Parallel modeling of sediment and radionuclide transport in rivers on multiprocessor systems and graphics processors

• Дата: 2025
• Автори: Sorokin, Maksym, Zheleznyak, Mark, Kivva, Sergii, Pylypenko, Oleksandr
• Формат: Стаття
• Мова: Ukrainian
• Опубліковано: Kyiv National University of Construction and Architecture 2025
• Теми: river contamination modeling; shallow water equations; finite volume method; parallel computing; GPU computing; MPI; OpenACC
• Онлайн доступ: https://es-journal.in.ua/article/view/343565
• Назва журналу: Environmental safety and natural resources

The study aims to develop and implement parallel algorithms for modeling sediment and pollutant transport in rivers within the COASTOX-UN modeling system, utilizing multiprocessor systems and graphics processing units (GPUs). The modeling system includes a hydrodynamic module (COASTOX-HD), a sediment transport module (COASTOX-SED), and a radionuclide transport module (COASTOX-RN), which can also be adapted for other pollutants. The methodology is based on numerical solutions of two-dimensional shallow water equations and advection-diffusion transport equations using the finite volume method on unstructured grids. Parallel computing is implemented through MPI (for distributed-memory systems) and OpenACC (for GPUs). The system was tested for simulating radionuclide transport in the Kyiv Reservoir during the 1999 spring flood and assessing organic pollutant concentrations in the Dnipro River near Kyiv following transboundary contamination of the Desna River in autumn 2024. Results demonstrate high computational efficiency of the developed algorithms. The combination of MPI and OpenACC technologies in the parallelized COASTOX-UN model enables simulations of sediment and pollutant transport on detailed grids for large water bodies, running efficiently on workstations, servers, and even gaming PCs/laptops with powerful GPUs. GPU-based computations outperform professional workstations in efficiency. The study highlights COASTOX-UN's potential for operational pollution forecasting during emergencies. The key innovation lies in adapting parallel computing algorithms for both CPUs and GPUs, significantly reducing computational costs without compromising accuracy. Future research will expand COASTOX-UN's functionality for additional pollutant types and further assess risks of anthropogenic impacts on aquatic ecosystems.