# An animal movement model incorporating home range and habitat selection

> How do you tell where a bear prefers to be from where it simply happens to wander? This 2008 paper built one model that estimates an animal’s home range and habitat preferences together from GPS collar data. They tested it on a female brown bear in Southeast Alaska during the salmon run, who strongly favoured large forest and avoided open, non-forested ground.

Summary by Kersey Lawrence, Original Wisdom. Updated 5 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2008-christ-ver-hoef-animal-movement-model-incorporating/

## Citation and access

- **Authors:** Aaron Christ, Jay Ver Hoef, Dale L. Zimmerman
- **Year published:** 2008 (published online 19 September 2007)
- **Journal or source:** Environmental and Ecological Statistics 15: 27–38
- **Full citation:** Christ, A., Ver Hoef, J., and Zimmerman, D.L. 2008. An animal movement model incorporating home range and habitat selection. Environmental and Ecological Statistics 15: 27–38.
- **DOI:** https://doi.org/10.1007/s10651-007-0036-x
- **Access status:** Hosted: U.S. government work, not subject to copyright in the United States
- **PDF:** https://www.originalwisdom.com/wp-content/uploads/2026/11/2008_Christ_and_Ver-Hoef_etal_an-animal-movement-model-incorporating-home-range-and-habitat-selection.pdf

## Tags

- **Species:** Mammals > Brown bear
- **Region:** North America > United States
- **Behaviour and ecology:** Behaviour; Habitat selection; Home range; Movement; Site fidelity
- **Methods and technology:** Bayesian statistics; Global Positioning System (GPS) collars; Markov chain Monte Carlo (MCMC); Resource selection function; Telemetry; Utilization distribution
- **Data type:** Existing data; Modelling or theoretical
- **Discipline:** Behaviour and ecology
- **Year published:** 2008

## Detailed keywords

- **Species:** Studied: brown bear (Ursus arctos), one adult female. Mentioned but not studied: salmon, as the food source during the study period.
- **Location and habitat:** Southeast Alaska, United States. Habitats: tidal flat, nonforest, clearcut, scrub forest, and small, medium and large forest, plus salmon streams.
- **Topics:** animal movement modelling, home range, utilization distribution, habitat selection and preference, resource selection, site fidelity, temporal autocorrelation in telemetry data, missing GPS locations.
- **Statistics or models:** A bivariate Gaussian random walk (a state-space vector autoregression) for movement, combined with a bivariate normal “fidelity” distribution for the home range and habitat preference terms. Bayesian estimation by Metropolis-Hastings Markov chain Monte Carlo (MCMC) in R, 50,000 iterations after burn-in. Flat priors, except normal priors on habitat parameters. Standardized preference coefficients comparable to resource selection coefficients. The 95% utilization distribution contours were drawn with a kernel smoother, with uncertainty bands from 100 posterior draws.
- **Data type:** A new statistical method, illustrated with existing GPS data. One female brown bear, with locations attempted every 20 minutes for 11 days during the October 2001 salmon run, gave 157 recorded locations. Based on part of A. Christ’s 2003 PhD dissertation at the University of Iowa.

## Synopsis

### 1. What they did

Wildlife biologists usually study two questions separately: where an animal lives (its home range), and which habitats it prefers within that area. The authors argue the two are linked, because animals move to reach the habitats they need. So they built a single model that estimates both at once. It also deals with a problem of frequent GPS locations: each location depends on the one before, so they are not independent.

The model combines three parts: a random step from the animal’s last location, a pull back toward the centre of its home range, and a preference for some habitats over others. It also copes with missed GPS locations.

They tested it on one female brown bear in Southeast Alaska. Her GPS collar was set to take a location every 20 minutes for 11 days during the October 2001 salmon run. Early-generation antenna and satellite problems meant many locations were missed, leaving 157. Habitat came from a map of 50 m squares with seven habitat types, plus distance to the nearest stream.

### 2. What they found

- Large forest was strongly preferred, and medium forest less so.

- Clearcut, scrub forest and small forest could not be told apart statistically.

- Nonforest was strongly avoided. The bear was never located there, although it was within reach.

- Tidal flats were left out, because they were too rarely within reach to estimate a preference.

- The likelihood of the bear’s next location dropped by about 1% for every 50 m away from a stream.

- The model also produced a home range map (a 95% contour) with uncertainty bands. The outer band was larger than the data alone suggest, because the locations were strongly linked over a short period.

### 3. What they concluded

- Movement, home range and habitat preference can be estimated together in one model. This avoids the researcher deciding, often arbitrarily, which habitat counts as “available” to the animal, and it accounts for linked GPS locations.

- Limitations: the method becomes slow with many locations or many habitat types. Home range shape and habitat effects can be confused when habitats run in long strips, as here. Some estimation issues remained under investigation.

- Next steps: a model covering several animals at once, to estimate selection for both individuals and populations. The authors provide their R code.

### 6. Other technologies, skills or innovations

- A GPS collar set for 20-minute locations; early antenna and satellite problems caused many missed locations

- A habitat map in a Geographic Information System (GIS) raster of 50 m squares, including distance to streams

- Bayesian Markov chain Monte Carlo (MCMC) estimation written in R

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

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

- The authors note that telemetry shows where an animal was, but not what it was doing there: feeding, sleeping or just passing through. Tracks and signs at those locations can supply exactly that, for example fishing sign at salmon streams or bed sites in large forest.

## Cite this summary

Lawrence, K. 2026. Summary of: An animal movement model incorporating home range and habitat selection (Christ and Ver Hoef et al., 2008). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2008-christ-ver-hoef-animal-movement-model-incorporating/