Potential of SAR-Derived Flood Maps for Hydrodynamic Model Calibration in Data Scarce Regions

Abstract

Synthetic aperture radar (SAR) observations of real world flood extents are often the only source of information in data scarce catchments and the only reliable resource for channels with subkilometer widths. Accordingly, this study aimed at evaluating the reliability of SAR-based flood maps for flood model performance assessment for an extremely data poor region in India. SAR images were converted to probabilistic flood maps by combining inundation extents obtained using visual interpretation, histogram thresholding, and texture-based classification. Flood extents simulated by a hydrodynamic model were compared with SAR-derived inundation extents using spatial objective functions to calibrate the lumped channel and floodplain friction parameters. The agreement between the modeled and observed flood extents showed an R2 value of 0.938 and a root mean squared error of 0.278 pixels for the validation. The results indicate that the proposed method has the potential to support flood model calibration and evaluation in ungauged basins.