Toward large scale ML-based habitat type prediction

César will present his PhD subject and initial results on predicting future trajectories of plant species and habitat distribution. Recent cyber-infras and new data sources offer opportunities to mobilise massive amounts of biological data that have never been analysed together. In particular, species occurrence data can be combined with various environmental and geographic rasters to derive models of species distribution and temporal evolution at the continental scale and to allow projections under climate and land use change scenarios. To learn these complex relationships and cope with data heterogeneity, we rely mainly on DL models and associated methodologies. Experiments are conducted, in the context of the GUARDEN European research project to predict EUNIS habitats, based on the European Vegetation Archive (EVA), and the CBN-Med database.