A predictive approach to determining the conservation status of species

Durée : 2020 - 2023
Programme : INRIA
Portée : Nationale

Site web
Species distribution modeling
Deep learning
Image time-series
Convolutional neural networks
Remote sensing
Data science

Assessing the conservation status and risk of extinction of species is a major challenge towards biodiversity management. Currently, this is done on a species-by-species basis, a process that requires strong expertise and takes considerable time. Our hypothesis is that it may be possible to infer species conservation status automatically by combining statistical learning, artificial intelligence and ecological modelling approaches. However, hard problems arise, in particular presence-only data, bias and class imbalance.


  • AMIS, Université Paul Valéry Montpellier
  • LIPHY, Université Grenoble Alpes


Estopinan, J., Servajean, M., Bonnet, P., Munoz, F., & Joly, A. (2022). Deep species distribution modeling from sentinel-2 image time-series: a global scale analysis on the orchid family. Frontiers in Plant Science13, 839327.