Population figures for wild species are quoted with a precision the underlying methods do not support, and understanding how they are produced explains why estimates change so much.
You cannot count them all
Except for a handful of species with very small ranges and populations, complete counts are impossible.
Which means every figure is an estimate produced by sampling and extrapolation, with a confidence interval that is frequently wide.
Reporting the central estimate without the interval is standard practice in coverage and it misrepresents what is known.
Mark and recapture
The foundational method.
Capture a sample, mark them, release them, then capture another sample later and observe what proportion carries marks.
If marked animals mix evenly with the population, the proportion recaptured indicates the total.
Which depends on assumptions — that marking does not affect survival or behaviour, that the population is closed during the study, that marks are not lost, and that every animal has equal capture probability.
Each assumption is violated to some degree, and more sophisticated models account for some of them.
Distance sampling
Observers travel transects and record animals seen and their distance from the line.
Since detection falls with distance, the pattern of observations allows estimation of how many were missed.
Which is used extensively for larger animals in open habitats, and it requires that animals on the line are always detected and that they do not move in response to the observer.
Camera trapping
Has transformed the field for elusive species.
Individual identification from markings — tiger stripes, leopard rosettes, whale flukes — allows mark-recapture without physical capture.
Which is why the species with the most reliable estimates are frequently those with individually distinctive patterns.
For species without distinctive markings, occupancy modelling estimates the proportion of an area occupied rather than absolute numbers.
Genetic methods
Samples from faeces, hair or environmental sources allow individual identification through genotyping.
Which permits mark-recapture analysis from material collected without seeing the animal at all.
Environmental DNA — genetic material shed into water or soil — can establish presence and increasingly abundance, and it is developing rapidly.
Why estimates change
A revised figure frequently reflects a better method rather than a changed population.
Which is a common source of confusion in coverage, where a higher estimate is reported as a recovery.
Distinguishing genuine trend from methodological change requires consistent methods over time, which is why long-running monitoring programmes are so valuable and so vulnerable to funding cuts.
The trend matters more than the number
An absolute figure with wide uncertainty is less informative than a trend measured consistently.
Which is why conservation assessments emphasise rate of decline over population size, and why a species can be listed as threatened despite substantial numbers.
Reading a figure
Ask what method produced it, over what area, and with what interval.
Ask whether it is comparable to the previous figure.
And treat single dramatic numbers with caution, since the methods rarely support that precision and the incentive to report a striking figure is considerable.
Detection probability
The concept underlying most modern methods, and it is the reason older counts are not comparable to newer ones.
Not every animal present is detected, and the proportion detected varies with habitat, weather, observer and season.
Which means a raw count is a minimum rather than an estimate, and comparing raw counts between surveys measures detection as much as abundance.
Occupancy and abundance models estimate detection probability explicitly from repeated visits, then correct for it.
That advance transformed the field, and it means many historical figures cannot be compared with current ones.
Aerial surveys
Used extensively for large mammals in open habitat, counting from aircraft along transects.
Which covers ground rapidly and undercounts consistently, since animals are missed under cover and in dense groups.
Correction factors derived from double-observer methods or from comparison with known populations address this partially.
Modelling and extrapolation
Estimates for large areas generally extrapolate from surveyed portions using habitat models.
Which introduces further uncertainty, since the relationship between habitat and density may not hold across the whole range.
Published estimates should state the method and the interval, and the primary sources generally do even where coverage does not.
Citizen science contributions
Volunteer recording produces enormous data volumes that professional survey cannot match.
Which is uneven in coverage and effort, and statistical methods now handle much of this, modelling observation effort explicitly.
Long-running structured schemes, where volunteers follow a defined protocol at fixed sites, produce data of quality comparable to professional monitoring at a fraction of the cost.
Which is why several national monitoring programmes depend on them entirely, and why their continuity matters.
Reporting
Population figures in coverage frequently omit the interval, the method and the area.
Which makes them uninterpretable, and the primary sources generally include all three.
Following a figure back to the paper that produced it takes a few minutes and is usually illuminating.