Digitalization progress in forestry

On May 18th, the Mistra Digital Forest program conference took place at a packed SLU in Umeå. Here you can read a summary with a focus on the morning presentations.

Photo: Emma Sandström, SLU

Welcome! How wonderful to meet in person again." With these words from Mistra Digital Forest's Sverker Danielsson, the spring program conference in Umeå began. Over three hours in the morning, participants would experience the breadth of research conducted within the program.

The first to be welcomed on stage was professor Ola Lindroos from SLU, who opened the day by discussing digitalization in forestry education. He emphasized the importance of developing educational software and training versions of the tools that students will encounter in their professional lives.

"Please consider this when developing decision support systems. It might accelerate the process of achieving even higher digitalization maturity within the sector," he said and handed over to the first session, which provided examples of research in remote sensing.

Ola Lindroos. Photo: Emma Sandström, SLU

Remote sensing

Christoffer Axelsson from SLU, began by discussing his post-doctoral project, where he utilized multispectral airborne laser scanning to classify tree species and estimate volume for individual trees. The multispectral instrument scans not just one but multiple wavelengths. The results demonstrate that the green, complementary wavelength is crucial for classifying deciduous trees, thereby contributing to more accurate results for the entire stand.

Christoffer Axelsson. Photo: Emma Sandström, SLU

After that Jörgen Wallerman spoke about estimating growth from airborne laser scanning. Access his full presentation here.

Jörgen Wallerman. Photo: Emma Sandström, SLU

"We have massive amounts of data, but we need new algorithms that can extract relevant forestry information and make use of all these measurement points," stated Kenneth Olofsson, SLU, in his presentation.

What he has set out to do is to obtain detailed information about timber quality using ground-based laser scanning. In his research, he uses laser-based sensors – commonly used in industries such as automotive – to capture stem profiles that accurately reveal the tree's diameter, volume, and geometric shape.

Johan Holmgren from SLU was also present to share how he has utilized mobile laser scanning in his work to estimate stem properties for larger areas. This becomes possible when aerial and mobile data of each tree in a selected area are matched and then serve as reference data for estimating the entire stand. The mobile laser scanning system was assembled using inexpensive components, and Johan highlighted the usefulness of the technology:

"Mobile laser scanning allows us to collect data wherever we are, not just on sample plots." By using it on the harvester, backpack, or vehicle, relevant data can be gathered.

Johan Holmgren. Photo: Emma Sandström, SLU

During the Q&A session that followed, an audience member asked about the market's interest in new digital solutions. It was emphasized how significant it is that major forestry companies are engaged in the program and that the research being conducted is relevant.

A question that arose regarding Christoffer Axelsson's research was whether it is possible to differentiate between deciduous trees using multispectral laser scanning. Christoffer stated that this matter has not been specifically investigated, but he believes it could be a more challenging task.

The speakers during the remote sensing session. Photo: Emma Sandström, SLU

Automation

”We are in a crisis”, stated Mikael Lundbäck from SLU. He was not referring to war or climate, but rather to the fact that forest technology has lagged behind other industries in terms of automation. As a newly graduated doctor, he has simulated and economically evaluated various combinations of remote control and automation of forwarders. The results show that the greatest potential for cost reduction lies in automating the loading and unloading processes, while the potential is slightly lower for automating the driving. Overall, the potential cost savings range from 6% to 19%.

Mikael emphasizes that the economic productivity is at stake, and therefore, there are strong reasons to discuss scenarios involving fewer operators. However, he also points out that if we choose to go down this path, it is essential to conduct research on social and environmental sustainability. "In order to pursue this direction, we need to address the social and environmental aspects as well," he added.

Next up on stage was Martin Servin from Umeå University. His team has trained an autonomous forestry machine in a simulated environment that can navigate safely, efficiently, and quickly in forest terrain. The technology used is a form of AI, where the machine is trained through trial and error. In the study, a swing-arm harvester is trained in a time-consuming process equivalent to 10,000 hours, but the simulation significantly speeds up the training process.

Martin Servin. Photo: Emma Sandström, SLU

"So far, it has been surprisingly easy to train AI control in a simulated environment," says Martin, emphasizing that the strategy is to train many small multi-agents that can be detached and combined, for example, with a human operator. Currently, work is underway to transfer the system to a physical forestry machine."

Biorefinery processes

Before it was time for lunch, the focus shifted to biorefinery processes. How can digital solutions be used to better match raw materials and end products? Maria Nordström, Skogforsk, och Tinh Sjökvist, Södra Innovation, illustrated examples. Their collaboration is directly related to Södra's new business area, biochar, and how raw materials with the "right" properties can contribute to increased resource efficiency. In a first step, the project's model determines the properties of the trees using harvesting data and age information. These properties are then "translated" into the various qualities required by sawmills, providing important preliminary information on how the incoming raw material can be best utilized.

Maria Nordström och Tinh Sjökvist. Photo: Emma Sandström, SLU

New technology provides us with access to much more well-defined wood fractions. This, in turn, results in better product control and higher-performing products. With that message, Gunnar Henriksson from KTH, began his presentation. The research he contributes to shows that the properties of pulp differ depending on which part of the tree is used. The study also revealed that the branches stood out, particularly from the rest of the spruce; "they resemble a combination of hardwood and softwood pulp," said Gunnar. But can the sorting of wood chips into different fractions be automated? Yes, that is what Mikael Thyrel from SLU means. He has developed a model using machine learning that can distinguish wood chips from three different wood fractions with 90% accuracy.

The speakers during the sessions on automation and biorefinery processes. Photo: Emma Sandström, SLU

Afterward, the program conference transitioned into an afternoon characterized by networking and meetings. Several companies were present during the day, sharing the ambition to make large datasets useful for the industry and forestry. These companies, exhibiting during the poster session, were accessible for the participants in Umeå to visit.

Access posters from the day (in Swedish)

Back to home page