Drones speed up execution of forest inventories

By Elton Alisson  |  FAPESP Innovative R&D – To plan production of a line of cosmetics derived from Astrocaryum vulgare, a fruit-bearing palm tree native to the Amazon Rainforest, engineers at Natura, Brazil’s leading cosmetics company, had to wait nine months to obtain the necessary data while a team of 16 local community members surveyed 1,200 hectares of forest in search of specimens in good enough condition to be sustainably managed for this purpose. A. vulgare is known as tucumã in Brazil and awara in Suriname, Guyana and other South American countries.

“This team of 16 people from seven forest-dwelling harvester communities we work with called on 300 farmers in the area, and only nine months later, based on the data they collected, were we able to estimate the supply of 1,000 tons of fruit, for example. That sounds like a lot, but it isn’t sufficient to produce the planned line of cosmetics derived from A. vulgare, currently one of our primary assets. A wait of almost a year to find out the productive potential of this fruit represented a major bottleneck,” said Rômulo Zamberlan, Natura’s head of advanced research, in a presentation during the Deep Tech Summit held in São Paulo city in November 2024.

To surmount this obstacle and optimize management of the supply chains involving the assets it uses to produce cosmetics, Natura is partnering with Bioverse, a São Paulo-based startup that maps forest cover and species prevalence.

Thanks to the technology developed by Bioverse, a deep tech that uses drones for imaging combined with artificial intelligence (AI) to identify and locate trees belonging to species from which raw material can be harvested efficiently, Natura succeeded in inventorying the area in only six months, locating six key species for its production process distributed in 40,000 hectares of forest.

“Trees belonging to the species of interest were identified in an area of 400 square kilometers [km²], equivalent to about a third of Irituia, a municipality in Pará [a state in the Brazilian Amazon]. It would have taken 25 years to do this mapping by the traditional method with the level of detail produced by this technology,” Zamberlan said.

All the data obtained via the digital inventory was provided to forest harvester community co-ops in the Amazon so that they can better plan production. “The co-ops use the data to decide where best to produce açaí or tucumã, for example. This makes the process more efficient for the communities,” he said.

Besides helping to improve estimates of productivity, the data obtained via the digital forest inventorying technology can be used to assess carbon stocks, quantify the value of the forest, and support forest regeneration plans, he explained. “The technology can map thousands of trees. We can use it to estimate the value of a standing forest,” he said.

VTOL drones

The digital forest inventory cannot yet be done via satellite because the Amazon Rainforest is made invisible from above by dense cloud cover for seven months of the year. In addition, the Amazon region has more than 4 billion trees and a huge variety of species, Zamberlan said. To circumvent these limitations, Bioverse has learned to use drones made by another startup, XMobots, located in São Carlos, São Paulo state.

Supported by the FAPESP Innovative Research in Small Business Program (PIPE), XMobots is part-owned by Embraer and specializes in developing and producing high-performance vertical takeoff and landing (VTOL) drones, as well as related technologies, such as multispectral sensors, gyrostabilized optronic sensors, AI-based data analysis software, and a drone service provision platform, among others (read more at: pesquisaparainovacao.fapesp.br/2464).

Bioverse uses one of its drones, the Nauru 500C, to produce forest inventories that identify species of interest for Natura. “It’s basically a small unmanned aerial vehicle that takes off vertically like a drone and flies like a normal aircraft, with a range of hundreds of kilometers,” Zamberlan said.

The images captured by the drone are analyzed by an AI-based system developed by Bioverse that identifies tree species and supplies their geographical coordinates. Based on this information, teams go into the field to validate the data. “The system already exhibits 90% accuracy in identifying species,” he said.

Natura currently has two innovation centers – one in Cajamar in metropolitan São Paulo, and another in Pará entitled Amazon Innovation Center – staffed by around 300 researchers with degrees in engineering, agronomy, chemistry, pharmacy, and data science, among others.

“Besides this contingent of researchers, we have a very large network of local and overseas collaborators. Most of our projects are partnerships with science and technology institutes or startups and companies of other kinds, both small and large,” he said.