Author(s): Qiuhua Liang, Yun Xing, Xiadong Ming, Xilin Xia, Huili Chen, Xue Tong, Gang Wang, Haibo Yang
Linked Author(s):
Keywords: Flood forecasting, intense rainfall, crowd-sourced data, remote sensing, high-performance computing
Abstract: Accurate real-time urban flood inundation forecasting has long been a technologically challenging task due to the complex urban environment and lack of high-resolution data and high-performance hydrodynamic models to accurately predict the highly transient flood hydrodynamics. In recent years, there has been a rapid development in data acquisition and high-performance computing technologies. Data from different sources are now widely available to support different aspects of urban flood modelling. A number of computationally efficient hydrodynamic models have been reported to support full-scale urban/catchment flood modelling at high resolutions. This paper develops and tests a (near) real-time flood forecasting system based on a GPU (i. e. , graphics processing unit) accelerated high-performance hydrodynamic inundation model and supported by data from open sources. The flood forecasting systems were tested in the Chinese city of Fuzhou for a recent severe urban flood event caused by Typhoon Meranti using a freely available 30m digital elevation model (DEM) data. The results show that the predicted inundation map, together with street photos and other information from the crowd sources (e. g. social media), may effectively indicate the inundation extent and flood depth of the city in (near) real time, potentially providing useful information for the public to improve preparedness and hence reduce flood impacts
Year: 2017