Carbon for soils, not soils for carbon

Abstract

The role of soil organic carbon (SOC) sequestration as a ‘win-win’ solution to both climate change and food insecurity receives an increasing promotion. The opportunity may be too good to be missed! Yet the tremendous complexity of the two issues at stake calls for a detailed and nuanced examination of any potential solution, no matter how appealing. Here, we critically re-examine the benefits of global SOC sequestration strategies on both climate change mitigation and food production. While estimated contributions of SOC sequestration to climate change vary, almost none take SOC saturation into account. Here, we show that including saturation in estimations decreases any potential contribution of SOC sequestration to climate change mitigation by 53%–81% towards 2100. In addition, reviewing more than 21 meta-analyses, we found that observed yield effects of increasing SOC are inconsistent, ranging from negative to neutral to positive. We find that the promise of a win-win outcome is confirmed only when specific land management practices are applied under specific conditions. Therefore, we argue that the existing knowledge base does not justify the current trend to set global agendas focusing first and foremost on SOC sequestration. Away from climate-smart soils, we need a shift towards soil-smart agriculture, adaptative and adapted to each local context, and where multiple soil functions are quantified concurrently. Only such comprehensive assessments will allow synergies for land sustainability to be maximised and agronomic requirements for food security to be fulfilled. This implies moving away from global targets for SOC in agricultural soils. SOC sequestration may occur along this pathway and contribute to climate change mitigation and should be regarded as a co-benefit.

1 INTRODUCTION

Keeping global warming below 1.5°C (consistent with art. 2 of the Paris Agreement) requires immediate action to reach net-zero carbon dioxide (CO₂) emissions in the coming decades (IPCC WG3, 2022). Net-zero CO₂ emissions are achieved when anthropogenic emissions are balanced by anthropogenic CO₂ removal. To limit global warming to 1.5°C, nearly all scenarios rely on the removal of CO₂ from the atmosphere. This is also the case for many scenarios limiting global warming to well below 2°C (IPCC WG1, 2021), confirming earlier assessments (van Vuuren et al., 2013).

In the discussions on carbon dioxide removal (CDR) techniques, SOC sequestration has gained significant societal and scientific attention in recent years. The exact climate change mitigation potential of SOC sequestration is strongly debated, but unlike other CDR techniques, SOC sequestration is claimed to be a ‘no regret’ or a ‘win-win’ option because its implementation provides numerous co-benefits and few trade-offs (IPCC WG1, 2021). Indeed, management interventions aiming at SOC sequestration have long been claimed to enhance soil fertility and productivity, increase soil biodiversity, improve water retention and purification, and reduce erosion, compaction, runoff and water pollution (Lal, 2004; Paustian et al., 1997, 2016; Smith, 2012). Improvement of these ‘soil functions’ would lead to larger and more stable yields globally, thus safeguarding global food security and adaptation of global agricultural production to climate change (Lal, 2004; Rumpel et al., 2019). Increasing enthusiasm is pushing towards ever more optimistic predictions, with, for example, claims that SOC sequestration would be key to supporting five (FAO, 2017), seven (FAO, 2019) and up to 12 Sustainable Development Goals (Smith et al., 2019). As a result, several recent papers (Amelung et al., 2020; Chabbi et al., 2017; Kopittke et al., 2022; Rumpel et al., 2019) and high-level international initiatives (the '4p1000 initiative’ launched during the COP21 in 2015, the Koronivia workshops held during the COP23 in 2018 and FAO's RECSOIL program) call for rapid upscaling and implementation of SOC sequestration practices to solve concurrently the challenges of climate change and food security.

Yet already more than 15 years ago, Janzen (2006) highlighted that a trade-off exists between storing C in soils for the benefit of climate change mitigation, and using soil C during soil organic matter (SOM) decomposition for the release of nutrients to support plant production. In parallel, both climate change (Smith, 2012) and food security (Giller, 2020) have been described as ‘wicked problems’, highlighting their daunting complexity and the formidable international social, political, economic and scientific interdisciplinary effort that is required to offer even partial solutions. In such a context, claims that a blanket solution can solve climate change and food insecurity concurrently should prompt some degree of scepticism. Like any ongoing scientific discussion, it needs relentless and rigorous examination.

Starting from this perspective, we re-examine the terminology and concepts underlying climate mitigation through SOC sequestration and the empirical evidence supporting the promised win-win with food security. We critically review estimates of the global SOC sequestration potential and quantify the potential contribution of SOC sequestration to reaching net-zero CO₂ emissions in a 1.5°C target climate scenario for the 2100 horizon. While we focus here on the technically achievable SOC sequestration potential, we also acknowledge that many large social, economic and political barriers exist to the implementation of SOC sequestration practices that cast yet more doubt on the global mitigation potential of SOC sequestration (Amundson & Biardeau, 2018; Bradford et al., 2019). Then, we review the existing evidence on the effect of increasing SOC stocks on above-ground plant production. Here, we focus on plant production as a key soil function because of its obvious link to food security, although we acknowledge that other functions are also key.

2 SOC SEQUESTRATION: CONCEPTS AND TERMINOLOGY

Soils are the largest reservoir of C in the terrestrial biosphere, containing 1700 Gt C in the top meter, that is, four times as much C as global vegetation, twice as much as the atmosphere and 160 times as much as the current annual anthropogenic CO₂ emission rate (Friedlingstein et al., 2022). Soils also contain a significant amount of inorganic C (Lal et al., 2021), but here we focus exclusively on SOC, claimed to hold the largest climate change mitigation potential and to benefit numerous soil functions.

Small changes in global SOC stocks (a few tenths of a percent) could lead to proportionally large contributions to global CO₂ emissions (tens of percent; Paustian et al., 2016). This makes soil both a prime candidate to lock up CO₂ from the atmosphere, and a major threat as poor land management practices and environmental changes release CO₂ from soils and enhance global warming. For a given area, net sequestration of SOC is the difference between inputs of C to the soil and the release of C into the atmosphere as CO₂ (or methane) from the decomposition of accumulated organic matter by soil heterotrophic organisms. SOC sequestration can therefore be defined as a state of the soil in which the soil C inputs exceed the soil C release, leading to an increase in SOC stocks. Therefore, it follows that SOC sequestration may be achieved by either increasing inputs to the soil or decreasing losses.

Although this definition is widely accepted, it is worth highlighting some common misconceptions. For example, SOC sequestration in a given soil is commonly implied to contribute to climate change mitigation (Chenu et al., 2019). However, this is only true when sequestration results from a net gain of C from the atmosphere (Powlson et al., 2011). As such, increasing C inputs from photosynthesis is the most effective climate change mitigation strategy. Increasing organic amendment application, which represents a transfer of C from one place to another, leads to a net atmospheric decrease only when the soil stabilisation of the added material is more effective at the application site than at the site from where it is exported (Kirschbaum et al., 2020). Furthermore, the emission of other greenhouse gases such as nitrous oxide (N₂O), can negate or exceed the benefit of sequestering SOC in some conditions (Guenet et al., 2021; Lugato et al., 2018). Agriculture is responsible for approximately 70% of global anthropogenic N₂O emissions (Tian et al., 2020), largely due to the inefficient use of nitrogen fertilisers. We, therefore, argue that a lack of consideration for nitrogen use efficiency is a missed opportunity to strongly reduce the greenhouse gas footprint of agriculture as much as it is a trade-off with SOC sequestration (Powlson et al., 2011). Another misconception arises from the lack of distinction between SOC stocks and changes in SOC stocks. It is critical to stress that while SOM and SOC remain the most commonly measured soil variables (Bünemann et al., 2018), they provide little insight into SOC sequestration and the role of soils in mitigating climate change. First, SOC stocks measured at a given point in time provide no information on a trend—SOC sequestration involves a dynamic change in SOC stocks that can be quantified only by monitoring the stocks over time or by quantifying the balance between soil C uptake and release. Second, ‘measuring SOC’ often entails measuring concentrations (in unit mass of C per unit mass of soil), which do not inform on the quantities of SOC stored in the soil. Concentrations need to be converted to stocks (in unit mass of C per unit of land area to a sufficient depth to avoid the effects of soil compaction). This is usually done by multiplying the concentration by the soil bulk density, which may change over time as a result of changing management practices. More accurate methods exist that account for changing bulk densities, such as the equivalent soil mass basis (Wendt & Hauser, 2013).

3 CLIMATE CHANGE MITIGATION POTENTIAL OF SOC SEQUESTRATION

3.1 Estimates of global soil organic carbon sequestration potential

Several methods can be used to quantify the global potential of soil to sequester atmospheric CO₂.

The first one is based on calculations of historical C losses from cultivated soils (comparing C stocks from ‘virgin’ and cultivated soils on similar soil types and land use) and ‘assuming a recovery of one-half to two-thirds of historical C losses as a reasonable upper limit’ (Paustian et al., 1997). Early estimates of historical C losses varied widely, from 40 Gt C (148 Gt CO₂) to as much as 537 Gt C (1989 Gt CO₂) (Lal, 2001). The most recent estimate (using data-driven statistical modelling) puts global C losses at 116 Gt C (430 Gt CO₂) over the last 12,000 years of human land use (Sanderman et al., 2017), suggesting a global SOC sequestration potential of 285 Gt CO₂ under the two-thirds recovery assumption.

The second and most common method uses ‘bottom-up’ approaches where average SOC sequestration rates are estimated for different management practices over varied experimental duration, from 5 to more than 80 years (Paustian et al., 1997). Then, these estimated rates are extrapolated to regional or global scale by multiplying the rate with the estimated land area available for a given practice. The management practices, referred to as natural climate solutions (NCS) (Griscom et al., 2017), include many agricultural practices such as no-till, cover-cropping, improved nutrient, water, and grazing management, agroforestry, as well as interventions in other land-use sectors such as forestry and wetland management and peat restoration. Bottom-up estimates vary widely in the literature, from 0.4 (IPCC, 2019) up to 9.1 Gt CO₂ year⁻¹ (Lal, 2018). The variety of underlying assumptions used to estimate the availability of land area for specific practices contributes greatly to this variability (Fuss et al., 2018). Moreover, different studies focus on different subsets of practices, from large ranges of NCS (Bossio et al., 2020), to agriculture and forestry (Minasny et al., 2017), agricultural lands only (IPCC WG3, 2022; Smith et al., 2008) or croplands only (Zomer et al., 2017). Whether biochar application should be included in estimates from agricultural management is also debated (Smith, 2016). Most recently, SOC sequestration in croplands, grasslands, agroforestry and biochar was estimated between 1.8 and 4.1 Gt CO₂ year⁻¹ (IPCC WG3, 2022). We also observe that depth is inconsistently reported in review papers. As mentioned above, SOC stocks (and therefore sequestration rates) are usually measured to a constant depth and then converted per unit surface area (implicitly to this constant depth). These reviews, therefore, potentially combine estimates of sequestration rates that are not directly comparable (i.e. sequestration rates estimated to a depth of 2 m cannot be averaged directly with sequestration rates estimated to 30 cm depth; Knotters et al., 2022). This also contributes to the discrepancies between global estimates. Bossio et al. (2020) calculated recently the contribution of soils to the total mitigation potential of 20 NCS across different land-use sectors with stringent limits on the area availabilities to safeguard biodiversity and food security (leaving sufficient area available for food production; Griscom et al., 2017). The authors found a global technical SOC sequestration potential of 3.3 Gt CO₂ year⁻¹, far less than previous estimates which did not constrain the available land area (Fuss et al., 2018), and with greater recognition of the need to use multiple solutions and address multiple objectives concurrently.

As an alternative approach, Janzen et al. (2022) argued that global SOC sequestration is most constrained by photosynthesis, the primary source of soil C inputs. The authors estimated global photosynthetic C inputs and, from this, estimated a global SOC sequestration potential of around 0.52 Gt CO₂ year⁻¹.

Interpreting global SOC sequestration estimates in terms of their climate change mitigation potential is problematic because of a number of well-known limitations as briefly mentioned in the previous section (Baveye et al., 2018; Powlson et al., 2011; Rumpel et al., 2019). In the following sections, we emphasise two aspects—saturation and non-permanence—that we deem particularly underrated and critical to the interpretation of SOC sequestration potentials.

3.2 Saturation

SOC sequestration saturates over time (Six et al., 2002; Smith, 2012). This ‘sink saturation’ is due to either a finite availability of mineral surfaces and other physico-chemical properties responsible for the persistence of C in soils, defining a maximum storage capacity, or to environmental constraints on C input stabilisation and decomposition rates leading to a new equilibrium value for SOC and defining an ‘effective’ storage capacity (Stewart et al., 2007). Therefore, converting estimated SOC sequestration rates into total SOC sequestration potential requires the time taken to reach a new equilibrium to be defined. Twenty years is often mentioned as a standard (Fuss et al., 2018; Smith et al., 2016), but other values are used, ranging from 5 to 50 years (Lal et al., 2018) and up to 85 years (Mayer et al., 2018). Furthermore, both the maximum and effective stabilisation capacities of soils depend on pedoclimatic conditions (Angers et al., 2011; McNally et al., 2017), suggesting that both the timespan of SOC accumulation and the rate of decrease in SOC sequestration are highly context dependent. A notable exception is peat and other organic soils such as mollisols, which are not bound to a maximum storage capacity, but require very specific conditions over large time spans to accumulate SOC. In mineral soils, texture is particularly influential, with sandy soils generally showing lower C storage capacities than heavier soils (Angers et al., 2011; McNally et al., 2017).

Most critically, the concept of sink saturation suggests not only a limited storage capacity but also that the rate of SOC sequestration decreases as SOC approaches the maximum or effective storage capacity (Poulton et al., 2018; Stewart et al., 2007). Saturation, therefore, implies not only that SOC sequestration is time-limited but also that sequestration rates decrease exponentially as soon as SOC stocks start to increase (Baveye et al., 2018). Most studies, however, with one notable exception (namely, Sommer & Bossio, 2014), assume a constant sequestration rate over the defined time period required to reach a new equilibrium (Bossio et al., 2020; Fuss et al., 2018). Some assessments provide only an annual rate in a specific year and compare this to other negative emission or emission reduction options. By doing so, the IPCC WG3 (2022) ranks SOC sequestration as the fourth most effective option in 2030, just behind the solar and wind energy transition. This is misleading and an unfair comparison because the potential of SOC sequestration will decline and reach zero over time, while it will be continued through time for most other options, including transitioning to non-CO₂ emitting energy production.

Here, we use the estimate of Bossio et al. (2020) as a starting value for SOC sequestration rates and illustrate the large overestimations in C sequestration potentials arising when C saturation is ignored. We compare the effect of three assumptions on the evolution of observed sequestration rates over time and illustrate their consequence for global SOC sequestration potential estimates: (1) a constant sequestration rate over time, (2) a rapid exponential decrease in sequestration rates with a 10-year exponential decay half-life; previously observed from converting croplands to grassland, which is arguably the fastest way to sequester SOC through land-use change (Baveye et al., 2018), and (3) a slow exponential decrease in sequestration rates: with a 30-year half-life (Figure 1a). The impact of these assumptions on global SOC sequestration potential is striking (Figure 1b): for the 2100 horizon, they are 257, 121 and 49 Gt CO₂ for constant, slowly decreasing and rapidly decreasing rates, respectively. This suggests that SOC sequestration potentials estimates that disregard saturation are overestimated by 53%–81%.

FIGURE 1

Visualisation of the impact of three different assumptions regarding C saturation on global soil organic carbon (SOC) sequestration and its potential to mitigate climate change to the horizon 2100. The three assumptions are as follows: An unrealistic constant sequestration rates over time, that is, no saturation limit (green dotted lines and frame); slow exponential decrease in sequestration rates (30 years exponential decay half-life; blue lines and frame); and fast exponential decrease in sequestration rates (10 years half-life; orange lines and frame). For the three assumptions, the temporal dynamics of global sequestration rates (a) and global SOC sequestration potential (b) are depicted for the example of the estimate from Bossio et al. (2020). The contribution of SOC sequestration to a pathway to 1.5°C target is represented for the whole range of literature estimates of sequestration rates in panels (c), (d) and (e), each panel corresponding to one assumption on saturation as indicated by the colours.

Finally, it is important to note that soils are not static: soil is formed at the soil-bedrock interface (saprolite) at the bottom of the soil column (mostly on centennial to millennial timescales), and soil is transported by erosion, exposing subsoil layers at the eroded site and burying topsoil layers where the eroded soil is deposited. Full accounting of these processes would be required to understand the carbon balance at larger temporal and spatial scale (Kirkels et al., 2014). Reviewing knowledge on the creation and redistribution of soils on SOC storage is beyond the scope of this article, but it is worth mentioning as it exposes yet another layer of complexity in the prediction of SOC sequestration.

3.3 Non-permanence

SOC sequestration is also reversible. This means that improved management practices must be maintained for SOC stocks to remain at their maximum potential after the new equilibrium has been reached (Smith, 2012). Although some authors argue that this is realistic (Rumpel et al., 2019), SOC remains vulnerable to losses and is sensitive to a range of environmental factors. This vulnerability remains poorly understood. Particularly, the reappraised role of aggregation, organo-mineral interactions and microbial eco-physiology in regulating soil C dynamics (Lehmann & Kleber, 2015; Liang et al., 2017) questions knowledge of SOC sensitivity to factors such as temperature (Bradford et al., 2016; Conant et al., 2011; Moinet et al., 2018, 2020, 2021) and water availability (Moinet et al., 2016). The large variability in the responses of SOC stocks to soil warming (van Gestel et al., 2018) suggests a context-dependent response and a highly uncertain future for newly sequestered SOC, even when best management practices are maintained. Arguably, the residence time of sequestered SOC is less than that of the C stored in fossil fuel reservoirs (millennia). This implies that, in the long-term, avoiding CO₂ emissions is much more effective at limiting global warming than removing atmospheric CO₂, a fact referred to as ‘asymmetry’ in IPCC terminology (IPCC WG1, 2021).

Grasping the potential impact of SOC sequestration on climate change necessitates a comparison with the reductions in CO₂ concentrations that are required. This has been done, for example, by comparing estimated SOC sequestration potentials to CO₂ emitted from anthropogenic activities (Minasny et al., 2017; Sommer & Bossio, 2014). Following this approach, the most optimistic global potential estimate of ca. 454 Gt CO₂ (Lal, 2018) represents a little under 12 years of global CO₂ emissions at the current rate of 39 Gt CO₂ year⁻¹, and 3.5 years of global CO₂ emissions under our assumption of a fast decreasing SOC sequestration rate. The more realistic estimate from Bossio et al. (2020) represents 1.2–4.2 years of global emissions. Such comparisons have been used to illustrate that SOC sequestration would allow ‘buying time’ while low-carbon technologies are being developed (Minasny et al., 2017). We contend that this is misleading because of the reversibility of SOC sequestration and the large uncertainties surrounding its timespan. SOC sequestration cannot substitute emission reductions, even temporarily, nor does it delay the need for immediate and aggressive cuts in greenhouse gas emissions if we are to keep global warming under 1.5 or well-below 2°C (Anderson et al., 2019).

As an alternative representation of the contribution of SOC sequestration to reaching the goals of the Paris agreement, we made some simple calculations. Current policies are expected to stabilise emission levels at the current rate (39 Gt CO₂ year⁻¹), while scenarios focusing on current trends typically show a modest increase (IPCC WG3, 2022). Under these circumstances, cumulative emissions would reach close to 3500 Gt CO₂ (range of 3000–4000 Gt CO₂) by the end of the century. The total carbon budget to meet the 1.5°C target is estimated at 500 Gt CO₂ from early 2020 to 2100 (IPCC WG1, 2021). This implies a total net reduction in emissions of around 3000 Gt CO₂. Avoiding positive emissions (e.g. by decarbonisation and demand reduction) will be the main pathway. CDR can contribute to meeting the target by offsetting positive emissions before net-zero is reached (e.g. from hard-to-abate sectors) or by contributing to net negative emissions. In this context, the possible contribution of SOC sequestration alone, under the unrealistic assumption of an unlimited sequestration potential, would range from 1.0% (IPCC, 2019) to 23.7% (Lal, 2018) of the efforts towards a 1.5°C target for the horizon 2100 (Figure 1c; taking 0.4 and 9.1 Gt CO₂ year⁻¹ as initial rates). Contributions decrease to 11.1% and 4.5% for the most optimistic SOC sequestration rate estimate (Lal et al., 2018) under the fast and slow assumptions, respectively; and to 4.0% and 1.6% for the more realistic initial SOC sequestration rate (3.3 Gt CO₂ year⁻¹ by Bossio et al., 2020; Figure 1d,e). The most recent estimated initial SOC sequestration rate (0.52 Gt CO₂ year⁻¹ by Janzen et al., 2022) would suggest an insignificant contribution of between 0% and 1%.

4 CONTRIBUTION OF SOC SEQUESTRATION TO SOIL FERTILITY AND CROP YIELDS

Crops do not take up C from the soil. Therefore, the mechanisms through which SOC affects crop biomass and yield are always indirect. Such mechanisms include an increased supply of nutrients from mineralisation of SOM, enhanced retention and buffering of nutrient stocks through increased cation exchange capacity, improved water infiltration and better soil structure (and thus increased water and nutrient availability), or through shifts in the composition and structure of soil organism communities which may positively impact nutrient cycling and pest and disease control (Johnston et al., 2009; Watts & Dexter, 1997).

While seemingly sound, the generality of this narrative has been contested, both theoretically (Janzen, 2006) and empirically (Terrer et al., 2021), as mentioned in the introduction. So, what is the empirical evidence for a causal link between SOC sequestration and improved crop biomass production and yields? In recent years, a large number of meta-analyses assessed the benefits of either SOC sequestration or management practices that lead to SOC sequestration, on crop yields. To shed more light on the premise that SOC sequestration for climate change mitigation will also increase plant production, we reviewed 21 of those meta-analyses, the methods they employed, the outcomes, the validity and the limitations of the studies (Table 1). This exercise led us to three main conclusions, successively elaborated in the following sections: (i) establishing causality between SOC sequestration and crop yield is problematic, (ii) the response variables used vary among the studies and all have limitations, and (iii) the outcomes vary with space, time and methods used. Altogether, the meta-analyses provide no convincing empirical support that there is a general positive correlation between SOC sequestration and improvements in crop yield.

TABLE 1. Overview of direction and size of mean effects of SOC sequestration on crop yields, depending on the assessment method used, across 21 meta-analyses

Correction of NPK supply?	Meta-analysis	Location	# data points	Management practice	Response variable	Significant mean effect
Observed yield effects—corrected for N, P and K contributions	Oelofse et al. (2015), figure 5a,b	Denmark	869	Spatial comparison	N recovery (wheat)	−0.17 (kg N/kg N)/% soil C
					N recovery (spring barley)	−0.16 (kg N/kg N)/% soil C
					Attainable yield	0 (t/ha)/% soil C
	Schjønning et al. (2018), figure 7	Denmark	975	Spatial comparison	Agronomic N use efficiency	−3.95 (kg grain yield/kg N) per 0.01 kg soil C/kg minerals
					Attainable yield	−630 kg grain yield/ha per 0.01 kg soil C/kg minerals
	Hijbeek et al. (2017)	Europe	107 (20)	Organic amendments	Δ attainable yield	0%
	Ravensbergen et al. (2021), figure 6a	Nigeria	676	Spatial comparison	N recovery	0 (kg N/kg N)/g/kg soil C
	Cooper et al. (2016)	N-America & Europe	1079 (26)	Reducing tillage	Δ yield (vs. deep inversion tillage)	−7.6%
					Δ yield (vs. shallow inversion tillage)	0%
Uncertain in some cases	Zavattaro et al. (2017)	Europe	310 (80)	Organic amendments	Δ yield	0%
	Chen et al. (2018), figure 2b, figure S1b	Global	329 (132)	Organic amendments	Δ yield (average over period)	0%
					Δ yield (last experimental year)	5.3%
	Dawe et al. (2003)	Asia	75 (25)	Organic amendments	Δ yield	0%
					Δ yield trends	0%
Observed yield effects -possibly including N, P or K contributions	Wei et al. (2016)	China	38 (32)	Organic amendments	Δ yield	7.6%
					Δ yield trends	26.9%
	Vanlauwe et al. (2011), figure 6a	SSA	721 (90)	Residues I, III, IV	Δ agronomic N use efficiency	−36%
				Residues II	Δ agronomic N use efficiency	0%
				FYM or compost	Δ agronomic N use efficiency	28%
	Jeffery et al. (2017)	Global	1125 (109)	Use of biochar	Δ yield (temperate)	0%
					Δ yield (tropics)	25%
	Luo et al. (2018)	Global	226 (106)	Organic amendments	Δ yield	27%
	Han et al. (2018)	China	75 (70)	Straw	Δ yield	13.4%
				Spatial comparison	Yield	267–414 kg grain yield/g soil C/kg soil
	Pan et al. (2009)	China	National statistics	Time series	Yield	0.43 t grain yield/% soil C
	Soussana et al. (2019), figure 1	Africa, Asia & L-A	151 (32)	Diverse	Δ yield	3.25% grain yield/% change in % soil C
	Alvarez et al. (2017), figure 5	Argentine Pampas	67 (62)	Legume cover crop	Δ yield (maize)	6.70%
				Non-legume cover crop	Δ yield (maize)	0%
				Cover crop	Δ yield (soybean)	−2.50%
	Kuyah et al. (2019)	SSA	397 (61)	Agroforestry (divers)	Δ yield	100%
	Li et al. (2021)	Upland China	622 (160)	Organic amendments	Δ yield (org only)	0%
					Δ yield (org + fert)	8.3%–8.7%
	Du et al. (2020)	China	774 (141)	Manure	Δ yield	7.60%
	Oldfield et al. (2019), figure 1	Global	834 (90)	Spatial comparison	Yield	equation Table 1 (Oldfield et al., 2019)
	Islam et al. (2022), figure 2	China	1071 (177)	Straw	Δ yield (monocropping)	11.80%
					Δ yield (double cropping)	5.50%

• Note: Red indicates a mean significant negative yield effect, orange indicates no significant yield effect and blue indicates a mean significant positive yield effect. Data points indicate the number of yield observations. If yield observations differ from the amount of field experimental locations, the latter is placed within brackets. An earlier (more limited) review of meta-analyses was published as a conference proceeding by Hijbeek et al. (2018).

4.1 Establishing causality is problematic

The most straightforward method to assess the effect of SOC on crop yields is a space for time substitution (as done by Han et al., 2018; Oelofse et al., 2015; Oldfield et al., 2019; Schjønning et al., 2018). Such studies compare crop production at several locations, some with larger SOC content than others, and assess differences in yields. Implicitly, a similar yield difference is assumed to occur when the SOC would have changed over time in one location. Interpretation is problematic, however, as soil texture and climate are strong confounding factors. Soil texture and climate affect SOC stocks (Burke et al., 1989; Feller & Beare, 1997; Hoyle et al., 2016; Miller et al., 2004) but also affect crop yields directly (van Ittersum & Rabbinge, 1997). Furthermore, if SOC can affect crop yield through the mechanisms mentioned earlier, the reverse may also be true: larger crop yields may lead to increased biomass inputs to the soil, through crop residues, roots and stubble, and may therefore positively affect the soil C balance. The direction of the causality in a correlation observed between SOC stocks and crop yields is therefore contested, both in spatial and time series observations (Figure 2).

FIGURE 2

Illustration of the complexity of unravelling causal relations between an increase in soil organic carbon (SOC) stocks and crop yields. Agricultural practices that increase SOC stocks may also enhance crop yields, suggesting that the evolution of SOC and yield in time may correlate with no causal link between them. Moreover, even if a causal link exists, determining which SOC or yield is the cause, and which is the effect, is problematic.

Correlating crop yields to a change in SOC stocks (or to a management practice that enhances SOC stocks) over time in the same location (as done by Alvarez et al., 2017; Chen et al., 2018; Cooper et al., 2016; Dawe et al., 2003; Du et al., 2020; Han et al., 2018; Islam et al., 2022; Jeffery et al., 2017; Kuyah et al., 2019; Li et al., 2021; Luo et al., 2018; Soussana et al., 2019; Wei et al., 2016; Zavattaro et al., 2017) is not subject to such confounding environmental factors but has other limitations. To increase SOC stocks (relative to a control treatment) in an experimental field, a change in management practices is required. Yet management changes not only affect SOC but may also influence crop yields directly (Figure 2). For example, cultivation of a cover crop can increase SOC stocks and, at the same time, can reduce N losses, thereby increasing soil N supply and possibly crop yields (Thapa et al., 2018). Another example is that of biochar, for which beneficial effects on crop yields are mainly observed in tropical acid soils (Jeffery et al., 2017), most likely due to the amelioration of aluminium toxicity (Shetty et al., 2021).

4.2 Limitations of response variables used

An additional complication that arises when assessing the contribution of SOC sequestration to crop yields is the composition of SOM, which changes over time. Often, adding organic amendments to soil is assumed to be representative of SOC sequestration (Dawe et al., 2003; Islam et al., 2022; Wei et al., 2016) as additional C inputs favour a positive soil C balance. However, fresh organic amendments are not equivalent to a long-term change in SOC stocks. Fresh organic amendments contain large amounts of nutrients (such as N, phosphorus [P] and potassium [K]), which are more readily available for crops than the nutrients contained in older, stabilised SOM. Fresh organic amendments may therefore have larger benefits for crop yields than older and more stable SOM. When analysing field experimental data, such NPK effects can be excluded by assessing yield response curves to N application in the presence of sufficient P and K supply and calculating the difference in attainable crop yields (as done by Hijbeek et al., 2017; Oelofse et al., 2015; Schjønning et al., 2018). In such cases, attainable crop yields are the plateau of the yield response curve. An increase in the plateau suggests improvements in other soil functions, beyond nutrient supply, that stimulate crop yields.

Alternatively, one can assess the influence of an increase in SOC on agronomic N use efficiency or N recovery (as done by Oelofse et al., 2015; Ravensbergen et al., 2021; Schjønning et al., 2018; Vanlauwe et al., 2011). These indicators assess the additional yield or crop N uptake per additional kg N applied. An increase would confirm the hypothesis that larger SOC stocks lead to improved soil structure, water holding capacity, or soil quality in general (although we acknowledge that the term is controversial; Bünemann et al., 2018), leading to more plant production per unit N applied. When experimental treatments receive sufficient P and K application, a change in agronomic N use efficiency is also representative of the long-term effect of SOC on crop yield.

Among the meta-analyses reviewed, some studies tried to correct for NPK effects on crop yields by assessing the ratio of N, P and/or K between treatments or aiming to standardise nutrient supply (Chen et al., 2018; Dawe et al., 2003; Zavattaro et al., 2017). At the same time, the authors acknowledged that this approach is ambiguous as a control treatment with equal NPK supply from mineral fertiliser is required, while relative nutrient mineralisation rates of organic amendments vary strongly with the quality of the amendment and across locations (Schröder, 2005). As a result, it is unclear whether findings from those studies represent the effect of old or fresh organic matter. In Table 1, such studies are classified as ‘uncertain in some cases’.

4.3 Conflicting evidence

Findings of the recent meta-analyses vary from negative to neutral to positive yield effects of increasing SOC (or organic amendments; Table 1). It is striking that neutral or negative effects are found much more often when studies correct for N, P and K supply (Table 1, top section). By contrast, a majority of the studies in which N, P and K supply was included found a positive effect of SOC on crop yields (Table 1, bottom section). Based on the previous section, we argue that these latter yield effects are not representative of expected yield increases when sequestering carbon over long periods of time in more persistent forms, as required to mitigate climate change.

In short, increasing SOC stocks does not always increase crop yields, and can also negatively affect them. Terrer et al. (2021) showed a trade-off between SOC sequestration and plant biomass production in a meta-analysis of 108 elevated CO₂ experiments across forests and grassland ecosystems, most likely due to competition for N between soil and plants. This was affirmed by another recent study (van der Pol et al., 2022), who found that converting a fallow-grain crop rotation to a continuous grain crop rotation increased grain yield and SOC stocks only when a legume crop was also included in the rotation (enhancing inputs from biological N₂ fixation). Competition for nutrients (particularly for N) is therefore a strong candidate explanation for observed negative yield effects.

Another critical point to stress is the context specificity of the relationship between SOC and yield. From the eight meta-analyses which gave insights into the longer-term effect of SOC sequestration on soil fertility and crop yields (i.e. correcting for NPK supply), four studies investigated the role of soil texture (Chen et al., 2018; Hijbeek et al., 2017; Oelofse et al., 2015; Zavattaro et al., 2017) and one investigated the role of climate (Chen et al., 2018). Considering soil type, three studies consistently found greater benefits on sandy soils. Only one study (Oelofse et al., 2015) found unclear effects across eight soil types, with borderline positive effects on coarse sandy loam and fine sandy clay loam soils but negative effects on fine sandy soils. A clearer picture was obtained by Zavattaro et al. (2017) who found significantly more benefits on light than on medium-textured soils (response ratio of 1.051 vs. 0.866). Similarly, whereas Hijbeek et al. (2017) found no overall yield effects of organic amendments, attainable crop yields were increased on soils with smaller clay content (an additional 0.13% increase in attainable yields when using organic amendments for 1% less clay particles). Finally, when assessing the yield change in the final years of experiments, Chen et al. (2018) found a positive yield difference for sandy soils, but not for any of the other soils investigated (clay, loam or silty soils) when standardising NPK supply. Using the same data selection, the authors also found a positive yield effect in tropical climates, with no significant yield differences in the final experimental year in the other climate zones (subtropical and continental).

Beyond pedo-climatic conditions, the initial SOC stocks are also critical. This was very clearly demonstrated in the meta-analysis of Oldfield et al. (2019), who showed that the yield effect of SOC saturates, with no further increase beyond a SOC content of approximately 2%.

5 SOIL-SMART AGRICULTURE

Two key points emerge from our review: (i) the climate mitigation potential of SOC sequestration is modest, at best, on a global scale, and context-specific, and (ii) correlations between SOC sequestration and crop yield are strongly context dependent and it is highly uncertain whether a causal link exists. Therefore, SOC sequestration is most certainly not a win-win option under all conditions.

Nevertheless, reaping the win-win between enhancing SOC and crop yields where and when possible is laudable. From this perspective, agendas that promote region-specific (or even more localised) assessments, as proposed recently (Amelung et al., 2020), seem appropriate. Yet even when and where SOC sequestration can go hand-in-hand with improved soil fertility and crop yields, strategies to promote one may not be the best suited to promote the other. The influence of soil type is particularly illustrative of this conflict (Rusinamhodzi et al., 2013). As discussed, clay soils have the largest potential to store SOC, and those that are depleted offer the greatest opportunity to increase SOC stocks (Angers et al., 2011; Kirschbaum et al., 2020; McNally et al., 2017), but it is in sandy soils that the effect of increasing SOC on crop yield may be the largest (Chen et al., 2018; Hijbeek et al., 2017; Zavattaro et al., 2017). In a context where organic resources are expensive and in high demand (Giller et al., 2008), a strategy targeted at maximising overall SOC storage for climate change mitigation would dictate that organic matter should preferably be diverted from sandy soils to clay soils (Kirschbaum et al., 2020), at the expense of the fertility of the sandy soils. By contrast, a primary focus on soil fertility and yield improvement would dictate organic matter to be diverted from clay soils to sandy soils, at the expense of overall SOC storage (Figure 3). Importantly, in the former case, an overall yield decrease would mean that more land is needed overall to maintain food production constant, leading to more deforestation or less reforestation and therefore SOC and above-ground C losses elsewhere.

FIGURE 3

Conceptual figure illustrating potential conflicts between soil organic carbon (SOC) sequestration and food production. The figure depicts two hypothetical cases in which crop residues are removed from one field after harvest to be applied as OM inputs to another crop field. In panel a, residues are transferred from a sandy soil to a clay soil. After some years, a new equilibrium for SOC stock is reached. The clay soil gains more SOC than the sandy soil loses, due to its higher C stabilisation capacity (Kirschbaum et al., 2020). Therefore the net overall effect is that C is sequestered, to the benefit of climate (provided that no additional N₂O or CH₄ emissions would arise). The clay soil also sees crop yield increasing, but not as much as the yield in the sandy soil decreases, due to the stronger yield effect of organic amendments in sandy than clay soils (Hijbeek et al., 2017; Zomer et al., 2017). The net effect for yield is that less crops are produced overall. The reciprocal transfer, in panel b, leads to mirrored effects: Small yield loss in the clay soil and high yield gain in the sandy soil, and large CO₂ emissions in clay soil and small SOC sequestration in the sandy soil with an overall SOC loss and aggravated climate change, but more food produced overall. Importantly, assuming that each field is owned by a different farmer, someone always loses. This clearly illustrates that local win-win scenarios can occur at the expense of fertility elsewhere.

Ensuring global food security in a stabilised climate is a tremendously complex endeavour. Our findings build on a wealth of knowledge showing the importance of local knowledge, of developing locally suited adaptative methods focusing on a wide set of environmental outcomes, and calling attention to social acceptability and economic viability (Giller et al., 2015; Pretty, 2018; Pretty et al., 2011). For example, while it is doubtful that increasing food production in Europe or North America would yield any benefit for global food security (Loos et al., 2014), the link between food production and food security is absolutely critical for smallholder African farmers (Giller et al., 2021), with an estimated 7% reduction in poverty for every 10% increase in yield (Pretty et al., 2011). The hard reality is that difficult choices must be made, and these choices have to be made locally.

Without a doubt, soils have a critical role to play, and SOM, with the carbon it contains, is fundamental to maintaining soil fertility and supporting plant production. Furthermore, SOM is central to the numerous functions delivered by soils (Kopittke et al., 2022). This, we do not question. However, our review demonstrates that the importance of SOC does not imply that increasing SOC stocks necessarily leads to an increase in yield. More generally, relationships between the multiple functions of soil are context-specific and numerous trade-offs exist (Schulte et al., 2014; Vazquez et al., 2021; Zwetsloot et al., 2021). It is increasingly recognised that greater scientific knowledge and concurrent management of these soil functions will be critical (Kopittke et al., 2022; Schulte et al., 2014). We conclude that, if we are to ensure food security in a changing climate, let alone other sustainable development goals, SOC sequestration is simply not up to the challenge. SOC sequestration is indeed one of many approaches, one small piece of a very large puzzle, and it should be treated as such. We argue that including SOC sequestration in the narrative of global climate change mitigation is ill-suited and we call for a soil-smart approach, which will not always be ‘climate-smart’.

ACKNOWLEDGMENTS

We thank Marie Zwetsloot, Jan-Willem van Groenigen, Jantiene Baartman and Rachel Creamer for internal reviews of draft manuscripts and suggestions for improvement. We thank Hein ten Berge, Andy Whitmore and Martin van Ittersum for the exchange of ideas on how to measure the effect of SOC sequestration on crop yields in earlier years. We also thank Thom Kuyper for creating the course ‘the carbon dilemma—a soil perspective’ at Wageningen University, and the MSc students that took part in this course for the lively class discussions which gave inspiration for the writing process.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Open Research

DATA AVAILABILITY STATEMENT

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.