Delta Hydro

CaMa-Flood-GPU

Active

GPU reimplementation of the CaMa-Flood global river routing model — continental flood simulation on PyTorch/Triton.

by Wuhan University / University of Tokyo

Pricing
Open Source
License
Apache-2.0
Commercial use
Permitted
Ecosystem
Python, PyTorch, CUDA
Platform
Linux, Windows

About CaMa-Flood-GPU

CaMa-Flood is the standard global river hydrodynamics model — Dai Yamazaki's catchment-based routing scheme with subgrid floodplain storage that underlies much of the global flood-hazard literature. CaMa-Flood-GPU, by Shengyu Kang (Wuhan University) with Jiabo Yin and Yamazaki himself, reimplements it in Python on PyTorch and Triton GPU kernels, keeping input compatibility with the original Fortran model while accelerating continental-scale runs dramatically.

It is research software in pre-release, changing quickly and best approached by teams who already know CaMa-Flood — but for large-domain, many-scenario flood routing (climate ensembles, global reanalysis), the GPU speedup changes what is feasible.

Where it fits in the workflow

TerrainCatchmentRainfallRunoffRoutingHydraulicsMappingReporting

Alternatives to CaMa-Flood-GPU

Ranked by the topics, categories and workflow stages they share with CaMa-Flood-GPU.