Author(s): Junyu Wei; Weihong Liao; Xiangyu Luo; Xiaohui Lei; Haocheng Huang; Hao Wang
Keywords: Modelling and visualisation tools; Flood simulation; High performance computing; Probabilistic forecast; Particle filter
Abstract: Modern large cities feature wide areas and strong spatial heterogeneity. The intensity and evolution of urban flood events are affected by multiple sources of uncertainty such as precipitation and underlying surface conditions. In order to comprehensively describe the uncertainty of input and output factors, a numerical model of ensemble forecast is constructed, and the probability distribution of forecast results is described with Monte Carlo samples. The complexity of a two-dimensional flood model grows exponentially with its spatial resolution, and ensemble forecast further exacerbates the demand for computational resources. Hence, advanced parallel computing technology must be adopted to ensure simulation efficiency.