Improving predictions of the location and use of warrens in sensitive rabbit populations.

The location and use of warrens is a key factor in the population dynamics of the European rabbit Oryctolagus cuniculus L., which can influence stability and persistence in the long term. Within the species' original distribution range, its numbers have declined sharply in recent decades, which is a serious problem for the conservation of Mediterranean ecosystems. Artificial warrens are commonly used to bolster rabbit populations and the characterization of natural warrens in those areas might improve their efficiency. In this study, we use binomial generalized linear models to identify the factors associated with the location and use of warrens in a low-density area and we evaluate the effect of including the spatial structure of the data in the model. In addition, we generate a map that predicts the most suitable areas for artificial warrens. Contrary to what was expected, habitat variables are only secondary factors, and the location and use of warrens is influenced mainly by spatial factors, such as proximity to nearby warrens. Furthermore, the aggregated spatial pattern of warrens suggests that, at the local scale, for example, the hunting estate, intra-specific interactions might be playing a primary role in these low-density populations. To identify the most suitable sites for artificial warrens and, thereby, improve the efficiency of artificial warren building for conservation purposes, information about spatial structures should be included in models predicting natural warrens of European rabbits. [ABSTRACT FROM AUTHOR]