# Estimating Site Occupancy and Abundance Using Indirect Detection Indices

> Tracking tubes, scent stations and snow track surveys are cheap ways to learn which animals are around, but how much can footprints really tell us about numbers? Two U.S. Geological Survey scientists built a model that turns track-station records into estimates of occupancy and abundance. They show that occupancy estimates hold up well, while one set of scent stations placed too close together counted 508 coyotes where a later mark-resight study estimated 73.

Summary by Kersey Lawrence, Original Wisdom. Updated 9 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2005-stanley-royle-estimating-site-occupancy-abundance/

## Citation and access

- **Authors:** Thomas R. Stanley and J. Andrew Royle
- **Year published:** 2005
- **Journal or source:** Journal of Wildlife Management
- **Full citation:** Stanley, T. R. and Royle, J. A. 2005. Estimating site occupancy and abundance using indirect detection indices. Journal of Wildlife Management 69(3): 874-883.
- **DOI:** 10.2193/0022-541X(2005)069[0874:ESOAAU]2.0.CO;2
- **Access status:** Hosted: U.S. government work, not subject to copyright in the United States
- **PDF:** https://www.originalwisdom.com/wp-content/uploads/2026/11/2005_Stanley_and_Royle_estimating-site-occupancy-and-abundance-using-indirect-detection-indices.pdf

## Tags

- **Species:** Mammals > Coyote; Mammals > Deer mouse; Mammals > Southern red-backed vole
- **Region:** North America > Canada; North America > United States
- **Tracking:** Sign surveys; Tracking; Tracks; Tracks and signs identification
- **Behaviour and ecology:** Behaviour
- **Methods and technology:** Akaike Information Criterion (AIC); Detectability; Occupancy modelling; Population estimation; Royle–Nichols abundance model; Simulation
- **Data type:** Existing data; Modelling or theoretical
- **Observer qualification:** Not needed
- **Discipline:** Wildlife tracking
- **Year published:** 2005

## Detailed keywords

- **Species:** Studied (example data sets): Clethrionomys gapperi, Peromyscus maniculatus and shrews (Sorex spp.) from tracking tubes; coyote (Canis latrans), deer, rodents and lagomorphs (rabbits and hares) from scent stations. The paper gives no common names for the small mammals and no species for the deer, rodents or lagomorphs. Mentioned from other studies: bobcat (Lynx rufus), fisher (Martes pennanti), marten (Martes americana).
- **Location and habitat:** Example 1: boreal forest region of northern Nova Scotia, Canada (tracking tube data from Nams and Gillis 2003). Example 2: Rocky Mountain Arsenal, Colorado, United States (scent station data from Hein and Andelt 1994).
- **Topics:** indirect detection indices (IDIs): scent stations, tracking tubes, tracking plates, snow track surveys and hair snares; detection stations that can be reset; site occupancy; abundance estimation; detection probability and the length of the sampling interval; overdispersion; station spacing and double counting; effective area sampled; study design for track-based surveys; trap-happy response (tube use rising over time); carcass attractants.
- **Statistics or models:** Maximum likelihood extension of the Royle and Nichols (2003) occupancy-abundance model (itself a generalization of the MacKenzie et al. 2002 site-occupancy model), adding the length of the sampling interval to detection probability; Poisson and negative binomial abundance models; logit links for detection covariates and log links for abundance covariates; Monte Carlo simulations (1,000 repetitions); Akaike Information Criterion (AIC) model selection; Hessian condition number; moment estimators for a Poisson-Poisson mixture model (proposed extension). Possible inconsistency: in Table 4 (scent stations) the β̂0 values (0.04 to 0.28) are small and positive, which looks more like detection probabilities than values on the logit scale described in the text; the paper does not explain this.
- **Data type:** Modelling or theoretical (new model and simulations), plus reanalysis of existing data: 12 lines of 80 tracking tubes (960 tubes) checked 5, 6 or 8 times over 4 weeks (Nams and Gillis 2003; dates not given), and 60 scent stations checked daily for 2 weeks (Hein and Andelt 1994; dates not given).

## Synopsis

### 1. What they did

Counting animals directly, by capture and recapture or by distance sampling, is often too slow, costly or difficult for large areas or elusive species. So researchers often use indices instead, including indirect ones based on what animals leave behind, such as tracks or hair. The trouble is that an index only reflects abundance if the link between the two is known and constant, which is almost never true.

The authors focused on a group of methods they call indirect detection indices (IDIs): detection stations that record visits over a period and can then be fully “reset” so old evidence is not counted again. Examples are scent stations (a lure in a circle of fine tracking material), tracking plates (sooted aluminium plates), tracking tubes (tunnels with a paper strip that records footprints), hair snares and snow track transects.

They extended an existing statistical model (Royle and Nichols 2003) so that the chance of detecting an animal at a station depends on both the number of animals there and the number of days between checks. They then:

- Tested the model with computer simulations, using 100 or 200 stations, each checked 6 times at intervals of 2 to 12 days, under different detection rates and abundances, with and without extra variation in abundance between stations.

- Reanalysed tracking tube data for small mammals in northern Nova Scotia, Canada (960 tubes), asking whether animals used the tubes more the longer they had been in place.

- Reanalysed scent station data from the Rocky Mountain Arsenal, Colorado (60 stations, about 1.1 km apart), asking whether a deer carcass at a station increased visits by coyotes, deer, rodents and lagomorphs.

### 2. What they found

Simulations

- Estimates of site occupancy (the share of stations where the species is present) were essentially unbiased and robust, whichever abundance model was used.

- Abundance estimates were sensitive to the model. When abundance varied more than the simple (Poisson) model allows, abundance was underestimated by about 30%. A model allowing for that extra variation (negative binomial) did much better, except when detection was very low, where estimates became unstable.

Tracking tubes

- For all three groups (Clethrionomys gapperi, Peromyscus maniculatus and Sorex shrews), the best model showed tube use increasing the longer tubes had been in place, in agreement with the original study.

- Estimated occupancy was 0.36 for C. gapperi, 0.44 for P. maniculatus and 0.56 for Sorex, with estimated totals of 566, 557 and 950 animals across the 960 tubes.

Scent stations

- Coyotes visited stations with a deer carcass more often; deer, rodents and lagomorphs did not, again in agreement with the original study.

- Estimated occupancy was 0.55 for lagomorphs, 0.96 for rodents, 0.88 for deer and 1.00 for coyotes.

- The abundance estimate for coyotes was 508, far above the 73 coyotes estimated by mark-resight in the same area less than a year later. The deer estimate also looked too high. The authors explain this by station spacing: coyotes can travel 4 km a night, but stations were only about 1.1 km apart, so the same animals were counted at several stations.

### 3. What they concluded

- The model gives a proper link between track-station data and occupancy and abundance, so these data no longer have to be treated as indices with an unknown relationship to the population.

- It offers a flexible way to test factors (such as habitat, time or attractants) that affect visits, occupancy and abundance.

- Occupancy estimates are reliable, but abundance estimates depend heavily on the model’s assumptions, so readers should examine those assumptions closely if abundance is the goal. Where possible, the focus should be on factors affecting abundance rather than absolute numbers.

- For a good study design: usually more than 100 stations; stations that can be completely reset at every check; more than 5 checks; and stations far enough apart that one animal cannot visit more than one station. One way is to space stations further apart than the average length of a home range plus 2 standard deviations. For snow track surveys along roads, they give the example of recording tracks crossing a 1.6 km road segment, then travelling about 8 km before the next segment.

- Abundance estimates could be checked by more intensive methods at a random subset of stations (double sampling).

- Future work: using counts of visits rather than just presence or absence, and the problem that home range size, and so the area each station samples, may change with density.

### 4. How tracking was or was not used

Track-based survey data are at the heart of the paper. The tracking tube data came from footprints left on paper strips as animals passed through, with species recorded where they could be determined, and the scent station data came from tracks in a circle of fine material around a lure. The authors did not collect field data themselves; they reanalysed other researchers’ data. The paper notes that a station’s visits can disappear before it is checked (for example, snow tracks blown away by wind), which would bias detection estimates. No trailing was involved.

### 5. Tracker or observer certification

Not reported. The paper does not say who identified the tracks in either data set. Species identification from tracks underlies all the results, and the original studies are not described on this point.

### 6. Other technologies, skills or innovations

- A maximum likelihood model that builds the length of the sampling interval into detection probability, so stations checked at different intervals can be analysed together.

- Monte Carlo simulation to test how well the model performs before using it on real data.

- A suggested check for unstable estimates: the condition number of the Hessian matrix.

### 7. Implications for conservation (from the paper)

- The model can help managers learn the distribution and abundance of species that are usually only detected indirectly, by tracks or hair, at small and large scales, and identify factors that influence them.

- The scent station example is a warning: stations placed too close together for wide-ranging animals can greatly overestimate abundance.

### 8. Implications for tracking (Original Wisdom's observations)

_These are Original Wisdom's observations, not the authors'._

- The model assumes every visit is recorded correctly and survives until the station is checked. Track identification and track survival (wind, rain, snow) are therefore built into every estimate, which makes the quality of track reading at stations as important as the statistics.

## Cite this summary

Lawrence, K. 2026. Summary of: Estimating Site Occupancy and Abundance Using Indirect Detection Indices (Stanley and Royle, 2005). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2005-stanley-royle-estimating-site-occupancy-abundance/