Harvester data and laser scanning can enable more precise forestry

Remote sensing—using data from satellites, airplanes or drones —can provide high-resolution geographical information that helps optimize forest management.

However, this requires detailed reference measurements of trees, and the availability of such reference data is currently limited, since manual inventories are both costly and time-consuming.

Scanner mounted on a car

In his thesis, Raul de Paula Pires explores new methods for collecting reference data to train AI models to correctly interpret remote sensing data.

“To fully implement precision forestry, we need far more reference data than we have today,” says Raul de Paula Pires.

One method involves mounting a lidar scanner on a vehicle to scan the forest on both sides of a forest road. Regularly scanning forests from vehicles before harvesting is probably not the future of forest inventory. Instead, vehicle-based scanning could be a useful way to collect reference data to validate and calibrate remote sensing from the air.

Harvester data is useful

Another, more accessible way to collect reference data for model training is to use harvester data. The harvesting head and operator already record information such as tree species, diameter, length, and quality. If the machine is also equipped with GPS, the data can be matched with remote sensing data to train the models—creating massive datasets very quickly.

“It’s free data—the harvester is already working in the forest every day. With little or no modification to the machine, you can get the position of every tree,” says Raul de Paula Pires.

Way to optimize forestry

With increasingly detailed data on trees and forests, entirely new opportunities for forest planning are emerging. It becomes possible to gain a comprehensive understanding of entire stands and the quality of individual trees. This enables forest management to be optimized while taking economic, environmental, and social values into account.

“It’s difficult to say exactly how much better forestry will become, but we can begin to integrate the many different interests and requirements placed on forest management,” says Raul de Paula Pires.

Raul de Paula Pires defence is on the 6th of november i Umeå, more information here.

Here you can access the dissertation Expanding data availability for tree-level remote sensing-based forest inventories.