Accelerating the Lagrangian particle tracking of residence time distributions and source water mixing towards large scales

摘要: Travel/residence time distributions (TTDs/RTDs) are important tools to evaluate the vulnerability of catchments contamination and understand many aspects catchment function behavior. In recent years, calculation TTDs/RTDs based on Lagrangian particle tracking approach together with integrated hydrologic modeling has become a popular counterpart analytical approaches lumped numerical models. As global water availability becomes more stressed due anthropogenic disturbance climate change, requirement large-scale long-term simulations for further pushes high computational costs tracking. Hence, speeding up an barrier advancement. this study, we accelerate program EcoSLIM, using combination distributed (e.g. MPI) multi-core accelerator (CUDA) simulations. EcoSLIM was developed be seamlessly paired ParFlow.CLM model calculations transient RTDs source mixing originally threaded OpenMP. This work extends implementation compare combinations MPI, CUDA Of these combinations, OpenMP-CUDA parallelism performed best moving from single-GPU multi-GPU. The multi-GPU shows strong scalability which increasingly efficient particles, demonstrating potential feasibility regional-scale, residence largely improves capability results also show advantages GPU-parallel traditional parallel-APIs (application programming interfaces) its widely next generation programs in subsurface environment modeling.