Spatial Statistical Methods in the Reconstruction of Badger Territories

For decades, Eurasian badgers (Meles meles) have been the object of several studies trying to explain their primitive social organization, feeding, territory and lately, their relationship with bovine tuberculosis which can cost £1bn over the next 10 years. Badgers spend the day sleeping in their setts and foraging during the night. They live in clans, sharing and defending a communal territory but foraging and feeding individually. Several attempts to explain what influences the size and shape of badgers’ territories have been made, considering, for example whether they are determined by the dispersion of resources or by the location of the main sett which consists of several holes with large spoil heaps and obvious paths emanating from and between sett entrances. Since badgers use communal latrines to mark their territories, another approach is to use statistical methods based on this information to delineate their territories.
A common method employed to reconstruct badger territories from latrines is the Min- imum Convex Polygon (MCP), another approach classifies the latrines as hinterland, boundary or extraterritorial excursions, based on elements surrounding them (fences, badgers paths, etc.). The use of extra information such as the presence of other latrines in the same direction from the main sett, can provide more robust models that can be used not only in a point estimation approach but in a sampling approach that uses the probability distribution fitted by a model and permits to quantify the uncertainty in the reconstruction of the territories.
This thesis consists of 7 chapters: Chapter 1 is this Introduction. Chapter 2 is the literature review which looks at badger ecological behaviour, techniques used to obtain information about the territories of the badgers and methods used to reconstruct them. Chapter 3 is the Unconditional Outlier Prediction Model (UOPM). It is an extension of an unpublished paper that uses a logistic regression to estimate the probability that a latrine is part of the territory or, alternatively, is an extraterritorial excursion. This information is used with the 100% MCP to make the reconstructions. Chapter 4 talks about the Conditional Outlier Prediction Model (COPM), which is an extension to the UOPM that uses Gibbs sampling in the reconstruction of the territories to allow for dependance between latrines. This model uses the 100% MCP of the sampled latrines which are not outliers, in order to map the reconstructed territory at each iteration. Chapter 5 presents the Unadjusted Ordinal Model (UOM). This model uses the original classification from the 2010 baitmarking Woodchester Park Badger Survey made by the Food and Environmental Research Agency (FERA), applying a cumulative ordinal model to estimate the probability distribution fitted by the model using a sampling approach to reconstruct the territories. All the previous chapters employ information from only the territory being reconstructed; Chapter 6 adjusts the probabilities obtained by a territory and the territories sharing at least one latrine with it, to reconstruct the territory using a sampling approach. The last chapter discusses the results of the methods proposed in this research.