Stefano Puliti on why forestry needs AI – and openness

This article is based on our podcast episode released on the 25th of March 2025. [Listen to the full episode here.]

In our recent podcast, we spoke with Stefano Puliti about his journey in forestry, which began in the countryside of Tuscany. Growing up surrounded by olive groves and forests, his path eventually led him north to Scandinavia.

"I did my bachelor's in Italy but then I realised that the Italian forestry world was a bit small," Stefano told us. He was drawn to Scandinavia where "forestry is a big thing," moving first to Finland, then Sweden, and finally settling in Norway.

Managing forests tree by tree with LiDAR

As a research scientist at the Norwegian Institute of Bioeconomic Research (NIBIO), Stefano now leads the SingleTree project. The team uses laser scanning (LiDAR) from helicopters, drones, and even airships to build 3D maps of forests at the individual tree level.

"We're seeing a shift towards reducing the area of intervention," Stefano explained.

"It's not only timber production anymore but suddenly it's a lot of functions that we have to answer and it's production but it's also biodiversity and making sure that the forests are healthy. We want to also understand more about the wood quality."

As forests are required to perform multiple functions, the detailed information that SingleTree provides becomes increasingly valuable.

A new way to work with point clouds

The project relies on point clouds – 3D maps made from millions of laser measurements. Traditional methods for processing this data had significant limitations, but AI has changed the approach.

"Deep learning allows us to skip all of that algorithmic pipeline," Stefano tells us. "Suddenly you have one model that can be applied to airborne data or to terrestrial laser scanning data, meaning that you don't have to decide which model to use for what."

Their recent paper, "Segment Any Tree," demonstrates this capability. The model can identify individual trees from point clouds regardless of forest type or data source.

The power of open data sets

A key part of Stefano's approach is making data freely available. Rather than keeping valuable datasets within research institutions, his team shares them widely.

"If research institutions want to be part of the next century, they have to come together. They have to be open. They have to start to collaborate and share data," he explains.

The results speak for themselves. "There were like 2,000 views and 1,000 downloads," Stefano mentioned about one of their datasets. "It increases the number of people who can work on these problems. And the more brains, the better solutions we're going to come up with."

From scientific theory to forestry practice

At the end of our podcast with Stefano, we couldn't help but ask him the same question he once asked his professor in Finland years ago: What's the next big thing in forestry technology?

"Making it work and bringing it into the real world. Scientists show, 'Oh, look, it works great,' but it really works great for some trees, in maybe a simple forest that they looked at. It's not really applied in the real world."

This implementation will require both technical improvements and cultural shifts in forestry.

“Within the next five to ten years, we will start to see a lot more of this automation in data capture and extracting information, but also in the machines, in the planning, in the sawmills, and linking all this together," Stefano predicts.

A changing view of forests

The evolution towards individual tree management reflects broader shifts in how we value forests.

"In the past five years, there's been a big change from companies just asking how much timber is out there to asking what the biodiversity is, because this is what they're being requested by forest owners," Stefano observes.

He concludes with a perspective that resonates deeply with us at Forsler. "Forests are not just containers of biomass or timber. There's a lot more in forests, and we should manage them accordingly."

If you’d like to learn more about  tree-level forest management, you should have a look at Stefano’s work here:

• SingleTree Project website
• Segment Any Tree
• SmartForest