- A recent study used machine learning to ask the question: What would land look like if humans didn’t exist?
- The results tell us all we need to know, a new op-ed argues: If we want to restore land effectively, we must understand the land.
- “This research is a reminder that ecological restoration isn’t about chasing a single, idealized pristine past without people. It’s about managing land in ways that are place-based,” the author writes.
- This article is a commentary. The views expressed are those of the author, not necessarily of Mongabay.
The sun is already sharp against the terraced slopes, the scent of thyme and rosemary rising from the rocky ground. Dry wind whistles through esparto grass, and somewhere in the distance is the soft clang of goat bells. A partridge erupts from the scrub. No roads. No fields. No fences. Just the land, as it might be without centuries of plowing, grazing or concrete.
We often imagine “pristine” nature as a wilderness untouched by people, but rarely do we picture ourselves in it. And yet, places like this have been shaped by humans for millennia through traditional practices and knowledge.
The difference now is that we have become separate from the needs of the land. The knowledge that was once passed down from centuries about knowing when to burn, when to rest the land, and when to intervene has become distant. Our disconnect from place is reshaping landscapes and undermining restoration. Reconnecting is essential for healthy, thriving land.
A recent study published in Nature Communications used a machine-learning model to ask the question: What would land look like if humans didn’t exist? And the results tell us all we need to know: If we want to restore land effectively, we must understand the land.
The paper covers forests, grasslands, shrublands and bare landscapes to explain that successful restoration is completely context-dependent, and that the wrong strategy, however well-intentioned, can damage the very system it’s meant to revive.
The tyranny of tree planting
One key finding: the study directly challenges assumptions that planting trees is the best way to restore land. While forests are vital in many places, the research shows that short-vegetation ecosystems like grasslands and shrublands are also natural, in ecologically valuable states, especially across arid and subtropical zones.
According to the authors’ modeling, just 43% of global landscapes would naturally support trees; 39% would be dominated by short vegetation, and 18% would remain largely bare. Yet, around the world, we’ve seen tree-planting projects rolled out in landscapes that were never meant to hold dense forests.
In Brazil’s Cerrado savanna, for example, experts have said that restoration should focus on native grasses and shrubs, not tree plantations. Get that wrong and you risk degrading biodiversity, reducing water availability, and even harming climate regulation, particularly if nonnative, fast-growing trees are used. Projects like Hyundai’s 2008 carbon offset project, which aimed to restore 1,200 hectares (about 3,000 acres) of forest in the Cerrado despite the main ecosystem being scrub-covered fields and grasslands, exemplify the flawed logic of conducting restoration while ignoring the local ecosystem. For this reason, the project was never certified and did not provide offsets.
The failure of the Hyundai case proves that to restore any landscape, you need to work with the specific ecology of the area, its community dynamics, and its socioeconomic realities.
Management matters
To create their map, the authors trained machine-learning models on data from more than 40,000 protected plots, factoring in climate, soil, fire frequency and wildlife grazing. What emerged was a detailed picture of each region’s natural potential of what would grow if the land regenerated without human interference.
One striking insight is the concept of “natural alternatives.” At least 675 million hectares (about 1.67 billion acres, nearly the size of the Amazon) could support more than one type of natural vegetation, depending on local conditions and management. This is a reminder that ecosystems are dynamic and complex, shaped by disturbances like fire and grazing as much as by the climate.
In fact, the study highlights that management decisions, such as how we handle fire or grazing, can have more impact on vegetation cover than predicted climate change impacts. In the Dinaric Alps of Southeast Europe and the West Sudanian savanna, factors of land management outweighed the influence of expected climate changes by 2050.
What does this tell us? Land and climate are locked in a feedback loop. Strip away vegetation, and the land loses carbon, water and resilience, feeding climate change. Restore vegetation through smart, place-based management, and the reverse happens: carbon is stored, biodiversity recovers, and landscapes become climate-resilient.
And we’ve seen this live, in action. In southwest Australia, the Noongar peoples’ traditional land management techniques have worked with the natural rhythms of the ecosystem for generations. Native seed collection and propagation, guided by traditional knowledge of species, soil and seasonal indicators, ensure that revegetation efforts match the land’s ecological needs. Controlled burns reduce the risk of destructive wildfires while stimulating the growth of native grasses and shrubs, which stabilize soils, store carbon and improve water infiltration. Paired with regenerative farming practices, this approach is restoring soil health, reconnecting habitat corridors, and creating livelihoods that keep people on Country.
In this landscape, the decisions about when, where and how to manage vegetation are rooted in thousands of years of observation and stewardship. It is clear that in order to restore land, we must center and support traditional Indigenous knowledge.
Right ecosystem in the right place
This research is a reminder that ecological restoration isn’t about chasing a single, idealized pristine past without people. It’s about managing land in ways that are place-based. The study’s data show us that vegetation cover is shaped as much by how we manage fire, grazing and water as by the climate itself. Aligning management choices with what the landscape truly needs is key to long-term resilience.
That means connecting to the soil beneath our feet, looking up at the canopy above our heads, and listening to the communities who know these systems best, especially traditional custodians whose knowledge has sustained them for thousands of years. When land management follows this knowledge, restoration works with nature’s rhythms, ensuring the right species thrive in the right places for the long term.
Restoration, at its best, begins with a simple question: What is this landscape meant to be, and what does it need now?
If we can answer that honestly, we stand a far better chance of restoring the right ecosystems, in the right places, in the right way.
Willem Ferwerda is founder of Commonland, a nonprofit organization that brings people, NGOs, businesses, and governments together to restore landscapes using a replicable, long-term approach.
Banner image: A staff member of Kinesi Nursery in Tanzania puts native plant seedlings in pots. Image courtesy of WeForest.
Related audio from Mongabay’s podcast: A discussion of the “right tree, right place, right community” approach to reforestation, and why local, community-based programs succeed:
See related content about restoration:
How to pick a tree-planting project? Mongabay launches transparency tool to help supporters decide
Why I quit the film industry to work on ecological restoration (commentary)
Citation:
Bastin, J.-F., Latte, N., Bogaert, J., Garcia, C. A., Berzaghi, F., Maestre, F. T., … Lejeune, P. (2025). Global alternatives of natural vegetation cover. Nature Communications, 16(1), 6484. doi:10.1038/s41467-025-61520-8
Credits
