Forests are among the Earth’s most powerful natural allies, sustaining life, regulating the climate, and maintaining the ecological systems on which humanity depends. Covering approximately 31% of the world’s land area, they store an estimated 662 gigatonnes of carbon and provide habitats for the majority of terrestrial biodiversity.
In simple terms, forests perform functions that are essential to the health of our planet. They protect soils, support wildlife, regulate rainfall, and maintain ecological balance. Yet, understanding and protecting these vital ecosystems requires more than simply identifying where trees grow. Not every tree-covered landscape is a natural forest, and confusing different forest types can conceal the true state of the world’s forests.
A centuries-old primary forest, shaped by natural processes and largely undisturbed by human activity, is ecologically different from a regenerating woodland, a commercial timber plantation, or an agricultural landscape dominated by oil palms and fruit trees. Although these landscapes may appear similar in satellite images, their biodiversity, carbon-storage capacity, ecological functions, and conservation value can vary dramatically. Treating them as equivalent risks overlooking the unique environmental importance of natural forests and distorting assessments of forest loss, restoration progress, and climate benefits.
This is where forest typology mapping becomes essential. Distinguishing natural forests from planted forests, plantations, and agricultural tree cover is fundamental to understanding what the world’s tree-covered landscapes actually represent. Until recently, however, no existing global dataset could distinguish these categories at such a high spatial resolution across the planet. The introduction of Forest Typology (ForTy) v1, a new global 10-metre-resolution dataset, marks an important step towards closing this knowledge gap. By revealing what lies behind the green canopy, this innovative mapping approach offers a clearer picture of world’s forests and provides valuable information for biodiversity conservation, carbon accounting, forest monitoring, and more informed environmental decision-making.
Forest Typology Mapping: What Is the Forest Typology (ForTy) Dataset?
The Forest Typology dataset is a global mapping system designed to distinguish forest types and other land uses that contain trees or woody vegetation. Developed using satellite observations, deep learning, and extensive reference data, it provides a 10-metre-resolution map of global land cover and forest types for 2020, distinguishing six categories aligned with definitions used by the Food and Agriculture Organization of the United Nations (FAO) and the European Union Deforestation Regulation (EUDR).
By revealing what lies beneath the broad label of forest cover, the forest typology mapping approach could give researchers, policymakers, and conservation organisations a clearer picture of where natural forests survive, where plantations dominate, and where agricultural tree crops have replaced natural ecosystems. The distinction matters because protecting the world’s forests begins with understanding what kind of forests we actually have.
Earlier global datasets have mapped forest cover, natural forests, primary forests, or planted forests. However, many have lacked the thematic detail needed to separate all these categories consistently across the entire world. ForTy v1 addresses this gap by classifying land into six categories:
- Primary Forest
- Naturally Regenerating Forest
- Planted Forest
- Plantation Forest
- Tree Crops and Agroforestry
- Other Land
The map uses a spatial resolution of 10 metres, meaning that each pixel represents an area of approximately 10 by 10 metres on the ground. This level of detail can help reveal differences that are difficult to identify using coarser global maps. The dataset represents a 2020 baseline, rather than a complete record of present-day forest conditions. Its value lies in providing a consistent reference against which later land-use changes can be assessed.
How Does Forest Typology Mapping Work?
Mapping forest types globally is more complicated than simply detecting green vegetation from space. Satellite sensors observe the physical characteristics of landscapes, such as vegetation structure, spectral signatures, and changes over time. However, they cannot always determine why a forest exists or how it was established. For example, a naturally regenerated forest and a planted forest containing similar tree species may look almost identical from above. Likewise, a timber plantation and an agricultural tree crop may have similar canopy patterns despite serving entirely different purposes.
To overcome these challenges, researchers developed a three-stage deep-learning pipeline using geospatial satellite embeddings and a combination of training techniques. The system draws on approximately 1.7 million globally distributed training samples. It combines information from existing forest maps, regional datasets, expert-labelled observations, and additional geographical evidence. The methodology includes:
– Deep learning: Identifies patterns in satellite-derived information to distinguish forest types.
– Weakly supervised learning: Allows the model to learn from reference data with different levels of detail.
– Active learning: Helps improve classification in underrepresented regions and difficult landscapes.
– Temporal analysis: Uses observations from different periods to improve consistency and reduce classification errors.
Rather than providing only rigid categorical labels, the dataset also supplies relative confidence scores for each class. These scores allow users to consider uncertainty and adjust classifications according to the requirements of their research or monitoring activities.
Six Categories That Reveal the Differences Between Forests
Understanding the six classes is essential to appreciating why this mapping system matters.
1. Primary Forest: Nature’s Living Heritage
Primary forests are naturally established forests of native species with no visible indications of human-caused disturbance under the study’s classification approach. These ecosystems often contain complex vegetation structures, mature trees, diverse habitats, and ecological relationships that have developed over long periods. They provide critical habitat for wildlife, store carbon, regulate local and regional water cycles, and support Indigenous peoples and local communities.
Although primary forests can experience natural disturbances such as wildfires, wind damage, or pest outbreaks, they may retain their classification when there is no evidence of human-caused land-use change. Identifying these forests is particularly important because their ecological complexity cannot necessarily be recreated through tree planting alone.
2. Naturally Regenerating Forests: Nature’s Power of Recovery
Naturally regenerating forests develop primarily through natural processes following disturbance or previous land use. They include forests recovering after logging, abandoned agricultural land returning to woodland, and secondary forests growing after substantial canopy removal.
These forests can range from young regrowth to mature secondary ecosystems that have developed over several decades. Their carbon storage, structure, and biodiversity vary according to factors such as age, climate, soil conditions, previous disturbance, and the surrounding landscape. When protected and allowed to recover, naturally regenerating forests can contribute substantially to ecosystem restoration, wildlife habitat, soil protection, and climate mitigation. However, their recovery potential should not be used to justify destroying primary forests. Protecting existing natural ecosystems remains essential.
3. Planted Forests: Trees Established by People
Planted forests are established predominantly through planting or seeding. They can serve a range of purposes, including timber production, restoration, erosion control, and other forestry objectives. Their environmental value depends on factors such as tree species, management practices, location, and the condition of the surrounding ecosystem.
Some planted forests can provide useful habitats and help restore degraded land. Others may offer fewer ecological benefits than naturally regenerating forests, particularly when they contain limited species diversity. Distinguishing planted forests from natural forests allows researchers to evaluate their respective contributions to biodiversity, carbon storage, and land management more accurately.
4. Plantation Forests: Intensively Managed Tree Production
Plantation forests are generally characterised by intensive management, regular tree spacing, even-aged stands, and one or a small number of tree species. They are frequently established for commercial timber, pulp, and other wood products.
ForTy v1 introduces an operational distinction based on rotation length: planted forests with rotations exceeding 40 years are classified as planted forests, while those with rotations of 40 years or less are classified as plantation forests.
This threshold is a modelling choice, not a universal FAO or EUDR definition. Actual management intensity cannot always be inferred directly from satellite observations, and the threshold may lead to misclassification in some regions.
Plantation forests can supply valuable wood products, but their ecological performance depends on how they are designed and managed. Large-scale conversion of natural forests into plantations can result in biodiversity loss and changes in carbon storage, even when the converted land remains covered by trees.
5. Tree Crops and Agroforestry: When Trees Serve Agriculture
Not every landscape dominated by trees qualifies as a forest under the relevant land-use definitions. This category includes agricultural tree crops such as oil palm, rubber, and fruit orchards, as well as agroforestry systems that integrate trees with agricultural production. Agroforestry can provide benefits like shade, improved soil protection, diversified farm production, and habitat for some species. However, agricultural tree cover is not automatically equivalent to a natural forest.
Under the EUDR framework discussed in the study, agricultural tree crops and agroforestry systems are excluded from the forest definition. This distinction matters when assessing deforestation risks, calculating forest areas, and determining whether agricultural commodities are associated with forest conversion. A landscape may remain green throughout the year while undergoing a major ecological transformation. Mapping tree crops separately helps reveal this difference.
6. Other Land: Beyond the Forest Canopy
The sixth category includes non-tree land uses, sparse woody vegetation, urban trees, small tree patches that do not meet the forest definition, and other wooded land. ForTy groups these varied landscapes into one broad class rather than distinguishing each land use separately.
This is an important limitation. Under the EUDR framework, other wooded land can have separate significance, including in assessments of forest degradation. Because ForTy does not map it as an independent category, additional information may be required to identify certain changes in canopy cover and forest condition.
How Accurate Is the New Global Forest Map?
The researchers evaluated the dataset using an independent, globally stratified sample of 9,100 sites. At the time of the study, 8,190 sites had been fully resolved through expert assessment and were used to estimate accuracy. Two trained experts independently assessed each site, helping provide a reference for evaluating the model’s predictions. The reported results were:
– 90.2% overall accuracy for the six-class classification.
– 94.8% overall accuracy for distinguishing natural forest from non-natural forest.
– 95.5% overall accuracy for distinguishing forest from non-forest.
These findings suggest that the dataset can provide useful information for global forest monitoring. However, accuracy figures should be interpreted carefully. They do not mean that every pixel or every forest type is classified correctly. Performance can vary by region, ecosystem, class, and the quality of the available reference information. Confidence scores and independent validation are therefore important when applying the map to specific conservation or regulatory decisions.
Why Does Forest Typology Mapping Matter for Climate and Biodiversity?
Better Carbon Accounting: Forests play a major role in the global carbon cycle. The study’s background notes that forests cover approximately 31% of the world’s land area and store around 662 gigatonnes of carbon. But carbon storage is not uniform across landscapes. Primary forests, secondary forests, planted forests, and agricultural tree crops can differ substantially in carbon stocks and how those stocks change over time. A detailed forest typology can improve carbon accounting by helping researchers distinguish the types of landscapes that are being lost, restored, or managed.
Stronger Biodiversity Conservation: Tree cover alone does not reveal the ecological quality of a landscape. A diverse natural forest and a single-species plantation may both appear as dense green canopies in satellite imagery, yet support very different communities of plants, insects, birds, and mammals.
Identifying remaining primary and naturally regenerating forests can help conservation planners prioritise areas where protection may deliver substantial biodiversity benefits. The dataset could also support assessments related to the Kunming–Montreal Global Biodiversity Framework, which includes the goal of conserving at least 30% of the world’s land and sea areas by 2030.
Improved Deforestation Monitoring: The European Union Deforestation Regulation makes distinctions between deforestation, forest degradation, and agricultural land use. A forest typology baseline can help monitor these processes by identifying the type of forest or tree-covered land present at a reference date.
For example, a transition from primary forest to an intensively managed plantation may involve substantial ecological degradation even if trees remain on land. Similarly, replacing natural forest with an agricultural tree crop can represent a major land-use change that a simple tree-cover map might fail to explain adequately.
ForTy’s 2020 baseline is designed to support this kind of assessment, although reliable monitoring also requires subsequent change-detection data and careful interpretation of the applicable legal definitions.
More Transparent Agricultural Supply Chains: Products such as cocoa, rubber, palm oil, coffee, and timber can be associated with different land-use histories. Separating agricultural tree crops from forest plantations and natural forests can help researchers and businesses assess potential deforestation risks more accurately. Such information may support supply-chain due diligence and conservation policies, but a global map alone cannot establish whether an individual product complies with the law. Parcel-level evidence, reliable supply-chain information, and applicable regulatory requirements remain necessary.
What Are the Limitations of the ForTy Dataset?
Despite its promising results, ForTy v1 should not be treated as a perfect or definitive representation of every forest worldwide.
First, the map represents conditions for 2020. It cannot, on its own, establish what has happened to a particular forest since then.
Second, forest type is partly a question of land use and management history, which cannot always be determined from satellite imagery. Some distinctions must be inferred from reference data and modelling.
Third, the 40-year rotation threshold used to distinguish planted forests from plantation forests is an operational simplification. It may not accurately reflect management practices across all countries and forestry systems.
Fourth, the Other Land class combines several different landscapes, including sparse woody vegetation and other wooded land. Users studying forest degradation may need additional datasets to separate these categories.
Finally, the reported accuracy measures represent performance across the validation sample. They should not be interpreted as a guarantee of equal accuracy in every biome or country. These limitations do not eliminate the dataset’s value. Instead, they highlight the importance of combining satellite mapping with field observations, regional knowledge, and other environmental information.
The Future of Forest Conservation: Look Beyond the Trees
Forests are more than collections of trunks, branches, and leaves. They are living systems shaped by ecological relationships, natural processes, human decisions, and the passage of time. A green landscape does not necessarily mean that a natural ecosystem remains intact. Nor does every planted forest lack ecological value. Understanding these differences requires better information about forest origins, management, structure, and land use.
The Forest Typology dataset represents a significant step towards that understanding by offering a globally consistent, high-resolution baseline for distinguishing natural forests, planted forests, plantations, agricultural tree crops, and other land. Its greatest contribution may be helping decision-makers move beyond the simplistic question of whether trees are present to ask a more important one: What kind of ecosystem is standing here, and what would be lost if it disappeared?
Protecting primary forests, allowing natural regeneration, restoring degraded landscapes, and managing plantations responsibly all have roles to play in a sustainable future. But these efforts must be guided by sound ecological evidence rather than tree-cover statistics alone.
At The Secrets of Nature, understanding the distinctions between forests is part of a broader responsibility to recognise nature’s complexity, respect its intrinsic value, and protect the living systems on which humanity depends. The better we understand our forests, the better equipped we become to protect them.
