Drones map forests in the Amazon

In total, more than 35 thousand hectares have already been mapped with the aim of collecting information on forest species

28.03.2023 | 14:44 (UTC -3)
Embrapa
In total, more than 35 thousand hectares have already been mapped with the aim of collecting information on forest species; Photo: Evandro Orfanó
In total, more than 35 thousand hectares have already been mapped with the aim of collecting information on forest species; Photo: Evandro Orfanó

A team from Embrapa Acre uses drones to map forest areas in the states of Acre, Rondônia and Amazonas. c With the support of artificial intelligence, the technology will optimize the carrying out of forest inventories, a fundamental step in timber and non-timber management plans.

The activity is part of the project “Geotechnologies applied to forest automation and spatialization of carbon stocks in native and modified land use in the Western Amazon (Geoflora)”, carried out in partnership with the JBS Fund for the Amazon.

The work has already been carried out in forest areas in the Chico Mendes Extractive Reserve, Antimary State Forest (Acre), Jacundá and Jamary national forests (Rondônia), municipal parks in Rio Branco and forest management areas on private properties in Acre and Amazonas . The second step of the activity is to systematize the information in a database, NetFlora. In addition to these areas, the project also foresees the mapping of forests with drones in other locations in the Amazon.

Evandro Orfanó, a researcher at Embrapa, explains that the information collected will be used to "train" algorithms in identifying forest species, a technique that will allow us to automate part of a forest inventory, in addition to providing greater precision and increased productivity for the activity. "Currently the performance of inventory teams in the field hardly exceeds 20 hectares per day, while with a drone that uses artificial intelligence it will be possible to exceed two thousand hectares."  

Still according to the researcher, the "trained" algorithms take around 20 milliseconds to locate, identify and size the tree crowns. However, to obtain this performance, a robust database is necessary that offers increasingly detailed visual characteristics of the species, by region, fruit production season and period of the year.

Support for forest concessions

Two of the areas flown over with drones are national forests under use concessions and with forest inventories audited by environmental agencies responsible for activities in the area. According to Orfanó, the management plans for these areas already use the Modeflora system, a factor that makes it possible to locate trees in the field and check the information collected with drones, a fundamental aspect to calibrate the 'training' of the algorithm and confirm what is seen from up", explains Orfanó.

José Humberto Chaves, general coordinator of Forest Monitoring and Auditing at SFB, sees practical applications of the technology. "The forest inventory carried out with a drone, in addition to reducing costs and providing greater quality to the management plan, makes it possible to adjust various aspects of management in the field, if necessary." 

The technology can be used by both the Brazilian Forest Service and environmental agencies to validate forest inventories presented by forest managers. "I believe that, soon, we will be able to use these tools to inspect forest inventories and, in this way, guarantee better control of the forest production estimated in the management plans. And who knows, in the future, this technology will allow us to create new mechanisms to identify areas of illegal deforestation", he adds.

Cultivar Newsletter

Receive the latest agriculture news by email

access whatsapp group