Linking resource selection and mortality modeling for population estimation of mountain lions in Montana

Abstract

To be most effective, the scale of wildlife management practices should match the range of a particular species’ movements. For this reason, combined with our inability to rigorously or regularly census mountain lion populations, several authors have suggested that mountain lions be managed in a source-sink or metapopulation framework. We used a combination of resource selection functions, mortality estimation, and dispersal modeling to estimate cougar population levels in Montana statewide and potential population level effects of planned harvest levels. Between 1980 and 2012, 236 independent mountain lions were collared and monitored for research in Montana. From these data we used 18,695 GPS locations collected during winter from 85 animals to develop a resource selection function (RSF), and 11,726 VHF and GPS locations from 142 animals along with the locations of 6343 mountain lions harvested from 1988–2011 to validate the RSF model. Our RSF model validated well in all portions of the State, although it appeared to perform better in Montana Fish, Wildlife and Parks (MFWP) Regions 1, 2, 4 and 6, than in Regions 3, 5, and 7. Our mean RSF based population estimate for the total population (kittens, juveniles, and adults) of mountain lions in Montana in 2005 was 3926, with almost 25% of the entire population in MFWP Region 1. Estimates based on a high and low reference population estimates produce a possible range of 2784 to 5156 mountain lions statewide. Based on a range of possible survival rates we estimated the mountain lion population in Montana to be stable to slightly increasing between 2005 and 2010 with lambda ranging from 0.999 (SD = 0.05) to 1.02 (SD = 0.03). We believe these population growth rates to be a conservative estimate of true population growth. Our model suggests that proposed changes to female harvest quotas for 2013–2015 will result in an annual statewide population decline of 3% and shows that, due to reduced dispersal, changes to harvest in one management unit may affect population growth in neighboring units where smaller or even no changes were made. Uncertainty regarding dispersal levels and initial population density may have a significant effect on predictions at a management unit scale (i.e. 2000 km 2 ), while at a regional scale (i.e. 50,000 km 2 ) large differences in initial population density result in relatively small changes in population growth rate, and uncertainty about dispersal may not be as influential. Doubling the presumed initial density from a low estimation of 2.19 total animals per 100 km 2 resulted in a difference in annual population growth rate of only 2.6% statewide when compared to high density of 4.04 total animals per 100 km 2 (low initial population estimate λ = 0.99, while high initial population estimate λ = 1.03). We suggest modeling tools such as this may be useful in harvest planning at a regional and statewide level.