A STUDY OF AUTOMATIC FLOOD-AREA DETECION USING ALOS-2 AND ANCILLARY DATA

In this study, we developed an automatic flood detection method using ALOS-2 and ancillary data. Our method can be automatically processed and provide easy-to-understand flood information rapidly for users working on-site disaster response. A Bayesian probabilistic model was used for integrating ancillary data (quasi-realtime flood simulation) and multiple SAR-derived images (amplitude and coherence). We also used a land-use/land-cover map to improve the flood classification result. We examined our methods using ALOS-2 data in the 2015 flood in Japan and obtained an accurate flood map in good agreement with reference data.