Genetic Monitoring for Managers


Type of GEM

Category 1a GeM Project Example: Estimating Abundance

Estimating abundance of grizzly bears in northwestern Montana

Kendall et al. 2009

Background: The grizzly bear (Ursus arctos) population in the Northern Continental Divide Ecosystem (NCDE) of northwestern Montana has been managed for recovery since being listed under the U.S. Endangered Species Act in 1975, yet no rigorous data had been available to evaluate the program's success. In 2002, members of the NCDE Managers Subcommittee identified an abundance estimate as their top research priority. Study design, planning, and initial field work were completed in 2002-2003, with ecosystem-wide sampling occurring in 2004.

Noninvasive genetic sampling (NGS) was selected as the preferred approach to estimate abundance in this area due to the remote, rugged nature of the terrain, and the low sightability of bears due to heavy forest cover. The USGS was selected to lead this project based on extensive experience with NGS in this region (Kendall et al. 2008).

To avoid issues of geographic closure violation, Kendall et al. (2009) sampled essentially all lands thought to be occupied by grizzlies in this population, approximately 31,400 km2 (7.8 million acres) of occupied grizzly bear habitat in the greater NCDE.

ndgbp study area montana 2
The blue area reflects the estimated extent of lands occupied by grizzlies associated with the NCDE in 2004, all of which was sampled in this study.

Lands were managed under numerous agencies, designations, and regimes, including Glacier National Park, parts of five national forests, parts of the two Indian Reservations, and significant amounts of state and private land.

They used two concurrent sampling methods to collect bear hair samples for genetic analysis and mark-recapture modeling of abundance. Boulanger et al. (2008) found through simulations based on empirical data that multiple sampling methods decreased the bias of estimates, while increasing the precision. Sampling was conducted between 15 June - 15 September, 2004.

ndgbp study area hairtapsThe primary sampling effort was systematically distributed hair traps (Woods et al. 1999). A hair trap consists of a single strand of barbed-wire strung around a series of trees or posts at a height of 50 cm with a liquid lure applied in the middle. One hair trap was set in each 49 km2 grid cell, four times. Hair traps were subjectively placed to maximize detection rates of bears, factoring in habitat quality, accessibility, proximity to other hair traps, while maintaining a safe distance from areas used by humans.

Crews deployed 2,558 hair traps and collected 20,785 hair samples.


ndgbp study area bearrubsThe secondary sampling method was to collect hair samples left by bears as they rub on trees, posts, power poles, etc. This is a natural behavior, and no attractant was used. Most bear rubs were located along trails and other recognized travel routes. Bear rubs were surveyed on approximately 80% of the study area; the majority of the unsampled area was east of the mountains on open grasslands.

Crews conducted 18,021 visits to 4,795 bear rubs (avg. 3.8 surveys/rub), and collected 12,956 hair samples.

To take advantage of three data types (hair trap, bear rubs, and bears known through management activities, they used a stepwise a priori approach to mark-recapture model development. To determine the best structure for each data type, they initially modeled hair trap and bear rub data separately, pooling the other two data types and using them as the first sample occasion for each exercise. For example, in the hair trap models, bear rub and physical capture detections were pooled as the first sample session followed by the four hair trap sessions. They then combined the most supported hair trap and bear rub models into a single analysis which included encounter histories for each of the 563 bears.

Support for candidate models was based on the sample size-adjusted Akaike Information Criterion for small sample sizes (AICc; Burnham and Anderson 2002). The model-averaged population estimates (based on their support in the data as indexed by AICc weights) was 765 (95% CI: 715-831) grizzly bears.

This study represents the first scientifically defensible abundance estimate for this threatened population. Beyond abundance, the large genetic dataset generated through NGS allowed the authors to investigate genetic connectivity, distribution, and relative density patterns, and demonstrates the utility of genetic monitoring at the population scale.