Author(s): Luis Cea; Ernest Bladé; Orlando García-Feal; Marcos Sanz; Moncho Gómez-Gesteira; Juan Fernando Farfán

Linked Author(s): Luis Cea Gómez

Keywords: shallow water equations, hydrological model, high performance computing, CUDA, Iber+

Abstract: In the last decade, an increasing number of studies have been published in which numerical models based on the two-dimensional shallow water equations are applied to compute rainfall-runoff transformation and surface runoff in rural and urban catchments. In this presentation, we discuss some critical aspects that should be considered to extend and apply a 2D shallow water model to distributed hydrological modelling, and we detail the physical processes, numerical methods and High Performance Computing (HPC) techniques that have been implemented in the freely available software Iber, to enable its application as a high-resolution hydrological model. The standard sequential version of Iber is coded in Fortran and it implements an unstructured finite volume solver for the 2D shallow water equations including several turbulence models. The solver implements several numerical schemes for the discretisation of the convective fluxes and source terms, as the scheme of Roe for standard open channel flow and river inundation applications, the DHD scheme for hydrological applications, and an upwind bed slope discretisation to guarantee the exact hydrostatic solution. In addition to the hydrodynamic module, which is the core of the model, Iber implements additional capabilities to compute: hydrological processes at the catchment scale, urban drainage processes and the interaction of the overland flow with the sewer system, sediment transport in rivers, soil erosion due to rainfall-runoff processes at the catchment scale, water quality in rivers and wood transport in rivers. Iber also includes a parallel version coded in C++ (Iber+) of most of its modules. Iber+ applies HPC techniques that take advantage of the parallelization functionalities of the CPU (using OpenMP) and the GPU (using Nvidia CUDA). In this line, several recent publications present implementations of shallow water codes that make use of OpenMP and CUDA and, depending on the complexity of the case, Iber+ can achieve speedups up to two orders of magnitude when compared with the standard version. After presenting the most significant recent developments implemented in Iber and the most immediate future work, we apply the model to 3 benchmark test cases: the laboratory experiment reported by Kirstetter et al. (2016), the analytical test case of a V-shaped schematic catchment presented by Overton and Brakensiek (1970) and the Copper Creek catchment in Colorado (Wainwright and Williams, 2017). For the 3 cases, the hydrographs, water depths and flow velocities computed with the model will be analysed and compared with experimental observations.

DOI: https://doi.org/10.3850/IAHR-39WC252171192022SS1085

Year: 2022