Terra Labs launches 2026 Centurion – the forest in continuous time and space
LiDAR data and orthophoto from Lantmäteriet. The point cloud is colored using data from Terra Labs Forest State Field.
A new generation of forest AI that changes how we describe, analyze and work with forests.
Since at least the 19th century, Swedish forestry has been organized around forest compartments. Forests are divided into areas described by attributes such as mean age, height, diameter and basal area, along with tree species composition. Site productivity and volume are derived from this information. This has been a practical way to inventory and plan the management of large forest holdings, but it also comes with limitations.
Two compartments with identical mean values can look entirely different in reality. And as the forest changes, the information does not necessarily keep up. Analyses and decisions are tied to predefined boundaries from the outset.
Today, Terra Labs launches 2026 Centurion, a new AI model that establishes a new foundation for forest inventory, analysis and planning. Centurion describes the forest as a continuous field in time and space, with high geographic detail throughout Sweden. Variation across the forest becomes available for analysis, both within compartments and across their boundaries.
All of this is built on open data, without requiring additional high-resolution imagery.
It marks a technological shift with implications far beyond traditional forest management.
From forest compartments to a continuous representation of the forest
Centurion is built on Terra Labs' Forest State Field technology, which describes forest conditions in both time and space.
Traditionally, forest attributes have been assigned to geographically discrete areas – forest compartments – for practical reasons. The information describes the compartment as a whole, even when forest conditions vary considerably within it.
With Centurion, the relationship is reversed. Forest attributes are described independently of compartment boundaries and can be summarized for any area.
This does not mean that forest compartments disappear. They still serve an important purpose in forest planning and management. The difference is that they no longer determine what information is available, which analyses can be performed or which decisions can be made.
And because Forest State Field also describes how the forest develops over time, growth, management needs and updates reflecting completed forestry operations can all be handled within the same model.
Compartment boundaries can remain. But the description of the forest no longer has to follow them. Management areas can be defined as needed.

Forest State Field prediction
After defoliation
A technical breakthrough delivering high detail from open data
A fundamental part of this technological shift is that Centurion requires only open data as input.
Terra Labs has achieved a technical breakthrough by combining deep learning methods for sensor fusion and learning from limited or incomplete annotations (self-supervised and weakly supervised learning) with high-quality proprietary ground truth data.
Applied to massive volumes of remote sensing data, this approach has produced in a new foundation model for Earth Intelligence, from which Forest State Field and our Centurion model have emerged.
Centurion can describe forest attributes with high geographic and temporal detail throughout Sweden. The difference is not access to more high-resolution data, but the model's ability to extract significantly more information from the data already available.
This makes the technology applicable at scale without requiring additional data collection. However, the available open data sources vary in how up to date they are and can be supplemented with high-resolution satellite imagery or other data sources when needed.
Four major advances in Centurion
1. Let the decision define the area
When forest data is no longer tied to predefined compartments, forest attributes can be analyzed with significantly greater geographic precision. The analysis and the decision define the area, rather than the other way around.
Identify a potential management area, a timber lot or a location that requires closer investigation. Centurion can describe the forest within that temporary area without changing existing compartment boundaries or collecting new data.
2. Fast calculations
Draw a temporary area for analysis. Expand it, divide it or compare it with another area. Centurion quickly calculates new estimates based on the updated geometry.
With results available immediately, users can test alternatives, compare outcomes and make decisions as they work. The analysis can evolve alongside the questions being asked.
3. Smarter ways to define forest stands
Centurion helps users identify and define areas based on the task at hand.
Specify what you want to investigate or achieve. The model analyzes forest attributes and proposes areas based on your objectives and criteria, such as areas with similar management needs. Adjust the criteria, and the model can suggest new boundaries.
The proposed areas can cover the entire area you've selected, or only the parts that meet your criteria. They can be used temporarily for analysis and planning.
AI provides the detailed analysis. Humans set the objectives and apply their professional forestry judgment.
Forest planners can spend less time manually identifying areas and defining their boundaries, and more time evaluating suitable management measures based on management objectives, local conditions and practical requirements.
4. Understand forest variation
One compartment may have evenly distributed trees with deciduous species scattered throughout. Another may have the same species proportions and mean values, but consist of dense clusters, sparse patches and deciduous trees concentrated in groups. These differences in spatial distribution can justify different management measures.
Centurion makes this variation measurable and understandable. For example, the model can quantify how much forest attributes vary within a stand and how that variation is distributed across the area. It can also identify areas that stand out and show where different forest attributes occur.
What was previously hidden behind a mean value becomes information that can be compared and used in decision-making.
The results can also be explained in natural language. This makes it easier to understand where forest conditions differ from the mean, why a compartment is described the way it is and what the variation could mean for management planning.
Discover what was previously difficult to measure
An experienced forestry professional can observe a great deal during a field visit. But systematically mapping the same level of detail across hundreds of thousands of hectares is an entirely different challenge.
This is where Centurion opens up new possibilities.
For example, the model can identify where other tree species occur within a compartment. Rather than simply knowing that a compartment contains ten percent birch, users can investigate where those birch trees are actually located, and whether they are concentrated in one area or distributed throughout the compartment.
Similarly, the model can analyze differences in stand density, site productivity and tree species composition. It can identify gradual transitions between forest types that are otherwise difficult to represent using traditional compartment boundaries.
At Terra Labs, we develop these capabilities under the concept of Beyond Human Level: using AI to systematically identify and quantify patterns that are difficult to map and keep up to date manually, particularly across large geographic areas.
Together with the Terra Labs AI agent, this information can be explored through natural-language questions across Sweden's forests.
Where are the deciduous trees? Where are suitable areas for conservation set-asides and low-productivity forest land? Where are management measures needed, today and in the future? Which areas deviate from the surrounding forest? Where are areas with similar conditions? Which parts of the forest match the needs of the industry or the specific requirements of an end customer?
Instead of manually combining map layers and searching for answers, users can ask a question and explore the results directly on the map.

From better forest data to better decisions
The real change is not simply about more detailed forest data, but what that data makes possible.
Consider thinning planning.
Traditionally, thinning needs are assessed using thinning templates applied to mean values for an entire compartment. In an uneven stand, this can mean that dense areas requiring thinning are masked by sparser sections.
Centurion can instead analyze thinning needs with high geographic accuracy and identify which parts of the forest meet the criteria, independently of compartment boundaries.
Rather than simply being told that a compartment should be thinned, users can see where thinning is needed, where it isn't, and how those needs vary across the area. This enables more precise management recommendations, better prioritization and lower costs.
The same principle can be applied throughout the forestry value chain:
Forest owners and valuers gain a more detailed understanding of what is actually present on a property and how variations affect its value.
Forest planners and managers can identify where management measures are needed within and across compartments, and prioritize field visits where they will provide the greatest value.
Timber buyers and the forest industry gain significantly better information for identifying areas of interest, calculating expected timber assortment yields, planning wood supply and meeting delivery commitments.
Revision, controlling and reporting can draw on a shared, detailed information base that can be updated as the forest changes.
When information can be used independently of compartment boundaries, every new question no longer needs to begin with a new inventory.
Terra Labs' forest data becomes a living source of information for decision-making, one that can be analyzed again as needs change and used by more people throughout an organization.
The forest is just the beginning
2026 Centurion is not only a paradigm shift for forestry. It is also an important step in Terra Labs' development of the next generation of Earth Intelligence: AI models capable of describing and understanding the physical world in time and space.
We are starting with forests, where substantial economic and ecological value depends on reliable information for decision-making, and where traditional inventory methods have long constrained what information is available.
But our ambition extends beyond forests. The same technological foundation can be used to understand and track changes in other parts of the physical world.
We are building a new foundation for how information about forests, and eventually other parts of our world, can be collected, kept up to date and put to use.
The forest has always changed. Now, the way we describe it can change too.