Author(s): Shih-Chieh Kao; Sudershan Gangrade; Mario Morales-Hernandez; Michael E. Kelleher; Matthew R. Norman; Youngsung Kim; Ganesh R. Ghimire; Juan-Manuel Perez-Garcia-de-Carellan; Jiyong Lee; Mariam Khanam
Linked Author(s):
Keywords: Computational hydrology; Flood inundation modeling; GPU acceleration; High-performance computing; Shallow-water equations
Abstract: We present TRITON v2, an open-source, GPU-accelerated shallow-water model for high-resolution flood inundation over large river basins and coastal domains. Implemented with Kokkos, TRITON v2 runs efficiently on heterogeneous GPU architectures and is coupled with a calibrated VIC–RAPID hydrologic system. Applications to Hurricane Harvey, Hurricane Helene, and the 2019 Missouri River Basin flood, including a 10 m basin-scale simulation on 552 GPUs, demonstrate accurate event reconstruction, strong scaling, and practical runtimes for continental-scale hazard assessment.
Year: 2026