Doc 4 of 9 · ← 3. State guides · Index · Next: 5. Daily operations →
ShellCast analysis depends on processed spatial inputs, not raw upstream datasets. For each state, organizations publish growing areas, leases, and related boundaries on their own schedules. Those exports change over time — geometry, seasonality rules (notably in Florida), file names, and attribute field names can all shift. ShellCast cannot read those files directly; they must be prepared so the pipeline can load shapefiles under data/pqpf/{nc,sc,fl}/inputs/, identify shellfish growing units (CMUs) and leases, and write forecasts to the database.
All three states need periodic input updates when upstream data changes. Today that work is manual (or ad hoc scripts on an operator machine). Fully automating fetch → process → deploy would be ideal but is not part of this repository.
Florida previously experimented with an automated ArcPy pipeline and optional Google Cloud Storage upload so the analysis server could pull fresh files. Upstream endpoint and schema changes made that path unreliable. Local data/pqpf/fl/inputs/ (same idea as NC and SC) is the practical choice for daily runs — see 02-CONFIGURATION.md ([gcp.bucket]).
North Carolina and South Carolina input updates follow similar manual steps (replace shapefiles under data/pqpf/nc/inputs/ and data/pqpf/sc/inputs/ when revisions are available).
File names, schemas, and the production inventory live in the local companion INPUT_DATA_INTERNAL.md (gitignored; not in the public repo).
Some deployments still support downloading prepared Florida inputs from a bucket before PQPF runs. Operators who use local files only should not call get_input_files() in the Florida analysis flow.
- Consider an open-source GIS stack for any future automated prep workflow.
NC
Shellfish Growing Area Classifications