Modelling the distribution of badgers Meles meles: comparing predictions from field‐based and remotely derived habitat data

Abstract

ABSTRACT Environmental heterogeneity is important in determining the distribution and abundance of organisms at various spatial scales. The ability to understand and predict distribution patterns is important for solving many management problems in conservation biology and wildlife epidemiology. The badger Meles meles is a highly adaptable, medium‐sized carnivore, distributed throughout temperate Eurasia, which shows a wide diversity of social and spatial organization. Within Britain, badgers are not only legally protected, but they also serve as a wildlife host for bovine tuberculosis Mycobacterium bovis. An evaluation of the role of badgers in the dynamics of this infection depends on understanding the responses of badgers to the environment at different spatial scales. The use of digital data to provide information on habitats for distribution models is becoming common. Digital data are increasingly accessible and are generally cheaper than field surveys. There has been little research, however, to compare the accuracy of models based on field‐derived and remotely derived data. In this paper, we make quantified comparisons between large‐scale presence/absence models for badgers in Britain, based on field‐surveyed habitat data and remotely derived digital data, comprising elevation, geology and soil. We developed four models: 1980s badger survey data using field‐based and digital data, and 1990s badger survey data using field‐based and digital data. We divided each of the four datasets into two subsets and used one subset for training (developing) the model and the other for testing it. All four training models had classification accuracies in excess of 69%. The models generated from digital data were slightly more accurate than those generated from field‐derived habitat data. The high classificatory ability of the digital‐based models suggests that the use of digital data may overcome many of the problems associated with field data in wildlife‐habitat modelling, such as cost and restricted geographical coverage, without any significant impact on model performance for some species. The more widespread use of digital data in wildlife‐habitat models should enhance their accuracy, repeatability and applicability and make them better‐suited as tools to aid policy‐ and decision‐making processes.