Publication Date:
2017-06-19
Description:
To increase the representation of physical processes in inundation modelling, current research approaches aim to integrate both hydrological and hydrodynamic models. A previous study by Hoch et al. (2017) showed that spatially explicit coupling approaches can outperform stand-alone runs by single-purpose models as they combine spatially distributed model forcing by hydrological models with more sophisticated routing schemes in hydrodynamic models. We here present GLOFRIM, a globally applicable computational framework for integrated hydrological-hydrodynamic modelling, to facilitate such coupling approaches and to cater for an ensemble of models to be coupled. It currently allows for coupling the global hydrological model PCR-GLOBWB with either Delft3D Flexible Mesh (DFM), solving the full shallow-water equations and allowing for spatially flexible meshing, or LISFLOOD-FP (LFP), solving the local inertia equations and running on regular grids. The main advantages of the framework are its open and free access, its global applicability, its versatility, and its extensibility with other hydrological or hydrodynamic models. Before applying GLOFRIM to an actual test case, we benchmarked both DFM and LFP for a synthetic test case. Results show that for sub-critical flow conditions, discharge response to the same input signal is near identical for both models, which agrees with previous studies. We subsequently applied the framework to the Amazon River basin to test the framework thoroughly and, in addition, to perform a first-ever benchmark of flexible and regular grids at the large-scale. Both DFM and LFP produce comparable results in terms of simulated discharge with LFP exhibiting slightly higher accuracy as expressed by a Kling-Gupta-Efficiency of 0.82 compared to 0.76 for DFM. However, benchmarking inundation extent between DFM and LFP over the entire study area, a critical success index of 0.46 was obtained, indicating that the models disagree as often as they agree. Differences between models in both simulated discharge and inundation extent is to a large extent attributable to the gridding techniques employed. In fact, the result show that the numerical scheme of the inundation model and the gridding technique can contribute as strongly to deviations in simulated inundation extent as, unlike the global flood model inter-comparison by Trigg et al. (2016), we control for model forcing and boundary conditions. This study shows that the presented computational framework is robust and widely applicable. GLOFRIM is designed as open access and to be easily extendable, and thus we hope that other large-scale hydrological and hydrodynamic models will be added, eventually capturing more locally relevant processes as well as allowing for more robust model inter-comparison, benchmarking, and ensemble simulations of flood hazard at the large scale.
Print ISSN:
1991-9611
Electronic ISSN:
1991-962X
Topics:
Geosciences
Permalink