Comparative Evaluation and Uncertainty Analysis of Depth–Damage Functions for Flood Loss Estimation Using 2D Hydraulic Modelling

Abstract Flood-induced economic losses in data-scarce regions are difficult to quantify reliably because hydraulic modelling, depth–damage relationships, and building-exposure data are rarely combined within a single, locally calibrated framework. This study developed an integrated methodology for estimating residential flood damage in the Poldokhtar urban floodplain on the Kashkan River, western Iran, a reach that experienced severe inundation in 2016 and 2019. The two-dimensional CCHE2D model was calibrated and validated against surveyed cross-sections and 30 field control points, achieving a velocity RMSE of 0.075 m s⁻¹ and no statistically significant difference between simulated and observed flood depths (paired t-test, p = 0.718). Four candidate depth–damage functions — two exponential forms, a piecewise linear function, and a cubic polynomial — were fitted to locally observed damage data; the exponential function S2 (Y = 1 − exp(− 0.94x)) provided the best performance (RMSE = 11.05%, MAE = 8.41%, CC = 0.96). Three approaches for converting relative damage into monetary loss were compared, showing that spatially disaggregated building information (Approach A2) reduced the relative error to 6.4%, compared with 31.1% and 41.3% for approaches based on generalized building values, and that a fixed regional recovery coefficient performed adequately only at shallow depths. A McNemar test further revealed a statistically significant asymmetry between simulated and satellite-derived flood extents (χ² = 14.02, p = 0.00018). The results demonstrate that coupling calibrated hydraulic simulation with locally derived depth–damage functions and spatially explicit exposure data substantially improves flood-loss estimation accuracy, offering a transferable framework for flood-risk assessment in other data-scarce Iranian floodplains.