In this latest post, Jesse Whittington shares insights into new models for monitoring wildlife, including grizzly bears. Don’t forget to watch the video of some bears captured on camera! Whittington and colleagues’ Open Access article, Generalized spatial mark–resight models with an application to grizzly bears is available in the Journal of Applied Ecology.

Wildlife managers around the world strive to estimate population abundance and the related metric density.  This basic information is required to understand ecological processes, assess the status of threatened species, monitor invasive species, and set harvest quotas.  Yet, estimates of abundance can be difficult, labour intensive, and expensive especially for species that occur in remote, rugged, and forested landscapes.  For example, estimates of grizzly bear density using DNA-based capture-recapture methods can cost millions of dollars.  We developed a new generalized spatial mark-resight model that in many cases can substantially reduce costs required to monitor grizzly bears and other wildlife populations.

Remote cameras are increasingly used to monitor wildlife populations because they are relatively inexpensive and non-invasive.  Remote cameras have been used to the estimate density of species such as tigers in which individuals can be identified by their distinctive colouration and coat patterns.  In these studies, repeat detections of individuals are required in spatial capture-recapture models to jointly estimate encounter rates at cameras, home range size, and density.  Remote cameras have rarely been used to estimate density of cryptic populations because of challenges associated with identifying individuals.  However, Chandler and Royle developed spatial mark-resight models for wildlife populations where some of the animals are identifiable.  These models have been successfully applied to species such as puma where some individuals were identified by natural marks and scars.

We found that these conventional spatial mark-resight models should rarely be used in radio-collar and tagging studies.  Conventional spatial mark-resight models assume that the marked (identifiable) and unmarked (not identifiable) animals share the same spatial distribution and thus average encounter rates both within the study area and in the surrounding state-space.  This assumption will likely be met when identifying individuals with natural marks.  However, when animals are captured for radio-collaring and tagging, the study area will have a higher density of marked animals compared to the surrounding state-space.  The differing spatial distribution of marked and unmarked animals will bias density estimates.

bears 3
Remote camera image of marked grizzly bear

We developed a generalized spatial mark-resight model that included sub-models for both the marking process (e.g. capture rates with traps) and the resighting process (e.g. detection rates on cameras), thus relaxing the assumption of equal spatial distributions.  Our simulation study clearly found that conventional spatial mark-resight models, which lacked the marking sub-model, produced negatively biased density estimates whenever animals were trapped and marked along a linear transect or on a grid of traps within the study area.  Conversely, the generalized spatial mark-resight model produced unbiased density estimates with nominal credible interval coverage in all scenarios.

bears 2
Remote camera image of unmarked grizzly bears

We applied the spatial mark-resight models to three years of grizzly bear remote camera and GPS radio-collar data collected in Banff, Kootenay, and Yoho National Parks, Canada.  Our remote cameras recorded 614 detections of 22 radio-collared grizzly bears and 1,354 detections of unmarked bears.  The conventional spatial mark-resight models erroneously estimated a 51% decline in density compared to previous DNA-based estimates.  Conversely, our new generalized spatial mark-resight model suggested the population was relatively stable, which is important for the conservation of grizzly bears.

We are excited about the ability to monitor wildlife populations using the combination of remote cameras, radio-collars or tags, and generalized spatial mark-resight models.  Given the explosion of remote cameras and the large number of radio-collaring studies, population density in many cases could be estimated with minimal additional costs.

The full paper, Generalized spatial mark–resight models with an application to grizzly bears is available in the Journal of Applied Ecology.