Case studies

European farms and farm-scale sites where somebody actually measured what a technique did, over years, with a source that is not the promoter's own marketing. Fifteen pass the bar, one of them a documented failure. Their results sharpen the technique pages rather than contradict them.

Regenerative farming is full of success stories. Very few of them come with measurements. For this page a case had to meet three tests: results measured on a real farm or a farm-scale site, over several years, by a source that is not the farm's or the promoter's own marketing. Almost everything we found failed at least one of them.

Fifteen cases pass: five from the first search, two Portuguese grazing studies, an 85-year English ley–arable experiment, one Spanish trial that found a technique did not work, four irrigation trials from single orchards to a five-site network, an English ley experiment that also tested a bought inoculant, and a 54-field Swiss trial network. All are research-station or university-run trials on working farms, which tells you something in itself: we did not find a single European farm that measured its own results to a standard we could publish. The site's advice to "measure something" has no exemplar yet.

The fifteen that pass are worth more than the hundred that do not. Two of them are the sharpest evidence in the library.

Fundulea, Romania: forty-seven years of no-till in continuous maize

Measured. Average maize yield, 2010 to 2020: no-till 2,100 kg/ha, disk till 3,500, plough 4,400. Over more than forty years, no-till raised total soil carbon by 21% and microbial biomass carbon by 121% against conventional tillage.

What it means for you. This is the no-till page's central claim made concrete. The global meta-analysis behind that page finds the yield penalty of no-till disappears only when it is combined with residue retention and rotation. Fundulea removed rotation and kept the penalty for four decades, at about half the crop, on land below the 700 mm line. The carbon gain is real and the yield loss is real. Anyone telling you one without the other is telling you half.

La Conchuela, Córdoba: seven years of soil loss in an olive grove

Measured. Runoff coefficient: cover crop 1.2%, tillage 3.1%, bare no-till 11.9%. Soil loss: 0.8 t/ha a year under the cover crop, 6.9 under bare no-till. Olive yields are not reported in the case summary.

What it means for you. Cover protects a sloping Mediterranean soil; the absence of tillage does not. A bare, untilled, herbicide-treated floor was the worst of the three treatments, losing nine times the soil of a spontaneous cover. That is what the mulching page argues, measured on a real grove, and it is what Spain's eco-scheme pays for on steep olive land (see the funding page).

Rainfed almonds, Granada and Almería: five farms and their accounts

Measured and derived (Spain, 2016 to 2018 data, 2018 prices). Average yield 350 kg/ha of shelled almond at €6.50/kg organic; subsidy €229.50/ha a year. Net revenue in an average year:

Tillage carried the highest erosion cost and still beat no-till and green manure on net present value. No-till would need a 45% price premium, about €644/ha a year, to match it; green manure 27%, about €387. Pioneer organic growers in the region actually receive 5 to 7%. A study the paper cites found an unmown understory cut almond yield by 63%.

What it means for you. Below 400 mm, a living cover you do not manage competes with your trees for water, and the accounts show it. Compost was the only alternative that beat tillage financially, and the authors call it the most promising practice to scale. It is also the one floor system this library does not yet have a page for. The cover crops page carries the rainfall gate this case confirms; the stress test notes the measured range.

LIFE Agromitiga, Andalusia: soil carbon on a farm network, preliminary

Measured. Rainfed cereals in Seville province: direct seeding sequestered on average 1.78 t of carbon per hectare a year, while tilled soils lost carbon. Jaén olive groves, top 5 cm: 0.92% carbon under traditional tillage, 1.43% under living cover. The project itself calls these first results, varying widely even within one zone.

What it means for you. Farm-scale, institutionally measured, and consistent with the technique pages. But the numbers are surface-depth carbon, reported by a project whose lead partner exists to promote the practice. We publish them with that flag attached, and the practice they measured is on the no-till page. The wider question of whether surface carbon gains are new carbon or moved carbon is on the claims page.

Marchfeld, Austria: three long-term sites on 560 mm, data published, not yet read

Measured. Baseline soil organic carbon 1.20%, 2.33% and 1.89% at the three sites. Yields by tillage system and year for 2018 to 2022, including the dry years 2018 and 2022, are in the published dataset. We have not yet read them. This is the best candidate anywhere in the file for a continental, sub-700 mm, multi-year comparison, and it is on the list. The practice it compares is on the no-till page.

Portugal, paired pastures under aridity stress: four years of rotation

Measured. Under rotation, soil organic carbon 27 to 67% higher, total nitrogen 67 to 77% higher, water-holding capacity 6 to 17% higher; no difference in bulk density or microbial biomass.

What it means for you. The one European measurement of a soil-water gain from a grazing system, on exactly the dry land this site is for. The caveat: the abstract does not report stocking rates, so the method and the stock numbers cannot yet be separated. See rotational grazing.

Alentejo, Portugal: a montado sheep farm before and after rotation, with two caveats

Measured. Soil carbon accrual of 2.47 to 2.92 t per hectare a year after conversion; farm emissions down 34%; lamb footprint from 16.2 to 12.7 kg CO₂e per kg liveweight.

What it means for you. Accepted because it is measured on a working farm, and listed with its two confounds stated: it compares the farm with its own past across different years, and stocking fell 10% at conversion, which the rotational grazing page explains. The carbon figure is about seven times the independent average for improved grazing. Read it as "a farm that destocked slightly and rotated did well", not as proof of the method.

Woburn, England: the ley–arable experiment since 1938

Measured. Topsoil organic carbon from 0.98% in 1938 to, by 2005 to 2009, 0.91% under continuous arable, 0.80% with roots or fallow, 1.27% and 1.24% under three-year grass and grass–clover leys, and about 1.40% under eight-year leys. Leys added 0.26 to 0.36 t of carbon per hectare a year while building, then plateaued. Wheat yields were higher after the leys with less fertiliser nitrogen (as reported).

What it means for you. The longest soil-carbon series on this site, and the European evidence for putting grass back into the arable rotation. It also shows the ceiling: once the soil reaches its new balance, "no further CO₂-C will be sequestered". See ley–arable rotation.

La Poveda, Madrid: the first documented failure on this page

Measured. Weed density down 80 to 90%. Barley yield no different in the wet years (2016, 2018) and 50 and 61% lower in the dry 2019. Spring forage 1,038 to 2,704 kg of dry matter a hectare; summer forage from near zero to 6,209 kg depending on rain. Two of four perennial species died in winter.

What it means for you. Every other case on this page found something that worked. This one found that Australian pasture cropping "appears not to be well adapted to the environmental conditions of central Spain", because two crops share one dry spring. A measured failure is worth as much as a measured success, and the page has needed one. See pasture cropping.

Alvarado, Extremadura: the same equipment, two very different years

Measured. In 2016 the supervised zones received about 298 to 306 mm and the automatic, sensor-corrected zones about 198 to 232 mm, a saving of roughly 24%. In 2017 the same system saved about 6%, at about 405 to 444 mm against 434 to 435 mm. Yields were comparable in both years.

What it means for you. This is the most useful single case on this page for anyone being sold a sensor system, because it is the same equipment, the same orchard and the same operators in consecutive seasons, and the answer moved by a factor of four. Any supplier quoting one percentage is quoting one year. Budget for the variance, not the brochure figure. See precision irrigation.

Andria, Apulia: a vineyard where the sensors agreed with the farmer

Measured. Water saved: 10% in the first year and 17% in the second. Yield attributes showed "slight or no significant differences" and quality "no significant differences". The vines under the decision-support system were less stressed in both years.

What it means for you. A modest, believable result, and the useful part is what it was measured against: the farm's own scheduling, not a straw-man of over-irrigation. A tenth to a sixth less water for the same crop is worth having where water is billed by the cubic metre, and worth nothing on the bill where it is billed by the hectare. Which of those you are in is on the costs page, and the technique is on the precision irrigation page.

Spen Farm, Tadcaster: a ley beat the bought biology, and then some

Measured. Wheat after the ley yielded 7.7 t/ha; continuous ploughed wheat 4.3 t/ha, and ploughed plus inoculant 3.5 t/ha. Root colonisation by native mycorrhizal fungi was 52% after the ley against 36% under the plough. Sooty ear mould was 90% lower after the ley. The inoculant "tended to reduce rather than increase mycorrhization", and the authors conclude it "provided no tangible benefits to farmers or the environment".

What it means for you. The ley reached 92 to 106% of the UK average wheat yield that year on about a quarter of the usual nitrogen, which is the headline. But read the middle row too: the bought inoculant was applied to the same field, in the same season, and did nothing, while the rotation that costs no money at all did everything. This case is the clearest illustration on the site of why the free version usually wins. See ley–arable rotation and mycorrhizal inoculation.

Five olive orchards across Spain: the deficit that cost nothing

Measured. The deficit "saves more than 20% of water on average, without reductions in olive fruit or extra virgin olive oil yield". Stem water potential never crossed the −3.5 MPa stress threshold. The oil chemistry moved, in sterols and the fatty-acid profile, but not in phenols, quality parameters or the sensory profile.

What it means for you. Five sites, one cultivar, one orchard system, and the cleanest water-saving result on this site: a fifth of the water gone with nothing given up, using no hardware at all. The catch is in the specificity. The saving came from cutting water in one named growth window, and the window differs by crop, cultivar and region. Copying the percentage without the timing is how this goes wrong. See precision irrigation.

Alcamo, Sicily: the technique worked and the wine got paler

Measured. Total water applied was 42 mm in 2005 and 28 mm in 2006. Yield rose from 2.68 to 3.36 kg a vine on the clay and from 2.41 to 3.09 on the loam. Titratable acidity fell from 7.15 to 6.41 g/L. Anthocyanins fell from 1,343 to 1,119 mg/kg, about 17%. Sugar was unchanged at roughly 22.8 °Brix.

What it means for you. A very small, well-timed irrigation lifted yield by roughly a quarter. It also took about a sixth of the colour out of the grapes. On a bulk contract that trade is good. On a red sold on colour and structure it is a loss, and no yield figure will tell you which you are looking at. This page needed a case where a technique worked and still cost the grower something, because most measured successes are reported without the column that moved the wrong way. See precision irrigation.

Fifty-four Swiss maize fields: a trial network, and a lottery with a predictor

Measured. Growth response ranged from −12% to +40%. A significant gain of 12 to 40% in dry biomass appeared in 14 of the 54 fields; two fields lost 12%; about a third showed no change or a decline. The abundance of soil-borne fungal pathogens, not soil nutrient status, was the best single predictor of success, and soil microbiome indicators plus a few chemistry values predicted 86% of the variation.

What it means for you. This is a trial network rather than one farm, and we include it because fifty-four working fields is closer to what a farmer faces than any single site on this page. The finding is uncomfortable for both sides of the argument: the product can work, substantially, and it does so in about one field in four, chosen by something a standard soil test does not measure. Anyone selling it an average is hiding the distribution. See mycorrhizal inoculation.

What the cases agree on

• No-till without rotation fails on yield. Fundulea's maize monoculture halved. The meta-analysis condition "with rotation and residue" is a rule, not a footnote.
• Surface cover, not tillage status, controls erosion and runoff. La Conchuela's bare no-till lost 6.9 t/ha a year; the cover crop lost 0.8.
• Below about 400 mm, an unmanaged living cover costs money in perennials. Almond net revenue fell from €1,126 to €415 per hectare. The alternative that paid was compost.
• Surface carbon gains are real and still need the depth caveat. Fundulea +21%, Agromitiga 1.78 t/ha a year, both measured near the surface.

Rejected, and why

The bar matters more than the list, so here is what fell below it.

What is missing

• Farm accounts for anything but almonds. The UK cover-crop costing on the costs page is plot work, not a farm.
• Restinclières, INRAE's walnut and wheat alley-cropping platform in southern France, monitored since 1995 with true controls. Its measured yields are behind access walls we could not pass. The agroforestry page uses its published summaries.
• The Marchfeld yields, downloadable and unread.
• Any Portuguese, Italian or Greek farm with measured results for these techniques.
• A farm that measured its own water: a pond level, soil moisture, a metered pump. None found. If you are that farm, the costs page explains how to send figures.

If a case here changes your reading of a technique, the technique page has been updated to say so. The decision tool now carries the Fundulea warning on no-till.