Modeling Individual Animal Histories with Multistate Capture–Recapture Models

摘要: Summary Many fields of science begin with a phase exploration and description, followed by investigations the processes that account for observed patterns. The ecology is no exception, recent decades have seen focus on understanding key underlying dynamics ecological systems. In population ecology, emphasis has shifted from state variable size to demographic responsible changes in this variable: birth, death, immigration, emigration. evolutionary some these same processes, rates birth are also determinants fitness. animal estimation variables their associated vital especially problematic because difficulties sampling such populations detecting individual animals. Indeed, early capture–recapture models were developed purpose estimating size, given reality all animals not caught or detected at any occasion. More recently, open draw inferences about survival face problems. paper multi‐state mark–recapture (MSMR), which first appeared 1970s but undergone substantial development last 15 years. These deal explicitly biological variation, different “states” (classes defined location, physiology, behavior, reproductive status, etc.) may probabilities detection. Animal transitions between states stochastic themselves interest. general proven be extremely useful provide way thinking remarkably wide range important processes. methods now stage refinement sophistication where they can readily used biologists tackle issues ecology. paper, we together information art multistate methods, explaining illustrating use. We modeling philosophy series principles how conduct analyses. cover features, anticipate ways expect develop years ahead. particular: – MSMR straightforward fashion biologists, thanks sound goodness‐of‐fit procedures, reliable parameter identifiability diagnostics, robust user‐friendly computer software.Constrained model selection procedures ANOVA‐like commonly over models, answer variety questions. as an example treatment meadow vole Microtus pennsylvanicus data. As random effects should integral part philosophy. Some simple approaches illustrated. States very way, example, combining several criteria, sites states, include nonobservable states. multisite recruitment roseate terns Sterna dougallii . appear natural framework sources information, viewed events organized into mutually exclusive alternatives. With available developments, becoming standard tool biology, shown rapid growth use literature. particular, ease constrained developed, less data hungry than was often feared. make it possible unify large array methodology, and, such, both step towards further unification “mother all” model, basis generalizations. Future developments concern generalizations reverse time approach estimation. “Multievent” accounting uncertainty determination, integrated state–space already full development.