26 million trees in Île-de-France

“This is data never before collected in this form and on such a scale in Île-de-France” emphasizes Laurie Gobled, director of information systems at the Institute. This data will in particular make it possible to feed the future 3D model proposed by the Île-de-France Region and already available for consultation in beta version on its Smart Services portal.

This innovative project, which covers an area of ​​12.000 km², was carried out using deep learning methods (sub-field of artificial intelligence) from a regional orthophotography carried out in 2021 for the update of the MOS repository. (Type of land use). The method consists of feeding an algorithm which then recognizes, by learning, the elements that are indicated to it.

After thirty days of server calculations, the Institute has a geographical layer of 26 million points, each of which corresponds to a tree over 3 meters high. These points are also filled in with the height of the tree, thanks to cross-referencing with the Institute's digital surface model (DSM) and the digital terrain model (DTM). This operation also made it possible to inventory low vegetation, between one and three meters, ie approximately 1.300.000 trees.

Beyond upgrading the regional 3D model, this new data will feed studies such as those on forest fires or on cool islands, update a vegetation index, identify the rows of trees along roads, etc.