Identifying biodiversity on a large scale

To recognize a plant, you need to know that it exists, the shapes of its flowers, seeds and leaves, the arrangement of these and many other things that enable you to identify it according to the time of year, its stage of development and its environment.

In a natural setting, a plant never lives in isolation. In a dense forest, each plant is usually accompanied by several others. In the Amazon rainforest, identifying biodiversity takes on a whole new dimension.

The XPRIZE Rainforest

The XPRIZE Rainforest is an international competition to identify the best possible biodiversity in an area of 100 hectares, i.e. 1 million square metres (10 km x 10 km) ... in three days and without any human intrusion.

To achieve this, we need to change the way we do things. The use of drones and aerial images comes to mind, but A.I.s capable of interpreting images need data to train on, and current data is that of individual plants, with little context. Nothing to do with the debauchery of leaves and tangled plants on display. All the more so as many plants are still completely unknown on a scientific level.

The Inria and CIRAD (Centre de coopération internationale en recherche agronomique pour le développement) researchers in charge of Pl@ntNet set to work in collaboration with the University of Sao Paulo. The team succeeded in documenting no fewer than 418 taxa (types of living beings), including 266 identified to species level, as well as three species never before described.

We were the best... but we weren't very good! We succeeded in fulfilling our mission, but Pl@ntNet cannot yet be used reliably in this way. On the other hand, the XPRIZE Rainforest competition served as a catalyst: it encourages us to move forward in ways we wanted to go, and it was an incredible experience."

For the full article: Pl@ntNet, the key to an ambitious competition in the Amazon rainforest

Illustration : DEZALB - Pixabay

Visit inria.fr