Input data

openAMUNDSEN-DA combines a standard openAMUNDSEN setup with explicit observation products and one data assimilation project configuration.

Setup, project and step configuration ownership in openAMUNDSEN-DA

The setup contains shared openAMUNDSEN inputs. One project defines the data assimilation period, observations and events; preparation materializes the step-specific inputs.

openAMUNDSEN inputs

The setup root contains the standard openAMUNDSEN configuration, static grids and meteorological forcing. Follow the upstream openAMUNDSEN input-data documentation for model-specific variables and formats.

openAMUNDSEN-DA additionally requires:

<setup>/
  <setup-name>.yml
  env/roi.gpkg
  grids/
  meteo/
  obs/
  projects/<project-name>/<project-name>.yml

output_data.grids.format in the setup YAML selects the only accepted model-grid reader. Supported inputs are grid-layout NetCDF and deterministic georeferenced GeoTIFF. ASCII output, in-memory output, NetCDF roi_pixel layout, mixed formats and discovery fallbacks are rejected.

Snow observation products

The project YAML declares every observation family, directory, file format, product tag, summary path and class mapping. Paths are setup-relative and may not escape the setup root.

Snow cover

Snow-cover input is GeoTIFF or NetCDF with an explicit class mapping. The preprocessor masks the ROI and configured exclusions and writes scf_summary.csv. When uncertainty-aware processing is enabled, NetCDF variables or GeoTIFF <stem>_uncertainty.tif sidecars must be present as configured.

Satellite products can reference a tracked acquisition manifest with columns product, source, product_identity, acquisition_time, time_source and time_quality. Acquisition time is resolved from the CF time coordinate, raster metadata, sidecar metadata, a configured filename parser, then this manifest. If none is available, preparation warns and records UTC midnight with time_quality=fallback_midnight.

Wet snow

Wet-snow input is GeoTIFF or NetCDF with explicit wet, valid and excluded class codes. The preprocessor writes wet_snow_summary.csv and, when configured, wet snow line diagnostics and profiles.

Station observations

Station snow depth and snow water equivalent use one CSV per station under the configured station directory. stations_da_metadata.csv declares station roles and uncertainty settings. Its station_id, x and y columns also support spatial subdomain selection when no legacy stations_snow_depth.csv is present. Station observations are read directly during project preparation; they do not have a raster-summary stage. A station event is supported only by an enabled DA station with a same-ID series and one finite, nonnegative observation within half the setup timestep of the model assimilation timestamp. Ties are invalid; a value elsewhere on the same date is not support.

Observation preprocessing

Run only the preprocessors required by the configured project:

openamundsen-da observations snow-cover PROJECT_DIR
openamundsen-da observations wet-snow PROJECT_DIR

PROJECT_DIR is <setup>/projects/<project>. Both commands read paths and class mappings from the project YAML. --overwrite replaces an existing summary; without it, an existing output is not silently changed. --json emits the same typed result as one machine-readable envelope.

Inspect summary dates, spatial support, invalid/cloud fractions and uncertainty statistics before selecting assimilation_events. openamundsen-da prepare then verifies that each configured event has the required observation and writes deterministic per-step inputs.

Prepared satellite rows record both the timezone-aware UTC observation time and the matched naive model-clock time. A date with several scenes requires the event’s observation_time selector. A unique scene may still use the date-only event form.

See Configuration for the YAML schema.