Combining remote sensing and ancillary data to improve species distribution models

Correlative distribution models are increasingly used in biological conservation. They often require the compilation of various environmental attributes (climate, topography, land cover,…), which can be very time-consuming. In the Lifewatch project, a database combining segmentation in homogeneous landscape units (“ecotopes”) and environmental attributes derived from regularly updated remote sensing data (land cover, potential solar energy,…) and other data sources (climate, topography,…) has been designed. Our aim was to assess the usefulness of this database for species distribution modelling and to propose further improvements. The distribution of ten species (four butterflies, three birds, two reptiles and one mammal) was modelled across Wallonia and model performance was assessed by Area Under the ROC Curve using the subsample approach. The models derived from ecotope segmentation and quantitative land cover attributes were compared to those obtained with a corresponding grid and a qualitative land cover database. The usefulness of ancillary data (soil properties, contextual attributes) was assessed to decide which variables should be integrated in the ecotope database. The results allowed us to assess and improve the relevance and accuracy of our database, which is a promising tool for species distribution modellers at the European scale.