Author(s): Xilin Xia; Qiuhua Liang; Xiaodong Ming
Linked Author(s): Qiuhua Liang
Keywords: Flood modelling; High-performance computing; Hydrodynamic model; Hydrological model; Storm Desmond
Abstract: Flooding is one of the most common types of natural hazards. The current practice of large-scale fluvial flood modelling relies on the use of hydrological models to predict upstream discharge hydrograph to drive inundation modelling at downstream. However, the oversimplified representation of both catchment topographies and hydraulics make hydrological models heavily rely on model parameterisation and calibration. This makes the modelling strategy unsuitable for prediction of extreme events featured with highly transient hydraulic processes, for which high-quality hydrological data is commonly missing to support model parameterisation and calibration. In this paper, the High-Performance Integrated hydrodynamic Modelling System (HiPIMS) has been adapted and applied to the whole 2500 km2 Eden catchment in the UK to reproduce the flood event caused by Storm Desmond in December, 2015. Without necessity of intensive calibration and using hydrographs as boundary conditions, satisfactory results have been obtained for both inundation extent and water level time series in channels, in comparison with observations. Accelerated by multiple modern graphic processing units (GPUs), the model runs more than 20times faster than real time for the simulation of the whole catchment at 20m resolution. The results successfully demonstrate HiPIMS as a promising tool for real-time flood forecasting and flood risk assessment.