Components of Integrated Population Models

Abstract

We give an overview of the data types and associated models typically combined in an integrated population model (IPM). These models are the components or “ingredients” of an IPM. We focus on key models for surveys of population counts, productivity, and capture-recapture. We start with an overview of the main features of these data types that focuses on information aggregation and the observation model. The specified level of mechanistic detail differs widely among the typical ingredients in the observation model of an IPM. Currently, observation models tend to be the most mechanistic in likelihoods adopted for capture-recapture data and more phenomenological in likelihoods for population counts and productivity data. We demonstrate implementation of mechanistic observation models for population counts in an IPM in two steps: first, by obtaining abundance estimates corrected for imperfect detection using capture-recapture, distance sampling, occupancy, or N -mixture modeling, and second, by incorporating these estimates into a Gaussian state-space model. We discuss the following models as the main components of an IPM: Gaussian and “demographic” state-space models for population counts, Poisson regression and nest survival models for productivity data, and Cormack-Jolly-Seber, ring-recovery, and multistate models for capture-recapture data, for which we cover both state-space and m-array implementations and the modeling of age effects, random effects, and covariates. We end with an introduction to basic spatial capture-recapture because this model type will become more important for individual-based spatial IPMs in the future. We show how to deal with the case where instead of actual population counts, we have estimates of population size and want to use those along with their SEs in an IPM. Finally, in the modeling of productivity, we give solutions to the right censoring and left truncation of the counts of young.