GPU-Accelerated High-Performance Parallel Computing for Fluviomorphological Modeling

Abstract

The characterization of spatial and temporal scales in fluvial environments presents significant challenges due to morphodynamic process complexity. GPU solvers offer accelerated parallel computing with notably reduced times, although their application is limited to certain problems. This work presents an in-house GPU code compared with OpenFOAM SedFoam (CPU) to analyze dune transport from the Paraná River in two-dimensional character with longitudinal profile. Results contribute to developing numerical tools extrapolable to other environments. This approach enables simulating multiple hydrosedimentological scenarios to identify determining factors in morphological changes and understand system response to hydrological and sedimentological parameter variations. Comparison between processing architectures will evaluate accuracy and computational performance.