Species distribution models

Abstract

Species distribution models are a group of methods often used to estimate\nconsequences of global change, to assess ecological status and for other ecological\napplications. The main idea behind species distribution models is that the\ngeographical distributions of species can, to a large part, be explained by\nenvironmental factors and that species distributions therefore can be predicted in\ntime or space. For robust and reliable applications, models need to be based on\nsound ecological principles, predictions need to be as accurate as possible, and\nmodel uncertainties need to be understood.\n\nTwo approaches are available for modelling entire species communities: (1) each\nspecies can be modelled individually and independently of other species or (2)\ncommunity information can be incorporated into the models. The first study in this\nthesis compares these two modelling approaches for predicting phytoplankton\nassemblages in lakes. The results showed that predictive accuracy was higher when\nspecies were modelled individually. The results also showed that phytoplankton can\nbe used for model-based assessment of ecological status. This finding is important\nbecause phytoplankton is required for assessing the ecological status of European\nwater bodies according to the European Water Framework Directive.\n\nDispersal barriers in the landscape or limited dispersal ability of species might be a\nreason for species being absent from suitable habitats, and these factors might\ntherefore affect model accuracy. The second study in this thesis examines the\ninfluence of dispersal and the spatial configuration of ecosystems on prediction\naccuracy of benthic invertebrate and phytoplankton distribution and assemblage\ncomposition. The results showed only a minor influence of spatial configuration and\nno effect of flight ability of invertebrates on model accuracy. However, the models\nused may partly account for dispersal constraints, since dispersal-related factors, such\nas lake surface area, are included as predictor variables. The result also showed that\ncomposition of littoral invertebrate assemblages was easier to predict at sites located\nin well-connected lake systems, possibly because the relatively unstable littoral zone\nnecessitates a need for species to re-colonize disturbed habitats from source\npopulations.