Permaculture has a storytelling problem. The movement is generous with inspiring before-and-after photos and thin on the kind of numbers a sceptic would accept. If you have ever watched a desert-greening video and thought "but did anyone actually measure that", this article is for you.
Below are projects with documented, quantified outcomes: hectares treated, tonnes of sediment stopped, water tables measured before and after, yields recorded across four decades. Where the evidence is strong, we say so. Where a famous project rests on its own press releases, we say that too. For the cultural and historical context behind these approaches, see our overview of how different cultures grow food.
Here is what this covers:
Key Takeaway
The strongest evidence for permaculture principles does not come from permaculture-branded farms. It comes from large restoration programmes and long-running agricultural trials that apply the same logic under different names, with budgets big enough to fund proper monitoring.
China's Loess Plateau Watershed Rehabilitation Project, by a wide margin. Run from 1994 to 2002 across nine Yellow River tributary watersheds, it treated 1,560,000 hectares (about 3,855,000 acres) of severely degraded land at a total cost of roughly $250 million, or about $65 per acre.
The interventions were textbook water-and-earth work: terraces on 223,700 acres, shrub planting on 336,100 acres, orchards on 76,300 acres, plus 149 key dams, 1,140 warping dams and 1,956 check dams.
| Indicator | Before | After |
| Annual grain output | 427,000 tonnes | 698,600 tonnes |
| Fruit production | 80,000 tonnes | 345,000 tonnes |
| Per capita net income | 360 yuan | 1,263 yuan |
| Population below poverty line | 59% (1993) | 27% (2001) |
| Land under perennial plants | 37% | 54% |
| Sediment into the Yellow River | Baseline | ~57M tonnes/yr less |
Source: World Bank Implementation Completion Report, 2004
The economic rate of return came in around 29%. A separate peer-reviewed analysis by Deng and colleagues found that China's broader Grain for Green programme had put roughly 8.6 million hectares (21.3 million acres) of the Loess Plateau under vegetation restoration since 2000, significantly reducing annual carbon loss through erosion.
Two caveats. The completion report is self-evaluative, written by the funder. And attributing the poverty drop entirely to terracing is generous, given that China's economy was growing fast through that decade.
The Rodale Institute's Farming Systems Trial in Pennsylvania has run side-by-side organic and conventional grain plots for more than 40 years. It is the longest-running comparison of its kind in the United States, and it produces the numbers permaculture advocates should be quoting instead of testimonials.
The headline result is about resilience rather than raw output. In normal years, organic and conventional yields are broadly comparable. In drought years, organic corn yields have run about 31% higher, with the gap in individual drought years measured between 28% and 34%.
The mechanism is not mysterious. Organic management built soil organic matter, which improved structure, which raised water infiltration by 25% to 50%. More rain gets in and stays there, so the crop has a buffer when rainfall fails.
Why This Works: Building Buffers, Not Yields
This is the permaculture argument stated in agronomic terms. Conventional systems optimize for peak output in a good year. Systems built on soil biology and organic matter optimize for the bad year, because that is when the difference between surviving and failing gets decided. The same reasoning drives everything from deep mulching to swales on contour: catch it, store it, slow it down.
The Abrha Weatsbha community in Ethiopia's Tigray region has the clearest small-scale numbers. Their method was mostly social: agree collectively to exclude grazing and woodcutting from degraded hillsides, then add tree planting and water harvesting structures.
The Equator Initiative case study reports that these measures resulted in more than 50% of rainwater being trapped to recharge groundwater rather than running off, with the water table rising from about 9 metres depth to between 2 and 4 metres.
That is a rise of roughly 5 to 7 metres (16 to 23 feet), which changes what is possible. Hand-dug wells became cheap. Dried-up springs started flowing. Irrigated high-value crops now come off two to three times a year regardless of rainfall.
In Mexico, Las Cañadas in Veracruz offers a different kind of record. The community reserve covers about 406 hectares (1,003 acres). The owners sold all their cattle and reforested 60 hectares (148 acres) with 50,000 native trees, letting the remaining meadows regenerate on their own. An independent seedling establishment study found oak survival around 83% outside the forest fragment, which is a genuinely measured number rather than a claim.
Get Our Free Permaculture Starter Guide
Join thousands of gardeners getting practical, evidence-based advice on what to plant together, how to build soil, and how to design a garden that keeps giving.
Read the Free GuideTwo of the most famous projects in the movement have the weakest public data. This is not an attack on either. It is a request that we stop citing them as proof.
Geoff Lawton's Greening the Desert site in Jordan's Dead Sea valley is the image most people carry of what permaculture can do. The project's own account of the sequel site describes design philosophy in qualitative terms. It gives no soil salinity readings, no yields, no survival rates, no groundwater data. The original site was around an acre. Reports that it struggled during a funding gap are anecdotal in both directions.
Mark Shepard's New Forest Farm in Wisconsin is the American equivalent. Roughly 106 acres (42.9 hectares) with an estimated 250,000 trees planted over 15 years. That tree count is impressive on its own. But there is no peer-reviewed study of the farm's yields, soil change or profitability, which makes it hard to answer the obvious question of whether the system beats the alternative.
Sadhana Forest reports 28,000 trees planted in Auroville, India and more than 80,000 in Anse-a-Pitres, Haiti, with survival rates above 80%. Those are useful figures. They are also self-reported and unverified.
Common Mistake to Avoid
Do not cite tree-planting numbers as restoration outcomes. Seedlings planted is the easiest thing in the world to count and the least meaningful. Survival at year five, canopy cover, water table change and yield are what actually matter, and they are much rarer in project reporting because they are harder and more expensive to measure.
Which leads to a practical filter you can apply to any project claim you encounter.
Ask who did the measuring
A funder, a university or a third-party evaluator is a different class of evidence from the project's own website. Neither is automatically wrong, but only one had an incentive to publish inconvenient findings.
Look for a baseline
An outcome without a documented starting point is a description, not a result. The Loess Plateau report is credible partly because it states what the land was doing in 1993 before anyone touched it.
Check the time horizon
Year-one photos prove almost nothing about a perennial system. Ask what the site looked like at year five and year ten, and whether anyone went back to check.
Separate inputs from outcomes
Trees planted and volunteer hours are inputs. Survival rates, water table depth, yield and income are outcomes. Projects reporting only inputs usually have a reason.
It is the cautionary tale, and worth understanding properly. Pitched in 2007 as a continuous tree barrier 8,000 km long and 15 km wide across the Sahel, it has since become a mosaic of restoration projects. The UNCCD now frames its ambition as restoring 100 million hectares (247 million acres), sequestering 250 million tonnes of carbon and creating 10 million green jobs by 2030.
Those are targets, not outcomes. Reporting drawing on The New Humanitarian's investigation found that 15 years in, only around 4% of the trees were actually in the ground.
The projects that have worked within the Great Green Wall footprint tend not to be the plantation ones. Farmer-managed natural regeneration, where farmers protect and prune tree stumps already in their fields rather than planting seedlings, has produced far better survival at far lower cost. That is a permaculture-compatible lesson: work with what the site is already trying to do.
The most useful review is Krebs and Bach's 2018 paper in Sustainability. Their conclusion is more supportive than sceptics expect and more qualified than advocates would like: there is scientific evidence supporting all twelve of Holmgren's permaculture design principles.
The catch is where that evidence comes from. It sits in agroecology, hydrology, forestry and soil science literature, not in trials of whole permaculture systems. "Catch and store energy" is supported by decades of rainwater harvesting research. "Use and value diversity" is backed by agroforestry meta-analyses. The individual principles hold up. The integrated designs built from them have almost never been measured.
That gap is structural rather than evasive. Polycultures are genuinely hard to evaluate with the plot-and-control methods agronomy relies on. Which is exactly why the Loess Plateau and Rodale numbers matter so much: they are the rare cases where somebody funded proper measurement over a long enough period.
Key Takeaway
The right conclusion is neither "permaculture is proven" nor "permaculture is unproven". It is that the underlying principles have solid science behind them, and the movement badly needs more projects that measure themselves. If you run a site, keeping records is the most useful contribution you can make.
Few that are permaculture-branded. The strongest peer-reviewed evidence sits in adjacent fields: Deng and colleagues on Loess Plateau land-use change, seedling establishment studies at Las Cañadas, and agroforestry meta-analyses. The Rodale Farming Systems Trial is rigorous and long-running but describes itself as organic rather than permaculture. Krebs and Bach's 2018 review found scientific support for all twelve permaculture principles while noting that support comes from studies of individual practices, not integrated whole-system designs.
The Loess Plateau Watershed Rehabilitation Project in China, treating 1,560,000 hectares (about 3,855,000 acres) between 1994 and 2002. It was never labeled permaculture, but its methods are the same ones permaculture teaches: contour terracing, check dams, converting steep annual cropland to perennial cover, and managing grazing. Documented outcomes include grain output rising from 427,000 to 698,600 tonnes and sediment into the Yellow River falling by roughly 57 million tonnes a year.
Publicly available data is thin, which is itself the answer. Most well-known permaculture farms earn a significant share of income from courses, consultancy and tours rather than production, and few publish farm accounts. The Rodale trial found organic grain systems competitive on yield and better in drought years, which supports the economic case indirectly. If you are weighing this for your own land, our piece on what a food forest is actually worth works through the numbers at household scale.
Because many are built around one person's tacit knowledge rather than documented systems. Complex polycultures need someone who understands why each element is where it is. When that person goes, maintenance decisions get made without the design logic behind them, and the system drifts. Projects that persist tend to have written management plans, multiple trained people, and an income stream that does not depend on the founder's personal reputation.
Sometimes, and much less often than planting campaigns imply. The Great Green Wall showed roughly 4% of intended trees in the ground after 15 years, and survival of those planted is frequently poor. Approaches that work with existing vegetation, such as farmer-managed natural regeneration and grazing exclosures, have consistently outperformed planting on cost and survival. Ethiopia's Abrha Weatsbha achieved a water table rise of 5 to 7 metres largely by agreeing to keep livestock off hillsides. Protection often beats planting.
Ready to Grow Smarter?
Get our free beginner's guide to permaculture gardening, 12 pages of practical tips you can use this weekend.
Read the Free Guide