Author(s): Laure Sicard; Pilar Garcia-Navarro; Sergio Martinez-Aranda
Linked Author(s): Pilar García-Navarro
Keywords: Scalar transport; Diffusion; GPU; Modeling; Convection
Abstract: Scalar transport models for Earth surface flows combine the 2D Shallow Water Equations (SWE-2D) and the 2D advective-diffusive transport equation for scalar properties. This kind of models have been used in various contexts, such as chemical tracing in environmental systems, understanding biological diffusion processes, sediment transport in rivers and coast, etc. Scalar transport models play a crucial role in simulating the movement of solutes in natural water mediums such as groundwater, rivers, estuaries and even within living organisms. The versatility and applicability of scalar transport models continue to expand with advancements in computational techniques. Nowadays the real challenge remains in producing efficient and fast scalar transport models thanks to the GPU-based computation technology. The generalized 2D transport equation for scalar advection-diffusion within a shallow water flow is expressed as: (1) where is the depth-average concentration of a scalar, h is the water depth, (u,v) are the depth-averaged flow velocity vector components and Ks is diffusion-dispersion tensor. The flow variables are computed using the 2D shallow water equations and the full system is solved using a Roe based finite volume technique. (Morales-Hernandez et al. 2018). Most of the GPU model implementations are designed for grid-type square-cells meshes. However, the performance of the numerical implementation of the hydrodynamic scalar transport in a GPU-based framework able to deal with arbitrary mesh topology is unaddressed.
Year: 2025