How “digital twins” could help predict the fate of a forest

In his office at Michigan State University, forestry professor David Carter shows off an image of a virtual forest on his laptop.

It’s not just any forest. It’s a computerized replica, or “digital twin,” of a loblolly pine stand, created using lidar, the laser scanning technology that self-driving cars use to map their surroundings.

Carter says virtual landscapes like these could allow forest managers to test different management strategies in a simulation before deploying them in the real world.

The concept is not new. Digital twins are now widely used across industries ranging from manufacturing to healthcare to run “what if” scenarios before taking action.

But now a similar revolution is taking place within forestry, opening up possibilities for more targeted treatments that minimize waste and bolster financial returns.

In some of his most recent work, Carter and his team have developed an AI tool that is now making this possible for pine plantations.

In the southeastern U.S., loblolly pines stretch across some 35 million acres, providing the raw material for everything from plywood to paper towels.

The seedlings are typically planted in neat rows and then thinned after 12 to 15 years, to make room for the best specimens to grow into more valuable products like utility poles, fence posts and flooring.

“It’s kind of like weeding your garden,” said Carter, a core faculty member in MSU’s Ecology, Evolution and Behavior program. The idea is to help the best plants flourish by reducing competition.

The first thinning in a pine plantation involves choosing an arbitrary starting row and removing every third or fourth row from there. But pine plantations are never truly uniform. Some trees are runty or misshapen; others arrow-straight and vigorous.

So Carter got to thinking: rather than choose a starting row on the fly, selected at random, what if landowners could try different starting rows virtually, ahead of time, and know which ones would maximize growth and profits down the line — before making the first cut?

In a study recently published in the Journal of Forestry, Carter and his team flew a drone equipped with a lidar scanner over a 7.5-acre stand of loblolly pines in central Virginia before and after thinning.

By emitting tens of thousands of laser pulses per second and measuring their return time, the technique generated a giant 3D cloud of points, essentially a digital copy, or twin, of the real-life forest below.

What’s exciting about lidar, Carter said, is it can “see” more of the trees.

Measuring every single tree over hundreds of acres would be a daunting task. If we sent teams out to do this on the ground they might measure 3% of a forest, taking a representative sample and extrapolating from there, he explained.

But with lidar and an AI method developed with co-author Matthew Sumnall of Virginia Tech, the researchers were able to map 90% of the 3,555 trees in the stand.

By comparing their computer models with on-the-ground measurements, they found the digital twins could predict the trunk diameter and volume of each tree with high accuracy.

That level of detail is powerful, Carter said, because it creates an opportunity to move away from a one-size-fits-all approach to forest management.

Back in his office, Carter uploaded the lidar data to an app he recently developed with MSU computer science M.A. Amith Reddy, configured a few settings, and hit run. By running simulations on the digital twin of the forest, the app uses machine learning to predict how the trees are likely grow in the future under different thinning scenarios.

Using their simulations, the team found that simply shifting the starting point by one row could preserve 15% more timber for future growth and higher profits.

Carter estimates the change could generate $70 per acre in additional earnings. Such gains might be of minimal importance in a given stand, but when compounded across thousands of acres, the savings could add up — especially for companies that are already using lidar for forest inventories.

“This shows that even a modest change in thinning practices could appreciably move the needle,” Carter said.

Carter added that the work has potential applications beyond loblolly pine plantations, such as evaluating forest restoration methods. But because all the trees in a pine plantation are the same age and species, “they’re sort of like the kindergarten for precision forestry,” Carter said.

“If you're going to try to use this technology and develop tools, they probably need to be able to work here first before we can go take them to more complicated scenarios,” Carter explained. “So pines are a nice test bed.”

The take home, he said, is that “now we can work with a finer paint brush when designing forests.”

This research was supported by the Acorn Alcinda Foundation.

CITATION: "Precision Forestry: Using Machine Learning and LiDAR to Inform Thinning in Pinus taeda Plantations – A Case Study," Erik Platt, David R. Carter, Amith Reddy, Timothy J. Albaugh, Rachel L. Cook, Otávio Campoe, Rafael Rubilar, Gunjan Barua & Matthew J. Sumnall. Journal of Forestry, April 20, 2026. DOI: 10.1007/s44392-026-00089-6