Assessment of the Suitability of a Global Hydrodynamic Model in Simulating a Regional-scale Extreme Flood at Finer Spatial Resolutions

Abstract:

Global river models (GRMs) are elemental for large-scale predictions and impact analyses. However, they have limited capability in providing accurate flood information at fine resolution for practical purposes. Hyperresolution (~1km resolution) modelling is believed to improve the representation of topographical constraints, which consequently result to better predictions of surface water flows and flood inundation at regional to global scales. While numerous studies have shown that finer resolutions improve the predictions of catchment-scale floods using local-scale hydrodynamic models, the impact of finer spatial resolution on predictions of large-scale floods using GRMs is rarely examined. In this study, we assessed the suitability of a state-of-the-art hydrodynamic GRM, CaMa-Flood, in the hyperresolution simulation of a regional-scale flood. The impacts of finer spatial resolution and representation of sub-grid processes on simulating the 2011 immense flooding in Chao Phraya River Basin, Thailand was investigated. River maps ranging from 30-arcsecond (~1km) to 5-arcminute (~10km) spatial resolutions were generated from 90m resolution HydroSHEDS maps and SRTM3 DEM. Simulations were executed in each spatial resolution with the new multi-directional downstream connectivity (MDC) scheme in CaMa-Flood turned on and off.

While the predictive capability of the model slightly improved with finer spatial resolution when MDC scheme is turned on, it significantly declined when MDC scheme is turned off; bias increased by 35% and NSE-coefficient decreased by 60%. These findings indicate that GRMs which assume single-downstream-grid flows are not suitable for hyperresolution modelling because of their limited capability to realistically represent floodplain connectivity. When simulating large-scale floods, MDC scheme is necessary for the following functions: provide additional storage for ovehrbank flows, enhance connectivity between floodplains which allow more realistic flood routing, and maintain flow capacity across varying spatial scales. These results explicate that simulating at hyperresolution alone will not necessarily increase the predictive capability of GRMs – more realistic representation of sub-grid flow physics is necessary in hyperresolution modelling.