Every product the course produces — every index, every comparison, every shared dataset — carries a short limitations statement. It is not boilerplate to be skipped; it is what makes the number trustworthy to anyone who later relies on it, because a number read without its limits is a number read wrongly. The statement reports, in plain language a non-specialist can act on, the terms of the uncertainty budget (Appendix A.6):
-
What it is. A relative index for detecting change over time at a place — not an absolute population, an absolute density, or a scientific-grade census.
-
The headline limit. Complete for conspicuous, recognizable reef-associated organisms; indicative only for small or cryptic ones (the conspicuousness limit, Module 5).
-
Position. The hub's location is from consumer-grade GPS, accurate to a few meters; this affects re-finding the place, not the count.
-
Counting. The count carries the two-observer spread as its counting-error range.
-
Visibility. An error in the detection radius enters the density roughly doubled (a 10% error in r is about a 20% error in density), so the calibration is the largest controllable error.
-
Representativeness. If the hub's reading is taken to stand for a wider area, that extrapolation is an added assumption, not a measurement.
-
The noise floor. The measured spread on unchanged structure, and the rule that change counts only beyond it.
-
Community structure (optional). Where the optional diversity layer is reported, it carries the same conspicuousness bias: the abundance-weighted indices (Shannon, inverse-Simpson) are robust, but richness is indicative only (Appendix A.9), and turnover is judged against its own compositional noise floor.
-
AI assistance (if used). If a model helped produce the count, the model and version are recorded, the human-confirmed count is the count of record, and the raw model count is logged so that model drift is never read as reef change (Module 9).
-
Conditions and comparability. The dive's conditions are recorded so that readings are compared like with like; a reading taken under conditions unlike the baseline's carries that difference as an added caveat. The survey is observational — it claims a trend, never a cause (Appendix C).
Limitations Statement — Template
A demographer completes this for each product; the bracketed fields are filled from the dataset:
This survey reports a relative index of reef-associated life at [hub reference], surveyed on [date(s)] in [visibility / conditions]. It is built to detect change over time at this place and is not an absolute population or density. The index is complete for conspicuous reef-associated organisms and indicative for small or cryptic ones. Hub position is from consumer-grade GPS (± a few meters). The reported index is [D̄ value] with a counting range of [two-observer spread]. The measured noise floor for this hub is [spread from re-occupation]; a future reading is treated as real change only if it differs by more than this. [If diversity reported: community diversity is Shannon [value] and inverse-Simpson [value] effective species; richness [value] is indicative only; turnover since the previous visit is Bray–Curtis [value] and Jaccard [value].] [If AI-assisted: the count was produced with model assistance ([model + version]); the human-confirmed count is the count of record.] [If extrapolated: this reading is taken to represent the surrounding area — an assumption beyond the surveyed stations.]
The governing principle of the whole analysis, and the line every demographer should be able to say plainly: claim the trend, state the budget — never claim the absolute population.