Introduction

Carbon exchanges between soil and atmosphere are of great concern for the understanding of the global carbon cycle and therefore for the understanding of climate change. Carbon from plants in the form of organic matter can be stabilized and stored in soil for thousands of years, in particular in subsoils (below 30 cm). However, the fate of subsoil carbon is not well understood. In addition to the fact that deep soils contain twice as much carbon as surface soils (Jobbagy and Jackson 2000), the organic matter is also older, as revealed by the universal decrease in 14C content with depth (Mathieu et al. 2015).

Soil carbon profiles behave as a dynamic system, which results from several processes acting together: the history of carbon inputs, the depth distribution of the latter, biodegradation processes, and various degrees of movement of dissolved or solid matter, including bioturbation that continuously buries surficial matter and brings deep soil material to the surface (Elzein and Balesdent 1995). In such systems, the organic matter contains a full age distribution from the youngest (recent plant- and root-derived products) to the very old, stabilized carbon. Statistically, the youngest organic matter has experienced fewer biodegradation cycles and is found closer to the surface.

The isotopic composition of carbon in SOM has often been used as a tool to understand organic matter dynamics through the 14C/12C and 13C/12C ratios (Campbell et al. 1967; Rafter and Stout 1970; Wang et al. 2018; Balesdent et al. 2018), where the 14C signal of organic matter is an indicator of its age, and the 14C content from nuclear bombs was further used to estimate the residence times of carbon in soils (Gaudinski et al. 2000; He et al. 2016), or the factors governing its stabilization (Mathieu et al. 2015). Several studies also recorded the 13C isotopic composition of soil carbon in order to investigate the stabilization of organic matter at depth (Poage and Feng 2004; Wynn et al. 2005), and an increase in δ13C values with depth has systematically been observed under C3 plant, on average by 1 to 3‰ in the first meter (Balesdent et al. 1993; Brüggemann et al. 2011). The reasons for this δ13C values enrichment are not yet fully known. Different mechanisms can be involved. Firstly, the absolute δ13C values of vegetation may have decreased with time (Boström et al. 2007). Post-photosynthesis fractionation may also occur since root materials are 13C-enriched compared to leaves (Gessler et al. 2007; Werth and Kuzyakov 2010). Secondly, true mass-dependent isotope fractionation may occur in the combined process of microbial respiration and microbial biosynthesis, so that residual products would be progressively 13C-enriched during the continuous degradation process, when compared to initial plant material as reviewed by Werth and Kuzyakov (2010). Decomposition may also select 13C-enriched or 13C-depleted components by differential microbial use efficiency or decay rate between molecules (Boström et al. 2007). Finally, the movement of organic matter within the soil profile (e.g. in the form of dissolved organic carbon) may lead to the accumulation of 13C-enriched or 13C-depleted compounds downwards (Kaiser et al. 2001). Incorporation of inorganic carbon atoms by heterotrophic microorganism (dark CO2 fixation) may also increase SOM 13C/12C isotope ratios by adding atoms that have the isotopic composition of atmospheric CO2, but the extent of the process has been considered as negligible (Šantrůčková et al. 2018).

Some of these studies focused on proposing mechanisms behind the 13C enrichment with depth and based their work on the assumption that vegetation in equilibrium with soil maintained constant δ13C values. However, there are evidences that the vegetation δ13C values has changed over time. The more recent change is due to fossil fuel burning and land use change during the last 150 years of Human activities, resulting in both an increase in pCO2 (the partial pressure of carbon dioxide in the atmosphere) and a dilution of the 13C of atmospheric CO2 by 13C depleted emitted CO2 (i.e. “Suess effect”, Keeling et al. 1995, 1979). The depletion in atmospheric δ13CO2 values leads to a more 13C-depleted vegetation biomass during photosynthesis (Farquhar et al. 1989). The “Suess effect” is considered now to be part of the 13C-depletion of SOM in surface soil (Boström et al. 2007; Breecker et al. 2015; Brunn et al. 2016, 2017).

In addition, elevation of pCO2 -from ca. 190 ppm prior to 17,500 BP (Before Present i.e. before 1950) up to 408 ppm in 2018 (US Department of Commerce 2017) - is expected to affect the fractionation process due to the change in stomatal conductance regulation in plants in order to balance the carbon input and to increase the intrinsic water-use efficiency (Keeling et al. 2017). Hence, several environmental studies have shown a relation between pCO2 and the isotope fractionation by C3 plant photosynthesis (Krishnamurthy and Epstein 1990; Van de Water et al. 1994; Feng and Epstein 1995; Pasquier-Cardin et al. 1999; Schubert and Jahren 2015; Voelker et al. 2016; Keeling et al. 2017). All but one (Kohn 2016) have concluded that the increase in pCO2 induces a decrease in δ13C values of C3-plant, leading to a decrease in the δ13C values of the vegetation ranging between 2.4 ‰ (Keeling et al. 2017) and 4.9 ‰ (Schubert and Jahren 2015) when CO2 increases from 190 to 400 ppm.

Our objective is to evaluate the impact of the changes of vegetation δ13C on the vertical distribution of δ13C values in SOM. We have first reconstructed a global trend of the isotopic composition of the vegetation through time by using atmospheric data and relations between pCO2 and δ13C values from four studies (Feng and Epstein 1995; Schubert and Jahren 2015; Voelker et al. 2016; Keeling et al. 2017). Then, we compiled data from 45 soil profiles around the world where δ13C and ∆14C values of SOM had been measured. Considering that organic matter is older in deep horizons, we hypothesized that the change in vegetation δ13C values with time is responsible for the soil 13C enrichment with depth. We have tested this hypothesis by comparing the global trend of vegetation δ13C with SOM δ13C values.

Material and methods

Definition and vocabulary

The isotopic composition of stable carbon is reported in δ13C values presented as per mil (‰) compared to Vienna Pee Dee Belemnite (V-PDB) international standard. We described in Table 1 the main terms we used in the following sections.

Table 1 Definition of the specific terms used in the text
Full size table

Atmospheric CO2 data

Atmospheric CO2 data are needed to reconstruct the isotopic composition of vegetation. The pre-bomb δ13Cair and ∆14Cair values of atmospheric CO2 were taken from Schmitt et al. (2012) and from Reimer et al. (2009) and the post-bomb from Francey et al. (1999) and Hua et al. (2013), respectively. The pCO2 values came from Francey et al. (1999), Schmitt et al. (2012) and from NOAA data (US Department of Commerce 2017) for the most recent dates. We performed linear interpolation when data were missing.

Vegetation δ13C values

The temporal change of isotopic composition of the vegetation was reconstructed from four studies by Schubert and Jahren (2015), Feng and Epstein (1995), Voelker et al. (2016) and Keeling et al. (2017) that found a relation of the δ13C values of the vegetation (δ13Cplant) with the atmospheric pCO2. They derived their relation from the Farquhar’s equation (Farquhar et al. 1989):

$$ {\delta}^{13}{C}_{plant}=\frac{\delta^{13}{C}_{air}-0.0044+\left(0.0017\ast \frac{C_i}{p{CO}_2}\right)}{0.0044-\left(0.0017\ast \frac{C_i}{p{CO}_2}\right)+1} $$
(1)

where Ci corresponds to the leaf intercellular space CO2 concentration and pCO2 to the atmospheric CO2 concentration (ppm).

The four equations we used from the different studies are summarized below:

  • From Schubert and Jahren (2015):

    $$ {\delta}^{13}{C}_{plant}=\frac{\begin{array}{c}{\delta}^{13}{C}_{air}-\left(28.26\ast 0.22\ast \left(\ p{CO}_2+23.9\right)\right]/\left[\right(28.26+\left(0.22\ast p{CO}_2+23.9\right)\Big]\\ {}\ \end{array}}{\left(28.26\ast 0.22\ast \left(\ p{CO}_2+23.9\right)\right]/\Big[\left(28.26+\left(0.22\ast p{CO}_2+23.9\right)\right)+1} $$
    (2)

  • From Feng and Epstein (1995), we used the linear relation between pCO2 and δ13C:

    $$ {\delta}^{13}{C}_{plant}={\delta}^{13}{C}_{air}-\left( constant-0.02\ast p{CO}_2\right) $$
    (3)

  • From Voelker et al. (2016), we calculated the term Ci/pCO2 = a + b*exp.(−0.0076*pCO2); (r2 = 0.43, p < 0.05) that we used in the Farquhar’s equation (Eq. 1)

  • From Keeling et al. (2017), we used the linear relation between pCO2 and δ13C:

    $$ {\delta}^{13}{C}_{plant}={\delta}^{13}{C}_{air}-\left( constant-0.014\ast p{CO}_2\right) $$
    (4)

Taking the absolute δ13Cplant values does not make sense because of spatial and interspecies variations, thereby we calculated for each model the difference: Δ13Cplant = δ13Cplant - δ13Cplant (Δ14C = 0).

From Eqs. 1, 2, 3 and 4 we reconstructed a global trend of the average variations of Δ13Cplant values over time. The current state of knowledge does not allow one scenario to be chosen over another, we decided to work with the mean value of Δ13Cplant from the four equations.

Soil database description

Forty-five soil profiles with both δ13CSOM and ∆14CSOM values measured from the surface to deep soil were extracted from the literature. These profiles were selected from the database of Mathieu et al. (2015) that also reports sampling year, location, soil type, land-use, soil layer identification, pedological properties, climatic data, carbon content, characteristic analyses and δ13C and ∆14C values of more than 300 soil profiles. From this database, only δ13C values measurements on bulk soil organic carbon by IRMS and C3 plant derived profiles were chosen. The 45 profiles encompass nine soil types (Luvisol, Gleysol, Podzol, Acrisol, Ferralsol, Chernozem, Andosol, Nitisol and Cambisol, IUSS Working Group WRB 2014) and three large ecosystems (grassland, forest, and arable land). Forty-four out of the 45 were sampled from 1959 to 2009 and the last one in 1900 (See Supplementary material). The location of the 45 profiles is presented in Fig. 1 as well as the mean annual temperature; mean annual precipitation and sampling year. The references from which the δ13C and ∆14C values were extracted and details of the data are presented in supplementary material.

Fig. 1
figure 1

Geographic distribution, range of values of mean annual temperature, mean annual precipitation, and sampling year of the 45 selected soil profiles

Full size image

Simulation of δ13C values of soil organic matter (SOM) from Δ14C values and sampling year

In order to compare the global change of δ13Cplant with δ13CSOM values through time, we first converted ∆14CSOM values into age. Then, to estimate how δ13CSOM values were impacted by the δ13Cplant values, we simulated changes in δ13CSOM values by integrating the δ13Cplant values for each profile depending on the sampling year.

The ∆14C values of SOM from the database were converted into mean calendar age according to Balesdent (1987): a mean age value α of SOM was calculated from ∆14CSOM values assuming an exponential distribution of ages, i.e., by solving Eq. 5:

$$ \frac{\Delta ^{14}C}{1000}+1=\frac{\int_{t=0}^{\infty}\left(1+{\Delta }^{14}{C}_{air\left(p-t\right)}\right)\ast {e}^{-\lambda \ast \left(t+m-p\right)}\ast {e}^{-\frac{t}{\alpha }}\ast dt}{\int_{t=0}^{\infty }{e}^{\frac{-t}{\alpha }}\ast dt}\kern7em $$
(5)

where p is the sampling year; ∆14Cair(p - t) is the atmospheric ∆14C values at the date (p - t); ∆14Cair values were obtained from atmospheric summer ∆14C values records in Hua et al. (2013) and Reimer et al. (2009) taking into account the hemisphere and atmospheric zone of the studied site (Hua et al. 2013); λ is the radioactive decay constant of 14C (ln(2)/λ = 5730 years); m is the date of 14C measurement and was fixed as 2 years before publication if unknown; α is the mean age of SOM at the sampling year, so that the calendar age of SOM is p – α. Eq. (2) has two solutions for p – α in the cases where the ∆14C values of SOM are higher than ∆14Cair values in the sampling year, i.e., corresponding to either young post-bomb carbon or a mixture of pre-bomb and bomb-peak carbon. In that case, the younger solution was chosen for litter layers, whereas the older one was chosen for organo-mineral horizons. A few litter samples with very high ∆14C values had no solution for p – α; in that case, the calendar age was set at ∆14Cair(p – α). Secondly, for each site we integrated the vegetation δ13Cplant values as a function of carbon mean age p – α and sampling year using the same exponential distribution of ages. Equation (6) is similar to Eq. (5) but with no radioactive decay.

$$ {\delta}^{13}{C}_{sim}=\frac{1}{\alpha}\ast {\int}_{t=0}^{\infty}\left({\delta}^{13}{C}_{plant\left(p-t\right)}\right)\ast {e}^{-\frac{t}{\alpha }}\ast dt $$
(6)

with δ13Csim, the SOM δ13C simulated by taking into account the mean value of δ13Cplant calculated with Eqs. 1,2,3 and 4.

Statistical analyses

To study the relation between δ13CSOM and ∆14CSOM values, we calculated for each profile the linear regression of the function (taking into account the litter values):

$$ {\delta}^{13}{C}_{SOM}=s\ast {\Delta }^{14}{C}_{SOM}+b $$
(7)

with b equal to δ13C(∆14C = 0).

To highlight the variables that affect the slope s of the function (7), we calculated linear regression with the software R (version 3.3.2.; lm function). The explanatory variables were “sampling year”, “mean annual precipitation”, “mean annual temperature”, “aridity index” (Trabucco and Zomer 2009), “elevation” and “soil type”. Significance is chosen when p < 0.05. The relation of the previous variables with the δ13C values at the depth 0 cm and with δ13C values of the litter was also tested with a linear model. The climatic data (precipitation, temperature, and aridity index) were taken from authors’ statements or from the geographical coordinates for the modern climate (New et al. 2002). Climatic variations during the last ten thousand years were not considered.

Results

∆14C and δ13C values relation in soil organic matter

The δ13CSOM values increased with depth on average from −27.3 ± 0.3‰ at 0 cm to −25.4 ± 0.8‰ at 100 cm (Fig.2a). The ∆14CSOM values decreased with depth for all the profiles, on average from 180 ± 42‰ to −307 ± 85‰ (Supplemental material), and accordingly, mean calendar age increased with depth (Fig.2b). There was a high variability among the profiles, especially at depths where fewer samples were measured, but similar patterns were found in each profile. The δ13C values at the depth 0 cm significantly depend on the sampling year. The mean absolute δ13CSOM values at the first 10 cm depth decreased with time: it was −26.9 ± 0.9‰ for profiles sampled between 1960 and 1973 and − 28.5 ± 0.5‰ between 2005 and 2010. Because of the covariation with other variables, we could not estimate a significant temporal trend in 13C gradient with depth. On the contrary there is a significant temporal trend in the relation of 13C- and 14C-gradients with depth. The slope s of the relation between δ13CSOM and ∆14CSOM values (Eq. 7) in a given profile varies from 0.002 ± 0.001 to −0.010 ± 0.002. We found that the slopes significantly depend on the sampling date and accordingly to the isotopic composition of the atmosphere δ13Cair values. The slope of the relation between δ13CSOM and ∆14CSOM values was more negative for profiles sampled more recently where δ13Cair values were also more negative (Fig. 3). As a result of the progressive incorporation of bomb-14C, ∆14CSOM values tend to increase with the sampling year in the topsoil, but much less in deep horizons. The negative correlation between s and sampling date therefore means that topsoil δ13CSOM values decrease, when ∆14CSOM values increase. The slope s was not significantly related with soil types.

Fig. 2
figure 2

a. Overall depth distribution of δ13CSOM values and b. Mean calendar age calculated with Eq. 5 in the 45 soil profiles

Full size image
Fig. 3
figure 3

Relation between the slope s and the sampling date of each profile, or the mean isotopic composition of the atmospheric (δ13Cair values) at the corresponding year; s is the slope of the linear regression of δ13CSOM values as a function of ∆14CSOM values of an individual profile. The red line and the equation on the graph correspond to the regression of the slopes s vs the sampling dates. Error bars represent one standard error of the estimated slope s in each profile. See Table 1 for definitions of variables and abbreviations

Full size image

The δ13CSOM values presented also a relation with the elevation where the soils were sampled. Mean annual temperature, latitude and aridity index did neither affect δ13C values of SOM, nor the slope of the relation between δ13CSOM and ∆14CSOM values.

Comparing δ13C values of soil (δ13CSOM) and δ13C values of vegetation (δ13Cplant)

The mean Δ13Cplant values decreased by 4.2 ± 0.6‰ between 20,000 BP and 2018 AD (Anno Domini). We found that the mean values of Δ13CSOM (moving average of 10 points) matched the reconstructed Δ13Cplant values through time (Fig. 4c).

Fig. 4
figure 4

a. Reconstruction of Δ13Cplant values from 20,000 BP to today from 4 different studies. b. δ13CSOM values as a function of the mean calendar age of each SOM sample and δ13Csim calculated with Eq. 6. The age was inferred from ∆14CSOM and sampling date, using Eq. 5. c. Mean values of Δ13Cplant and Δ13CSOM (moving average of 10 points). The light-blue lines represent two standard errors of the estimated mean value Δ13Cplant from the four scenarios; it does not take into account the error of each individual scenario. The light-red lines represent the confidence interval of the mean value Δ13CSOM (95%). Note that the x axis is divided into 4 scales. See Table 1 for definitions of variables and abbreviations

Full size image

To test the hypothesis that the vertical δ13CSOM gradient in soil is due to the historical change in vegetation δ13C values, we calculated for each sample the simulated δ13CSOM values (δ13Csim) by taking the isotopic composition of the vegetation (mean value of δ13Cplant) at the calendar year of SOM (Eq. 3). The observed ∆13CSOM value of each sample was then compared to the predicted equivalent difference, ∆13Csim (Table 1). The observed ∆13CSOM values relative to the simulated ∆13Csim values are shown in Fig. 5. The simulation explains 40% of the variance. Few observed δ13CSOM values presented a high enrichments (> 2.5‰) and are older than 3000 BP (Fig. 5).

Fig. 5
figure 5

Observed ∆13CSOM versus ∆13Csim values, predicted from the sole hypothesis of a past change in δ13Cplant values. Highlighting of layers older than 3000 BP in red and litter in blue. The red line is the linear regression. See Table 1 for definitions of variables and abbreviations

Full size image

Discussion

δ13C values of SOM are derived from δ13C values of vegetation

The almost systematic 13C-enrichment that accompanies carbon ageing with depth can be the result of temporal change in the initial composition of carbon or isotopic effects associated to soil processes. In this study, we analyzed soil 13C gradients as a function of carbon age (and not depth per se), and isolated the sole effect of the change in initial δ13Cplant values on the resulting δ13CSOM values. Both changes in pCO2 and δ13C values of atmospheric CO2 induce changes in vegetation δ13C values. In our panel of soil profiles, the average vertical 13C gradient (Fig. 2) is similar to the expected change in vegetation δ13C values (Fig. 4). Variations of δ13CSOM values with time (Figs. 4b, c) clearly mimic the simulated changes in δ13Cplant values. For example, from the four studies, we calculated an average decrease in δ13Cplant values of 2.4 ± 0.6‰ from 5000 BP to −50 BP (i.e. 2000 AD) and we observed an average decrease in δ13CSOM values of 2.5 ± 1.5‰ over the same period of time (Fig. 4c). Here, the values of observed ∆13CSOM coincide with the calculated values of ∆13Cplant over a broad time range, i.e., both between 1000 and 3000 BP and between −30 and − 50 BP (i.e. 1980 and 2000 AD, Fig. 4). The average change of δ13Cplant values with time explains the average increase of δ13CSOM values with depth for the studied soil profiles. This point is supported by the relation of the gradients s with the sampling date (Fig. 3) that indicates both the impact of pCO2 and δ13Catm values, with soils sampled before the 1970’s having a weak gradient of δ13C values. Moreover, by comparing ∆13CSOM values of eight soil profiles, representing the major soil types, with the mean variation of ∆13Cplant values over time (Fig. 6), we showed that the average change in δ13CSOM values is attributed to the average change in δ13Cplant values with time for very different types of soils. The mean δ13C value of C3 plants thus varied with major transitions from high values at the deglaciation (ca. 12,000 BP) and with an exponential drop associated to the newly suggested geological epoch of the Anthropocene (ca. from 0 BP = 1950 AD, Steffen et al. 2016).

Fig. 6
figure 6

Δ13CSOM values of 8 selected soil profiles (black dot) representing the major soil types, climates and land cover of the database compared to the global trend Δ13Cplant values (in blue) in function of the age (year before 2020). The Y-axis is the mean calendar age of SOM inferred from ∆14CSOM values and sampling date from 2020. The light-blue lines represent two standard errors of the estimated mean value Δ13Cplant. See Table 1 for definitions of variables and abbreviations

Full size image

Considering that the variation in δ13Cplant values with time induce the change in the isotopic signature of the soil carbon input, we were able to derive its impact on SOM δ13C values. Simulated data, δ13Csim values, mimicked the measured data, δ13CSOM values in Fig. 4b and showed a relation with a slope close to 1 (Fig. 5) although δ13CSOM values are enriched compared to δ13Csim values for the old layers. The δ13CSOM and δ13Csim values in Fig. 5 were obtained independently: δ13Csim values were deduced from the ecophysiological literature (Eq. 6), whereas δ13CSOM values were obtained from the soil database (Supplementary material). To express δ13CSOM values variations with time, and thus to derive age in years from ∆14CSOM values, we chose an exponential hypothesis to express the fact that mean age of SOM (Eq. 5) is the mixture of young and older compounds in each sample. This is an over-simplification of the soil carbon demography, but is in accordance with the observed smoothing of the bomb-14C peak of the 1960s. The latter is delayed and diluted in SOM as observed by O’Brien and Stout (1978) and Trumbore (2009). The exponential hypothesis has no final impact on the general relation between δ13CSOM and δ13Csim values, since δ13Csim values and calendar age from ∆14C values were calculated using the same distribution of ages. Consequently, the relation revealed in Fig. 5 between δ13CSOM and δ13Csim values is real and not an artifact of data handling: 40% of soil organic matter isotopic signal with depth is explained by the variations of isotopic composition of vegetation with time by only considering average changes in atmospheric CO2. The relation of δ13CSOM values with elevation also suggests the impact of pCO2 on δ13Cplant values but was not taken into account in our linear models that only consider mean global reconstructed pCO2.

In addition to changes in pCO2, and subsequent δ13Cplant values, δ13C values enrichment in SOM might be caused by post-photosynthesis fractionation observed in different plant organs (Badeck et al. 2005; Brüggemann et al. 2011). In fact, C3 plants roots are 13C-enriched by 1.2 ± 0.6 ‰ compared to the shoots (Klumpp et al. 2005; Badeck et al. 2005; Werth and Kuzyakov 2006). Moreover, the proportion of root-derived C inputs is expected to be higher at depth (e.g., from <40% of inputs at 0 cm to more than 80% at 100 cm) depending on the contribution of dissolved organic carbon in deep horizons (Balesdent et al. 2011). The proportion of root-derived C could even be higher in boreal environments where 50 to 70% of organic matter are derived from roots and root-associated microorganisms, especially ectomycorrhizal fungi (Clemmensen et al. 2013). The 13C root enrichment was not included in our simulation and could be the reason for the deviation from unity in the relation with δ13CSOM values (Figs. 4b and 5). Indeed, using this assumption, a 13C root enrichment of 1.2 ± 0.6 ‰ (Werth and Kuzyakov 2006) would contribute at least to an additional SOM enrichment, accounting for 0.5 ± 0.2‰ in the 0 and 1 m soil depth. Notably, this value corresponds to the observed difference between the overall depth gradient of ∆13CSOM and ∆13Csim values (Figs. 4 and 5). Therefore, we suggest that more than the half of the 13C-enrichment in SOM is directly derived by δ13Cplant values, including root derived carbon.

Variations in δ13C values due to plant and soil diversity

Our results have shown that changes over time of δ13Cplant values explain at least 40% of SOM δ13C values variations (Fig. 5). However, the reason of the uncertainty around the trend (Figs. 4 and 5) was not explained by our model. First of all, we have chosen to use a global δ13Cplant value over time based on the average of Eqs. 1, 2, 3 and 4 excluding the effect of local climate or different types of vegetation. Therefore, the variance around the change in δ13C values with time might be due to the spatial variation associated with differences in plant species and the temporal variation of vegetation element (e.g. shrubs, grass, forests), that may have occurred on a given site. Indeed, photosynthetic δ13C values discrimination in the plant is sensitive to various environmental conditions such as light and water (Farquhar et al. 1989) and to vegetation type (Brugnoli and Farquhar 2000), on the plant genotype (Roussel et al. 2009) and nutritional status. Balesdent et al. (1993) have for instance reported that the variance of δ13C values of the different plant species in a single forest was almost as large as the variance of the vegetation δ13C values in the world. Plant δ13C values were mainly determined by local pedoclimate; i.e. soil microclimates controlled by local temperature, water content and aeration of the soil. Since plant species react differently to environmental changes with different fractionation factors (Ehleringer and Cerling 1995; Voelker et al. 2016), the δ13C values of vegetation as a function of age record the local changes, also observed in SOM.

The relation of climatic variables and ecosystem type with s values (Fig. 3) were not detectable in our dataset but could be partly responsible for the variance around the trend.

Different soil types and pedogenesis also had an impact on the variation in δ13C values. Beyond the date of sampling, organic matter represented a relative 13C depletion of 2‰ on average compared to the δ13Cplant values in soils, in which the pedogenesis is controlled by percolating processes or low clay content (Podzol, Cambisol) in intermediate age (between 100 and 1000 BP). For instance, δ13CSOM values of the Podzols in the database have decreased on average by only 0.98‰ between 2500 BP and − 40 BP compared to the 1.82‰ on average for the other soils. Therefore, soil processes such as interaction with the inorganic phase and/or organic carbon dissolution (Kaiser et al. 2001) could create variance in the generally predicted δ13C values profile. Wynn et al. (2005) found that fine-textured soils lead to the selective accumulation of 13C-enriched components of SOM (carbohydrates, bases, amino acids) and also protect microbial compounds. The texture of the soil selected in this database was not always reported impeding to confirm this statement. However, soil texture is likely one of the cause of variations and differences in 13C enrichment among soil profiles.

δ13CSOM values enrichment due to microbial processes

From our results, δ13Cplant values cannot explain the whole 13C gradient in SOM (Fig. 5). Previous explanations of the δ13C values enrichment with depth during soil processes were kinetic discrimination during respiration, preferential consumption of compounds and contribution of sorption processes.

The first hypothesis is the preference of 12C for respiration by micro-organisms (Agren et al. 1996; Ekblad et al. 2002). Since organic matter is mainly composed of microbe-derived products more than from plant-derived molecules (Bol et al. 2009), this discrimination during respiration should affect δ13CSOM values when age of SOM increases. Laboratory experiments on respired CO2 from soils have resulted in contrasted fractionation. Werth and Kuzyakov (2010) by synthetizing several results have found a δ13C values difference between respired CO2 and soil microbial biomass between +4.6 ‰ and − 3.2‰. This large variation could be due to the different methods used for CO2 sampling or to different carbon sources used by micro-organisms (Boström et al. 2007) and did not support the hypothesis. The nature of substrate may affect more the isotopic composition of CO2 than the fractionation process itself during respiration (Fernandez and Cadisch 2003). Since micro-organisms preferentially mineralize carbon that is more accessible which corresponds to younger carbon in accordance with the continuous litter quality theory (Agren et al. 1996; Lehmann and Kleber 2015); we suggest that a part of the respired CO2 is 13C-depleted compared to SOM because fresh carbon (young carbon) is 13C-depleted. Therefore, the isotope discrimination associated to the respiration would have a lower contribution to the depth gradients, in accordance with the conclusion of Boström et al. (2007) or Breecker et al. (2015) and in contrast with the hypothesis of Agren et al. (1996).

The microbial biomass is also 13C-enriched by 1.2 ± 2.6 ‰ compared to SOM (Werth and Kuzyakov 2010). This suggests a preferential utilization of 13C-enriched compounds by micro-organisms because hardly decomposable compounds are generally 13C-depleted such as lignin and lipids compared to proteins or starch (Bowling et al. 2008). However, lignin is degraded quickly in soils (Kögel-Knabner 2000). Microbial biomass is able to decompose any kind of organic compounds which are accessible but the protection of some organic compounds by organo-mineral interaction could induce preferential use during biodegradation (Lützow et al. 2006). In addition, microorganisms are mainly composed of root-derived carbon in soils (Kramer et al. 2010; Schmidt et al. 2011; Clemmensen et al. 2013), generally enriched compared to the aboveground vegetation leading to an enrichment with depth of carbon biomass. Hence, true mass-dependent isotope fractionation may be compensated by isotope effects in opposite direction, such as preferential use or absorption of compounds with varying δ13C values during stabilization or degradation. 13C-enrichment in soil during biodegradation processes cannot be systematic. However, δ13CSOM values are generally enriched relative to δ13Csim values for the old layers (> 3000 yr BP, Figs. 4 and 5); suggesting a 13C-enrichment of SOM relative to the global trend δ13Cplant values due to microbial processes during the time course of decomposition. But in addition to the discrimination associated with biodegradation and to the post-photosynthesis fractionation, we propose another hypothesis for the very high values of Δδ13CSOM in Figs. 4 and 5, i.e., the contribution of old organic matter from Pleistocene vegetation which is 13C-enriched (Fig. 4). As mentioned, a number of processes might have led to accumulation of 13C-enriched past vegetation and subsequent stabilization of older carbon at depth; such as successive degradation stages of SOM (Lehmann and Kleber 2015), the mineral protection of organic compounds (Basile-Doelsch et al. 2015) and the lack of energy needed for micro-organisms to decompose organic matter in deep layers (Fontaine et al. 2007). Another possible source of carbon in subsoils is dissolved organic carbon coming from desorption and dissolution of organic compounds during microbial processes (Kaiser et al. 2001) and becoming older with deep transport. Further research is needed regarding the quantification of 13C enrichment in SOM in subsoils due to microbial processes.

Conclusion

13C enrichment of soil organic matter with depth is commonly recorded in soil profiles worldwide. Disentangling its origin before using this variable as conservative to yield information of the composition of soil organic matter mixture or any soil process is a prerequisite aiming at defining organic matter dynamics. Here, atmospheric and paleoclimatic data of CO2, four physiological models to reconstruct past δ13C values of vegetation and soil radiocarbon data revealed that the variation of past vegetation δ13C values is an important reason of the average δ13C values enrichment in soil with depth. Around the global trend, three other mechanisms may affect δ13C values of SOM: i) The increasing contribution of root-derived carbon with increasing depth results in a general 13C enrichment in soil carbon with increasing depth - ii) Soil processes which may lead to weaker or stronger 13C gradients for example by accumulation of 13C-depleted components in low-clay soils such as podzols and cambisols- iii) The discrimination associated with microbial processes which may lead to the accumulation of 13C-enriched compounds during the decomposition of SOM.

We suggest that above and below ground vegetation δ13C values, all together, could even be the only reason of the 13C variations in SOM with minor alterations due to soil processes.

Finally, considering that the stable isotopic composition of soil carbon is related to the absolute age of organic matter, the large change in C3 plant isotopic composition associated with the Anthropocene may provide an indication of the age of soil carbon in topsoils in a “13C dating” approach. The change in pCO2 during the Anthropocene is sufficient to estimate the age within the last 100 years. Since the mean residence time of SOM in topsoil lies between decades and centuries, dating SOM in that range is of particular interest. Moreover, “13C-dating” in combination with radiocarbon dating is accordingly a potential tool to separate the ∆14C values before and after the bomb signal. This method may be complementary to radiocarbon dating.