openAMUNDSEN-DA

openAMUNDSEN-DA is the data assimilation layer for the distributed snow and hydrological model openAMUNDSEN. It prepares configured snow observations, executes sequential ensemble updates and writes reproducible diagnostics and compact gridded results.

The project is a v0.9 research preview. A scientific manuscript describing the openAMUNDSEN-DA framework and its Rofental application is in preparation. This repository documents the software interface and operational workflow.

Documentation

• Installation
• Input data
• Configuration
• Running the model
• Output data
• Example data sets
• How to Use
• CLI reference

Install

Install the Python package on Python 3.11–3.14:

python -m pip install openamundsen-da

For the complete geospatial runtime, use the multi-architecture container:

docker pull ghcr.io/openamundsen/openamundsen-da:0.9.4

Use exact versions or image digests for reproducible work. The moving edge image tracks the latest green main commit and is not a release.

Command-line workflow

The installed distribution exposes one command, openamundsen-da:

openamundsen-da observations snow-cover PROJECT_DIR
openamundsen-da observations wet-snow PROJECT_DIR
openamundsen-da prepare PROJECT_DIR
openamundsen-da run PROJECT_DIR --max-workers 24
openamundsen-da render PROJECT_DIR --max-workers 24
openamundsen-da clean PROJECT_DIR

clean is preview-only unless --apply is supplied. A successful single-domain run already removes package-owned restart state after all configured outputs pass validation; member grids remain available.

Large domains use explicit data assimilation stages:

openamundsen-da subdomains prepare PROJECT_DIR
openamundsen-da subdomains run PROJECT_DIR
openamundsen-da subdomains merge PROJECT_DIR
openamundsen-da subdomains render PROJECT_DIR

The separate openamundsen-da subdomains model prepare|run|merge branch tiles one ordinary openAMUNDSEN simulation without projects or assimilation.

Python interface

from openamundsen_da import prepare_project, run_project

project = "/data/rofental/projects/project_2022_2023"
prepare_project(project)
result = run_project(project, max_workers=24)
print(result.manifest_path)

The supported Python surface contains the six top-level workflow operations documented in the API reference. Subdomain modules and lower-level scientific routines are internal interfaces.

Configuration boundary

• <setup-name>.yml is pure openAMUNDSEN configuration and shared setup data.
• <project-name>.yml owns the time span, observations and data assimilation.
• generated step YAML files own one assimilation window and step-local model overrides.

Model grid inputs are selected exclusively by output_data.grids.format and support grid-layout NetCDF or deterministic georeferenced GeoTIFF. The compact data assimilation result is always results/grids/da_output_grids.nc.

Development

Build and validate the distribution, then build the same wheel-based image used for release:

python -m pip install build twine
bash scripts/ci/build_distribution.sh
docker build -t openamundsen-da:local .

Run tests through the repository wrappers:

bash scripts/ci/run_lint.sh
bash scripts/ci/run_unit_tests.sh
bash scripts/ci/run_integration_tests.sh
bash scripts/ci/run_integration_tests_subdomain.sh

See tests/README.md for the exact validation contracts.

License

openAMUNDSEN-DA is released under the MIT License, including commercial use subject to the license terms.