Porous Shallow Water Modeling for Urban Floods in the Zhoushan City, China

Typhoon-induced intense rainfall and urban flooding have endangered the city of Zhoushan every year, urging efficient and accurate flooding prediction. Here, two models (the classical shallow water model that approximates complex buildings by locally refined meshes, and the porous shallow water model that adopts the concept of porosity) are developed and compared for the city of Zhoushan. Specifically, in the porous shallow water model, the building effects on flow storage and conveyance are modeled by the volumetric and edge porosities for each grid, and those on flow resistance are considered by adding extra drag in the flow momentum. Both models are developed under the framework of finite volume method using unstructured triangular grids, along with the Harten-Lax-van Leer-Contact (HLLC) approximate Riemann solver for flux computation and a flexible dry-wet treatment that guarantee model accuracy in dealing with complex flow regimes and topography. The pluvial flooding is simulated during the Super Typhoon Lekima in a 46 km(2) mountain-bounded urban area, where efficient and accurate flooding prediction is challenged by local complex building geometry and mountainous topography. It is shown that the computed water depth and flow velocity of the two models agree with each other quite well. For a 2.8-day prediction, the computational cost is 120 min for the porous model using 12 cores of the Intel(R) Xeon(R) Platinum 8173M CPU @ 2.00 GHz processor, whereas it is as high as 17,154 min for the classical shallow water model. It indicates a speed-up of 143 times and sufficient pre-warning time by using the porous shallow water model, without appreciable loss in the quantitative accuracy.

Automated summary

Intense rainfall induced by typhoons poses a threat to the city of Zhoushan each year, driving the need for accurate flooding prediction. Two models, a classical shallow water model and a porous shallow water model, were developed and compared for the city, with the porous model showing significantly faster computational speed and improved efficiency without compromising accuracy in predicting floods in complex urban areas with varying topographies and building geometries.