Calibration framework for complex 2D hydrodynamic models: Use of satellite-derived flood extent and water depth data, and evaluation with various performance metrics

Abstract

A methodology for calibrating a 2D hydrodynamic model using remotely sensed data is presented. The methodology was tested in Spercheios river basin, Central Greece, whereas two independent flood events -in February 2015 and December 2021- were considered for calibration and validation, respectively. Model performance was assessed with respect to the predicted inundation extent and water depths, benchmarked against a SAR-derived flood map and FwDET (Floodwater Depth Estimation Tool) estimates. The methodology combined Global Sensitivity Analysis (GSA) to screen the most influential parameters with a grid-search optimization approach to calibrate them, employing various performance metrics, i.e., the Critical Success Index (CSI), Success Index (SI), Hit Rate, False Alarm Ratio, Error Bias, Accuracy Index, and RMSE. The calibration revealed discrepancies due to metric selection, highlighting the importance of carefully selecting a suitable objective function. SI was found to more representatively reflect optimal model response and offered higher flexibility compared to CSI, suggesting its greater suitability for calibration. Although the RMSE is considered a rather detailed metric offering a cell-by-cell flood depth evaluation, the identification of its relative efficiency was hampered by the limitations of FwDET in accurately estimating water depths. Specifically, the algorithm’s reliability was found to decrease as the extent of inundated areas increased, while being heavily affected by the micro-topography of the area. The calibration results indicated equifinality issues stemming from the complex and non-linear nature of the model and the spatial interrelationships among the model variables. GSA results proved effective in identifying potential equifinality challenges early in the process.