The Open-Source Storm Water Management Model (SWMM2D) — Next-Generation Computational Engine

OpenSWMM is an open-source hydrology, hydraulic, and water-quality engine built on EPA SWMM. It provides a reentrant C++20 implementation, a public C API, Python bindings, and a command-line solver. The legacy SWMM 5.x solver and output reader remain available alongside the new engine.

Engine documentation · Python documentation · SWMMVis GUI

Current implementation

The source version is 6.0.0-alpha.4. The following summarizes the implementation as of October 21, 2026; feature availability depends on the build and selected solver.

Engine and interfaces: independent simulation handles, structure-of-arrays state, domain-specific C APIs, Python model building and editing, runtime forcing and controls, and plugin-based input/output.
Hydrology: rainfall-runoff, infiltration, conventional LID controls, two-zone subcatchment groundwater, degree-day and UEB energy-balance snow, temperature-dependent RDII abstraction recovery, and runtime climate and potential evapotranspiration forcing.
1D hydraulics: steady, kinematic-wave and dynamic-wave routing, plus explicit finite-volume Saint-Venant routing (FLOW_ROUTING FV). Dynamic-wave extensions include semi-implicit node continuity, Anderson acceleration, and dynamic Preissmann slots. Street/inlet capture, variable-speed pumps and additional storage geometries are supported.
2D surface flow and groundwater: finite-volume surface routing with local time stepping and conservative 1D exchange; mesh aquifers with hydraulic/quality boundaries, bulk or SIGMA layer chemistry, reactions and porous soil heat. Qualified standalone CPU SSP-RK2 profiles support water, chemistry, age, heat, accepted surface latent cooling, surface boundaries and copied runtime forcing. Managed numerical, binary or mesh-HDF5 checkpoints preserve native weather and immutable hourly meteorology, calendars, pending batches and accepted receipts, including owned water-only degree-day/UEB snow and independent networks with water-only or qualified native LEGACY pollutant state. See the transport profiles and limits.
Water quality: legacy routing, Eulerian advection-reaction-dispersion (ARD), Lagrangian parcel transport (LARD), multispecies reaction expressions and integrators, treatment expressions, and water age. Solver and process combinations have different support limits.
Heat transport: network and watershed temperature calculations, surface energy exchange, radiation and sediment exchange. The experimental coupled mesh thermal model adds porous-ground heat storage and transport, channel/ground contact, reciprocal UEB snow/soil exchange, and supported mesh-connected Richards LID heat exchange. See the thermal implementation notes for activation and qualification limits.
LIDs as storage nodes: layered controls connected through ordinary hydraulic links, independent outlet controls, layer treatment, and classical or Richards-column flow. See the storage-node LID manual for routing and transport restrictions.
Persistence and coupling: SWMM input files, optional GeoPackage and HDF5 model I/O, binary and 2D HDF5 results, and hot-start/checkpoint support, including native coupled thermal state for supported configurations. Optional HydroCouple components provide metadata inspection and qualified CPU 1D execution and outfall transfers.

Kokkos OpenMP, CUDA, HIP and SYCL surface-solver plugins load separately from the core engine. The default source build includes the OpenMP plugin; device backends need their matching toolchains. An optional host OpenMP operator computes raw CLOSED_FORM groundwater recharge; water, transport, reactions and heat retain CPU ownership. Other groundwater acceleration remains outstanding; backend support is not interchangeable. See the 2D CPU implementation and qualification.

Quick start with Python

pip install openswmm

from openswmm.engine import Solver, Nodes
 
with Solver("model.inp", "model.rpt", "model.out") as solver:
    nodes = Nodes(solver)
    while solver.step():
        depth = nodes.get_depth("J1")

Use a node ID from your model in place of J1. The Python guide covers model construction, parameter editing, runtime forcing, results and checkpointing. Published packages may lag the source implementation described above.

Build from source

Requires CMake 3.21+, C17/C++20 compilers, Ninja for the Linux/macOS presets, and vcpkg. Dependencies and optional features are declared in vcpkg.json.

git clone https://github.com/HydroCouple/openswmm.engine.git
cd openswmm.engine
git clone https://github.com/microsoft/vcpkg.git vcpkg
./vcpkg/bootstrap-vcpkg.sh
export VCPKG_ROOT="$PWD/vcpkg"
 
cmake --preset=Linux
cmake --build --preset=Linux
./build/linux/bin/Release/openswmm model.inp model.rpt model.out

Use Darwin on macOS or Windows from a Visual Studio Developer Prompt; on Windows, bootstrap vcpkg with bootstrap-vcpkg.bat and set VCPKG_ROOT in that shell. Build folders follow the platform preset. The legacy executable is openswmm-legacy.

To build and run the test suite:

cmake --preset=Linux-tests
cmake --build --preset=Linux-tests
ctest --test-dir build/linux-tests --output-on-failure

Platform presets also provide -debug and -tests-release variants. See CMakePresets.json and CMake options for optional 2D, GPU, GeoPackage, HDF5, Python and HydroCouple builds. For Python source builds, use pip install ./python.

Documentation

Engine manual: input sections, files, reports and programmatic interfaces.
Reference manuals: process formulations, numerical methods and API reference.
Release notes and roadmap: release history and development direction.

Contributing

Contributions are welcome — bug reports, fixes, new features, docs, tests, and benchmarks.

Read CONTRIBUTING.md for the development workflow and the Code of Conduct.
Fork the repo and create a feature branch.
Ensure C++ (ctest) and Python (pytest) tests pass.
Follow existing style and naming.
Open a PR against develop.

Contributor License Agreement

First-time contributors must sign the project CLA before a pull request can be merged. The CLA grants the project a perpetual, royalty-free copyright and patent license to your contributions and preserves the project's ability to relicense in the future; you retain full copyright ownership of your work.

Signing is automated through CLA Assistant — when you open your first PR, a bot comments with a one-click sign-in link. The CLA covers all subsequent contributions, so you only sign once. Corporate contributors should additionally submit a CCLA per CLA §6.

License

Apache License, Version 2.0 — see [LICENSE](LICENSE) and [NOTICE](NOTICE). Original EPA SWMM material is in the public domain under 17 USC § 105 and is not subject to the Apache license grant.

Acknowledgements

OpenSWMM builds on the EPA Storm Water Management Model. See docs/authors.md for the full contributor list.