Cooperative Fish and Wildlife Research Units Program: Alaska
Education, Research and Technical Assistance for Managing Our Natural Resources

Alaska Project

Evaluating Sources of Bias in Caribou Population Surveys

August 2021 - December 2022


Participating Agencies

  • HQ

Understanding how to efficiently monitor wildlife is critical for proper harvest and habitat management. Monitoring of caribou herds in Alaska is almost exclusively accomplished through aerial surveys conducted during the pre-calving season. These surveys rely on locating radio-marked animals in order to find large aggregations of caribou across the landscape and using photo-census techniques to estimate group size. These data are then used in a specific mark-recapture framework, known as a Rivest model, to generate estimates of unseen caribou and, thus, total herd size. However, this approach is likely sensitive to grouping behavior (i.e. the distribution of group sizes among the entire herd) and both the total number and distribution of radio collared animals throughout the herd. The degree to which these factors influence the precision and accuracy of caribou abundance estimates is unknown and could have substantial impacts on our understanding and resulting management of caribou herds.
We will use empirical data on observed group sizes and collar distribution from 11 aerial surveys of the Western Arctic Herd and 12 aerial surveys from the Teshekpuk herd conducted between 1984 and 2019 to parameterize a series of simulations to evaluate accuracy and precision of Rivest estimates. We will use a parametric bootstrapping approach to simulate 10,000 realizations of true population size, group size distribution, and collar distribution from which we will use empirical estimates of detection probabilities to randomly detect both marked (with collars) and unmarked (without collars) groups. We will then use these simulated surveys results in a Rivest framework to generate estimates of abundance that we will compare to true (simulated) abundance for bias and confidence interval coverage.
By identifying the effects of grouping behavior and sample size on the reliability of Rivest estimates we hope to improve survey approaches and resulting inferences. Understanding situations in which Rivest estimates may be unreliable could 1) reduce surveys efforts by eliminating potential surveys conducted under non-ideal circumstances, and 2) lead to improved sampling protocols by identifying optimal sampling situations. Because caribou populations are known to exhibit substantial fluctuations, it is critical that inferences regarding population dynamics be derived from reliable estimates of abundance through time.