External And Upstream-Source Checks

The external module compares independent representations at two boundaries.

GridMET Mirror

The packaged annual NetCDF came from the University of Idaho GridMET service. The module requests a tiny raw-value block for the same variable, year, date, and grid indices from the independent NCAR/GDEX THREDDS mirror. It compares the packed integers before scale/offset decoding, which avoids false agreement from rounding after conversion.

from cubedynamics.validation import ValidationPaths
from cubedynamics.validation.external import run_external_validation

paths = ValidationPaths.discover()
result = run_external_validation(
    paths,
    fire_id=20657,
    variable="tmmx",
    external_network=True,
)

The published run compared 27 raw cells and found zero residual for every one. Network access is optional so offline validation stays reproducible; a skipped mirror query is recorded explicitly in result.json.

FIRED Daily And Event Products

The event-level FIRED polygon is compared with the union of its 29 daily progression polygons in EPSG:5070. The sampled event has intersection-over-union 0.9999999827 and a Hausdorff distance of 0.000125 m, showing that the two upstream layers identify the same footprint to numerical projection tolerance.

External GridMET and FIRED checks

Sources And Reuse

GridMET is a roughly 4 km, daily surface meteorology product described by Abatzoglou (2013) and distributed through the University of Idaho GridMET catalog.
The independent comparison uses the NCAR/GDEX GridMET THREDDS collection.
FIRED source, citation, and cache guidance are documented on the FIRED dataset page.

The validation suite stores only small derived QA samples and plots in the repository. Users of the underlying source data remain responsible for the upstream citation and reuse terms.