Atmospheric forcing uncertainty in flash flood hydrodynamics: Coupled NWP, radar and gauge comparisons in an orographically complex mountain catchment

This study quantifies how rainfall forcing affects the spatial structure of flash flood depth, velocity, and integrated hazard in the Kırkgöze-Çipak Basin, Eastern Anatolia, Türkiye. Four rainfall scenarios were propagated through an InfoWorks ICM Rain-on-Grid model calibrated and independently validated with separate 2021 events: WRF and IFS numerical weather prediction outputs, weather radar estimates, and Thiessen gauge interpolation. Two events were analyzed: a dispersed event on 8 April 2022 and a concentrated convective event on 25 May 2020. The scenarios produced broadly consistent responses for the dispersed event, with area-weighted metrics differing by less than a factor of three. For the convective event, inter-source differences widened substantially. The area with depth above 0.5 m ranged from 2.158 km2 for WRF to 0.238 km2 for Radar, while the extreme hazard area ranged from 1.481 km2 to 0.008 km2. Cross-event analysis identified a directional reversal in extreme hazard area between NWP-derived and ground-based products, event-specific inter-source spread ratios of approximately three to twelve, spatial compression of the Radar-derived response, and differing depth and velocity responses across product families. These ratios are interpreted as empirical indicators for the analyzed events rather than generalized amplification factors. A matched infiltration sensitivity experiment for the convective event reduced all threshold-based hydrodynamic responses while preserving the rainfall-product ordering and slightly increasing inter-source variability. IFS and Traditional showed the strongest pairwise spatial agreement, indicating inter-product similarity. The findings support event-dependent multi-source evaluation across a larger historical database.