We first identified the nine key soil regenerative practices (Table 1) by conducting a literature review and mapping practices from three foundational sources: two pioneer European initiatives beginning to identify emerging soil-improving cropping systems and soil carbon sequestering practices that combat biodiversity decline, desertification, erosion and more, namely (1) the European SoilCare project (Oenema et al., 2017) and (2) the 4 per 100 initiative (Pellerin et al., 2020), as well as (3) a study defining key elements of regenerative farming (LaCanne and Lundgren, 2018). The criteria for selecting practices to quantify included:

We then further revised our practice categories and nomenclature according to practices found in the literature review.

To quantify the observed impact on annual soil carbon sequestration rates of these nine regenerative practices, we performed an observational synthesis similar to Georgiou et al. (2022) from studies that included soil C sequestration data for each practice. To locate studies, we ran two Google Scholar queries. In the first query, we used the following search terms: (“vineyard” OR “viticulture” OR “wine” OR “grape*” OR “permanent crop”) AND (“management practice” OR “cropping system”) AND (“carbon sequestration” OR “soil organic carbon”) AND (“Kg C”). For the second query, we used the following terms: (“vitis vinifera”) AND (“practice” OR “manage*”) AND (“soil organic carbon”) AND (“Switzerland”). The overall goal was to find soil C sequestration figures relevant to viticulture, with the second query having a secondary focus on Switzerland to find studies for a separate but related qualitative examination of barriers and enablers to adopting regenerative practices among 20 French-speaking Swiss winegrowers (Villat, 2021). As with all literature searches, our search terms constrained the articles we found; see the ‘Future research’ section for our recommendations to build on the present work in the future.

The studies were then sorted for relevance and scanned to determine if they actually included a comparable C sequestration measure. Of the relevant studies from which a C sequestration figure could be extracted, 70% came from the first query and 30% came from the second query. Where relevant and possible, the data from meta studies was expanded and studies that were referenced were added. Additional sources recommended by entities such as the Changins School of Viticulture and Oenology, the Swiss Research Institute of Organic Agriculture (FiBL), and the University of Geneva’s Institute for Environmental Sciences were also included.

Observational, experimental, as well as modeling studies were sampled, coming primarily from North and South America, Europe, and China. There were also a few samples from Southeast Asia, Africa, Australia, and New Zealand. Studies included measured soil organic C at depths varying from 10 to 100 cm (averaging 21 cm across all studies gathered). The average study length was 15 years.

To be included, studies had to report the change in annual soil carbon sequestration upon adopting a regenerative practice. Studies were eliminated if they did not report a soil organic C sequestration rate directly (or a baseline measurement which enabled calculating a sequestration rate over time), or if the causal effect of the practice could not be dissociated from other practices. For a baseline, we looked for studies that indicated initial soil organic C before a regenerative practice was applied on control plots, or that measured the difference in soil organic C before and after the application of a regenerative practice. We did not include studies that simply indicated an absolute soil organic C value after applying the practice, because the effect of the regenerative practice could not be calculated. In some cases, a proxy practice was used, such as manure integration (a byproduct of animal integration), since there were few studies on animal integration directly.

For each relevant study, we noted the document identifier (DOI), author, year, title of the study, depth of the soil sample, length of the study in years, and which of the nine soil regenerative practices the study referred to. We also classified land use type as either “arable cropland” or “woody perennials” (Gattinger et al., 2019). Arable cropland included figures from arable, vegetable and mixed land for annual cropping systems as well as from grasslands. The term ‘mixed land’ is used to refer to studies or meta studies on the combined effect of a certain practice on soil organic C across a number of different sites, plots or studies with different types of annual crops. Woody perennial cropland included perennial agroecosystems, such as vineyards and coffee agroecosystems, which recent research indicates can be assessed for soil threats in the same way (Diti et al., 2020). Woody perennials are analyzed separately because they have a longer lifespan than annual arable crops and have a greater average gain in soil C over time (Ehrlén and Lehtilä, 2002; Ledo et al., 2020).

The carbon sequestration rate for each study was noted according to practice and standardized to tons of carbon per hectare per year (t C/ha/yr). The rate value came from averages calculated in independent studies. We compared the carbon sequestration rate for each practice within each of the two land use types using an ANOVA, followed by t-tests to compare means within practices. We also identified meta-analyses of average global carbon sequestration rates using soil-organic-C-enhancing practices for both arable and woody perennial agriculture, which we used as a point of comparison (Zomer et al., 2017; Payen et al., 2021).

On arable land, agroforestry and the use of double cover crops (one legume and one non-legume) had mean sequestration rates of 1.22 t C/ha/yr. (N = 14) and 1.20 t C/ha/yr. (N = 2), respectively (Figure 2). Combining a (non-legume) cover crop and no-tillage had a mean of 1.01 t C/ha/yr. (though also with a low number of samples, N = 6). Non-chemical pest management had a mean of 0.89 (N = 4), animal integration 0.67 (N = 8), and cover cropping 0.58 t C/ha/yr. (N = 15). Non-chemical fertilizer, no-till, and legume cover-cropping had similar sequestration means, averaging 0.48 t C/ha/yr. together (N = 187, N = 25, and N = 3, respectively). The widest range in values across all practices was for non-chemical fertilizers (N = 187), from emitting 3.8 to sequestering 5.9 t C/ha/yr., and was the only practice with negative outliers (Figure 2), indicating a source of C emissions in some cases.

For woody perennial land use, animal integration had a mean sequestration rate of 2.05 t C/ha/yr. (N = 4), and non-chemical pest management 1.86 t C/ha/yr. (N = 1) (Figure 3). Combining a cover crop and no-tillage (N = 2) had a mean sequestration rate of 1.43 t C/ha/yr., while the use of a cover crop alone (N = 6) had a mean of 1.31 t C/ha/yr. The use of a legume cover crop (N = 2) and no-tillage alone (N = 5) had a mean sequestration rate of 0.75 and 0.73 t C/ha/yr., respectively, while agroforestry (N = 5) had a mean of 0.67 t C/ha/yr. Non-chemical fertilizer had a mean sequestration rate of 0.04 t C/ha/yr. (N = 56), and included the lowest estimate (emitting 6.5 t C/ha/yr).

We found generally higher sequestration rates for regenerative practices in woody perennials (average across practices: 1.10) compared with arable land use (average across practices: 0.76) (Figure 4). Paired t-tests between arable and woody perennial land use revealed that although only the averages for non-chemical fertilizer were significantly different, animal integration yielded more than twice the mean carbon sequestration in woody perennial as in arable cropland (Figure 4). Non-chemical pest management sequestered 71% and the use of a cover crop sequestered 77% more in woody perennial than arable cropland.

The average sequestration rate from our dataset was lower for woody perennials than the 2.05 t C/ha/yr. reported as the global average by Payen et al. (2021) (Figure 4), whose study was based on a literature search of worldwide soil organic C sequestration in experimental studies using the following vineyard carbon sequestration practices: cover crops, no-tillage, amendments, biochar application, hedging and agroforestry (Payen et al., 2021). However, our data was higher for arable croplands than the global average of 0.56 t C/ha/yr. reported by Zomer et al. (2017) (Figure 4) whose study was based on soil organic C sequestration data found for cropland in global geospatial datasets, and using current status and a 20-year modeling scenario to show the potential increase from adopting practices including organic manures, cover cropping, mulching, conservation tillage, agroforestry, and rotational grazing (Zomer et al., 2017).

Our results indicate that combining practices that foster biodiversity rather than applying them individually may increase carbon sequestration. For example, for both arable and woody perennial land, combining the two practices of cover crop and no-tillage yielded a higher mean belowground sequestration value, at 1.01 t C/ha/yr. and 1.43 t C/ha/yr., than a cover crop (leguminous or not) or no tillage alone (Figures 2, 3). For arable land, double cover crops (one legume and one non-legume) yielded a higher mean sequestration rate (1.20 t C/ha/yr) than either (non-legume) cover cropping (0.58) or legume cover crops (0.41) individually (Figure 2).

For the seven regenerative practices we were able to quantify, we found that on average, all practices sequestered carbon. The only practice to show a statistically significant difference between land uses was non-chemical fertilizer, where sequestration was higher on arable land (Figure 4).

The high C sequestration for animal integration for woody perennials (Figure 3), the figures for which came mostly from studies of vineyards, aligns with findings of applying animal manure sequestering 20% more soil organic C than applying crop residues (Bolinder et al., 2020). Previous studies have shown additional benefits from integrating animals beyond carbon sequestration, including pest and fire prevention and reduced fertilizer and pesticide requirements from cattle grazing (Barbosa et al., 2019), as well as reduced nitrate leaching into groundwater, pests, and erosion, and increased water retention and nitrogen fixation from including perennial forages and incorporating animal manure (Asbjornsen et al., 2014). However, the overall impacts of integrated crops and animals largely depend on the interaction between grazing management and intensity (Lazcano et al., 2022).

For arable croplands, our finding that agroforestry sequesters a high amount of soil organic C fits with previous literature. This previous literature shows that combining woody perennials with other crops can increase soil carbon storage via net primary production, and through the return of tree litter to the earth, root and rhizosphere decomposition, and non-tillage of areas around woody plants (Chenu et al., 2014). This is also the case for traditional polyculture agroecosystems, such as olive-grapevine groves in Italy, where soil C was higher than in either crop grown intensively alone as a monoculture (Brunori et al., 2020). Increasing plant diversity can increase microbial biomass and therefore soil organic C (Chen et al., 2019). Allowing biodiversity to regenerate can also help strengthen vineyard pest resilience by regulating the ecosystem and make it difficult for pests to establish themselves (Guo et al., 2019).

Our study indicated that combining practices may increase C sequestration. For example, a combination of both a cover crop and no-tillage can be beneficial for mycorrhizal fungi development, thereby contributing to carbon sequestration. Symbiotic neighboring plants such as cover crops, interrow grass cover and certain weeds can allow arbuscular mycorrhizae to develop (Trouvelot et al., 2015). These mycorrhizae contribute not only to C reserve mobilization but also to plant development and prevention of soil erosion and water loss, which can otherwise be a limiting factor in biomass accumulation (including soil C) (Trouvelot et al., 2015). Meanwhile, no tillage has been found to allow mycorrhiza to form with less disturbance, even when rotating crops (Moitinho et al., 2020).

Our results for arable land indicate that a double cover crop (combining one legume and one non-legume) sequesters more soil C than either individually, and further studies may be able to show that the result also applies to viticulture, for the result is in line with recent studies that have shown that mixing grass and legume cover crops in vineyards worldwide with different soil types increases not only nitrogen but also soil organic carbon stocks (Ball et al., 2020).

Combining regenerative practices can help farmers combat climate change and biodiversity loss and help them move to what Soloviev and Landua (2016) call the second, multifunctional stage of regenerative farming. In this stage, an optimal mix of multifunctional practices, crop species, and carbon-sequestering strategies is adopted for the farm’s ecology, focused on improving the vitality of entire living ecosystems in soil and beyond (Soloviev and Landua, 2016). Indeed, regenerative practices may work best in concert (Figure 4) (Trouvelot et al., 2015; Teague, 2018), and combining soil regenerative practices that foster biodiversity rather than applying them individually may sequester more carbon. The long-term Jena Experiment shows that carbon storage strongly increases with increasing plant species richness (Weisser et al., 2017). Applying animal integration, non-chemical pest management, diverse cover crops, agroforestry, non-chemical fertilizer, and redesigning the system at the landscape level can all contribute to this species richness. Research is also emerging on the greater resilience to climate change of perennial plant communities with high species diversity, due to the greater probability of including species with traits adapted to climate change (Asbjornsen et al., 2014).

Woody perennial (including vineyard) soil generally exhibited higher increases in carbon sequestration with regenerative practices than arable agricultural soil, although the only statistically significant difference for a practice between land uses was for non-chemical fertilizer, which sequestered more carbon in arable cropland (Figure 4). Considering our sample size was an order of magnitude larger for this practice thanks to data from a meta study on organic farming (Gattinger et al., 2019), significant differences were easier to detect. However, it should be noted that Gattinger et al. (2019) considered organic fertilizer in the form of slurry or stacked manure to be non-chemical fertilizer, whereas for our analysis, we considered only compost or waste products from crop residues as non-chemical fertilizer, and considered manure application as the practice of animal integration. Indeed, regenerative agriculture as a closed-loop system would tend to use only manure produced on the farm, but organic agriculture [the focus of the Gattinger et al. (2019) study] allows the import of external sources of manure and slurry, which may have contributed to the net C emissions we observed for the negative outliers for organic farms using imported non-chemical fertilizer in Figures 2, 3. Regardless of the source, the amount of fertilizer incorporated must be carefully monitored to avoid problems related to overuse including nutrient loss, soil acidification or alkalinization, reduction of microbial communities, and water contamination (Han et al., 2016).

The combination of no-tillage and the use of a cover crop appeared to have a slightly greater effect in woody perennials (Figure 4). While arable crops often have their roots disturbed or removed and have to start from scratch each year, woody perennials like grapevines have the time to develop symbiotic arbuscular mycorrhizae. These mycorrhizae enhance the vine’s ability to obtain water and nutrients from the soil, and to collect carbon to store below ground (Trouvelot et al., 2015). No tillage in combination with cover crops that allow mycorrhizae to colonize roots may therefore contribute to higher carbon sequestration, especially in perennial crops like grapes that stay rooted in the ground longer.

Regenerative practices not only potentially work in concert to sequester more soil organic C, but are part of a whole system that includes biodiversity, human health and well-being, long-term economic health and more. Drawing together our findings, we visualized a framework for how the different soil regenerative practices and their extractive opposites could affect outcomes such as biodiversity, soil carbon sequestration, long-term agricultural productivity, and human health and wellbeing (Figure 5). The different regenerative practices we quantified here act as levers to increase the rate of soil organic C sequestration (inflow). Meanwhile, their opposites, for which we do not have C emission data, are hypothesized to act as levers to increase the rate of carbon emissions into the atmosphere (outflow). For example, as the reinforcing green arrows from animal integration show, well-managed animal integration contributes positively to outcomes such as biodiversity and increasing the carbon sink rate (which in turn augments the soil carbon stock). In contrast, high nitrogen fertilizer input tends to decrease biodiversity, as shown by the detracting red arrow. Note the reinforcing loop with two positive flows between the outcomes of long-term vineyard productivity and human health and well-being, showing that the two are mutually dependent on one another, notably when the human in question is the grower.

In sum, carbon sequestration and soil management practices are pieces of a bigger regenerative ecosystem. Even though reversing biodiversity loss is central to regenerative agriculture, it is often disregarded in favor of other soil health indicators (such as carbon sequestration) when discussing recommended practices (Giller et al., 2021). However, adopting soil regenerative practices has the potential to help regenerate degraded biodiversity, increase soil carbon sequestration for climate mitigation, ensure long-term farm or vineyard productivity, and secure human health and wellbeing. On arable croplands, agroforestry may sequester slightly more carbon, and in woody perennials, animal integration may sequester slightly more than other practices. Overall, we found evidence that any of the seven soil regenerative practices we studied can increase soil C sequestration; combining practices may further strengthen carbon sequestration and other elements of a healthy ecosystem. Because woody perennials, such as grapevines, sequester more soil organic C than arable land uses, they may be a particularly favorable crop to lead the way for regenerative agriculture.

To make best use of limited resources, winegrowers and other farmers might consider a sequenced adoption approach for regenerative practices that balances the effectiveness of carbon sequestration with the perceived ease of adoption (Villat, 2021). Visualizing the average soil C sequestration of each practice (from the average values combining arable and woody perennials from Figure 4) and the perceived ease of adoption of the practice based on interviews with 20 winegrowers in Switzerland (Villat, 2021), we recommend growers start with potential quick wins, where soil C sequestration potential is highest and the practices are perceived as easiest to adopt (upper right quadrant, Figure 6). This quadrant includes the combined use of a cover crop and no-tillage, non-chemical pest management, the use of a cover crop, and the use of a double cover crop (combining a legume and a non-legume). The top left quadrant, including agroforestry and animal integration, represents practices that need promotion, or need to be made easier, for they are perceived as more difficult to adopt but represent high soil C sequestration potential. Incentives or strategies for enabling these practices are therefore important. The C sequestration potential of redesigning the system at the landscape level and reduced traffic is unknown and farmers perceive them as difficult to adopt; further research is needed to quantify their outcomes before these processes are prioritized. Finally, practices that are perceived as easy to adopt but whose carbon sequestration potential is relatively lower need less time spent on incentives or promotion (lower right quadrant).

Based on current research and evidence, we advise farmers and policymakers to prioritize the adoption of effective and feasible agricultural practices. These may include a combination of using cover crops and no-till methods, implementing non-chemical pest management strategies, and utilizing double (legume and non-legume) cover crops. Additionally, attention should be directed toward the support and development of animal integration methods, which have been proven to be effective in terms of carbon sequestration but may be met with resistance from farmers. It is crucial to recognize the potential benefits of these practices and provide adequate resources to facilitate their implementation.

Although robust measurements are needed to back up the claimed benefits of regenerative agriculture, available data are often not sufficient to do so. Data demonstrating the beneficial outcomes for many soil health practices remains a challenge (Mattila et al., 2022). It is notable that the practices with the largest sample sizes had the lowest estimated C sequestration rates (Figures 2, 4), so it is important to be cautious about the quantitative potential for regenerative practices to sequester carbon until more data are available.

In this study, we were able to provide a robust estimate of the impact of non-chemical fertilizer on soil carbon sequestration (N = 243), and reasonably robust estimates for three practices in arable crops: no-till (N = 25), cover crop (N = 15), and agroforestry (N = 14) (Table 2). Three of the remaining single practices (legume cover crop, animal integration, and non-chemical pest management) and two combined practices [combining a legume and non-legume cover crop (double cover crop) and combining cover crop and no-tillage] were only quantified in a handful of cases, particularly for woody perennial systems (Table 2). Two of the nine regenerative practices we identified from the literature (low traffic and redesigning the system at the landscape level) did not have any quantification of their effect on soil carbon outcomes available in the literature. More studies are needed to assess carbon sequestration potential for these practices, to achieve greater statistical power and higher confidence in sequestration rates. There is a particular need for more studies in woody perennial systems such as vineyards, and particularly from long-term experimental and monitoring studies and randomized controlled trials, which despite difficulties in detecting small changes in soil carbon pools in agricultural soils, provide a necessary and valuable complement to modeling estimates. Now that we have identified relevant regenerative practices, we recommend future synthesis studies search the existing literature for observed sequestration rates of specific regenerative practices by name, building off of the practices listed in Table 2 (e.g., searching for “animal integration,” “low traffic,” etc., and related synonyms).

Ideally, we would have analyzed soil C rates according to soil depth measured, length of study, bulk density, and soil type. However, 28% of studies did not report soil depth at which soil C was measured, and 22% did not report study length, despite reliable soil C figures only accruing over several years. Many studies did not report bulk density, making it difficult to estimate the extent of the effect of bulk density on reported soil C accrual rates. Soil type was also not always reported, either. Often these were studies measuring a number of other factors aside from soil C. Moreover, in the studies synthesized we found a range of 14 different units used to estimate soil C, from t C ha⁻¹ y⁻¹ to SOC % increase per year, to grams per kilogram increases, all which had to be converted to the t C ha⁻¹ y⁻¹ standard for comparability. For better comparability, we recommend that future studies measuring soil C use the international measurement unit “t C ha⁻¹ y⁻¹” and report as much information as possible, but at least measurement depth, bulk density and soil type.