The CSC site and experimental design have previously been described in Hawes et al. (2018). Briefly, the CSC long-term platform is a 42-hectare block of six arable fields, established in 2009 at Balruddery Farm near Dundee, Scotland (56.48 latitude, 3.13 longitude) (Figure 2). The farm is in a temperate maritime arable environment, with an average annual rainfall of 800 mm, an annual accumulated temperature of 1,100–1,375 day-degrees C (above 5.6 °C) and a mean annual potential water deficit of 50–75 mm. The area is moderately exposed (2.6–4.4 m s⁻¹ wind speed) and has moderate winters of 50–110 day degrees C of accumulated frost. The soils are imperfectly draining Balrownie Series with an average pH of 5.7. Topsoil depths range from 25 cm to 40 cm, textures from sandy loam to sandy silt loam and stone contents of 10–20% volume.

The CSC experiment is a split-field comparison of regenerative cropping against standard agronomic practice over multiple 6-year rotations of potatoes, winter wheat, winter barley, winter oilseed rape, spring sown faba beans and spring barley. Each field was divided into two, with conventional and regenerative treatments randomly allocated to each half and then planted with one of the six crops of the rotation in 2010 (Figure 3).

Replication of the 12 half-field plots (2 treatments × 6 crops) is year-on-year, providing 6 replicates of each crop/treatment combination per rotation. This necessitates more complex statistical analysis to handle potential confounding effect of year and field. However, it allows for large plot sizes (approximately 200 × 100 m blocks) which, unlike small, plot-scale experiments, enables the use of commercial farm machinery for sowing, management and harvest. Data generated on soil, biodiversity, crop and economics are therefore representative of field scale commercial situations, thus allowing realistic estimates of costs and benefits that are relevant to real-world commercial situations.

The conventional cropping system is ploughed annually following standard cultivation practices for the region and each of the crops in the rotation. Cereal straw is baled and removed. Fertiliser inputs are based on Scottish Nitrate Vulnerable Zone recommendations (for N) and an estimate of offtake (for P and K). Crop protection treatments (fungicides, insecticides and herbicides) are prescriptive and prophylactic, following label recommendations and standard commercial agronomy (Hawes et al., 2018). Specific products, active ingredients and input rates vary year on year depending on the crop requirements, field conditions and weather, but agronomic decisions are made with the goal to optimise crop yield and profit margin as in standard commercial, intensive cropping practice.

The regenerative system aims to enhance soil health and biodiversity while maintaining commercially viable crop yields. The goal for soil management is to improve physical structure to minimise erosion and runoff losses, provide better conditions for root growth, and sustain diverse communities of soil organisms to support carbon turnover, nutrient cycling and more efficient resource uptake via fungal and microbial associations. To achieve these multiple targets, the regenerative system at the CSC combines six management strategies into a single cropping system:

As with the conventional treatment, this regenerative system is flexible rather than fixed, based on indicator monitoring and an iterative process of performance review against management targets. This enables improvements in sustainability outcomes over time through modification of existing interventions and incorporation of new best practice options within the overall regenerative strategy. Full details of all agronomic interventions can be made available on request.

Long-term trends in indicators of soil, biodiversity and crop production are monitored to assess whole-systems impact of the regenerative system relative to conventional practice. Sampling was carried out annually between 2011 and 2024 following standardised assessment protocols at GPS-located sample points across the 12 treatment x crop combinations shown in Figure 3. Due to resource constraints, not all indicators were measured in all years at all sample locations. A list of the indicators reported here is given in Table 1, cross-referenced to the nodes in the agroecological network in Figure 1, along with their frequency and intensity of sampling.

Soil carbon (%C) and aggregate stability (%WSA) were measured as indicators of soil structure, both predicted to increase in the regenerative cropping systems relative to standard practice. Soil carbon was measured across the six fields in March before spring crops were sown from 2012 to 2022 (10 years data). Approximately 1.5 kg of bulk soil was sampled to a depth of 15 cm at each of 350 permanent GPS sample locations every year. Samples were sieved to 10 mm and stones exceeding 10 mm in diameter were removed. A sub-sample of sieved soil was then put through a 2 mm sieve, dried at 70 °C for 24 h, milled using a Retsch mill and 5 mg weighed for analysis using an Exeter Analytical CE440 Elemental Analyser (EAI, Coventry, UK) which determines the carbon, hydrogen, and nitrogen content in organic and inorganic compounds by combustion.

Aggregate stability was measured as the proportion of 2-8 mm diameter soil aggregates that remain intact when immersed in water (Le Bissonais, 2016). Eight bulk soil samples were collected from each treatment across the six CSC fields in March 2025. Soil from each of the 96 sample locations was air dried and a 5 g sub-sample was passed through an 8 mm sieve. The 8 mm sieved soil subsequently retained on a 2 mm sieve was used to test for water stable aggregates (%WSA). In a wet sieving apparatus (Eijelkamp Soil and Water, Netherlands), 2 mm-mesh sieves filled with the air-died soil are mechanically moved up and downward (up-down strokes = 1.3 cm) in a can filled with distilled water for 3 min ± 5 s (34 times/min). Unstable aggregates disintegrate and pass through the 2 mm sieve into the water-filled can under the sieve. The cans with soil from unstable aggregates were then removed and replaced by new water-filled cans. All stable aggregates retained on the sieve mesh were then fully disintegrated and soil from the stable aggregates was collected into the second set of water cans. Note that gravel, stones, plant residuals and roots remain on the sieve and only the soil from the aggregates passed into the water cans. After drying the two sets of cans with the soil from unstable and stable aggregates, %WSA is calculated as the proportion of stable aggregates in the sample (weight of the stable aggregates divided by the sum of both the stable and unstable aggregate fractions).

Earthworms were sampled from 9 locations in each treatment across all 6 fields in spring (April, after spring crops were sown) and autumn (October/November, after autumn crops were sown) each year from 2017 to 2024 (108 samples each year for 7 years). Sampling was carried out when soil was damp but not waterlogged or frozen. A 20 × 20 × 20 cm (8,000 cm³) soil pit was dug out at each location, placed on a white tray and hand sorted for earthworms. Juveniles were counted and returned to the pit. Adult earthworms were separated into three ecological functional groups following Stroud (2019): epigeic (surface active red coloured, small <8 mm long, found in leaf litter), anecic (topsoil dwellers, small-medium sized, pink, grey, green or mottled yellow), or endogeic deep vertical burrows, large >8 mm with red or black head. Numbers in each group were recorded, specimens photographed for future reference and then returned to the pit.

Soil microarthropods were collected from two pooled soil cores, using cylinders of 10 cm depth by 5 cm diameter (sample volume 392 cm³), taken at 10 GPS sample locations in each half of fields L (winter wheat) and K (winter barley), one- and two-years following potato cultivation, respectively. All 50 soil samples were collected within a week of each other at the end of March 2025. Soil was stored in a cool box after sampling and placed in Tullgren funnels (Capinera, 2008) within 4 h of collection. Soil microarthropod extraction followed similar procedures to Crossley and Blair (1991) but used 25 Watt bulb Tullgren funnel banks. Extraction into collecting jars of 70% ethanol occurred over 6 days and counts of Acari and Collembola were conducted under a stereo microscope.

Decomposition rates were estimated using the teabag method (Keuskamp et al., 2013) where 3 replicate green tea (Lipton EAN no.: 8722700 055525) and Rooibos tea (Lipton EAN no.: 8722700 188,438) teabags were dried at 70 °C for 48 h, weighed, labelled and buried 10 cm deep in 2 rows of 3 tea bags at 9 GPS sample locations in each of the 2 treatments across the 6 crops in the rotation. Sampling was carried out during the cropping seasons 2017, 2018, 2019, 2021 and 2024, 6 to 13 years from conversion to regenerative cropping. Tea bags were installed in April after spring crops were sown and retrieved approximately 90 days later after crop harvest. Loose soil and roots were gently removed, and the tea bags were then dried at 70 °C for 48 h before re-weighing. Initial weight, final weight and exact number of days of burial were used to calculate rate of litter decomposition for each tea bag following the formula described in Keuskamp et al. (2013). Higher values represent faster decomposition rates by microbes. This procedure measures decomposition by microbial activity since the tea is contained within a mesh too small to allow access by microarthropods or earthworms.

Taxonomic and functional diversity of species in the seedbank are an indicator of the long-term filtering effects of management on biodiversity. At the CSC, the arable weed seedbank was monitored by seedling emergence from soil samples taken across all 350 GPS sample locations in March over a 13-year period (2011–2024). Soil samples were sieved to 10 mm and approximately 1.2 L of the sieved sample placed in a 15 × 21 cm seed tray to a depth of 4 cm. Seed trays were arranged randomly in an unheated glasshouse on benches fitted with capillary matting, kept moist via an overhead automatic misting system. The volume and depth of soil and the general conditions for emergence, were similar to those in historical studies (Roberts, 1958) and more recent assessments carried out at the Hutton (Heard et al., 2003; Hawes et al., 2010). As seedlings emerged, they were identified to species, recorded and removed. The first flush of seedlings was considered complete after about 4 weeks, when there was no further emergence for a period of 14 days. The soil was then re-sieved, placed back in the seed trays and the procedure repeated for a second flush of emergence to ensure that the majority of viable seeds in each sample were recorded. The total number of weed seedlings per species per sample was converted to number per m² field surface to a depth of 0.15 m using the conversion factor 31.75 based on tray surface area.

The arable weed flora, germinating from annually disturbed soil seedbanks, provide resources for above- and below-ground foodwebs, supporting arable biodiversity within cropped fields. Weed densities were assessed in June each year from 2011 to 2024. The number of individual plants per species were recorded in 18 × 0.5m² quadrats per half field (providing 216 data points per year for 13 years).

Pollinating insects were sampled using nine sets of three 20 cm diameter coloured pan traps (blue, white and yellow) within the cropped area of each field half (integrated and conventional management) across 6 crops in June 2017, 2018, 2020, 2021, 2023, and 2024. Pan traps were positioned at the height of the crop canopy and filled with 100 mL per trap of dilute Teepol® detergent. Traps were set at 10 a.m. on a sunny day, not less than15 °C with a wind speed under 5 on the Beaufort scale. Traps were collected at 4 p.m. on the day of sampling and the insects sampled from the three traps at each location were combined into a single sample and stored in 70% ethanol until processing. Numbers of specimens in all pollinator groups (bumblebees, solitary bees, honey bees and hoverflies) were counted and recorded.

Carabids were sampled by pitfall trapping in April every year from 2011 to 2017 and again in 2024. In each field half, nine pitfall traps were installed 10 m apart in a line perpendicular to the field margin into the cropped area, starting 20 m from the field edge. This provides a sample number of 108 for the cropped area each year across all 6 fields. Pitfall traps comprised white plastic cups 8 cm deep and 7 cm diameter containing 80 mL antifreeze and inserted into the ground such that the rim of the cup was level with the ground surface. A rain cover was positioned over the top of each trap raised 5 cm above the ground surface on wire legs to reduce the risk of overflowing in heavy rain. Traps were left to run for 14 days before collection. All arthropods sampled were then transferred to tubes of 70% ethanol for storage, carabids were identified to species and counted.

Finally, agronomic indicators included crop yield and carbon footprint. Crop yields were recorded as weight of grain from the combine and potato offtake by tonne box for each 18 m wide tramline (five in each half of Road field, and six in each half of the remaining five fields). Data were converted to tonnes ha⁻¹ yr.⁻¹ based on sown area. Agronomic impacts on carbon footprint were estimated using the Cool Farm Tool (coolfarm.org version 2.29.0). This online calculator estimates CO₂ equivalents (CO_2e in kg ha⁻¹ yr.⁻¹) for all inputs, fuel use and agronomic operations to assess relative differences between crops and treatments in on-farm (direct) greenhouse gas emissions and soil carbon sequestration. Input data comprised basic field information (cultivated area, soil type), soil organic matter, fertiliser type (NPK, compost etc), fertiliser and compost application rates, crop protection inputs (fungicide, insecticide and herbicide type, application rates and % active ingredient), and field operations energy use (machine type (drill, plough, sprayer, combine etc) and number of operations). Based on these data, the carbon footprint was calculated for every crop, year and treatment providing 144 comparisons over two crop rotations.

All datasets were double-punched, checked, validated and entered into an SQL relational database for archiving. Data were analysed using Genstat 23rd edition and data used in this analysis will be made available via Zenodo.

To account for year-on-year replication in trial design, a Generalised Linear Mixed Model (GLMM) was used to test for treatment and crop or field effects under the composite null hypotheses:

For (1), the effect of management treatment on each indicator can vary with field and the cumulative effects of previous cropping and management history are more important than current crop type in determining the treatment response. For these indicator variables, Treatment and Field were therefore declared as fixed effects, and Year and Crop included as random effects to account for any confounding effect of year and crop on the interaction. The application of this model is indicated by (F) in Table 2. For (2), the current crop type and its management is more important than the field itself in determining immediate responses of each indicator to treatment. In these cases, Treatment and Crop were declared as fixed effects, and Year and Field as random effects. This model is applied to the indicators marked (C) in Table 2. Aggregate stability and the abundance of mites and collembola were only measured in a single year. In these cases, the confounding effect of crop and field cannot be distinguished (indicated with * in Table 2). This remains a limitation for interpretation of results for these three indicators until further years’ data can be included in the analysis.

One-sample Kolmogorov–Smirnov (K-S) tests were applied to each indicator dataset to determine degree of conformity to a normal distribution. For all indicators K-S p-values were <0.05 suggesting significant difference from a normal distribution. Apart from yield, no data transformations were performed. Instead, a Poisson distribution with log-link function was applied in each GLMM analysis to avoid assumptions of normality. To account for inherent differences in tonnes ha⁻¹ offtake of tubers, grain and oilseed, yield comparisons were made by converting t ha⁻¹ for each harvested strip in each year to the proportion of the maximum yield recorded for that crop across all years and both treatments. This relativised data was used to compare the effect of cropping system on overall crop productivity. Actual yields for each crop over time are shown separately.

Back-transformed mean values for each indicator by treatment, along with standard errors of the treatment difference, degrees of freedom, the F-statistics and probabilities for fixed effects and their interactions are given in Table 2.

To test for associations between weed density and pollinator activity, and between weed density and crop yield, Generalised Linear Regression models were applied with a Poisson distribution and log-link function. The fitted models were Dicots + Treatment + dicots.treatment with crop and year as random effects and pollinators or relativised yield as the response variables.

Finally, the whole-systems response to cropping system, incorporating all indicators together, was assessed using Canonical Variance Analysis (Genstat 23rd edition) with Treatment combined with crop rotational stage as the grouping factor. Rotational stage was defined on the basis of results from the GLMM where significant interaction between treatment and crop type indicated different responses with time since potato cultivation for many of the indicators measured. Rotation stage 1 was therefore classified as the potato crop and following winter wheat crop (sown into ploughed field soil). Rotation stage 2 were the remaining cereal, oilseed and bean crops (direct drilled 3 years or more after potato cultivation). This gave 4 groups (conventional vs. regenerative each divided into two rotational stages, early and late).

Soil structure is determined largely by soil type, level of disturbance and organic matter content. Reduced tillage and organic matter inputs in the regenerative treatment resulted in soils with higher carbon content, averaging 3.4% in the regenerative treatments compared to 2.4% under standard practice across all six fields. Significant differences were also detected between fields with K, L and M fields having less %C than N, O and P fields (Figure 3), possibly due to topography resulting in more erosion risk in the former compared to the latter. No significant interaction between treatment and field was detected over the ten-year period.

Increased soil carbon and reduced disturbance by tillage resulting in significantly greater aggregate stability (%WSA) in the regenerative soils (63.4%) compared to conventional (35.7%). Although this indicator was only sampled in 1 year, %WSA was significantly lower in the potato field and following winter wheat crop than in crops with a longer period since the last potato cultivation (Figure 4A). Intensive soil disturbance for potato planting and harvest had a negative impact on soil structural quality, bringing the %WSA down to just 6–7% in the first 2 years following conventional potato cultivation. This effect was less in the regenerative soils (16 %WSA in the first year recovering to 42% in the second year and >90% thereafter), indicating the potential for regenerative practices to mitigate some of the negative effects of potato planting and harvest on soil structure. Further monitoring is required to separate field from crop effects, but this interpretation is supported by Bartlová et al. (2015) who demonstrated positive associations between reduced tillage and %WSA, humic content and crop yields. They proposed a 5-point classification scale of structural quality from very low to very high based on %WSA. Averaged across all fields, CSC conventional soils fall into the “medium” category (34.1–50 %WSA) and the regenerative soils had a %WSA falling within the “very high” category (>66 %WSA) after 3 years of no-till. This suggests that regenerative practices incorporating more organic matter and minimising soil disturbance result in better structural quality.

Structural quality is important for water infiltration and water holding capacity, lessening risk of both waterlogging and drought, minimising erosion losses and improving the soil conditions for root growth (temperature, moisture and strength) (Franzluebbers, 2002). Informal observations at the CSC indicate that water logging over winter in the regenerative treatments may be less than in conventionally managed soil. However, direct measurement of root growth parameters in relation to these variables are needed to test this hypothesis and determine whether improvements in the regenerative system are agronomically significant in terms of improving crop resilience to extreme weather events.

Soil structural quality not only improves conditions for plant growth but also provides a better environment for soil microbes and arthropods: organic matter fuels below-ground foodwebs and increased soil pore size diversity creates environmental conditions for a diversity of soil organisms responsible for a range of soil functions (Hartmann and Six, 2023). Beneficial soil organisms include arthropod detritivores which break down dead plant and animal matter, microbial decomposers which are important in nutrient cycling processes, and symbiotic fungi and bacteria which improve plant nutrient availability and uptake. Microbial foodweb diversity also helps regulate pathogen populations by competition with neutral antagonists, potentially enhancing the pest suppressive properties of the soil (Bonanomi et al., 2018). In cultivated soils, these functions are reduced compared to natural, undisturbed ecosystems (Creamer et al., 2015). However, the impact of agricultural intensification on soil biota can be minimised in less intensively managed systems, depending on specific management options in relation to species’ functional traits [life history, body size, burrowing depth, habitat and food preferences (Postma-Blaauw et al., 2010)].

Earthworms contribute to soil functions such as structural maintenance, nutrient cycling, organic matter formation, decomposition and water infiltration (Frazão et al., 2017) and can increase crop yields up to 25% (van Groenigen et al., 2014). Non-inversion tillage can increase earthworm abundance to levels comparable to those found in undisturbed field margins (van Groenigen et al., 2014). Less intensive management options particularly reduced tillage and retention of crop residue are therefore needed to enhance earthworm populations and earthworm-related functions within arable fields (Stroud et al., 2017).

At the CSC, total number of earthworms averaged 9.1 per sample (225 m⁻²) in conventional and 14.3 (358 m⁻²) in regenerative soils across crops, years and seasons, but with differences between treatment response and overall abundance according to crop sequence. The most abundant category of earthworms were juveniles sampled in autumn which made up 60% of all earthworms sampled. This group showed the largest difference between treatments and crops through the rotation (Figure 4B). Fields cultivated for potato and those in the following winter wheat crops had the lowest number of earthworms and no significant difference between cropping systems. Two years after potato, in the winter barley crops, earthworm numbers were still significantly lower than in beans and spring barley at the end of the crop rotation but showing recovery in the regenerative system and a significant difference between treatments. By the fourth year after potato cultivation, the number of earthworms appeared to have recovered from disturbance during potato planting and harvest, and numbers sampled were higher in the regenerative system and remained higher than the conventional crops for the rest of the crop rotation.

Annual average yield for potato crops from 2011 to 2023 was 43.9 tonnes ha⁻¹ in the regenerative cropping system and 44.3 tonnes ha⁻¹ under conventional management. This difference was not statistically significant when tested using a linear mixed model on data from the first six-year rotation (Hawes et al., 2018). Potato yields were comparable to the national average for Scotland (40–55 t ha⁻¹) indicating good overall performance in both management systems and no apparent yield penalty in the regenerative treatment.

Winter wheat yields were significantly lower in the regenerative cropping system in the first rotation, showing a consistent 1 t ha⁻¹ yr.⁻¹ yield penalty with an average of 7.1 t ha⁻¹ compared to 8.3 t ha⁻¹ under standard commercial practice (Hawes et al., 2018). In the second rotation, mineral fertiliser requirement was calculated from soil nitrogen supply rather than applied at a standard reduced rate. This helped reduce the yield gap between treatments from 2017 onwards to an average of 7.4 t ha⁻¹ (regenerative) and 7.7 t ha⁻¹ (conventional). The ten-year winter wheat national average is 8.6 t ha⁻¹, comparable to the conventional cropping system over all years.

Winter barley yields averaged 8.3 t ha⁻¹ in conventional treatments compared to 7.2 t ha⁻¹ in regenerative systems, but despite the difference in the first rotation being non-significant (Hawes et al., 2018), variability was high and conventional crops appeared to perform better in 2014, 2019, 2021 and 2022 compared to those in regenerative systems. In the second rotation, from 2017, this could be due to a build-up of winter wheat volunteers with direct drilling, resulting in competitive effects in some years. Further investigation is required to determine possible causes. Both systems produced yields comparable to the national average of 7.5 t ha⁻¹.

Yields of oilseed were not statistically different between treatments in the first rotation (Hawes et al., 2018) and, over the whole period, averaged 3.7 and 3.2 t ha⁻¹ in the conventional and regenerative treatments, respectively. However, lower yields were recorded in some years for regenerative crops due to poorer establishment when direct drilled through the crop residue from the previous cereal crop and, in some fields, disproportionate pest damage where pigeons favoured the no-till side of the field. Issues with establishment were resolved in later years using non-inversion till rather than direct drilling into stubble, providing better seed-soil contact.

Regenerative management had no significant impact on bean yields in the first crop rotation, averaging 5.5 t ha⁻¹ compared to 5.8 t ha⁻¹ for conventional crops (Hawes et al., 2018). However, the move from non-inversion tillage to direct drilling in the second rotation may have contributed to a decline in the bean yield in three of the 6 years to just 3.3 t ha⁻¹ relative to standard crop management. The combination of direct drilling, dry soil conditions during establishment and weedy conditions in some fields can detrimentally affect final yield.

Spring barley yields were 6.1 t ha⁻¹ in the regenerative system compared to 6.5 t ha⁻¹ in the conventional and a national average of 6.7 t ha⁻¹. This difference was not statistically significant for the first rotation (Hawes et al., 2018). Lower yields in regenerative cropping in some years were most likely caused by competition with the clover companion crop (in 2014), and with weeds (in 2019) where under-sown clover limits the weed control options available. Soil compaction could also be an issue in the second rotation where direct drilling replaced non-inversion tillage. Spring barley is the last crop in the rotation before potato and therefore the longest in no-till. Some varietal effects were detected: cv. Sassy has a better rooting structure than other varieties tested and performed better in the regenerative system than the conventional, particularly in dry seasons.

To enable direct comparison across all indicators, mean yields per crop, treatment and year were converted to a relative value across all crops (Table 2). On average, yields in regenerative systems were 64% of the maximum compared to 72% in conventional systems (p < 0.01). No differences in relative yields were detected between fields and there was no treatment.field interaction. Weed competition can have a significant negative effect on crop yield, and reduced tillage systems are at risk of generating a weed burden over time. The CSC regenerative treatments had significantly greater weed densities than the conventional crops, but a Generalised Linear Regression model showed no negative correlation between weed density and relativised crop yield (p = 0.94, deviance ratio 0.14, d.f.126). This, together with the positive association between weed densities and pollinator numbers, suggests that occasional tillage and the targeted weed management approach used in the regenerative system at the CSC may be sufficient to minimise the trade-off between biodiversity and crop yield.

The small yield penalties observed for regenerative crops were likely to be due to a range of extraneous factors including, in some years, extreme weather (drought conditions in spring, water logging in winter), pest damage (pigeon grazing) and establishment issues (direct drilled OSR and beans). These uncontrolled variables represent risk factors faced by commercial growers in the real-world and highlight the importance of experimental testing at commercially realistic spatial and temporal scales to provide quantitative data on the actual risks, costs and benefits of alternative cropping strategies for growers. Based on these results, improvements to system design should focus on: crop residue management (to improve OSR seedling establishment), ameliorating compaction issues after 4 or 5 years with no tillage (cereals), further reducing reliance on mineral fertilisers (e.g., by increased use of legume companion crops), better integration of IPM strategies, and screening for crop varieties that perform better in no-till and reduced input systems.

The regenerative system had a lower carbon footprint than the conventional system (Table 2) due to increased soil carbon storage, reduced mineral fertiliser and crop protection inputs and, apart from potato crops, less tillage disturbance. With the exception of potato, all the regenerative crops produced a negative carbon footprint through carbon sequestration. This carbon footprint improved through the rotation with lowest values in the spring sown faba beans and spring barley crops after 4 years no-till (Figure 4G). Conventionally cropped beans had a lower carbon footprint than other conventional crops. The highest CO₂e across all crops was recorded for conventionally managed potato. The regenerative system therefore had less environmental impact than the conventional and an overall positive effect on greenhouse gas emission reductions.

Canonical Variates Analysis of all indicators together show a separation between conventional and regenerative systems and between crops through the rotation (Figure 7). The first axis (explaining 65% of the variation in the indicator data) separates the late-stage regenerative crops (min-till oilseed and direct drilled beans and spring barley) from all conventional ploughed crops and from early-stage regenerative crops (fields ploughed for potato and the following winter cereals planted into disturbed soil). The second axis (explaining a further 31% variation) separates the early-stage regenerative crops from all annually ploughed conventional crops. A clear distinction was therefore demonstrated between conventionally managed potato and winter cereal crops with low values for most indicators, and the regenerative spring sown beans and barley that had the longest time since potato cultivation and highest values for most indicators. Intermediate between these two extremes were the conventional spring crops (at the upper end of the conventionally managed range), and the regenerative potato and winter crops (at the lower end of the regenerative range). This suggests that, over multiple rotations, the combination of practices that comprise the CSC regenerative system can have a positive impact on soil health and biodiversity relative to standard commercial practice despite periodic disturbance for potato cropping.