DEGALI — Dense Gas Dispersion for Liquid Hydrogen

DEGALI stands for Dense Gas Dispersion for Liquid Hydrogen. It is a modern Python model built from a verified reimplementation of the US EPA DEGADIS 2.1 dense-gas dispersion model and extended for cryogenic hydrogen releases.

Alpha research software. The DEGADIS 2.1 compatibility path is strongly regression-tested. The liquid-hydrogen extensions are suitable for research, scenario comparison and sensitivity studies, but are not independently certified for regulatory separation distances or safety-critical design.

한국어 요약: 현재 공개본은 원본 DEGADIS 재현 경로와 액화수소 연구 확장을 함께 제공합니다. PRESLHY 수평 야외제트 범위에서 정량 비교를 마쳤지만, 미완성 TKE/압력/열폭 폐쇄와 제한된 독립 검증 때문에 설비 인허가 판단의 단독 근거로 사용하면 안 됩니다. 자세한 판정은 1차 결과를 참고하십시오.

What is included

• Python ports of all six DEGADIS 2.1 programs. The independently obtained Fortran oracle is retained locally and is not redistributed.
• Steady, transient, receptor-dose and jet-to-ground workflows.
• Legacy and modern thermodynamic backends.
• Opt-in LH2 source flashing, buoyant trajectory, air condensation/freezing, component enthalpy, ground interaction and thermal-profile research paths.
• Experimental finite-TKE, independent thermal-width and Reynolds-stress operators. These are deliberately marked as research-only until their physical closure inputs and downstream field performance are validated.

Current evidence

The port reproduces the five EPA reference cases and was checked locally against a source-built Fortran implementation. The first frozen LH2 assessment uses seven horizontal outdoor PRESLHY trials, 38 concentration sections, 17 vertical profiles and 41 temperature sensors.

An opt-in thermal-profile candidate lowers minimum-temperature MAE to 16.47 K but worsens other acceptance metrics, so it is not the default. These samples are correlated observations from one campaign, not 38 or 41 independent validation experiments.

See model status, claim grading, and data/reproduction notes.

Installation

Create a Python 3.10 or newer environment and install from a clone:

python -m pip install -e ".[test]"

CoolProp is optional for the DEGADIS 2.1 compatibility path and required for most cryogenic-hydrogen calculations:

python -m pip install -e ".[coolprop]"

Quick start

from degali import run_steady, run_transient, run_jet_to_ground

profile, source = run_steady("B9.INP")
print(profile.distance_to(0.05))

transient = run_transient("B9T.INP")
profile, jet, source = run_jet_to_ground("EX2.INO", "EX2.IN")

The command-line interface exposes the same main workflows:

degali steady B9.INP
degali transient B9T.INP --snapshot 60 --snapshot 120
degali dose B9T.INP --at 200 --at 400 --at 800
degali jet EX2.INO --bridge EX2.IN

Verification

Fast development checks:

python -m pytest -m "not slow" -q

The original Fortran oracle and raw REDIPHEM, SMEDIS and PRESLHY files are not redistributed. Tests that need them skip with an explicit message unless the documented environment variables are configured. Their provenance, reduction method and aggregate validation results are documented. See publication scope.

The consolidated technical reference describes the governing physics, LH2 extensions, validation process, quantitative results and remaining model-form limitations.

Repository layout

src/degali/  model and validation code
tests/       self-contained numerical, physical and regression tests
tools/       research and release verification utilities
docs/        derivations, provenance, audits, limitations and results

Scope and safety

The current validated domain does not cover arbitrary equipment conditions, obstacles, indoor releases, pool spreading, downward/strongly wind-steered jets, or general transient source behaviour. Always compare safety-critical results with independent experiments and an accepted consequence-analysis workflow. See security and safety reporting.

Citation and license

Citation metadata are provided in CITATION.cff. Zenodo's GitHub integration assigns the DOI after the first release; the DOI is added to the citation metadata in the first post-release metadata update.

The complete GitHub, Zenodo DOI and PyPI release sequence is documented in the publication guide.

The DEGALI Python implementation is available under the MIT License. Original Fortran implementations and external experimental datasets are not bundled or redistributed.

The chronological development log that previously occupied this front page is preserved in docs/README-development-log-2026-09-06.md and CHANGELOG.md.