# Data sources and modeling boundaries

The repository ships deterministic, project-authored fixtures so that tests,
examples, and benchmarks can run without network access or a private dataset.
The fixture names reference Paracel and Spratly reef examples used by the
project, but the coordinates, areas, flow parameters, and benchmark connector
are computational inputs maintained in the source tree rather than a copied
external survey dataset.

These fixtures are suitable for algorithm demonstrations, regression testing,
and reproducibility checks. They are not field-calibrated observations, a
claim about current ocean conditions, or a substitute for a quality-controlled
ecological data product.

For an applied study, replace the fixture arrays with documented observations
and record the source, license, coordinate reference system, temporal window,
quality-control procedure, and parameter calibration method. The simulation
and network APIs accept those replacements without changing the analysis
stages.

The benchmark intentionally uses a zero-advection diffusion control. This
isolates deterministic software behavior and keeps CI execution inexpensive;
it should not be interpreted as a physical forecast.