Abstract
Carbon cycle feedbacks represent large uncertainties in climate change projections, and the response of soil carbon to climate change contributes the greatest uncertainty to this. Future changes in soil carbon depend on changes in litter and root inputs from plants and especially on reductions in the turnover time of soil carbon (τ_{s}) with warming. An approximation to the latter term for the top one metre of soil (ΔC_{s,τ}) can be diagnosed from projections made with the CMIP6 and CMIP5 Earth System Models (ESMs), and is found to span a large range even at 2 °C of global warming (−196 ± 117 PgC). Here, we present a constraint on ΔC_{s,τ}, which makes use of current heterotrophic respiration and the spatial variability of τ_{s} inferred from observations. This spatial emergent constraint allows us to halve the uncertainty in ΔC_{s,τ} at 2 °C to −232 ± 52 PgC.
Similar content being viewed by others
Introduction
Climate–carbon cycle feedbacks^{1} must be understood and quantified if the Paris Agreement Targets are to be met^{2}. Changes in soil carbon represent a particularly large uncertainty^{3,4,5,6,7}, with the potential to significantly reduce the carbon budget for climate stabilisation at 2 °C global warming^{8}. Previous studies have investigated the response of soil carbon to climate change based on both observational studies^{9} and Earth System Models (ESMs)^{10}. ESMs are coupled models which simulate both climate and carbon cycle processes. Projects such as the Coupled Model Intercomparison Project (CMIP)^{11,12}, have allowed for consistent comparison of the response of soil carbon under climate change from existing stateoftheart ESMs. However, the uncertainty due to the soil carbon feedback did not reduce significantly between the CMIP3 and CMIP5 model generations^{6}, or with the latest CMIP6 models (see Fig. 1 and Supplementary Fig. 1), such that the projected change in global soil carbon still varies significantly amongst models^{13}.
This study uses an alternative method to obtain a constraint on the ESM projections of soil carbon change. In previous studies, emergent constraints based on temporal trends and variations have been used successfully to reduce uncertainty in climate change projections^{14}. Our approach follows the method used in Chadburn et al.^{15}, where a spatial temperature sensitivity is used to constrain the future response to climate change—which we term as a spatial emergent constraint. Our study combines the Chadburn et al.^{15} method with the soil carbon turnover analysis of Koven et al.^{16} to get a constraint on the sensitivity of soil carbon turnover to global warming.
Soil carbon (C_{s}) is increased by the flux of organic carbon into the soil from plant litter and roots, and decreased by the breakdown of that organic matter by soil microbes which releases CO_{2} to the atmosphere as the heterotrophic respiration flux (R_{h}). If the vegetation carbon is at steadystate, litterfall will equal the Net Primary Production of plants (NPP). If the soil carbon is also near to a steadystate—and in the absence of significant fire fluxes and other nonrespiratory carbon losses—the litterfall, NPP, and R_{h} will be approximately equal to one another. Even over the historical period, when atmospheric CO_{2} has been increasing and there has been a net land carbon sink, this approximation holds well (see Supplementary Fig. 4).
In order to separate the effects of changes in NPP from the effects of climate change on R_{h}, we define an effective turnover time^{17} for soil carbon as τ_{s} = C_{s}/R_{h}. The turnover time of soil carbon is known to be especially dependent on temperature^{3}. A common assumption is that τ_{s} decreases by about 7% per °C of warming (equivalent to assuming that q_{10} = 2)^{18}. However, this sensitivity differs between models, and also between models and observations.
We can write a longterm change in soil carbon (ΔC_{s}), as the sum of a term arising from changes in litterfall (ΔC_{s,L}), and a term arising from changes in the turnover time of soil carbon (ΔC_{s,τ}):
Model projections of the first term (ΔC_{s,L}) differ primarily because of differences in the extent of CO_{2}fertilisation of NPP, and associated nutrient limitations. The second term (ΔC_{s,τ}) differs across models because of differences in the predicted future warming, and because of differences in the sensitivity of soil carbon decomposition to temperature (which includes an influence from faster equilibration of fastturnover compared to slowturnover carbon pools under changing inputs^{13}). This study provides an observational constraint on the latter uncertainty. As the vast majority of the CMIP6 and CMIP5 models do not yet represent vertically resolved deep soil carbon in permafrost or peatlands, we focus our constraint on carbon change in the top 1 metre of soil. To ensure a fair likeforlike comparison we also exclude the two CMIP6 models that do represent verticallyresolved soil carbon (CESM2 and NorESM2), although this has a negligible effect on our overall result. Our study therefore applies to soil carbon loss in the top 1 metre of soil only. Below we show that it is possible to significantly reduce the uncertainty in this key feedback to climate change using currentday spatial data to constrain the sensitivity to future warming.
Results and discussion
Proof of concept
For each ESM, we begin by calculating the effective τ_{s} using timeaveraged (1995–2005) values of C_{s} and R_{h} at each gridpoint, and applying our definition of τ_{s} = C_{s}/R_{h}. We do likewise for observational datasets of soil carbon in the top 1 metre^{19,20} and timeaveraged (2001–2010) heterotrophic respiration^{21}, as shown in Fig. 2. Figure 2c shows the map of inferred values of τ_{s} from these observations, with a notable increase from approximately 7 years in the warm tropics to over 100 years in the cooler high northern latitudes.
Similar maps can be diagnosed for each of year of data, for each ESM, and for each future scenario, giving time and space varying values of τ_{s} for each model run. This allows us to estimate ΔC_{s,τ}, via the last term on the right of Eq. (1). For each ESM, the R_{h,0} value is taken as the mean over the decade 1995–2005, to overlap with the time period of the observations and to maintain consistency across CMIP generations. Individual gridpoint τ_{s} values are calculated for each year before calculating areaweighted global totals of ΔC_{s,τ}. The uncertainty of ΔC_{s,τ} stems from the uncertainty in soil carbon turnover (τ_{s}), and the uncertainty due to differing climate sensitivities of the models. In this study, we aim to quantify and constrain the uncertainty in τ_{s}. To isolate the latter uncertainty, we consider ΔC_{s,τ} for differing levels of global mean warming in each model. The resulting dependence of global total ΔC_{s,τ} on global warming is shown in Fig. 1a, for each of the ESMs considered in both CMIP6 and CMIP5 (seven CMIP6 ESMs and nine CMIP5 ESMs), and for three Shared Socioeconomic Pathways (SSP): SSP126, SSP245 and SSP585 (CMIP6)^{22}, or the equivalent Representative Concentration Pathways (RCP): RCP2.6, RCP4.5 and RCP8.5 (CMIP5)^{23}. In all cases ΔC_{s,τ} is negative, which is consistent with the soil carbon turnover time decreasing with warming. The more surprising thing to note is the huge range in the projections, with a spread at 2 °C global mean warming of approximately 400 PgC, regardless of future SSP/RCP scenario. Figure 1b plots the fractional change in soil carbon ΔC_{s,τ}/C_{s,0}, showing that there is a large range of effective q_{10} sensitivities between the model projections.
Unfortunately, we do not have timevarying observational datasets of C_{s} and R_{h} that might allow us to directly constrain this projection uncertainty. Instead we explore whether the observed spatial variability in τ_{s} (as shown in Fig. 2c) provides some observational constraint on the sensitivity of τ_{s} to temperature. In doing so, we are motivated by Chadburn et al.^{15} who used the correlation between the observed geographical distributions of permafrost and air temperature to constrain projections of future permafrost area under global warming. Similarly, we use ESMs to test whether the spatial variation in τ_{s} reveals the sensitivity of soil carbon turnover to temperature. The spatial patterns of τ_{s} in CMIP5 simulations and observations were previously shown in Koven et al.^{16}, and here we test whether such relationships can be used to estimate the response of soil carbon to future climate change, using a combination of CMIP6 and CMIP5 models.
Figure 3a is a scatter plot of \(\mathrm{log}\,{\tau }_{{\rm{s}}}\) against temperature, using the τ_{s} values shown in Fig. 2c and mean temperatures from the WFDEI dataset over the period 2001–2010^{24}. The thick blackdotted line is a quadratic fit through these points. Also shown for comparison are equivalent quadratic fits for each model (coloured lines), using the model \(\mathrm{log}\,{\tau }_{{\rm{s}}}\) and mean nearsurface air temperature (T) values for each gridpoint, over an overlapping period with the observations (1995–2005). There is a spread in the individual data points due to variation in soil moisture, soil type, and other soil parameters^{25}. The model specific spread in the data can be seen for the CMIP6 and CMIP5 models in Supplementary Figs. 2 and 3, respectively. Although models do not account for every possible factor contributing to this spread, the spread of points in the models is generally similar to the observations. However, differences between the bestfit functions relating τ_{s} to T are evident between the models, and between the models and the observations^{16}.
This suggests that we may be able to constrain ΔC_{s,τ} using the observed τ_{s} vs. T fit from the observations, but only if we can show that such functions can be used to predict ΔC_{s,τ} under climate change. In order to test that premise, we attempt to reconstruct the timevarying ΔC_{s,τ} projection for each model using the timeinvariant τ_{s} vs. T fit across spatial points (Fig. 3a), and the timeinvariant R_{h,0} field. The change in soil carbon turnover time (Δτ_{s}(t)) for a given model run is estimated at each point basedon the τ_{s} vs. T curve, and the timevarying projection of T at that point. A local estimate of the subsequent change in soil carbon can then be made basedon the farthest righthand term of Eq. (1) (R_{h,0} Δτ_{s}), which can be integrated up to provide an estimated change in global soil carbon in the top 1 metre (ΔC_{s,τ}).
Figure 3b shows the result of this test for all models and all respective SSP/RCP scenarios. The axes of this plot show equivalent variables which represent the global ΔC_{s,τ} between the mean value for 2090–2100 and the mean value for 1995–2005. The yaxis represents the actual values for each model as shown in Fig. 1, and the xaxis represents our estimate derived from spatial variability (as in Fig. 3a). As hoped, actual vs. estimated values cluster tightly around a onetoone line with an r^{2} correlation coefficient value of 0.90. Although some hotclimate regions will inevitably experience temperatures beyond those covered by currentday spatial variability, these tend to be regions with low soil carbon, so this does not have a major impact on the success of our method.
Spatial emergent constraint
This gives us confidence to use the τ_{s} vs. T fit and R_{h,0} from observations to constrain future projections of ΔC_{s,τ}. To remove the uncertainty in future ΔC_{s,τ} due to the climate sensitivity of the models, we investigate a common amount of global mean warming in each model. Figure 4a is similar to Fig. 3b but instead for the more policyrelevant case of 2 °C of global warming. As before, the yaxis represents the modelled ΔC_{s,τ}, and the xaxis is our estimate derived from spatial variability. Once again, the actual and estimated values of ΔC_{s,τ} cluster around the onetoone line (with r^{2} = 0.87). The model range arises partly from differences in the initial field of heterotrophic respiration (R_{h,0}), and partly from differences in Δτ_{s} (compare first row to penultimate row of Table 1).
The vertical green line in Fig. 4a represents the mean estimate when the τ_{s} vs. T relationship and the R_{h,0} field from the model are replaced with the equivalents from the observations. The spread shown by the shaded area represents the relatively small impact on ΔC_{s,τ} of differences in modelled spatial climate change patterns at 2 °C of global warming. In order to estimate the remaining uncertainty in ΔC_{s,τ}, we treat this spread as equivalent to an observational uncertainty in an emergent constraint approach^{26}. We apply a standard statistical approach^{27,28} to estimate the probability density function of the yaxis variable (model ΔC_{s,τ}), accounting for both this observational spread and the quality of the emergent relationship. To test the robustness to the choice of observations we have repeated the analysis with different datasets that represent heterotrophic respiration, which produces stronglyoverlapping emergent constraints, and completing the analysis with both CMIP6 and CMIP5 models shows that the result is also robust to the choice of model ensemble (see Table 1).
Figure 4b shows the resulting emergent constraint (blue line), and compares to the unweighted histogram of model values (grey blocks), and a Gaussian fit to that prior distribution (black line). The spatial emergent constraint reduces the uncertainty in ΔC_{s,τ} at 2 °C of global warming from −196 ± 117 PgC to −232 ± 52 PgC (where these are mean values plus and minus one standard deviation for the top 1 metre). This same method can be applied to find constrained values of ΔC_{s,τ} for other values of global warming. Figure 4c shows the constrained range of ΔC_{s,τ} as a function of global warming. This rules out the most extreme projections but nonetheless suggests substantial soil carbon losses due to climate change even in the absence of losses of deeper permafrost carbon.
Methods
Obtaining spatial relationships
In this section we explain how the quadratic relationships representing the spatial \(\mathrm{log}\,{\tau }_{{\rm{s}}}\)temperature sensitivity shown in Fig. 3a (and Supplementary Figs. 2, 3 and 6) were derived, for both the Earth System Models (ESMs) in CMIP6 and CMIP5, and using the observational data. This is similar to the method used in Koven et al.^{16}.
Obtaining spatial relationships for CMIP models
The CMIP6 models used in this study are shown in the Table 2, and the CMIP5 models used in this study are shown in Table 3.
To obtain model specific spatial \(\mathrm{log}\,{\tau }_{{\rm{s}}}\)temperature relationships, the following method was used. A reference time period was considered (1995–2005), this was taken as the end of the CMIP5 historical simulation to be consistent across CMIP generations and to best match the observational data time frame considered. Then, monthly model output data was time averaged over this period, for the output variables ‘soil carbon content’ (C_{s}) in kg m^{−2}, ‘heterotrophic respiration carbon flux’ (R_{h}) in kg m^{−2}s^{−1}, and ‘air temperature’ in K. The variables C_{s} and R_{h} were used to obtain values for soil carbon turnover time (τ_{s}) in years, using the equation τ_{s} = C_{s}/(R_{h} × 86400 × 365). The model temperature variable units were converted from K to °C.
For each model, these values of \(\mathrm{log}\,{\tau }_{{\rm{s}}}\) were plotted against the corresponding spatial temperature data to obtain the spatial \(\mathrm{log}\,{\tau }_{{\rm{s}}}\)temperature plot. Then, quadratic fits (using the python package numpy polyfit) are calculated for each model, which represent the spatial \(\mathrm{log}\,{\tau }_{{\rm{s}}}\) relationship and sensitivity to temperature. These model specific relationships are shown by the coloured lines in Fig. 3a in the main manuscript, and in Supplementary Fig. 2 for CMIP6 and in Supplementary Fig. 3 for CMIP5.
Obtaining spatial relationships for observations
Following Koven et al.^{16}, we estimated observational soil carbon data (to a depth of 1 m) by combining the Harmonized World Soils Database (HWSD)^{19} and Northern Circumpolar Soil Carbon Database (NCSCD)^{20} soil carbon datasets, where NCSCD was used where overlap occurs. To calculate soil carbon turnover time, τ_{s}, using the following equation: τ_{s} = C_{s}/R_{h}, we require a global observational dataset for heterotrophic respiration. In the main manuscript, CARDAMOM (2001–2010) heterotrophic respiration (R_{h}) is used^{21}. We completed a sensitivity study on the choice of observational heterotrophic respiration dataset, see below. The WFDEI dataset is used for our observational air temperatures (2001–2010)^{24}. Then, these datasets can be used to obtain the observational \(\mathrm{log}\,{\tau }_{{\rm{s}}}\)temperature relationship, using the same quadratic fitting as with the models. This represents the ‘real world’ spatial temperature sensitivity of \(\mathrm{log}\,{\tau }_{{\rm{s}}}\), and is shown by the thickdottedblack line in Fig. 3a of the main manuscript. A comparison of the derived observational relationships can be seen in Supplementary Fig. 6.
Observational sensitivity study
We completed a sensitivity study to investigate our constraint dependence on the choice of observational heterotrophic respiration dataset (CARDAMOM (2001–2010)^{21}). The other observational datasets considered are as follows: NDP08 ‘Interannual Variability in Global Soil Respiration on a 0.5 Degree Grid Cell Basis’ dataset (1980–1994)^{29}, ‘Global spatiotemporal distribution of soil respiration modelled using a global database’^{30}, and MODIS net primary productivity (NPP) (2000–2014)^{31}. Supplementary Fig. 4 shows scatter plots showing onetoone comparisons of these observational datasets against one another, and Supplementary Fig. 5 shows the corresponding comparisons of the equivalent \(\mathrm{log}\,{\tau }_{{\rm{s}}}\) values calculated from each dataset.
The CARDAMOM R_{h} dataset is used in the main manuscript for the following two main reasons: firstly, we calculate τ_{s} using heterotrophic respiration which allows for consistency between models and observations, and secondly, the dataset does not use a prescribed q_{10} sensitivity^{21}. Instead, the CARDAMOM R_{h} dataset was derived by explicitly assimilating observations into a processbased diagnostic landsurface model. To test the robustness of our results, we also repeated our analysis with MODIS NPP and Raich 2002, for both CMIP6 and CMIP5 together, and as separate model ensembles. Supplementary Fig. 6 shows the observational \(\mathrm{log}\,{\tau }_{{\rm{s}}}\)temperature relationships, derived using each of these observational datasets. The results are presented in Table 1 which shows the constrained values of ΔC_{s,τ} at 2 °C global mean warming.
We decided not to complete the paper analysis using the Hashimoto dataset since not only is it inconsistent with the three other datasets considered, it also shows an arbitrary maximum respiration level (Supplementary Fig. 4), which likely results from the assumed temperaturedependence of soil respiration in this dataset which takes a quadratic form^{30}. The quadratic form is justified based on a sitelevel study in which it is used to fit temporal dynamics. However, the parameters for the quadratic function that are fitted in the Hashimoto study are very different from those in the sitelevel study, which therefore suggests that the same relationship does not apply to the global distribution of mean annual soil respiration.
Equation for the soil carbon turnover time component of soil carbon change
The equation used in this study for the component of the change in soil carbon (ΔC_{s}) due to the change in soil carbon turnover time (Δτ_{s}) was derived in the following way. Starting with the equation for soil carbon (based on the definition of τ_{s}):
As discussed in the main manuscript, we can write this change in soil carbon (ΔC_{s}), as the sum of a term arising from changes in litterfall (ΔC_{s,L}), and a term arising from changes in the turnover time of soil carbon (ΔC_{s,τ}):
Hence, the equation for the component of soil carbon change due to the change in τ_{s} is:
In this study we use R_{h} from the reference period (‘present day’), which we call R_{h,0}, to allow us to investigate the response of ΔC_{s,τ} as a result of the response of τ_{s} to climate change.
Modelled future temperature
The proof of principle figure (Fig. 3b) considers ΔC_{s,τ} between the end of the 21st century (2090–2100), for each future SSP scenario (SSP126, SSP245, SSP585)^{22} or equivalent future RCP scenario (RCP2.6, RCP4.5 and RCP8.5)^{23}, and our reference period from the historical simulation (1995–2005), for each CMIP6 ESM and CMIP5 ESM, respectively.
To consider specific °C of global warming (Fig. 4), the future spatial temperature profiles at these specific global mean warming levels, for example: 1 °C, 2 °C and 3 °C global mean warming, were calculated as follows. The temperature change is calculated from our reference period (1995–2005), and then a 5year rolling mean of global mean temperature is taken to remove some of the interannual variability. Once the year that the given temperature increase has been reached is obtained, a time average including −5 and +5 years is taken, and the spatial temperature distribution of that model averaged over the deduced time period is used for the calculations of future τ_{s}.
Anomaly correction for future temperature projections
To remove uncertainty due to errors in the models’ historical simulation, a spatial future temperature anomaly was projected using each model and each respective future SSP/RCP scenario separately. To calculate this, the temperature at the reference time frame (1995–2005), which overlaps the WFDEI observational temperature data time frame (2001–2010), is subtracted from the future temperature profile for each model (as calculated above), to calculate the temperature change. Then, this temperature anomaly is added onto the observational temperature dataset to give a modelderived future ‘observational’ temperature for each model.
Proof of concept for our method
Our method relies on the idea that the spatial temperature sensitivity can be used to project and constrain the temporal sensitivity of τ_{s} to temperature, and subsequently global warming. To test the robustness of this method, ΔC_{s,τ} calculated using model Δτ_{s}, and temperature sensitivity relationshipderived Δτ_{s}, are compared.
The change in soil carbon turnover time (Δτ_{s}) was either calculated using model output data to obtain modelderived Δτ_{s} as follows:
where,
Or calculated using the derived quadratic \(\mathrm{log}\,{\tau }_{{\rm{s}}}\)temperature relationships to obtain relationshipderived Δτ_{s}, which is based on the following equation:
where, T is near surface air temperature, and T^{f} represents a future temperature, and T^{h} represents historical (present day) temperature from our reference period (1995–2005). The exponentials (\(\exp\)) are taken to turn \(\mathrm{log}\,{\tau }_{{\rm{s}}}\) values to τ_{s} values. p(T) represents the quadratic \(\mathrm{log}\,{\tau }_{{\rm{s}}}\)temperature relationship as a function of temperature to obtain our estimated \(\mathrm{log}\,{\tau }_{{\rm{s}}}\).
These Δτ_{s} values are then put back into the Eq. (4) (with modelspecific R_{h,0}) to obtain the corresponding ΔC_{s,τ} values. The proof of principle figure (Fig. 3b) investigates the robustness of our method, where projections of model and relationshipderived values of ΔC_{s,τ} are compared, and an r^{2} value of 0.90 is obtained. The correlation of the data was also tested when investigating different levels of global mean warming to obtain the constrained values (Fig. 4). The r^{2} values for were as follows: 1 °C is 0.84, 2 °C is 0.87 and 3 °C is 0.87.
Calculating constrained values
To obtain the constrained values of ΔC_{s,τ}, the modelderived future ‘observational’ temperature for each model is used together with the observational derived \(\mathrm{log}\,{\tau }_{{\rm{s}}}\)temperature relationship, to project values for future τ_{s}. Then this together with relationshipderived historical τ_{s} deduced using the observational temperature dataset, can be used to calculate Δτ_{s}. Finally global ΔC_{s,τ} can be obtained by multiplying Δτ_{s} by the observational dataset for R_{h,0} (using Eq. (4)), and then calculating a weightedglobal total. As each modelderived future ‘observational’ temperature is considered separately, we obtain a range of projected observationalconstrained ΔC_{s,τ} values.
We have now obtained a set of x and y values, corresponding to the relationshipderived and modelled values of ΔC_{s,τ}, respectively, for each ESM. Where we have an x and y value for each model, representing the modelled ΔC_{s,τ} (y values), and the model specific relationshipderived ΔC_{s,τ} (x values). We also have an x_{obs} value representing the mean observationalconstrained ΔC_{s,τ} value, and a corresponding standard deviation due to the uncertainty in the modelled spatial profiles of future temperatures. We follow the method used in Cox et al. 2018, which can be seen in the ‘Leastsquares linear regression’ section and the ‘Calculation of the PDF for ECS’ section of the methods from this study^{32}. Using this method, we obtain an emergent relationship between our x and y data points, which we can use together with our x_{obs} and corresponding standard deviation to produce a constraint on our yaxis. This is shown in Fig. 4a. From this we obtain a constrained probability density function on ΔC_{s,τ}, with a corresponding uncertainty bounds which we consider at the 68% confidence limits (±1 standard deviation). Figure 4b show the probability density functions representing the distribution of the range of projections, before and after the constraint.
This method allows us to calculate a constrained probability density function on ΔC_{s,τ} at each °C of global mean warming, using the data seen in Fig. 4a for 2 °C warming, and our corresponding constrained values for 1 °C and 3 °C warming. Figure 4c shows the resultant constrained mean value of ΔC_{s,τ} obtained for each °C of global mean warming, and the corresponding uncertainty bounds at the 68% confidence limits (±1 standard deviation).
Calculating effective q_{10} for change in soil carbon
Simple models of soil carbon turnover are often based on just a q_{10} function, which means that τ_{s} depends on temperature as follows:
We compared the results for ΔC_{s,τ} that would be derived from a simple q_{10} function with our emergent constraint results for ΔC_{s,τ}, to estimate an effective q_{10} sensitivity of heterotrophic respiration.
To do this, we can obtain an equation for Δτ_{s} derived from Eq. (8). This is done by considering the following, where τ_{s,0} is an initial τ_{s}, we can substitute in τ_{s} in temperature sensitivity form to obtain an equation for Δτ_{s} in temperature sensitivity form:
Then, we can substitute this Δτ_{s} into Eq. (4) and simplify to obtain an equation relating ΔC_{s,τ} and ΔT:
This equation was used to calculate different ΔC_{s,τ}–ΔT sensitivity curves based on different values on q_{10}, for example q_{10} = 2, with different amounts of global mean warming to represent ΔT, and initial observational soil carbon stocks C_{s,0}. These curves can be seen on Figs. 1b and 4c. Note that there is no direct relationship between the effective q_{10} for soil carbon change shown in Figs. 1b and 4c, and the spatial τ_{s}–T relationships in Fig. 3a. Our q_{10} value is an effective q_{10} value that indicates the sensitivity of global soil carbon (in the top 1 metre) to global mean temperature.
Data availability
The datasets analysed during this study are available online: CMIP5 model output [https://esgfnode.llnl.gov/search/CMIP5/], CMIP6 model output [https://esgfnode.llnl.gov/search/cmip6/], The WFDEI Meteorological Forcing Data [https://rda.ucar.edu/datasets/ds314.2/], CARDAMOM Heterotrophic Respiration [https://datashare.is.ed.ac.uk/handle/10283/875], MODIS Net Primary Production [https://lpdaac.usgs.gov/products/mod17a3v055/], Raich et al. 2002 Soil Respiration [https://cdiac.essdive.lbl.gov/epubs/ndp/ndp081/ndp081.html], Hashimoto et al. 2015 Heterotrophic Respiration [http://cse.ffpri.affrc.go.jp/shojih/data/index.html], and the datasets for observational Soil Carbon [https://github.com/rebeccamayvarney/soiltau_ec].
Code availability
The Python code used to complete the analysis and produce the figures in this study is available in the following online repository [https://github.com/rebeccamayvarney/soiltau_ec].
References
Ciais, P. et al. Carbon and other biogeochemical cycles. In Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Book section 6, (eds. Stocker, T. et al.) 465–570 (Cambridge University Press, Cambridge, New York, 2013).
UNFCCC. Adoption of the Paris Agreement. fccc/cp/2015/l.9/rev.1. http://unfccc.int/resource/docs/2015/cop21/eng/l09r01.pdf (2015).
Jenkinson, D. S., Adams, D. & Wild, A. Model estimates of CO_{2} emissions from soil in response to global warming. Nature 351, 304 (1991).
Cox, P. M., Betts, R. A., Jones, C. D., Spall, S. A. & Totterdell, I. J. Acceleration of global warming due to carboncycle feedbacks in a coupled climate model. Nature 408, 184 (2000).
Bradford, M. A. et al. Managing uncertainty in soil carbon feedbacks to climate change. Nat. Clim. Change 6, 751 (2016).
ToddBrown, K. et al. Changes in soil organic carbon storage predicted by earth system models during the 21st century. Biogeosciences 11, 2341–2356 (2014).
Arora, V. K. et al. Carbon–concentration and carbon–climate feedbacks in CMIP5 earth system models. J. Clim. 26, 5289–5314 (2013).
Jones, C. et al. Twentyfirstcentury compatible CO_{2} emissions and airborne fraction simulated by CMIP5 earth system models under four representative concentration pathways. J. Clim. 26, 4398–4413 (2013).
Knorr, W., Prentice, I. C., House, J. & Holland, E. Longterm sensitivity of soil carbon turnover to warming. Nature 433, 298 (2005).
ToddBrown, K. et al. Causes of variation in soil carbon simulations from CMIP5 earth system models and comparison with observations. Biogeosciences 10, 1717–1736 (2013).
Eyring, V. et al. Overview of the coupled model intercomparison project phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev. 9, 1937–1958 (2016).
Taylor, K. E., Stouffer, R. J. & Meehl, G. A. An overview of CMIP5 and the experiment design. Bull. Am. Meteorol. Soc. 93, 485–498 (2012).
Koven, C. D. et al. Controls on terrestrial carbon feedbacks by productivity versus turnover in the CMIP5 earth system models. Biogeosciences 12, 5211–5228 (2015).
Hall, A., Cox, P., Huntingford, C. & Klein, S. Progressing emergent constraints on future climate change. Nat. Clim. Change 9, 269–278 (2019).
Chadburn, S. et al. An observationbased constraint on permafrost loss as a function of global warming. Nat. Clim. Change 7, 340 (2017).
Koven, C. D., Hugelius, G., Lawrence, D. M. & Wieder, W. R. Higher climatological temperature sensitivity of soil carbon in cold than warm climates. Nat. Clim. Change 7, 817 (2017).
Carvalhais, N. et al. Global covariation of carbon turnover times with climate in terrestrial ecosystems. Nature 514, 213 (2014).
Davidson, E. A. & Janssens, I. A. Temperature sensitivity of soil carbon decomposition and feedbacks to climate change. Nature 440, 165 (2006).
FAO, I. & ISRIC, I. Jrc: Harmonized world soil database (version 1.2). FAO, Rome, Italy and IIASA, (Laxenburg, Austria, 2012).
Hugelius, G. et al. The northern circumpolar soil carbon database: spatially distributed datasets of soil coverage and soil carbon storage in the northern permafrost regions. Earth System Science Data 5, 3–13 (2013).
Bloom, A. et al. CARDAMOM 20012010 global carbon modeldata fusion (mdf) analysis. (2015).
O’Neill, B. C. et al. The scenario model intercomparison project (scenariomip) for CMIP6 (2016).
Meinshausen, M. et al. The rcp greenhouse gas concentrations and their extensions from 1765 to 2300. Clim. Change 109, 213 (2011).
Weedon, G. P. et al. The wfdei meteorological forcing data set: watch forcing data methodology applied to erainterim reanalysis data. Water Res. Res. 50, 7505–7514 (2014).
Exbrayat, J.F., Pitman, A. J., Zhang, Q., Abramowitz, G. & Wang, Y.P. Examining soil carbon uncertainty in a global model: response of microbial decomposition to temperature, moisture and nutrient limitation. Biogeosciences 10, 7095–7108 (2013).
Eyring, V. et al. Taking climate model evaluation to the next level. Nat. Clim. Change 9, 102–110 (2019).
Cox, P. M. et al. Sensitivity of tropical carbon to climate change constrained by carbon dioxide variability. Nature 494, 341 (2013).
Wenzel, S., Cox, P. M., Eyring, V. & Friedlingstein, P. Projected land photosynthesis constrained by changes in the seasonal cycle of atmospheric co 2. Nature 538, 499 (2016).
Raich, J. W., Potter, C. S. & Bhagawati, D. Interannual variability in global soil respiration, 1980–94. Global Change Biol. 8, 800–812 (2002).
Hashimoto, S. et al. Global spatiotemporal distribution of soil respiration modeled using a global database. Biogeosci. 12, 4121–4132 (2015).
Zhao, M., Heinsch, F. A., Nemani, R. R. & Running, S. W. Improvements of the modis terrestrial gross and net primary production global data set. Remote Sens. Environ. 95, 164–176 (2005).
Cox, P. M., Huntingford, C. & Williamson, M. S. Emergent constraint on equilibrium climate sensitivity from global temperature variability. Nature 553, 319 (2018).
Acknowledgements
This work was supported by the European Research Council ‘Emergent Constraints on ClimateLand feedbacks in the Earth System (ECCLES)’ project, grant agreement number 742472 (R.M.V. and P.M.C.). S.E.C. was supported by a Natural Environment Research Council independent research fellowship, grant no. NE/R015791/1. P.F. was supported by the European Union’s Framework Programme Horizon 2020 for Research and Innovation under Grant Agreement No. 821003, ClimateCarbon Interactions in the Current Century (4C) project. E.J.B. was funded by the European Commission’s Horizon 2020 Framework Programme, under Grant Agreement number 641816, the ‘Coordinated Research in Earth Systems and Climate: Experiments, Knowledge, Dissemination and Outreach (CRESCENDO)’ project (11/2015–10/2020) and the Met Office Hadley Centre Climate Programme funded by BEIS and Defra. C.D.K. acknowledges support from the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, through the RUBISCO SFA and the Early Career Research Program. G.H. was supported by the Swedish Research Council (201804516) and the EU H2020 project Nunataryuk (773421). We also acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modelling groups (listed in Tables 2 and 3 in the “Methods” section) for producing and making available their model output.
Author information
Authors and Affiliations
Contributions
R.M.V., S.E.C., and P.M.C. designed the study and drafted the manuscript, and R.M.V. did the analysis and produced the figures. P.F. provided insightful guidance on the best way to derive and represent soil carbon turnover. C.D.K. provided the initial code and gave advice on the differing representation of soil carbon in the CMIP6 model ensemble. E.J.B. provided observational datasets for: heterotrophic respiration, soil respiration, net primary production and temperature. G.H. provided observational soil carbon data. All coauthors provided guidance on the study at various times and suggested edits to the draft manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Peer review information Nature Communications thanks the anonymous reviewers for their contributions to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Varney, R.M., Chadburn, S.E., Friedlingstein, P. et al. A spatial emergent constraint on the sensitivity of soil carbon turnover to global warming. Nat Commun 11, 5544 (2020). https://doi.org/10.1038/s41467020192088
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467020192088
This article is cited by

Projected soil carbon loss with warming in constrained Earth system models
Nature Communications (2024)

Emergent constraints on carbon budgets as a function of global warming
Nature Communications (2024)

Biophysical Controls on Soil Carbon Cycling in a Northern Hardwood Forest
Ecosystems (2024)

The legacy of one hundred years of climate change for organic carbon stocks in global agricultural topsoils
Scientific Reports (2023)

Divergent datadriven estimates of global soil respiration
Communications Earth & Environment (2023)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.