New insights into the uncertainty of spatial models in physical geography

Spatial modelling has an important role in contemporary physical geography. Models are increasingly used in a range of applications in biogeography and geomorphology. Sample size, geographical attributes of species and modelling algorithms have been recognized as major sources of uncertainty in spatial modelling. In this thesis, particular attention was paid to evaluating the effects of these uncertainty sources on the quality of models. In addition, the ability of consensus methods to improve the accuracy of an ensemble of predictions was assessed. The results of this thesis demonstrated that ca. 200 observations is a minimum sample size in periglacial modelling exercises. Above this sample size the predictive accuracy of the models reached a “plateau”.