State space and movement specification in open population spatial capture–recapture models

Abstract

With continued global changes, such as climate change, biodiversity loss, and habitat fragmentation, the need for assessment of long‐term population dynamics and population monitoring of threatened species is growing. One powerful way to estimate population size and dynamics is through capture–recapture methods. Spatial capture (SCR) models for open populations make efficient use of capture–recapture data, while being robust to design changes. Relatively few studies have implemented open SCR models, and to date, very few have explored potential issues in defining these models. We develop a series of simulation studies to examine the effects of the state‐space definition and between‐primary‐period movement models on demographic parameter estimation. We demonstrate the implications on a 10‐year camera‐trap study of tigers in India. The results of our simulation study show that movement biases survival estimates in open SCR models when little is known about between‐primary‐period movements of animals. The size of the state‐space delineation can also bias the estimates of survival in certain cases.We found that both the state‐space definition and the between‐primary‐period movement specification affected survival estimates in the analysis of the tiger dataset (posterior mean estimates of survival ranged from 0.71 to 0.89). In general, we suggest that open SCR models can provide an efficient and flexible framework for long‐term monitoring of populations; however, in many cases, realistic modeling of between‐primary‐period movements is crucial for unbiased estimates of survival and density.

Sex-specific population dynamics and demography of capercaillie (Tetrao urogallus L.) in a patchy environment

“…This activity center model removes a desirable feature of open population SCR models with explicit movement models—that exposure to capture in year t+1 depends on the individual activity center location in year t. Removing this dependence erodes the ability to separate movement from survival and recruitment. However, Gardner et al (2018) demonstrate that survival (and by implication, recruitment) can be estimated without bias using an independent activity center model when the state space extent is known and sampled exhaustively. We believe our inferences under the patch-constrained independent activity center model should be reliable for this data set because 1) we estimated that we detected >90% of all individuals in each year surveyed, 2) the scat survey covered the entire state space—the area in which cappercaillie were documented to occur between 1978 and 2009, and 3) we observed evidence of very little movement between patches that would allow individuals to enter and leave this set of 8 patches (discussed further below).…”

Bayesian model selection for spatial capture–recapture models

“…It is specially of concern in the context of the models we study here in that investigators may be forced to integrate out S (activity centers of individuals) in order to construct tractable likelihoods and thus oversimplifying ecological reality. For example, by retaining the activity centers it is possible for investigators to confront open SCR models (Gardner, Sollmann, Kumar, Jathanna, & Karanth, ). We therefore hope that our study motivates the continued use of Bayesian methods by investigators for all its advantages, instead of opting out of them merely for the sake of using simple model selection methods.…”

Estimating abundance with interruptions in data collection using open population spatial capture-recapture models

“…Movement of ACs between occasions is an important feature of OPSCR models and a miss-specified movement process can have important consequences for inferences (Ergon and Gardner 2014, Gardner et al 2018). For the purpose of this study, we developed a Markovian movement model assuming distance between consecutive individual ACs being normally distributed.…”

“…In order to distinguish between temporary emigration and mortality, we integrated a movement model in the OPSCR model allowing shifts of individual activity centers between occasions. This is an important component of the OPSCR model as it can improve survival estimates and can take into account the impact of animals moving within and out of the sampled area (Ergon and Gardner 2014, Gardner et al 2018). It is a particularly important feature of the model in the context of sampling interruption, as it helps propagating spatial locations of individual across occasions.…”

“…OPSCR models do not only use information from individual detection collected during several occasions (such as CR models), but also use the spatial information contained in the detections and model movement of individuals between occasions (Ergon and Gardner 2014, Royle et al 2014, Bischof et al 2016a). In OPSCR, modelling individual movement between occasions allows estimating the probability of the individual being alive but outside of the study area, which should facilitate abundance estimates (Ergon and Gardner 2014, Gardner et al 2018), especially during the gap years. The use of data collected over multiple occasions and propagation of individual information on spatial location and demographic status across time steps should help OPSCR bridge gaps in data collection, allowing inferences about abundance during occasions with sampling interruption.…”