Introduction

Organic-rich soils, such as peatlands, have developed over centuries to millennia under water-logged conditions, where aerobic microbial mineralisation of decaying plant material is inhibited by low oxygen (O2) diffusion1,2,3. The OC content of peat soils can exceed 45% (wt/wt) and represents a terrestrial carbon (C) pool of approximately 600 Gt C, compared to 875 Gt of CO2-C in the contemporary atmosphere4,5. Therefore, the dynamics of CO2 uptake and release from organic-rich soils are critical to past and future atmospheric CO2 levels and climate change trajectories6,7,8.

Millions of hectares of organic-rich soils are managed for agricultural purposes, mainly as productive cropland or grasslands4,9. This involves artificial drainage and other practices (liming, fertilisation, tillage) that alter the physical and biochemical properties of the soils10,11. In particular, drainage increases the supply of O2 to the soil microorganisms, leading to high CO2 emissions from the mineralisation of the accumulated organic matter11,12,. Globally, drained peat soils alone emit 2 Gt CO2 equivalents per year4, resulting in large proportions of soils being in transition from true peat to organo-mineral soils with lower OC content13,14,15,16.

As part of large-scale efforts to mitigate climate change, rewetting of drained organic soils is an emerging strategy to reduce CO2 emissions from the agricultural sector17,18. Rewetting involves restoring natural hydrological regimes by disconnecting drainage pipes and ditches, thereby creating wet anaerobic soils that minimise aerobic mineralisation and CO2 emissions3,17. Comprehensive data syntheses show that rewetting contributes to net climate benefits17,19 although the anaerobic conditions can increase methane (CH4) emissions due to acetoclastic and hydrogenotrophic CH4 production combined with restricted microbial CH4 oxidation (methanotrophy) near the soil surface20,21.

National reporting of CO2 emissions from drained and rewetted organic soils is made to the UNFCCC as part of the land use, land-use change and forestry (LULUCF) sector, following the guidelines of the Intergovernmental Panel on Climate Change (IPCC)22,23. For Denmark, the national inventory report for 2022 includes direct emissions of 1.9 and 2.6 Mt CO2 from drained soils with 6–12% and >12% OC, respectively, under the land use categories of cropland and grassland24. This is calculated from national emission factors (i.e., area-scaled net CO2 emissions) for soils with >12% OC12,24, and assuming that the emission factors for soils with 6–12% OC are half of those for soils with >12% OC24. It is highly uncertain, however, whether and how soil OC content controls area-scaled CO2 emissions from drained organic soils16,25,26,27,28,29. Indeed, there is a global lack of data documenting the CO2 emissions from soils with 6–12% OC, and Denmark is the only country to have tentatively assigned an emission factor to these soils24. This is partly related to the inconsistent and complex definitions of OC-rich soils, which vary between countries and organisations15,30. The Danish classification defines soils with >6% OC as organic soils31, while other countries and the IPCC use thresholds of >12% OC or even higher15,16,23,27,30. According to IPCC, soils with 6–12% OC can be excluded from reporting of greenhouse gas (GHG) emissions, and their GHG balance can be estimated like that of mineral soils27,28. However, evidence from a few exisisting field studies suggests that soils with around 10% OC16,28 and as low as 5% OC13 can emit CO2 at similar rates to soils with >12% OC. This highlights the need for benchmarking of CO2 emission factors for soils with different OC contents, including the 6–12% OC range, to avoid systematic underestimation of national GHG emissions reported under LULUCF.

To address this knowledge gap, we compared the area-scaled CO2 emissions (mg CO2-C h−1 m−2) from organic soils with a wide range of OC contents (6.2–52.1%, n = 103), which was facilitated by a comprehensive study under controlled aerobic conditions with undisturbed soil cores sampled across Denmark (Supplementary Fig. 1).

Results and discussion

Area-scaled CO2 emissions

Area-scaled CO2 emissions at standardised water potentials (pF 2) were not significantly different for classes with different OC content, defined by intervals of 6% OC (Fig. 1a and Supplementary Table 1; medians, P = 0.14, χ2 = 8.36, df = 5; means, P = 0.17, F = 1.60, df = 5). This was also concluded when comparing emissions from soils with 6–12% OC and >12% OC (Fig. 1b; medians, P = 0.21, χ2 = 1.55, df = 1; means, P = 0.31, F = 1.02, df = 1). These emission data complement German field studies on drained organic soils, where (i) similar CO2 emissions were reported for soils with 9.2–10.9% and 15.6–17.2% OC under grassland and arable management16, (ii) grasslands with OC contents of around 10% were found to emit as much CO2 as soils with >30% OC28 and (iii) even drained grasslands with 5% OC could emit as much CO2 as true peat soils13. The larger data set presented here confirms that OC content in the soil (wt/wt) is a poor predictor of area-specific CO2 emissions, and therefore, it is questionable to assign an emission factor to 6–12% OC soils that is half of the emission factor for soils with >12% OC24. In fact, the mean CO2 emission from 6-12% OC soils was only 8.9% lower than that from >12% OC soils (Fig. 1b), and the 95% confidence interval for the group means overlapped strongly (9.4–14.1 and 11.8–14.3 mg CO2-C h−1 m−2 for the 6–12% and >12% OC groups, respectively), making it uncertain to assign different emission factors. On the other hand, the present data also suggest that disaggregation of emission factors for soils with even higher OC contents may not be justified (Fig. 1a).

Fig. 1: Area-scaled CO2 emission from soils in classes of different organic carbon content.
figure 1

Area-scaled emission rates of CO2 from soil cores in classes of increasing organic carbon (OC) content (wt/wt). CO2 emissions were measured at standardised water potential (pF 2) and were not significantly different (P > 0.05) between individual OC classes (a) or between soils with 6-12% OC and >12% OC (b). Boxes show the interquartile range (IQR) with mean (diamond) and median (line). Whiskers extend to the lowest and highest values within the 1.5× IQR range; outliers are shown as dots.

Linear regression analyses (Fig. 2a) confirmed the weak and non-significant association between OC content and area-scaled CO2 emission rates (r2 = 0.008, F (1, 101) = 0.779, P = 0.38). However, the role of OC as an explanatory variable for area-scaled CO2 emissions could be stronger for soil OC density in the soil (wt/vol) than for OC content, as OC density integrates the amount of OC available for microbial mineralisation, i.e., taking into account the soil bulk density15. Nevertheless, OC density (Fig. 2b) also had a weak and non-significant effect on area-scaled CO2 emissions (r2 = 0.014, F (1, 101) = 1.469, P = 0.23). This is consistent with field studies, where OC stock was a non-significant predictor of CO2 emissions, even when specifically considering the amount of OC that was exposed to aerobic mineralisation, referred to as effective16,28 or aerated13 OC stock, i.e., depending on the drainage depth.

Fig. 2: CO2 emission from soils at individual organic carbon (OC) content and OC density.
figure 2

Area-scaled (a, b), OC-specific (c, d), and soil weight-specific (e, f) emission rates of CO2 from individual soil cores (n = 103). Emissions are plotted against soil OC content (a, c, e) and soil OC density (b, d, f) as explanatory variables. Coefficients of determination (r2) and significance (P) of linear regressions are shown in each panel.

Soil weight-specific and OC-specific emissions

Incubation studies with organic soils have previously investigated CO2 emission rates as a function of OC content25,26,27. These studies were carried out with physically mixed or sieved samples and consequently reported soil weight-specific and/or OC-specific CO2 emission rates (i.e., rates scaled by soil weight or OC quantity rather than soil area). We also calculated such rates from the present data, and in agreement with previous studies, soil weight-specific CO2 emission rates increased systematically with increasing OC content (Fig. 2e and Supplementary Fig. 2a)26,27. Indeed, soils with 6–12% OC had median and mean weight-specific emissions that were almost half of those (47–51%) for soils with >12% OC (Supplementary Fig. 2b). The contrasting relationship between OC content and area-scaled or soil weight-specific CO2 emissions is mainly explained by the fundamental negative relationship between OC content and soil bulk density15,32 (Supplementary Fig. 3), which means soils with lower OC content have higher mass. Thus, for the same CO2 emissions per unit area, soil cores or sites of similar (but arbitrary) depth will have lower soil weight-specific emissions for soils with, e.g., 6–12% OC than for soils with >12% OC. Therefore, in terms of practical CO2 emission factors, comparisons of weight-specific CO2 emissions are misleading because the role of bulk density is ignored. Nevertheless, soil weight-specific emissions may be relevant for analyses in other studies aiming at a detailed understanding of the soil physical and biochemical drivers of CO2 emissions.

Carbon dioxide emissions scaled by the amount of OC (OC-specific CO2 emissions) were negatively correlated with OC content and OC density (Fig. 2c, d), meaning that the OC in more degraded soils (i.e., with lower OC content) appeared to be more susceptible to mineralisation. Such negative relationships between measures of OC and OC-specific CO2 emissions corroborate results from studies on disturbed soils. Incubation studies with sieved, air-dried samples (7.6–52.6% OC) indicated that the susceptibility of OC to mineralisation increased with decreasing OC content in German organic soils26 and similar results were reported for Finnish soils (3.1–49.4% OC) incubated under anoxic and physically disturbed conditions27. The present results, obtained with structurally intact organic soils, support the possibility of positive feedback whereby organic soils in advanced stages of decomposition (i.e., with low OC content) may show an increased mineralisation rate26,33,34. However, other studies have emphasised the importance of OC quality and intrinsic decomposability for rates of CO2 emission with the assumption that lower rates would be maintained by soils in advanced stages of decomposition, where the more easily decomposed fractions would already be depleted35,36,37. Yet, recalcitrant compounds, such as aromatics and phenolics, were not found to be enriched in drained organic topsoils under agriculture in Germany33, and simple chemical characteristics of OC as proxies for biochemical decomposability were unable to explain the variability in CO2 emissions in a comprehensive study of drained organic soils in Switzerland25. In summary, the role of OC quality needs to be further clarified, but the present data do suggest that the remaining OC content as such is not a strong predictor of area-scaled CO2 emission rates (Figs. 1 and 2a).

We speculate that rate-limiting properties for CO2 emissions may differ among our 103 organic soils, including a combination of inorganic geochemical properties38, but also properties related to the microbiome and soil physical conditions at the study sites.

Role of soil water content

While OC measures were poor predictors of area-scaled CO2 emissions, soil volumetric water content (VWC) was the most significant single predictor (r2 = 0.34, F (1, 101) = 51.41, P < 0.001) across in situ water contents (Supplementary Fig. 4) and even when standardised to pF 2, i.e., across a more narrow range of VWC (Supplementary Table 2 and 3). Single chemical attributes are generally poor or insufficient proxies for CO2 emissions25,38, but several studies have identified water table depth as an overarching driver of CO2 emissions under field conditions, where higher water tables reduce CO2 emissions by limiting O2 availability19,39,40,41. Nevertheless, CO2 emissions increased with higher VWC in the present range of 27–87% (Supplementary Fig. 4). Thus, while the diffusive O2 supply may limit CO2 emissions under fully water-logged conditions, increased intermediate water availability may stimulate aerobic microbial activity by increasing (i) soluble substrate diffusion, (ii) microbial mobility and (iii) intracellular water potential in drained organic soils32,42,43. Therefore, as found also for mineral soils44, partial rewetting of organic topsoil does not necessarily result in reduced CO2 emissions. This role of VWC was consistent with the absence of CH4 emissions in our soil incubations, which could arise if the higher water contents inhibited aerobic mineralisation.

Given the importance of soil water availability for microbial metabolism, we studied CO2 emissions at a standardised water potential of pF 2. However, measurements of CO2 emissions at in situ water contents corroborated the rates measured at pF 2 (Supplementary Fig. 5) and confirmed the lack of association between OC content and area-scaled CO2 emissions over separate intervals of VWC (Supplementary Fig. 6). Overall, these results support the conclusion that OC content and OC density are ineffective predictors of area-scaled CO2 emissions and that CO2 emission factors may not be different between soils with 6-12% and >12% OC.

Implications for national GHG inventories

Adjustments to the national emission factors accepted under the UNFCCC are not trivial and require scientific documentation beyond the scope of this study. However, the technical implications of potentially similar CO2 emission factors for soils with 6–12% and >12% OC are imperative. As an example, such parity suggests that the Danish National Inventory Report24 may underestimate the direct CO2 emissions from managed croplands and grasslands on organic soils by 40% (Supplementary Table 4). Based on the National Inventory Report24 and analyses by the Danish Council on Climate Change45, the Danish government has decided to support the rewetting of 100,000 ha of drained organic agricultural soils by 2030, including both soils with 6–12% OC and soils with >12% OC. Strategies and the climate effect of rewetting the soils with >12% OC have been analysed19, but the most efficient strategy for the whole rewetting initiative will fundamentally depend on how the emissions from soils with 6–12% OC are conceived since these are likely to be currently underestimated45.

We infer that a similar or even greater underestimation of area-scaled CO2 emissions for 6–12% OC soils exists in many other countries with large proportions of organic soils in transition from true peat to organo-mineral soils due to extensive agricultural management. Global mapping of agricultural soils with 6–12% OC soils is underdeveloped, and national area estimates outside of Denmark are not readily available. However, agricultural cropland and grassland soils with 6–12% OC (n = 477) may be even more common than those with >12% OC (n = 196) as indicated by point data from the LUCAS (Land Use/Cover Area frame Survey) module of the European Soil Database, which contains observations from more than 20,000 sites stratified across the EU in a 2 by 2 km grid46. Improved soil mapping and development of emission factors for 6–12% OC soils is needed to improve national GHG accounting for organic soils under the UNFCCC, but also to allow for better estimates of the climate benefits of global initiatives to rewet drained organic soils. The present study, under controlled conditions, allowed for the relative comparison of CO2 emissions from organic topsoils with a wide range of OC contents and showed that area-scaled emissions are not controlled by the OC content as such. While these results provide valuable insights, they also highlight the need for targeted field research to refine CO2 emission factors for soils with 6–12% OC, which we recommend should be given high priority.

Methods

Study sites, soil sampling, and analyses

Denmark (43,000 km2), situated in the northern part of Europe, has a temperate to cold climate with warm summer and no dry season (Cfb, Dfb in Köppen-Geiger classification)47. The mean monthly temperatures (1991–2020) range from 1.5 °C in February to 16.9 °C in August, and the mean annual precipitation is 759 mm48. The flat landscape (mean, 31 m a.s.l) is divided into an eastern part with loamy Weichselian moraines and a western part with sandy glacial outwash plains and Saalian moraines47,49. The area of soils with 6–12% and >12% OC was previously (2021) reported at 162,000 and 129,000 ha, respectively, of which 55–60% was used for agriculture50.

The sampling sites for the study (n = 103) were selected to represent organic soils with a wide range of OC contents (>6% OC) from all regions of Denmark (Supplementary Fig. 1). Sampling positions were georeferenced (using differential GPS), the peat depth was measured using a soil auger, and the depth to the groundwater table was measured using piezometers. The land use was registered, and the topsoil was characterised according to the degree of decomposition (Supplementary Table 5).

Soil cores and bulk samples were taken from the topsoil at a depth of between 10 and 15 cm. Two soil cores were sampled in 100-cm3 stainless steel rings (height, 3.5 cm; diameter, 6.1 cm) and stored in undisturbed intact layering at 2 °C for biological incubation experiments. Bulk soils were air-dried and sieved (2 mm) for measurement of total C and nitrogen (N) by dry combustion (Vario Max Cube, Elementar Analysensysteme, GmbH). All soils were free of inorganic carbonates as tested by HCl effervescence51, and total C was rated as OC. Soil pH was measured using a glass electrode in a 1:5 (vol/vol) soil/water suspension (Supplementary Table 3).

Incubations and measurement of CO2 emissions

Soil cores in the metal rings (n = 206) were trimmed to the volume of 100 cm3. From each site, soil cores were prepared for incubation at a standardised matric potential (ψm) of −100 hPa (referred to as pF 2) and at the in situ water content. Adjustment to pF 2 by controlled wetting and drying on sandboxes52 was achieved in 2 weeks and allowed the effects of OC on CO2 emissions to be studied without the interaction of randomly different soil water status at the sampling times. However, CO2 emissions were also measured at in situ water contents to substantiate the rate measurements with pF 2 adjustment and to assess the role of OC content in CO2 emissions at discrete water content classes.

Weights were recorded for all soil cores before they were sealed at the bottom with a plastic lid and incubated in airtight 1-L glass jars fitted with a rubber septum for gas sampling. The soils were temperature equilibrated and incubated in the dark at 15 °C for one week. Then, at intervals of 5–7 days, 10-mL headspace gas samples were withdrawn using a syringe with a hypodermic needle and transferred to evacuated 6-mL Exetainer vials (Labco Ltd., Lampeter, UK) for analysis of CO2 (and CH4) using an Agilent 7890 gas chromatography GC system (Agilent, Nærum, Denmark) configured and calibrated as previously described53. After each gas sampling, the jars were opened for 10 min to equilibrate with atmospheric air, and after closing, a gas sample was taken from each jar to document the new baseline concentrations. Four empty glass jars were included as controls. After the final gas sampling (60 days), weights and dry weights of all soil cores were recorded to calculate soil bulk density. Relative water losses during the 60-day incubation period were approximately 4.7% (median, n = 206) and were not compensated for in the intact soil cores due to the risk of preferential rewetting of the surface layer. The headspace concentration of O2 after 7 days of soil incubation was checked in control experiments using gas chromatography54 and was 19.4 ± 0.4% O2 (mean ± standard deviation, n = 10), i.e., close to the atmospheric concentration.

Calculations and statistics

Emissions of CO2 during each of the 5–7 day incubation intervals were calculated as:

$${{{{{{\rm{CO}}}}}}}_{2}\,{{{{{\rm{emission}}}}}}\,(\mu {{{{{\rm{g}}}}}}\,{{{{{{\rm{CO}}}}}}}_{2}{{\rm{-}}}{{{{{\rm{C}}}}}})=({{{{{{\rm{C}}}}}}}_{{{{{{\rm{t}}}}}}}-{{{{{{\rm{C}}}}}}}_{0})\times {V}_{h}\times M\times {V}_{m}^{-1}$$
(1)

where C0 and Ct are the CO2 headspace concentrations (µL L−1) at the start and end of the incubation interval, respectively, Vh is the effective headspace volume (0.9 L), M is the molar mass of carbon (12 g mol−1), and Vm is the molar gas volume at 15 °C (23.6 L mol−1). Cumulative CO2 emissions over the 60-day incubation period (ΣCO2) were calculated by summing the individual 5–7 day CO2 emissions. Area-scaled, soil-weight-specific, and OC-specific CO2 emission rates were calculated as:

$${{{{{\rm{Area}}}}}}{{\rm{-}}}{{{{{\rm{scaled}}}}}}\,{{{{{\rm{emission}}}}}}\,{{{{{\rm{rate}}}}}}\,({{{{{\rm{mg}}}}}}\,{{{{{{\rm{CO}}}}}}}_{2}{{\rm{-}}}C\,{{\rm{h}}}^{-1}\,{{{{{{\rm{m}}}}}}}^{-2})={\Sigma {{{{{\rm{CO}}}}}}}_{2}\times {{{{{{\rm{time}}}}}}}^{-1}\times {A}^{-1}$$
(2)
$$ {{{{{\rm{Soil}}}}}}\,{{{{{\rm{weight}}}}}}{{\rm{-}}}{{{{{\rm{specific}}}}}}\,{{{{{\rm{emission}}}}}}\,{{{{{\rm{rate}}}}}}\,({{{{{\rm{mg}}}}}}\,{{{{{{\rm{CO}}}}}}}_{2}{{\rm{-}}}{{{{{{\rm{C}}}}}}\,{{{{{\rm{h}}}}}}}^{-1}\,{{{{{{\rm{kg}}}}}}}^{-1}\,{{{{{\rm{soil}}}}}}) \\ ={\Sigma {{{{{\rm{CO}}}}}}}_{2} \times {{{{{{\rm{time}}}}}}}^{-1}\times {{{{{\rm{soil}}}}}}\,{{{{{{\rm{dw}}}}}}}^{-1}$$
(3)
$$ {{{{{\rm{OC}}}}}}{{\rm{-}}}{{{{{\rm{specific}}}}}}\,{{{{{\rm{emission}}}}}}\,{{{{{\rm{rate}}}}}}\,({{{{{\rm{mg}}}}}}\,{{{{{{\rm{CO}}}}}}}_{2}{{\rm{-}}}{{{{{\rm{C}}}}}}\,{{{{{{\rm{h}}}}}}}^{-1}\,{{{{{{\rm{kg}}}}}}}^{-1}\,{{{{{\rm{OC}}}}}})\\ ={\Sigma {{{{{\rm{CO}}}}}}}_{2} \times {{{{{{\rm{time}}}}}}}^{-1}\times {{{{{\rm{OC}}}}}}\,{{{{{{\rm{mass}}}}}}}^{-1}$$
(4)

where ΣCO2 is the 60-day cumulative CO2-C emission, time is the total incubation time (1440 h), A is the surface area of the soil cores (29.2 cm2), soil dw is the dry weight of the individual soil cores, and OC mass is the weight (kg) of OC for the individual 100 cm3 soil core.

Statistical analyses were performed using R version 4.3.055. Differences in mean CO2 emission rates for different classes of OC content were tested by one-way analysis of variance (ANOVA) using the F-statistic56. Assumptions of normal distribution and homogeneity of variance were tested by Shapiro–Wilk and Levene’s tests, respectively, and by visual inspection of residuals against fitted values and histogram of residuals56. In the case of non-normal distribution and/or heteroscedasticity, data were log or square root transformed for ANOVA analysis. Reported means are results extracted from the statistical models after back-transforming from the log or square root scale, using the emmeans function (i.e., emmeans (model, ~OC content, type = ”response”)). Significant ANOVA tests (P < 0.05) were followed by post-hoc pairwise comparisons using the emmeans function with Tukey adjusted P values56,57. In addition to tests on means, differences in median CO2 emission rates between OC classes were tested using the non-parametric Kruskal-Wallis test, which computes P-values that are based on the χ2 distribution56.

Ordinary least squares regressions between CO2 emissions and soil parameters, as explanatory variables were performed using the lm function in R. Coefficients of determination (r2), were used as a measure of the explained variation in CO2 emissions56. In case of non-normal distribution and/or heteroscedasticity, data were log or square root transformed for the analyses. The fitted lines in regression plots are a linear approximation of the analysis performed in log or square root scale.

Correlations between soil parameters were analysed using the cor function in the corrplot package in R. As the data transformations did not meet the assumptions of normal distribution, the non-parametric Spearman’s rank correlation procedure was used56.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.