-
Notifications
You must be signed in to change notification settings - Fork 0
Home
The Tiger Hillslope Link Model (HLM) is a complete rewrite of the original HLM concept developed by the Iowa Flood Center. This implementation of the runoff generation component is GPU-accelerated (built with CUDA) and parallelized across compute nodes using MPI. This architecture makes it suitable for large hydrologic domains spanning hundreds of thousands to millions of hillslopes.
The code operates in parallel at two levels. At the node level, MPI is used to distribute subsets of hillslopes to individual ranks. Each rank is assigned a single GPU device, and within each rank, all hillslope systems are solved concurrently using CUDA kernels. Rank 0 serves as the coordinator and is responsible for reading the model configuration, loading spatial parameters, distributing data across ranks, and managing global output writing. All input forcings (e.g., temperature and precipitation) are ingested in NetCDF format and mapped to the appropriate spatial units using a lookup CSV. The model also accepts a YAML configuration file for global settings and a CSV file for hillslope-specific parameters.
The code is open to everyone, and we are actively developing and testing new features. Each official release will be versioned and archived with a DOI via Zenodo. Once the model software and accompanying validation papers are accepted, they will be linked to the corresponding Zenodo entries to ensure transparency and reproducibility.
Documentation is available through the following wiki pages.
Getting Started
User Guide
Programmer Guide