Earth observation workspace

Real-case setup

The gates a topic must pass before it stops being a learning resource. Each gate names the evidence that closes it. None of these gates is recorded as passed in this repository.

Current state, stated plainly

The pipeline is implemented and tested on synthetic data. No real acquisition has been processed into a published case, so every topic remains a learning resource and the comparison viewer will only show whatever site an operator has actually installed.

  1. 1 · Candidate location

    Team

    Required

    • A specific bounding box, small enough for a restrained tile budget.
    • A reason this place matters to the topic.
    • Known terrain and land-cover context.

    Evidence to record: A short written rationale plus the bbox written into the site YAML.

    The bbox is fixed; changing it later invalidates statistics already quoted.

    Status: not recorded

  2. 2 · Actual acquisitions

    Team, using the catalogue

    Required

    • Real granule names from the same track, frame, direction and maturity.
    • Both pair dates present in that stack.
    • Neither date inside a known instrument gap.

    Evidence to record: `python -m nisar.cli discover` output saved with the site configuration.

    A granule name is recorded verbatim; nothing is inferred from a scene identifier.

    Status: not recorded

  3. 3 · Resource approval

    Team

    Required

    • Granule sizes checked before download.
    • Free disk and staging overhead confirmed.
    • Explicit approval for the download volume and any cost.

    Evidence to record: A note of the expected gigabytes, the free space, and who approved it.

    The team has agreed the volume. A download is not started by assumption.

    Status: not recorded

  4. 4 · Scientific checks

    A reviewer with radar background

    Required

    • Mask and validity semantics understood for the actual product.
    • Grid alignment between the two dates inspected, not assumed.
    • Threshold sensitivity examined rather than one threshold quoted.

    Evidence to record: Written findings, including what remains uncertain.

    The reviewer signs the scope and the residual doubts.

    Status: not recorded

  5. 5 · Build and verification

    Team

    Required

    • `build-site` completes on the real granules.
    • `verify` passes.
    • Product tree, statistics and time series inspected for plausibility.

    Evidence to record: Command output and the site id, recorded in `docs/IMPLEMENTATION_VERIFICATION.md`.

    A passing build proves the software ran, not that the interpretation is right.

    Status: not recorded

  6. 6 · Interpretation review

    Reviewer plus team

    Required

    • Observation statements separated from hypotheses.
    • Independent context gathered for the same dates where possible.
    • Limitations written into the site caveat.

    Evidence to record: A signed review naming the reviewer, the case, and the limits.

    No claim of flooding, damage, subsidence or prediction is made from this pipeline alone.

    Status: not recorded

  7. 7 · Imagery configuration

    Team

    Required

    • A URL-restricted public Mapbox token in an untracked environment file.
    • The token never committed, and never the private secret token.
    • Basemap and raster layers checked in a real browser.

    Evidence to record: A screenshot of the rendered comparison, plus the token's restriction settings.

    Rendering is verified by a person, not assumed from a successful build.

    Status: not recorded

  8. 8 · Publication

    Team

    Required

    • Caveat visible on the page a visitor sees.
    • Data provenance reachable from the interface.
    • Submission rules re-read from the current official guides.

    Evidence to record: The public URL, and a statement of what is deployed and when it was checked.

    Nothing is described as validated unless gate 6 produced a record.

    Status: not recorded