A hierarchical approach for estimating state‐specific mortality and state transition in dispersing animals with incomplete death records. Issue 4 (1st February 2023)

• Record Type: Journal Article
• Authors: Hodel, Florian H.; Behr, Dominik M.; Cozzi, Gabriele; Ozgul, Arpat
• Is Part Of: Methods in ecology and evolution. Volume 14:Issue 4(2023)
• Page Start: 1074
• Page End: 1091
• Publication Date: 2023-02-01
• Subjects: Bayesian inference -- capture–recapture -- capture‐recovery -- interval censoring -- multistate mark‐resighting -- Siler model -- state‐based mortality -- true survival; Ecology -- Periodicals; Evolution -- Periodicals; 577
• DOI: 10.1111/2041-210X.14069
• Languages: English
• ISSNs: 2041-210X

Abstract

Abstract: Unbiased mortality estimates are fundamental for testing ecological and evolutionary theory as well as for developing effective conservation actions. However, mortality estimates are often confounded by dispersal, especially in studies where dead‐recovery is not possible. In such instances, missing individuals (i.e. individuals with unobserved time of death) may have died or permanently emigrated from a study area, making inferences about their fate difficult. Mortality before and during dispersal, as well as the decision to disperse, usually depend on a suite of individual, social and environmental covariates, which in turn can be used to draw conclusions about the fate of missing individuals. Here, we propose a Bayesian hierarchical model that takes into account time‐varying covariates to estimate transitions between life‐history states and mortality in each state using mark‐resighting data with missing individuals. Specifically, our framework estimates mortality rates in two states (resident and dispersing state) by treating the fate of missing individuals as a latent (i.e. unobserved) variable that is statistically inferred based on information from individuals with a known fate and given the individual, social and environmental conditions at the time of disappearance. Our model also estimates rates of state transition (i.e. emigration) to assess whether a missing individual was more likely to have died or survived due to unobserved emigration from the study area. We used simulations to check the validity of our model and assessed its performance with data of varying degrees of uncertainty. Our modelling framework provided accurate mortality and emigration estimates for simulated data of different sample sizes, proportions of missing individuals, and resighting intervals. Variation in sample size appeared to affect the precision of estimated parameters the most. Our approach offers a solution to estimating unbiased mortality of both resident and dispersing individuals as well as the probability of emigration using mark‐resighting data with incomplete death records. Conditional on the availability of data on known‐fate individuals and relevant time‐varying covariates, our model can reconstruct the fate (death or emigration) of missing individuals. The modularity of our framework allows mortality analyses to be tailored to a variety of species‐specific life histories.