Cooperative Ecosystem Studies Unit, Rocky Mountain CESU

archived

The purpose of the Northern Rocky Mountain Science Center (NOROCK) funding opportunity is to develop a new method for classifying bear sightings using probabilistic methods. The proposed approach will be based on modeling observations of bear, their movements and the numbers of cubs. In this approach, the true sighting history for bears is treated as an unobserved (latent) random variable that must be predicted. The current method sin effect selects one of the possible true sighting histories and then treats this as if it were known. In the proposed approach, the prediction of the latent sighting histories is an intermediate step in the estimation of bear abundance N. Importantly, the uncertainty in prediction which potential sighting history is the true one is carried over into the quantified uncertainty in N. Thus, the proposed method will be able to () correct for bias in estimation resulting from an arbitrary classification of bears, and (2) correct for underestimation of uncertainty in N resulting from unmodeled uncertainty in the determination of the true sighting history. To develop a hierarchical model for (1) information on locations and time of sightings of radio-collared bears and (2) locations and times from sightings of bears from observational flights and ground surveys of our study area, including both collared and un-collared bears. To fit this model using Bayesian model fitting procedures leading to inference about N. Key steps in the development and fitting of this joint model are the extraction of relevant data, the development of the algebraic structure of the joint model, and the writing of computer code for fitting the model to the data. It is assumed that the data will be provided by the relevant agencies supporting this research. Began an initial step towards developing the model by completing an initial model; however, further work that is required is the development of a model for changes in the numbers of cubs during the survey period and a spatial model describing the distribution of females with cubs of the year within the greater Yellowstone. A third stage of this project will involve developing a Markov chain Monte Carlo updater for Bayesian fitting of the model developed.

High