Chapter 19 - Spatial Mark-Resight Models

Abstract

Throughout most of this book we have dealt with the situation where all individuals in the population are uniquely identifiable. In Chapter 18 we introduced and developed an SCR model for non-identifiable individuals. These two extremes are common in the study of animal populations with non-invasive sampling methods. However, there is also an intermediate situation where part of the population is tagged or otherwise marked, while some individuals are unidentifiable. So-called mark-resight models can be used to estimate population size and density by combining data from both the marked and unmarked individuals. In mark-resight studies a sample of individuals is captured and marked during a single marking event. Marking is followed by resighting surveys, in which both the detection of marked and unmarked animals is recorded. In this chapter we consider mark-resight within a spatial context and develop a spatial mark-resight (SMR) model. We first provide some background information on (non-spatial) mark-resight and the types of data such surveys can provide. The main distinctions hinge on whether or not we know the number of marked individuals, and whether we can correctly determine if an individual is marked or not. An essential assumption of mark-resight models is that the marked individuals are a representative sample of the study population. While this is also an implicit assumption of capture-recapture models, in mark-resight models this means that the process of marking individuals requires careful consideration in order to produce a random sample. In the context of SMR models, this assumption refers not only to the demographic composition, but also to the spatial distribution of the marked individuals in the state-space S. Throughout the central part of this chapter we will make the assumption that marked animals are a random sample from the population in S. We explore models for both known and unknown numbers of marked individuals, and for imperfect individual identification of marks, and approaches to incorporate telemetry location data. We present 2 applications: First, we apply the models with unknown number of marked individuals to a study of Canada geese from North Carolina. Second, we give an example involving radio telemetry applied to a study of racoons on the Outer Banks. We conclude the chapter by presenting some general strategies for addressing a situation where marked individuals are not a random spatial sample from the state-space. We expect that real life studies rarely meet this assumption, and the effect of this violation, as well as the development of models that relax the assumption, needs to be the focus of future work.