates globally on a sensor footprint basis. The system downscales flooded fraction to make a high-resolution Boolean flood extent depiction that is both spatially continuous and consistent with the lower resolution data (see Figure). The downscaling step is based on a relative floodability index derived from higher resolution topographic and hydrological data and processed to represent the minimum total water fraction threshold above which each grid point is expected to be flooded given microwave-derived water fraction inputs. We have completed daily, 100-m resolution flood maps for Africa for the 9.3-year AMSR-E period and will soon complete global flood maps fo the same period. We will present animations of daily flood extents during major events and discuss: validation of the flood maps against imagery derived from MODIS and Landsat data; analyses of minimum detectable flood size; statistical analyses of flooding over time; applications for this novel historical dataset; and prospects for near real-time flood mapping.