The Gold Standard of Ecology | Abundance–Occupancy Relationships, Rarity and Sampling Scale

1. Define abundance and occupancy on the same frame

Back to contents

Abundance may be a count, density, biomass or modelled intensity; occupancy is the proportion or number of sampling units used by a species. Both require a stated area, time window and unit. Mean abundance can be calculated across all sites or only occupied sites, producing different relationships. Before fitting a line, make every denominator and inclusion rule explicit.

2. Separate interspecific and intraspecific questions

Back to contents

An interspecific relationship compares species: are widespread species also locally abundant? An intraspecific relationship follows one species across places or times: do its occupancy and abundance rise or fall together? The mechanisms and dependence structure differ. Pooling them can turn a community comparison into a trend claim that the data do not support. State which points represent species, years, regions or populations.

3. Treat grain, extent and duration as part of the result

Back to contents

Larger cells are more likely to contain a species, larger extents include more environmental variation, and longer surveys create more opportunities for detection or temporary use. These changes can bend, flatten or tighten the relationship even when populations are unchanged. Compare species and years on common supports, or model the scale explicitly. A relationship estimated at one scale should not be transferred to another without validation.

4. Model detection before interpreting absence

Back to contents

Observed occupancy combines ecological use with the chance of recording the species. Rare, cryptic or seasonal organisms may be missed, while abundant conspicuous species are easier to detect. Repeat visits, replicated devices or integrated observation models can separate these processes. If detection rises with local abundance, raw presence–absence data can produce a positive relationship partly through observation rather than ecology.

5. Define rarity as more than one low number

Back to contents

Rarity can mean small geographic range, low local abundance, narrow habitat breadth or some combination. A species can be widespread but thinly distributed, or locally dense within a tiny range. Conservation priority should not be inferred from one corner of an abundance–occupancy plot alone. Report uncertainty, trend, detectability and exposure to threats alongside the chosen rarity dimensions.

6. Worked example: woodland birds under two grid systems

Back to contents

A monitoring programme summarises the same bird observations in one-kilometre and ten-kilometre cells. At one kilometre, several low-density species occupy few cells and the abundance–occupancy curve is steep. At ten kilometres, most species appear in many cells, occupancy approaches its ceiling and the curve flattens. One scarce bird also has low detection; repeat-visit modelling raises its estimated occupancy without changing count-based abundance. The community did not become less rare when the grid changed—the measurement support changed.

7. Test whether the relationship supports the decision

Back to contents

Refit the model across reasonable grains, extents, seasons and abundance definitions. Use spatial or temporal blocks for validation and inspect influential common species separately from rare species. Simulate detection and aggregation under a stationary community to see how much pattern the design can create. A monitoring shortcut is defensible only if changes in occupancy track the abundance quantity relevant to the decision.

8. Changed-context transfer check: camera surveys of mobile mammals

Back to contents

Move from breeding-bird squares to camera stations used by wide-ranging mammals. The transferable questions remain unit, duration, detection and abundance definition. The changed context makes occupancy closer to site use than permanent presence, and repeated photographs are not independent individuals. Longer deployments can raise recorded occupancy at constant density. The model must distinguish movement through a station from population size.

9. Know the limits and common misconceptions

Back to contents

A positive abundance–occupancy relationship does not prove a single causal mechanism, and it does not mean every widespread species is abundant. Saturation at high occupancy can hide abundance change. Range contraction and local density decline may occur at different rates. Cross-species analyses also inherit phylogeny, body size and habitat differences. Occupancy is useful, but it cannot automatically replace direct abundance evidence.

10. Audit both axes and the observation process

Back to contents

Publish site geometry, grain, extent, duration, effort, detection histories, abundance estimator, occupancy definition, inclusion rules, transformations and code. Show raw and modelled values with uncertainty. Repeat the analysis on common footprints and alternative supports, then document which species drive the slope. An audit should let readers see whether rarity is biological, observational or created by the map.

11. Frequently asked questions

Back to contents

Does high occupancy guarantee high abundance? No. Can occupancy track population change? Sometimes, but the relationship must be validated at the relevant scale. Why do larger cells raise occupancy? They include more space and opportunities for presence. Should non-detections count as absences? Only after accounting for observation. Is rarity one category? No; range, abundance, habitat breadth and trend describe different dimensions.

12. The gold standard keeps scale attached to rarity

Back to contents

A credible abundance–occupancy analysis defines both axes on the same sampling frame, distinguishes within-species trends from among-species patterns and models detectability. It tests how the result changes with grain, extent and duration before using occupancy as a proxy for abundance or rarity. The relationship becomes informative when its measurement support is part of the claim, not hidden beneath it.