An efficient framework applied in unsteady -condition flood modelling using sparse grid stochastic collocation method
Ying Huang, Xiaosheng Qin
Wednesday 1 july 2015
11:15 - 11:30h
at South America (level 0)
Themes: (T) Flood risk management and adaptation, (ST) Flood risk assessment
Parallel session: 9J. Floodrisk - Assessment
An efficient framework based on the stochastic collocation (SG) method and sparse grid quadrature, called sparse grid stochastic collocation (SGSC) method, is applied in the uncertainty quantification of flood inundation modelling under joint effect from Manning’s roughness coefficient (n) and hydraulic conductivity (Ks), which are relatively sensitive parameters in affecting flood simulation results (such as flow velocities and flow depths). In this study, as uncertain inputs, n is assumed in a uniform distribution and Ks is in a lognormal distribution, respectively, of which the sampling space is based on the collocation points constructed by sparse grid quadrature. Based on these constructed collocation points, output fields over the modelling domain can be represented by Lagrange polynomials and their moments can be obtained by corresponding collocation points and weights. To demonstrate the applicability of the proposed method, a 2D flood inundation case is selected and an appropriate surrogate model by SGSC is built up. The means and standard deviations of the flow-depth and velocities distributions over the modelling domain are compared with those simulated by traditional Monte Carlo (MC) simulation. The simulation results show that SGSC approach can efficiently deal with the multi-uncertainty problems during the flood numerical modelling and significantly reduce the computational burden caused by repetitive runs of numerical models, which has been a common problem in MC simulation.
