Machine learning has been used to generate highly detailed maps of more than 100 million individual trees from 24 sites across the United States.

The maps provide data about individual tree species and conditions, which can greatly aid conservation efforts and other ecological projects.

The research, conducted by a team led by Ben Weinstein of the University of Florida (UF), Oregon, has been published in the open-access journal PLOS Biology.

Although ecologists have long collected data on tree species by surveying small plots of land and extrapolating those findings, this technique cannot account for variability across an entire forest.

Similarly, there are other methods that can cover broader areas, but they often struggle to categorise individual trees.

To generate their large and highly detailed maps, the UF researchers trained a deep neural network using images of the tree canopy and other sensor data gathered by aircraft.

These training data covered 40,000 individual trees and, like all the data used in the study, were provided by the US National Ecological Observatory Network.

The neural network was able to classify most common tree species with 75% to 85% accuracy. In addition, the algorithm was also able to provide other analyses, such as reporting which trees are alive or dead.

The researchers found that the technique had the highest accuracy in areas with more open space in the tree canopy and performed best when categorising conifer species, such as pines, cedars and redwoods. The network also performed best in areas with lower species diversity.

“The diversity of overlapping datasets will foster richer areas of understanding for forest ecology and ecosystem functioning,” the researchers said.