Brown bear population trends from demographic and monitoring-based estimators

A primary goal of monitoring wildlife populations is the estimation of population growth rate, ë. Two common methods by which biologists estimate λ are demographic studies of marked individuals, which tend to be expensive and labor-intensive, and estimators derived from time series of population indices. We compare grizzly bear (Ursus arctos) population growth rates in the Banff ecosystem (Alberta, Canada) from a published demographic study to estimates from concurrent monitoring of an index of population size, the number of females with cubs-of-the-year (F_cub). We estimated population trends by transforming the index into 2 population estimators (bias-corrected Chao and summation), and used each to estimate λ. The 95% confidence intervals of λ from the 2 monitoring-based estimators overlapped the point estimate of the demographic study. Precision of the bias-corrected Chao estimator was very low (95% CI of λ = 0.572-1.679); its application to the time-series used here is essentially fruitless. Precision of the summation estimator (95% CI of λ = 0.847-1.137) and the demographic study (0.99-1.09) were higher, but the CI of the former at least could be artificially narrow. Because all estimates were close to 1.00, the long-term fate of this population may depend critically on subtle changes in growth rate and on environmental stochasticity. Given that long-term demographic studies are not feasible in this system, population monitoring may be a worthwhile way to assess population dynamics. However, given the low power of many monitoring techniques to detect trends and the low precision of the F _cub estimators in particular, long time-series and explicit measures to remove sampling variance should be employed to increase trend estimate precision.