The generator is a stated model, not a fit. Each station carries four coefficients: vessel density, retention, anthropogenic nitrogen load, and legacy mercury. Producer tissue concentration rises with vessel density multiplied by retention, because leachate concentrates where boats are and water stays. Producer δ15N rises with the nitrogen load. Every level above the producer is that number multiplied by a transfer factor.
The three transfer regimes are the part worth arguing with, because they are the claim the buyer is being sold, and they are set by the evidence ledger rather than by taste. Mercury magnifies, which is the best-established result in the field. Zinc magnifies too, which is the correction this page shipped wrong in v0.1 to v0.4. Copper has no settled direction at all in a synthesis of 9,929 samples, so it is carried as unresolved with a wide band, and closing that band on one coast is the project. Crustaceans are the trap in the middle: a decapod holds copper and zinc near a constant for haemocyanin, so its box reports physiology. Those edges are still drawn, because hiding them would hide the trap, and they are held out of the transfer-regime statistic, which is the same conclusion stated twice.
| Element | Where it comes from | What replaces it |
|---|---|---|
| station coefficients | Hand-set, matching the eight stations already on the Lorica map | Measured vessel counts, LiveOcean residence time, the DGT array |
| δ15N baseline | Modelled from nitrogen load | EA-IRMS on the same Ulva tissue already going to the metals lab |
| trophic position | Post-style 3.4‰ enrichment per level | Site-fitted enrichment from the co-collected column |
| transfer factors | Directions set by Sun et al. 2020 per analyte (ledger T2 to T4), with a saturation term on copper that is this project's hypothesis and not a citation (T6) | Fitted per station per analyte from co-collected pairs, with confidence intervals |
| producer condition | Modelled growth against the light and temperature field | Tagged-thallus growth, which the protocol already collects for growth-dilution correction |
| assemblage composition | Modelled from vessel density and retention | Multi-marker metabarcoding, already input community_composition |
Where the claims are checked. Every regime, threshold and site fact on this page carries a row in projects/biofouling.evidence.md, with a verdict of VERIFIED, PARTLY-TRUE, INCORRECT, UNVERIFIABLE or UNVERIFIED and the correction where the claim was wrong. That ledger is what caught the zinc direction. The nine design-document claims in DESIGN.md Appendix A remain unchecked.
The rule this tool has to respect. DESIGN.md section 7.3 refuses arithmetic between accumulator classes, and a transfer factor is a cross-class ratio. The resolution used here: a factor is computed only within one station on one collection date, is carried with its uncertainty, and is compared between stations as rank order. Raw cross-class arithmetic stays forbidden.