Multi-index image differencing method for flood water detection

This study presents a remote sensing methodology to detect floods with a change detection approach based on image differencing of several water-related indexes. The proposed methodology is expected to integrate the strengths of each individual index and considers the agreement level among outputs obtained by different indexes as an indicator of overall uncertainty. The analysis of data frequency distribution is used to obtain thresholds to implement data slicing and production of thematic maps. By considering different magnitudes of change, the proposed method is expected to be sensitive to detect different types of flood-related changes, including the detection of recent tracks of water presence. This is particularly interesting for those situations whenever it is impossible to obtain cloud-free satellite images immediately after a flood event, which is often the case given the limitations of optical sensors. The methodology has been applied to a fluvial flood event occurred in the surrounding of a natural lagoon in the Aveiro region (Portugal). Landsat 7 ETM+ and Landsat 8 OLI surface reflectance products were used as inputs. Sentinel 1 GRD data was used for comparison purposes. Results indicate an overall consistency, which allows us to expect the proposed method is replicable for other events and areas.