Beillouin et al. (2023). A global meta-analysis of soil organic carbon in the Anthropocene

Article

https://doi.org/10.1038/s41467-023-39338-z

A global meta-analysis of soil organic carbon

in the Anthropocene

Damien Beillouin

1,2

, Marc Corbeels

3,4

, Julien Demenois

3,5,6

, David Berre

3,7,8

,

Annie Boyer

9

, Abigail Fallot

10,11

, Frédéric Feder

12,13

& Rémi Cardinael

3,14,15

Anthropogenic activities profoundly impact soil organic carbon (SOC),
affecting its contribution to ecosystem services such as climate regulation.
Here, we conducted a thorough review of the impacts of land-use change, land
management, and climate change on SOC. Using second-order meta-analysis,
we synthesized

ﬁ

ndings from 230

ﬁ

rst-order meta-analyses comprising over

25,000 primary studies. We show that (i) land conversion for crop production
leads to high SOC loss, that can be partially restored through land manage-
ment practices, particularly by introducing trees and incorporating exogenous
carbon in the form of biochar or organic amendments, (ii) land management
practices that are implemented in forests generally result in depletion of SOC,
and (iii) indirect effects of climate change, such as through wild

ﬁ

res, have a

greater impact on SOC than direct climate change effects (e.g., from rising
temperatures). The

ﬁ

ndings of our study provide strong evidence to assist

decision-makers in safeguarding SOC stocks and promoting land management
practices for SOC restoration. Furthermore, they serve as a crucial research
roadmap, identifying areas that require attention to

ﬁ

ll the knowledge gaps

concerning the factors driving changes in SOC.

Soil organic matter (SOM), mainly composed of carbon, is a critical
component of soils

1

. It plays a major role in regulating soil health

2

,

3

and

other ecosystem services such as biodiversity conservation and food
production

4

. Moreover, it is a key contributor to the global carbon

cycle and climate regulation

5

. The global soil organic carbon (SOC)

stocks to 2 m of soil depth are estimated at approximately 2400 Gt C

6

,

which is three times the amount of carbon in the atmosphere

7

. Small

changes in SOC stocks can, therefore, signi

ﬁ

cantly impact atmospheric

carbon dioxide (CO

2

) levels and climate change

Achieving global net zero CO

2

emissions by 2050 is crucial for

limiting global warming to 1.5 °C by the end of the century

10

. This

requires avoiding, reducing, and offsetting greenhouse gas (GHG)
emissions in all sectors, including the agriculture, forestry, and other

land uses (AFOLU) sector

11

. For instance, it is critical to preserve

carbon-rich ecosystems like peatlands, old-growth forests, wetlands,
and mangroves, which together hold at least 260 Gt of

‘

irrecoverable

’

carbon

12

. On the other hand, negative emission technologies can offset

excess GHG emissions

10

, and natural climate solutions such as SOC

restoration can play a crucial role in this process, while also offering
additional bene

ﬁ

ts such as biodiversity conservation

13

–

15

. SOC pre-

servation and restoration alone can contribute up to 25% of the
potential of natural climate solutions, with 40% of this potential
coming from SOC preservation and 60% from the restoration of
depleted SOC stocks

16

.

Numerous factors, hereinafter referred to as

‘

drivers

’

, directly or

indirectly impact SOC levels. Land-use change is a major SOC driver at

Received: 4 November 2022

Accepted: 5 June 2023

Check for updates

1

CIRAD, UPR HortSys, Montpellier, France.

2

HortSys, Univ Montpellier, CIRAD, Montpellier, France.

3

AIDA, Univ Montpellier, CIRAD, Montpellier, France.

4

IITA,

Nairobi, Kenya.

5

CIRAD, UPR AIDA, Turrialba, Costa Rica.

6

CATIE, Centro Agronómico Tropical de Investigación y Enseñanza, Turrialba, Costa Rica.

7

CIRAD,

UPR AIDA, Bobo-Dioulasso, Burkina Faso.

8

CIRDES, USPAE, Bobo-Dioulasso, Burkina Faso.

9

CIRAD, DGDRS, DIST, Montpellier, France.

10

CIRAD, UMR SENS,

Montpellier, France.

11

SENS, Univ Montpellier, CIRAD, Montpellier, France.

12

CIRAD, UPR Recyclage et Risque, Montpellier, France.

13

Recyclage et Risque, Univ

Montpellier, CIRAD, Montpellier, France.

14

CIRAD, UPR AIDA, Harare, Zimbabwe.

15

Department of Plant Production Sciences and Technologies, University of

Zimbabwe, Harare, Zimbabwe.

e-mail:

damien.beillouin@cirad.fr

Nature Communications

| (2023) 14:3700

1

12

3

4

5

6

78

9

0

()

:,;

12

3

4

5

6

7

8

9

0

()

:,;

the global scale

17

,

18

. According to Winkler et al. (2021)

19

, almost a third

of the global land area has undergone land-use change in the last six
decades (1960

–

2019). The conversion of natural ecosystems to agri-

cultural land is estimated to have resulted in a carbon debt of 116 Gt in
the top 2 m soil layer

19

. For example, the conversion of primary forest

to cropland in the tropics caused a 25

–

30% loss of SOC stocks

20

.

Afforestation of cropland can partially restore these stocks

21

, yet SOC

restoration is generally slower than depletion. Land management is
another crucial driver of SOC change, and several agricultural practices
such as manure application

, no-till farming

, cover cropping

and

agroforestry

26

,

27

have been proposed to increase SOC stocks

28

. Finally,

climate change can also have a signi

ﬁ

cant negative impact on global

SOC stocks, by increasing SOC mineralization due to higher
temperatures

29

, or by decreasing carbon inputs to the soil as a result of

less favorable plant growth conditions linked to more variable and
extreme weather events

30

. Greater efforts in land-use and land man-

agement that turn soils into future carbon sinks are therefore
required

31

.

Thousands of experiments have investigated the impact of drivers

of SOC change, and the

ﬁ

ndings are being consolidated in a growing

number of meta-analyses

32

,

33

. Yet a comprehensive understanding of

the global effects of land-use change, land management, and climate
change on SOC is still lacking. Each of these

ﬁ

rst-order meta-analyses is

restricted in its scope and often focuses on a limited number of land
use interventions or geographical regions

34

. Furthermore, their results

can be highly variable and sometimes contradictory, which to a certain
extent is related to methodological issues and the number of experi-
ments synthesized. A second-order meta-analysis approach combines
and synthesizes results from multiple

ﬁ

rst-order meta-analyses in a

systematic and quantitative way, and enhances the statistical power of
the analysis by increasing sample size. It also allows for a more rigorous
critical analysis, if the quality of the

ﬁ

rst-order meta-analyses is con-

sidered. Second-order meta-analysis methods have seldom been
applied to drivers of SOC change (but see Bolinder et al. 2020

35

, Young

et al. 2021

36

, and Lessmann et al. 2020

37

on the effects of some speci

ﬁ

c

land-use interventions on SOC). A comprehensive synthesis of results
of previous meta-analyses on SOC can facilitate evidence-based

decision-making and priority-setting aimed at increasing SOC on a
global scale

38

,

39

. Combined with local knowledge, it can also contribute

to identifying the best land-use and land management practices for
SOC preservation and restoration at local and regional scales.

Here, we conducted a second-order meta-analysis that included

over 220 speci

ﬁ

c drivers of SOC change, representing the direct and

indirect effects of human interventions including land-use change
(e.g., conversion of forest to cropland), land management (e.g.,
mineral fertilization), and climate change (e.g., warming). Our analysis
combined 230

ﬁ

rst-order meta-analyses (Fig. S1), that synthesized the

results of more than 25,000 primary studies and 190,200 paired
comparison data. In order to determine the most appropriate meta-
analytical models, we examined the random structure of the models,
the inclusion of quality scores, and the redundancy of primary studies
(

see Methods

). Additionally, we compared the estimates obtained

through frequentist versus Bayesian inference methods to ensure the
robustness of our conclusions (Fig. S2). We found no evidence of
publication bias, as indicated by the Egger

’

s test and the trim-and-

ﬁ

ll

method (Figs. S3-5). To align with the metrics used in the retrieved

ﬁ

rst-order meta-analyses, our study examined SOC expressed as either

a stock (Mg C ha

−

1

) or as a concentration (g C kg

−

1

soil) (Fig. S6).

Results and discussion

Large but variable impacts of land-use change and land
management, uncertain impacts of climate change

Our results revealed that the overall effects of land-use change and land
management on SOC were 7

–

10 times larger than the direct effects of

climate change (i.e., excluding the indirect effects of wild

ﬁ

re and snow

cover change, Fig.

1

). Both negative and positive effects of land-use

change and land management practices were found, thereby high-
lighting the opportunities of increasing SOC but also the risks of its
depletion.

Of the 60 types of land-use change analyzed, 25% presented a

decrease in SOC that was greater than 24%, while another 25% showed
an increase that was higher than 15%. Among the 143 identi

ﬁ

ed land

management practices, 25% displayed a decrease in SOC that was
greater than 1%, while another 25% presented an increase that was

Croplands (n= 34530)

Forest lands (n= 11988)

Graslands (n=9131)

Other lands (n=4681)

Wetlands (n=1267)

Land-use :

Land-use change

(n= 10139)

Land management

(n=40729)

Climate change

(n=1800)

Soil organic carbon (SOC) change (%)

-50

0

50

Fig. 1 | The distribution of changes in soil organic carbon (SOC) resulting from
land-use change, land management, and climate change.

Each dot represents a

mean effect size in terms of SOC change for a sub-category of the three main drivers
of SOC change per land-use type. Croplands are shown as orange dots, forest lands
as dark green dots, grasslands as light green dots, wetlands as blue dots, and other
lands as gray dots. Dots of sub-categories of land-use change are colored according

to the initial (i.e., previous) land-use type. The dot sizes are proportional to the
number of paired data used to calculate the mean effect sizes. Violin plots repre-
sent the distribution of values within each of the three main categories, with the 25,
50, and 75% quantiles denoted by vertical black bars. The number of paired data (n)
for each land-use type is shown in the bubbles on the right side of the plot. An
interactive version of the plot is available at

https://rpubs.com/dbeillouin/Figure1

.

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

2

higher than 23%. Thus, large but at the same time highly variable
impacts can be expected, with some of the considered land-use
changes or land management practices proving to be highly effective
in increasing SOC, while others are not.

On the other hand, the direct effects of climate change (i.e.,

warming, drought, and CO

2

enrichment) were relatively small, as

indicated by the 25

th

and 75

th

percentiles of respectively

−

2.4% and

+4.0% SOC change. The largest climate-related SOC changes were
associated with the indirect effects of climate change (e.g., wild

ﬁ

res,

declining snow cover) (Fig.

1

).

The effect of land management practices varied markedly

according to the land-use type: 83% of the cropland management
practices analyzed resulted in a signi

ﬁ

cant increase in SOC, and 70% of

all land management practices that led to a signi

ﬁ

cant gain in SOC

occurred in croplands (Fig.

2

). On the other hand, in almost half of the

cases where management practices were applied to forests, there was a
signi

ﬁ

cant decrease in SOC. Importantly, wetlands showed the largest

decrease in SOC across all land-use types in response to management
interventions (i.e., up to

−

60% for practices that lead to wetland

degradation).

Furthermore, our results con

ﬁ

rmed the high potential to rebuild

SOC in croplands

40

, which is, however, largely associated with the

generally low initial SOC levels observed in croplands

17

,

41

. On the other

hand, in the case of forest lands, the challenge is to maintain SOC levels
by avoiding forest conversion and degradation

13

; few solutions cur-

rently exist to increase SOC in forest lands. Finally, the number of
studies for a given land-use or land management intervention may not
necessarily re

ﬂ

ect the importance for SOC preservation. For example,

there are relatively few studies on wetlands despite their crucial role as
a major storehouse of carbon for climate change mitigation

16

.

The importance of carbon-rich ecosystems

Across all studies, the conversion of forest lands, grasslands, and
wetlands to croplands consistently resulted in large SOC loss, with
a mean change of

−

25%, con

ﬁ

dence intervals (CI): [

−

34,

−

16]),

−

16%, CI

[

−

30, 1.4], and

−

25% CI [

−

32,

−

17], respectively (Fig.

2

). These large

losses of SOC, combined with the extensive areas converted to crop-
lands observed in the last decades (~ 1.0 million km

2

over the last 60

years

19

), have thus substantially contributed to the atmospheric CO

2

increase

2

.

Our

ﬁ

gures may underestimate the actual SOC losses from the

conversion of forest lands because most of the underlying primary
studies

—

and thus most of the resulting meta-analyses

—

quanti

ﬁ

ed

these losses within a time frame ranging from a few years to a max-
imum of a few decades following conversion, whereas the time to
reach a new SOC equilibrium after a land-use change is much longer
(e.g., estimated to be about 80 years for the conversion of grassland to
cropland

42

). This underlines the importance of long-term experiments

and maintaining ongoing monitoring of SOC

43

. Natural ecosystems are

also known to contain more stable soil carbon stores than
agroecosystems

44

, and their preservation is therefore crucial for near-

term climate change mitigation

45

.

The degree of SOC loss following land conversion to croplands

varied depending on the type of cropping system established, with
lower SOC losses observed after forest conversion to croplands culti-
vated under agroforestry practices (

−

12%, CI [

−

19,

−

4.9]), or with per-

ennial crops (

−

7.2%, CI [

−

18, +4.3]), compared to those cultivated with

annual crops (

−

32%, CI [

−

38,

−

24]). Similar results were found for the

conversion of grasslands to croplands, where SOC losses were higher
for croplands cultivated with only annual crops (

−

19%, CI [

−

27,

−

10]),

compared to agroforestry or the inclusion of perennial crops (+1.7%, CI

Fig. 2 | Percentage change in soil organic carbon (SOC) due to land-use change.

The arrows represent the effect of a land-use change on SOC, with the

ﬁ

nal land-use

on the arrowhead side. The arrow sizes are proportional to the magnitude of the
SOC change in reference to the initial land-use, with negative effects highlighted in
orange and positive effects highlighted in green. The mean SOC changes are noted
alongside the arrows. An asterisk indicates a signi

ﬁ

cant effect. The numbers in the

disks around each land-use type represent the number of data pairs used to

calculate the SOC change. The sizes of land-use pictograms are proportional to
their respective mean SOC stock values (t ha

−

1

), as provided by FAO and ITPS

(2020)

83

. Given that the mean initial SOC levels vary across different land-use types,

the effect sizes (and the sizes of the arrows) expressed as percent change should be
interpreted in relation to these initial levels. Details by sub-category of land-use
type are available at

https://rpubs.com/dbeillouin/Figure2

.

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

3

[

−

5.9, +9,9] and

−

4.7%, CI [

−

9.2, +0,1], respectively). Introducing per-

ennial crop species in croplands appears to be a promising approach
for mitigating the negative effects of cropland expansion on SOC.
Perennial crops are known to intercept higher amounts of solar
radiation throughout the growing season and produce more biomass
compared to annual crops

45

, as well as allocate more resources to

belowground plant parts and have permanent and deeper root
systems

46

. These characteristics make them more conducive to

increase SOC compared to annual crops. In addition to their bene

ﬁ

ts

for SOC accumulation, perennial crops are recognized for providing a
range of other ecosystem services

47

. Yet, both the extent of SOC loss

following the conversion of forests or grasslands to croplands and the
potential mitigation effect of perennial crops depend to a large extent
on local pedoclimatic conditions

48

,

49

.

The conversion of croplands to forests or grasslands resulted in

signi

ﬁ

cant SOC gains of +57% (CI [ + 30, +90] and +26% (CI [ + 8,2, +47],

respectively (Fig.

2

). This potential for SOC increase is particularly

pronounced for degraded croplands (with low SOC levels), and gen-
erally in tropical regions

50

. However, it is widely acknowledged that

ecosystem restoration often fails to fully recover the functions of the
undisturbed ecosystems

51

,

52

, including soil carbon sequestration

53

. It

should be noted that a high percentage increase in SOC following the
conversion of croplands to grasslands or to forests does not neces-
sarily indicate that the SOC levels in natural grassland or forestland can
be easily achieved. The high levels of SOC increase should be inter-
preted with caution, as the initial levels of SOC in croplands are
generally low.

Finally, the conversion of forest lands to grasslands and the con-

version of grasslands to forest lands did not lead to signi

ﬁ

cant SOC

changes (

−

15%, CI [

−

28, +1.4], and +1.3% CI [

−

6.4, +9.6], respectively).

The effect of these two types of land-use change on SOC is considered
to be particularly determined by local soil and climate conditions

54

.

Soil organic carbon increase through land management
practices

Exogenous carbon inputs resulted in the largest increases in SOC in
croplands and grasslands. Biochar led to a mean SOC gain of +67%,
CI [ + 31, +112] in croplands and +32%, CI [ + 26, +38] in grasslands.
Organic amendments applied in croplands resulted in a SOC gain of
+29%, CI [ + 15, +45], and +34%, CI [ + 21, +48] in grasslands (Fig.

3

,

see details by sub-types of amendment in the online Table asso-
ciated to Fig.

3

and in Supplementary Dataset 1). Biochar applica-

tion is regarded as having a high climate change mitigation
potential

55

, which is supported by our

ﬁ

ndings. However, we have

also demonstrated a signi

ﬁ

cant variability in the effect of biochar,

which can likely be attributed, in part, to the diverse application
rates and physicochemical properties of biochars

52

. The applica-

tion of biochar has long-lasting effects that persist for a longer
period than the biomass it is derived from, resulting in most of the
CO

2

removal bene

ﬁ

ts associated with its use. This is even true when

taking into account GHG emissions that occur during its produc-
tion and handling

56

. On the other hand, the scarcity of biomass in

some regions or its competition with livestock feeding, particularly
in sub-Sahara African countries

56

, can hamper the large-scale pro-

duction of biochar. Besides, possible adverse effects of biochar on
soil properties and biodiversity should be considered

57

. It is

also essential to emphasize that the effectiveness of exogenous
carbon inputs in mitigating climate change may be restricted
depending on the alternative fate of the amendments

58

,

59

, and the

feasibility to produce them. On the other hand, enhanced soil
health and plant productivity resulting from organic amendments
can further increase carbon inputs from plant growth and reinforce
the positive effects on SOC. The use of mineral fertilizers may
involve similar mechanisms. However, our results show that
application of mineral fertilizer resulted in smaller effects on SOC:

+9.4%, CI [ + 6.3, +12.6] in croplands and +3.6%, CI [0.5, +6.7] in
grasslands. Fertilizer use is considered by some authors as a main
contributor to soil carbon sequestration (an estimated global
increase of 70

–

88 Mt C yr

−

1

, Lessmann et al. 2022

35

), especially in

low fertile soils

60

. Interestingly, our study suggests that partial or

total substitution of mineral fertilization with organic amendments
in cropland leads to an increase in SOC of +34%, CI [ + 20, +49].

Agroforestry signi

ﬁ

cantly increased SOC in croplands by +20%, CI

[ + 17, +23]. The integration of trees in croplands resulted in an average
SOC increase of 33%, CI [ + 24, +43] for multi-strata systems, 32%, CI
[ + 9.0, +60] for parklands, and 21%, CI [ + 14, +30] for alley cropping,
19%, CI [

−

4.9 + 50] for improved fallows, and 17%, CI [ + 13, +22] for

hedgerows (see the online interactive Table associated to Fig.

3

).

Growing trees in grasslands (i.e., silvopasture) also resulted in a sig-
ni

ﬁ

cant SOC increase of +26%, CI [ + 11, +42]). The use of agroforestry is

regularly brought up by policymakers as a climate change mitigation
action, e.g., 40% of the 147 non-Annex I countries under the Kyoto
Protocol propose agroforestry as a solution in their Nationally Deter-
mined Contributions

61

. Yet much of the SOC storage potential asso-

ciated with the integration of trees in croplands and grasslands occur in
countries where agroforestry is not considered as a climate change
mitigation option

62

. Moreover, including trees in croplands could

increase competition for resources with crops and often requires sup-
plementary nutrient inputs in the short term

63

. It is worth noting that

other types of crop diversi

ﬁ

cation had limited effects on SOC in crop-

lands. For instance, crop rotation resulted, on average, in a 6.5% SOC
increase (CI [

−

0.9, +14]), and mixtures of crop species showed a 0.2%

change (CI [0.1, +0.3]), on average. Interestingly, replacing a mono-
speci

ﬁ

c forest with a mixture of tree species resulted in a signi

ﬁ

cant

9.0% SOC increase (CI [ + 5.4, +12.7]). The number of meta-analyses
investigating the effects of some of these management practices on
SOC is, however, limited.

Several other land management practices showed smaller but sig-

ni

ﬁ

cant positive changes in SOC, including no-till farming (+9,3%, CI

[ + 5,6, +13]; see the online interactive Table associated to Fig.

3

), reduced

tillage intensity (+12%, CI [ + 0.1, +24], crop residue retention (+13%, CI
[ + 9.8, +16]), and perennial energy cropping (+12, CI [ + 6.8, +17]). Several
of the above practices are frequently implemented in a combined form.
An example of combining several practices is the use of reduced tillage,
crop residue retention and crop diversi

ﬁ

cation, which are the funda-

mental principles of conservation agriculture. However, the number of

ﬁ

rst-order meta-analyses dealing with combined and more complex

agricultural practices remains limited (Supplementary Dataset 3). Simi-
larly, organic agriculture involves varying degrees of organic amend-
ments and crop diversi

ﬁ

cation, among other practices. Our results

indicate a large effect of this farming system on SOC ( + 35%, CI
[ + 11, +64]).

In contrast to the aforementioned practices, some other land

management practices had signi

ﬁ

cant negative effects on SOC, such as

prescribed

ﬁ

re in forest lands (

−

21%, CI [

−

34,

−

4.9]). It is worth noting

that prescribed burning is sometimes recommended to reduce carbon
losses from possible future wild

ﬁ

res, but its carbon costs (i.e., GHG

emissions and SOC losses) may outweigh the reduced wild

ﬁ

re

emissions

64

. Some other forest management practices (i.e., converting

secondary forest to plantation;

−

23%, CI [

−

24,

−

21%]), forest harvesting

(

−

8.1%, CI [

−

13,

−

3.1]) also negatively impacted SOC. In grasslands,

increased intensity of grazing (

−

9.9%, CI [

−

18,

−

0.5]) and the presence

of grazing compared to no grazing (

−

7.1%, CI [

−

31,

−

3.9]) had negative

impacts on SOC.

Indirect effects of climate change could have a pronounced
negative effect on SOC

Direct effects of climate change such as warming, drought or rainfall
increase, and CO

2

enrichment, had either uncertain or small impacts

on SOC change when analyzed individually (Fig.

4

). Speci

ﬁ

cally,

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

4

Fig. 3 | Percentage change in soil organic carbon (SOC) due to land manage-
ment practices.

Results are detailed for for croplands (

a

) forest lands (

b

) and

grasslands (

c

) Diamonds and lines of the main plots represent the mean effect sizes

and 95% con

ﬁ

dence intervals (CIs), respectively. The dot sizes are proportional to

the number of paired data analyzed. The histograms above each plot represent the
probability distribution of effects for all practices combined. The bar plots on the

right side of each main plot represent the number of primary studies for each
practice (bars) and the percentage of primary studies used by at least two meta-
analyses (lines). Forest degradation includes the transformation of primary forest
to secondary forest and secondary forest to plantation forest. Details of the
information used to make this graph as well as details by sub-category of land
management are available at

https://rpubs.com/dbeillouin/Figure3

.

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

5

drought did not signi

ﬁ

cantly affect SOC (mean effect across all land-

use types: +0.8%, CI [

−

1.9, +3.7]), the same was observed for warming

(

−

0.3%, CI [

−

2.3, +1.8]), and rainfall increase (+1.4%, CI [

−

2.3, +5.3]). CO

2

enrichment, on the contrary, had a signi

ﬁ

cant effect on SOC for two of

the four land-use types analyzed, and a mean effect of +4.3%, CI [ + 3.5,
+5.2], across all land-use types.

Considering the combined direct effects of climate change, there

was a signi

ﬁ

cant positive impact on SOC change, with a mean effect of

+ 4.6%, [+3.2, +6.0]. This effect can likely be associated with the pre-
sence of CO

2

enrichment in the majority of the examined combina-

tions. Available data on the individual and combined direct effects of
climate change on SOC are, however, very limited. Further studies are
needed to better understand the individual and combined effects of
climate change on SOC

65

, especially in the different land-use types and

under different climatic conditions. This would allow to re

ﬁ

ne our

understanding of the mechanisms controlling SOC dynamics and
develop more accurate projections of SOC changes in response to
global climate change. In particular, there is a need to better under-
stand the responses of the different SOC fractions to climate change
effects

42

(and see Supplementary Dataset 2), including the various

confounding factors affecting SOC decomposition rates

66

. For exam-

ple, climate change simultaneously affects SOC input through its effect
on both plant biomass production and SOC decomposition, i.e.,
warming can lead to increased plant biomass, but it can also enhance
decomposition;

14

elevated CO

2

levels can stimulate plant growth while

it is expected that they have a minimal effect on SOC decomposition

67

.

On the other hand, our

ﬁ

ndings indicate that indirect effects

associated with climate change, such as wild

ﬁ

res or a decrease in snow

cover, may have a more substantial impact on SOC than the direct
climate change effects. The underlying data are, however, scarce and
the available results should be con

ﬁ

rmed in further experimental

studies. It is worth noting that the effect of wild

ﬁ

re on SOC is in line

with the effect observed for prescribed

ﬁ

res in forests (as shown in

Fig.

3

), as well as with the effect of

ﬁ

res of

‘

non-classi

ﬁ

ed land types

’

(as

detailed in the online Table associated to Fig.

4

). These indirect effects

associated with climate change should not be overlooked, given that,
for instance, the occurrence of forest

ﬁ

res has doubled in the last 40

years

68

, and has notably increased in many biomes

69

. Other indirect

effects of climate change, such as the effect of

ﬂ

ooding and changes in

snow cover, have thus far received limited attention in existing meta-
analyses, and there is a lack of synthesized results on these subjects.

In summary, our comprehensive second-order meta-analysis,

encompassing over 220 types of land-use change, land management
practices, and climate change factors impacting SOC, has identi

ﬁ

ed

both the main factors associated with SOC loss and various options
that have the potential to maintain or increase SOC levels on a global

scale. The preservation of SOC can be effectively achieved through the
protection of natural ecosystems and the introduction of perennial
crops in croplands. However, it is important to note that worldwide
efforts to preserve SOC may encounter local challenges arising from
the indirect effects of climate change, which may have a more sub-
stantial impact on SOC compared to direct climate change effects.

In our analysis, we have identi

ﬁ

ed signi

ﬁ

cant knowledge gaps that

require further research, namely the neglected ecosystems such as
wetlands and the impact of climate change on SOC. Our global second-
order meta-analysis is a valuable tool for scientists conducting future
research on SOC preservation and restoration. The results of our
study, including the ranking of effective land management practices
for protecting or increasing SOC, can serve as a valuable reference for
decision-makers involved in climate change mitigation efforts. How-
ever, it is essential to exercise caution when implementing these
results in local contexts. This is due to the unique pedo-climatic con-
ditions of different regions and the limitations in the representative-
ness of our results in certain environments.

Methods

We conducted a second-order meta-analysis to examine the impact of
drivers of SOC change, namely land-use change, land management and
climate change, across various land-use types. Our analysis synthe-
sized the

ﬁ

ndings of 230

ﬁ

rst-order meta-analyses on SOC that were

conducted worldwide (see Fig. S1 for details) on croplands, forest
lands, grasslands, wetlands, and other lands. We assessed the quality of
the

ﬁ

rst-order meta-analyses and the potential overlap of primary

studies between the meta-analyses to produce reliable estimates of
effects.

Systematic literature search

We

ﬁ

rst performed a systematic search for peer-reviewed meta-ana-

lyses on bulk SOC stocks or concentrations using various databases,
including the Web of Science Core Collection, Scopus, CAB Abstracts,
and Agricola (through the OVID platform), and the Google Scholar
search engine. The search was done on January 9th, 2020, and updated
on July 10

th

, 2022, using the following search string: (

“

meta*analysis

”

OR

“

systematic review

”

) AND (

“

soil organic carbon

”

OR

“

SOC

”

OR

“

soil

organic matter

”

OR

“

SOM

”

OR

“

soil carbon

”

) in title, abstract, and

keywords

ﬁ

elds. We screened the titles and abstracts of the 1005

identi

ﬁ

ed papers for their potential inclusion in our study. The full text

of the retained papers was then independently reviewed by at least two
co-authors of this study for the following criteria. A paper had to (i)
analyze the effect of one or several factors on bulk SOC stocks or
concentrations, (ii) present a statistical formal analysis of at least two
primary studies on SOC, (iii) present indicators of precision of the

Land use :

Other lands (n= 727)
Forest lands (n= 549)

Grasslands (n= 260)

Croplands (n= 161)

Wetlands (n= 14)

Combination of at leat 2

climatic factors

(n=232)

Warming

(n=641)

Rainfall increase

(n=641)

CO

2

enrichment

(n=580)

-20

0

20

40

Soil organic carbon (SOC) change (%)

a. Direct effects

Drought

(n=157)

Wildfire

(n=957)

Snow cover

decrease

(n=6)

Soil organic carbon (SOC) change (%)

b. Indirect effects

-75

-50

-25

0

Fig. 4 | Percentage change in soil organic carbon (SOC) due climate change.

Results are detailed for direct (

a

) and indirect factors (

b

) of climate change. Dia-

monds and lines represent the estimated mean effect sizes and the 95% con

ﬁ

dence

intervals (CIs), respectively. The number of paired data (n) for each category

of land-use type is shown in the bubbles on the right side of the plot. Details of the
information used to make this graph as well as details by sub-category of climate
change factor are available at

https://rpubs.com/dbeillouin/Figure4

.

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

6

effect sizes (standard errors or con

ﬁ

dence intervals). At the end, 230

meta-analyses were retained for our second-order meta-analysis. Fur-
ther details on the methods used can be found in a related data paper

34

and in an evidence map

33

.

Data extraction and coding

All the effect sizes of the 230 meta-analyses included in our study were
then extracted from the text, tables, or

ﬁ

gures (using Plot Digitizer

http://plotdigitizer.sourceforge.net/

) and recorded in an Excel

ﬁ

le. The

land-use type associated with each effect size was documented
according to IPCC standards

70

. We also extracted the metrics (e.g.,

ratio, percentage change), possible transformations (e.g., logarithm),
con

ﬁ

dence intervals or other indicators of variability, and the number

of primary studies and observations that were used to calculate the
effect sizes. The list of primary studies (and their DOIs, when available)
used in each meta-analysis was retrieved, thereby allowing us to
identify the number of common primary studies between each pair of
meta-analyses. Finally, we characterized the meta-analyses included in
our study by eight quality criteria related to the literature search,
statistical analyses, and potential bias analysis (see Beillouin et al.,
2021

34

for a precise description of these criteria). A poor methodology

used to retrieve the primary studies and analyze the data can indeed
result in biased and misleading results

71

.

Effect sizes (and con

ﬁ

dence intervals) expressed as standardized

mean differences or percentage changes in the

ﬁ

rst-order meta-ana-

lyses were converted into ratios to ensure the comparability of the
results between the meta-analyses

72

and were log-transformed to

ensure normality. After applying the meta-analytical models (see
below), we transformed the log ratios and their associated con

ﬁ

dence

intervals and expressed them in percent change to facilitate inter-
pretation. We categorized the drivers of SOC change reported in the
meta-analyses into three main categories: land-use change, land
management, and climate change. In case a meta-analysis presented
subgroup analyses, only independent effect sizes, i.e., based on dif-
ferent sets of primary studies, were retained.

Pairwise meta-analysis

To estimate the effect of factors on SOC, we tested several meta-
analytical models, varying in the structure of their random effects, the
inclusion or not of the quality score of the

ﬁ

rst-order meta-analyses,

and of the redundancy of the primary studies between the meta-
analyses. The best model for each factor (whose results are presented
in the main text) was then chosen based on the Akaike information
criterion (AIC).

The most complex model is a three-level meta-analytical model,

including a variance

–

covariance matrix considering the precision, the

quality and the redundancy of the

ﬁ

rst-order meta-analyses. This

model is written as follows:

log

ð

Y

ij

Þ

=

μ

+

b

i

+

φ

ij

+

ε

ij

ð

1

Þ

with

b

i

∼

N

ð

0,

τ

2

Þ

,

φ

ij

∼

N

0,

υ

2

and

ε

ij

∼

N

0,

σ

ij

2

,where

Y

ij

is the j

th

effect-size of the i

th

meta-analysis (one meta-analysis could present

several effect-sizes),

μ

is the mean estimated effect (shown as the

diamonds in the Figures),

b

i

is the random meta-analysis effect (i.e.,

between-cluster heterogeneity),

φ

ij

is the random effect size effect

within the i

th

meta-analysis (i.e., within-cluster heterogeneity), and

ε

ij

is

the random estimation error associated with the j

th

effect size of the i

th

meta-analysis (i.e., the sampling error). Here, the clusters represent the
meta-analyses included in our study. The three-level model implies the
estimation of two heterogeneity variance parameters (here noted

τ

2

and

υ

2

).

We weighted each effect size by the inverse of its variance, as

recommended by Marín-Martínez and Sánchez-Meca (2010), and
reduced the weight of the lower-quality meta-analyses according to

Doi et al. (2015)

73

,

74

. As a quality proxy, we used the percentage of the

eight quality criteria met by a meta-analysis (see Beillouin et al., 2022

33

for a detailed explanation of the criteria). We also considered the non-
independence between the effect sizes of different meta-analyses by
calculating a variance-covariance matrix based on a pseudo correlation
between meta-analyses

75

. The proxy of the correlation between each

pair of meta-analyses was estimated as (2×m)/(n

1

+ n

2

), where m is the

number of common primary studies between each pair of meta-ana-
lyses, and n

1

and n

2

represent the total number of primary studies in

the two respective meta-analyses.

The other meta-analytic models that were tested corresponded to

(i) a model with a simpli

ﬁ

ed random structure, i.e., without the

between-cluster heterogeneity (

b

i

); (ii) a three-level hierarchical model

without considering the redundancy of primary studies, and (iii) a
three-level hierarchical model without considering both the redun-
dancy of primary studies and the quality of the

ﬁ

rst-order meta-

analyses.

Analysis of the results and sensitivity analyses

The model results were summarized by the median as a point esti-
mate and the 95% con

ﬁ

dence interval as a measure of uncertainty of

the point estimate. Potential publication bias was assessed with
funnel plots and Egger

’

s test

76

. The funnel plots assume that studies

with high precision will be plotted near the average mean effect, and
studies with low precision will be spread evenly on both sides.
Egger

’

s test gives the degree of funnel plot asymmetry as measured

by the intercept from regression of standard normal deviates
against precision. The sensitivity of the results against publication
bias was tested with the Rosenthal fail-safe number, i.e., the number
of additional studies with a mean null result necessary to provide a
non-signi

ﬁ

cant global estimated effect (see Fig. S3-5). We also esti-

mated the mean effect considering the missing studies based on the
trim-and-

ﬁ

ll methodology

77

.

Finally, we tested the robustness of our model results by com-

paring the results obtained using frequentist and Bayesian statistics.
Frequentist model parameters were estimated by maximum likelihood
using the

metafor

R package

78

. For the Bayesian approach, we used

weakly informative prior scenarios, following the recommendations of
Williams, Rast, and Bürkner (2018)

79

. Speci

ﬁ

cally, the true pooled

effect size prior was set to a normal distribution with a mean of 0 and a
variance of 1. The variance priors of the models were set to a half-
Cauchy distribution with location parameter set at 0 and a scale
parameter set at 0.5. The posterior distribution was approximated
through Markov chain Monte Carlo (MCMC) simulation methods using
Just Another Gibbs Sampler (JAGS) software (version 4.3.0) through
the

brms

package

80

. The MCMC algorithm was run with three Markov

chains, each including 20,000 iterations after a burn-in period of
8,000 iterations. In addition, the chains were thinned by storing one
out of ten iterations to reduce autocorrelations in the subsequent
sample. Convergence was assessed via three different criteria: (i) the
potential scale reduction factor,

R

ˆ

, whose values must be equal or

close to 1, (ii) the effective number of independent simulations draws,

neff

, which must be >100, (iii) graphically, by drawing trace plots and

assessing whether the simulated values of the chains overlapped. The
best Bayesian model for each factor was chosen based on the wAIC

81

.

Comparisons of the model results between the frequentist and Baye-
sian approach are available in Fig. S 2.

All statistical analyses were conducted with R software (version

3.0.2), dplyr for data management, and ggplot2 for data visualization.
All scripts used in this study are available in the MetaSynthesis R

82

.

Data availability

The data are available under the repository

https://doi.org/10.18167/

DVN1/KKPLR8

and described in the corresponding DataPaper: Beil-

louin et al. 2021

34

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

7

Code availability

The codes used for the analyses are available in the MetaSynthesis
package

82

References

1.

Banwart, S. A., Nikolaidis, N. P., Zhu, Y.-G., Peacock, C. L. & Sparks,
D. L. Soil functions: connecting earth

’

s critical zone.

Annu. Rev.

Earth Planet. Sci.

47

, 333

–

359 (2019).

2.

Lal, R. Soil health and carbon management.

Food Energy Secur.

5

,

212

–

222 (2016).

3.

Bünemann, E. K. et al. Soil quality: a critical review.

Soil Biol. Bio-

chem.

120

, 105

–

125 (2018).

4.

Old

ﬁ

eld, E. E., Bradford, M. A. & Wood, S. A. Global meta-analysis of

the relationship between soil organic matter and crop yields.

SOIL

5

, 15

–

32 (2019).

5.

Lal, R. Soil carbon sequestration impacts on global climate change
and food security.

Science

304

, 1623

–

1627 (2004).

6.

Batjes, N. H. Total carbon and nitrogen in the soils of the world.

Eur.

J. Soil Sci.

47

, 151

–

163 (1996).

7.

Friedlingstein, P. et al. Global carbon budget 2020.

Earth Syst. Sci.

Data

12

, 3269

–

3340 (2020).

8.

Rumpel, C. et al. The 4p1000 initiative: opportunities, limitations
and challenges for implementing soil organic carbon sequestration
as a sustainable development strategy.

Ambio

49

, 350

–

360 (2020).

9.

Amelung, W. et al. Towards a global-scale soil climate mitigation
strategy.

Nat. Commun.

11

, 5427 (2020).

10. IPCC 2018, V. P. Summary for Policymakers. In: Global Warming of

1.5 °C. An IPCC Special Report on the impacts of global warming of
1.5 °C above pre-industrial levels and related global greenhouse
gas emission pathways, in the context of strengthening the global
response to the threat of climate change, sustainable development,
and efforts to eradicate poverty. (2018).

11.

Anderson, C. M. et al. Natural climate solutions are not enough.

Science

363

, 933

–

934 (2019).

12. Goldstein, A. et al. Protecting irrecoverable carbon in Earth

’

s eco-

systems.

Nat. Clim. Change

10

, 287

–

295 (2020).

13. Cook-Patton, S. C. et al. Protect, manage and then restore lands for

climate mitigation.

Nat. Clim. Change

11

, 1027

–

1034 (2021).

14. Griscom, B. W. et al. Natural climate solutions.

Proc. Natl. Acad. Sci.

USA

114

, 11645

–

11650 (2017).

15. Girardin, C. A. J. et al. Nature-based solutions can help cool the

planet

—

if we act now.

Nature

593

, 191

–

194 (2021).

16. Bossio, D. A. et al. The role of soil carbon in natural climate solu-

tions.

Nat. Sustain.

3

, 391

–

398 (2020).

17. Wiesmeier, M. et al. Soil organic carbon storage as a key function of

soils - A review of drivers and indicators at various scales.

Geoderma

333

, 149

–

162 (2019).

18. Ramesh, T. et al. Soil organic carbon dynamics: Impact of land use

changes and management practices: a review. in

Advances in

Agronomy

156

1

–

107 (Elsevier, 2019).

19. Winkler, K., Fuchs, R., Rounsevell, M. & Herold, M. Global land use

changes are four times greater than previously estimated.

Nat.

Commun.

12

, 2501 (2021).

20. Don, A., Schumacher, J. & Freibauer, A. Impact of tropical land-use

change on soil organic carbon stocks - a meta-analysis: soil organic
carbon and land-use change.

Glob. Change Biol.

17

,

1658

–

1670 (2011).

21. Laganiã

ˆ

Re, J. & Angers, D. A. Carbon accumulation in agricultural

soils after afforestation: a meta-analysis: soc accumulation follow-
ing afforestation.

Glob. Change Biol.

16

, 439

–

453 (2010).

22. Maillard, É. & Angers, D. A. Animal manure application and soil

organic carbon stocks: a meta-analysis.

Glob. Change Biol.

20

,

666

–

679 (2014).

23. Luo, Z., Wang, E. & Sun, O. J. Can no-tillage stimulate carbon

sequestration in agricultural soils? A meta-analysis of paired
experiments.

Agric. Ecosyst. Environ.

139

, 224

–

231 (2010).

24. Mehra, P. et al. A review of tillage practices and their potential to

impact the soil carbon dynamics.

Adv. Agronomy

150

, 185

–

230

(Elsevier, 2018).

25. Poeplau, C. & Don, A. Carbon sequestration in agricultural soils via

cultivation of cover crops

–

A meta-analysis.

Agric. Ecosyst. Environ.

200

, 33

–

41 (2015).

26. Cardinael, R. et al. Revisiting IPCC Tier 1 coef

ﬁ

cients for soil organic

and biomass carbon storage in agroforestry systems.

Environ. Res.

Lett.

13

, 124020 (2018).

27. Cardinael, R. et al. Increased soil organic carbon stocks under

agroforestry: a survey of six different sites in France.

Agric. Ecosyst.

Environ.

236

, 243

–

255 (2017).

28. Dignac, M.-F. et al. Increasing soil carbon storage: mechanisms,

effects of agricultural practices and proxies. A review.

Agron. Sus-

tain. Dev.

37

, 14 (2017).

29. Crowther, T. W. et al. Quantifying global soil carbon losses in

response to warming.

Nature

540

, 104

–

108 (2016).

30. Reichstein, M. et al. Climate extremes and the carbon cycle.

Nature

500

, 287

–

295 (2013).

31. Bruni, E. et al. Additional carbon inputs to reach a 4 per 1000

objective in Europe: feasibility and projected impacts of climate
change based on Century simulations of long-term arable experi-
ments.

Biogeosciences

18

, 3981

–

4004 (2021).

32. Xu, S., Sheng, C. & Tian, C. Changing soil carbon: in

ﬂ

uencing fac-

tors, sequestration strategy and research direction.

Carbon Balance

Manag.

15

, 2 (2020).

33. Beillouin, D. et al. A global overview of studies about land man-

agement, land

‐

use change, and climate change effects on soil

organic carbon.

Glob. Change Biol.

28

, 1690

–

1702 (2022).

34. Beillouin, D. et al. A global database of management, land use

change and climate change effects on soil organic carbon.

Sci. Data

9

, 228 (2021).

35. Bolinder, M. A. et al. The effect of crop residues, cover crops,

manures and nitrogen fertilization on soil organic carbon changes
in agroecosystems: a synthesis of reviews.

Mitig. Adapt. Strateg.

Glob. Change

25

, 929

–

952 (2020).

36. Young, M. D., Ros, G. H. & de Vries, W. Impacts of agronomic

measures on crop, soil, and environmental indicators: a review and
synthesis of meta-analysis.

Agric. Ecosyst. Environ.

319

,

107551 (2021).

37. Lessmann, M., Ros, G. H., Young, M. D. & Vries, W. Global variation in

soil carbon sequestration potential through improved cropland
management.

Glob. Change Biol.

28

, 1162

–

1177 (2022).

38. Button, K. & Nijkamp, P. Environmental policy assessment and the

usefulness of meta-analysis.

Socioecon. Plann. Sci.

31

,

231

–

240 (1997).

39. Pigott, T. D. & Polanin, J. R. Methodological guidance paper: high-

quality meta-analysis in a systematic review.

Rev. Educ. Res.

90

,

24

–

46 (2020).

40. Chenu, C. et al. Increasing organic stocks in agricultural soils:

knowledge gaps and potential innovations.

Soil Tillage Res.

188

,

41

–

52 (2019).

41. Batjes, N. H. Technologically achievable soil organic carbon

sequestration in world croplands and grasslands.

Land Degrad.

Dev.

30

, 25

–

32 (2019).

42. Rocci, K. S., Lavallee, J. M., Stewart, C. E. & Cotrufo, M. F. Soil

organic carbon response to global environmental change depends
on its distribution between mineral-associated and particulate
organic matter: a meta-analysis.

Sci. Total Environ.

793

,

148569 (2021).

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

8

43. Funk, J. M. et al. Securing the climate bene

ﬁ

ts of stable forests.

Clim. Policy

19

, 845

–

860 (2019).

44. Keith, H. et al. Evaluating nature-based solutions for climate miti-

gation and conservation requires comprehensive carbon account-
ing.

Sci. Total Environ.

769

, 144341 (2021).

45. Manevski, K., Lærke, P. E., Jiao, X., Santhome, S. & Jørgensen, U.

Biomass productivity and radiation utilisation of innovative crop-
ping systems for biore

ﬁ

nery.

Agric. For. Meteorol.

233

,

250

–

264 (2017).

46. Thorup-Kristensen, K. et al. Digging deeper for agricultural resour-

ces, the value of deep rooting.

Trends Plant Sci.

25

, 406

–

417 (2020).

47. Beillouin, D., Ben

‐

Ari, T., Malézieux, E., Seufert, V. & Makowski, D.

Positive but variable effects of crop diversi

ﬁ

cation on biodiversity

and ecosystem services.

Glob. Change Biol.

27

, 4697

–

4710 (2021).

48. Li, W. et al. Temporal response of soil organic carbon after

grassland-related land-use change.

Glob. Change Biol.

24

,

4731

–

4746 (2018).

49. Poeplau, C. et al. Temporal dynamics of soil organic carbon after

land-use change in the temperate zone - carbon response functions
as a model approach: soil organic carbon and land-use change.

Glob. Change Biol.

17

, 2415

–

2427 (2011).

50. Walker, W. S. et al. The global potential for increased storage of

carbon on land.

Proc. Natl. Acad. Sci. USA

119

, e2111312119

(2022).

51. Li, R. et al. Time and space catch up with restoration programs that

ignore ecosystem service trade-offs.

Sci. Adv.

7

, eabf8650 (2021).

52. Holl, K. D.

Primer of ecological restoration

. (Island Press, 2020).

53. Veldkamp, E., Schmidt, M., Powers, J. S. & Corre, M. D. Deforesta-

tion and reforestation impacts on soils in the tropics.

Nat. Rev. Earth

Environ.

1

, 590

–

605 (2020).

54. Powers, J. S., Corre, M. D., Twine, T. E. & Veldkamp, E. Geographic

bias of

ﬁ

eld observations of soil carbon stocks with tropical land-

use changes precludes spatial extrapolation.

Proc. Natl. Acad. Sci.

USA

108

, 6318

–

6322 (2011).

55. Lehmann, J. et al. Biochar in climate change mitigation.

Nat. Geosci.

14

, 883

–

892 (2021).

56. Giller, K. E., Witter, E., Corbeels, M. & Tittonell, P. Conservation

agriculture and smallholder farming in Africa: the heretics

’

view.

Field Crops Res.

114

, 23

–

34 (2009).

57. Brtnicky, M. et al. A critical review of the possible adverse effects of

biochar in the soil environment.

Sci. Total Environ.

796

,

148756 (2021).

58. Powlson, D. S., Whitmore, A. P. & Goulding, K. W. Soil carbon

sequestration to mitigate climate change: a critical re-examination
to identify the true and the false.

Eur. J. Soil Sci.

62

, 42

–

55 (2011).

59. Butterbach-Bahl, K. et al. Livestock enclosures in drylands of Sub-

Saharan Africa are overlooked hotspots of N

2

O emissions.

Nat.

Commun.

11

, 4644 (2020).

60. Chowdhury, S. et al. Role of cultural and nutrient management

practices in carbon sequestration in agricultural soil. in

Advances in

Agronomy

166

131

–

196 (Elsevier, 2021).

61. Rosenstock, T. S. et al. Making trees count: Measurement and

reporting of agroforestry in UNFCCC national communications of
non-Annex I countries.

Agric. Ecosyst. Environ.

284

, 106569

(2019).

62. Chapman, M. et al. Large climate mitigation potential from adding

trees to agricultural lands.

Glob. Change Biol.

26

,

4357

–

4365 (2020).

63. Fanin, N. et al. Effects of mixing tree species and water availability

on soil organic carbon stocks are depth dependent in a temperate
podzol.

Eur. J. Soil Sci

.

73

, e13133 (2022).

64. Bowman, D. M. J. S., Williamson, G. J., Price, O. F., Ndalila, M. N. &

Bradstock, R. A. Australian forests, mega

ﬁ

res and the risk of

dwindling carbon stocks.

Plant Cell Environ.

44

, 347

–

355

(2021).

65. 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

–

822 (2017).

66. Davidson, K. E. et al. Livestock grazing alters multiple ecosystem

properties and services in salt marshes: a meta-analysis.

J. Appl.

Ecol.

54

, 1395

–

1405 (2017).

67. van Groenigen, K. J., Qi, X., Osenberg, C. W., Luo, Y. & Hungate, B. A.

Faster decomposition under increased atmospheric CO

2

limits soil

carbon storage.

Science

344

, 508

–

509 (2014).

68. Mansoor, S. et al. Elevation in wild

ﬁ

re frequencies with respect to

the climate change.

J. Environ. Manage.

301

, 113769 (2022).

69. Duane, A., Castellnou, M. & Brotons, L. Towards a comprehensive

look at global drivers of novel extreme wild

ﬁ

re events.

Clim.

Change

165

, 43 (2021).

70. IPCC 2019. Re

ﬁ

nement to the 2006 IPCC Guidelines for National

Greenhouse Gas Inventories. In: Buendia, E., et al. Eds., Agriculture,
Forestry and Other Land Use (AFOLU), préparé par le Programme
pour les inventaires nationaux des gaz à effet de serre 4, 110 p. in.

71. Thornton, A. Publication bias in meta-analysis its causes and con-

sequences.

J. Clin. Epidemiol.

53

, 207

–

216 (2000).

72. Borenstein, M., Hedges, L. V., Higgins, J. P. T. & Rothstein, H. R.

Introduction to Meta-Analysis

. (John Wiley & Sons, Ltd, 2009).

https://doi.org/10.1002/9780470743386

.

73. Doi, S. A. R., Barendregt, J. J., Khan, S., Thalib, L. & Williams, G. M.

Advances in the meta-analysis of heterogeneous clinical trials II: the
quality effects model.

Contemp. Clin. Trials

45

, 123

–

129 (2015).

74. Marín-Martínez, F. & Sánchez-Meca, J. Weighting by inverse var-

iance or by sample size in random-effects meta-analysis.

Educ.

Psychol. Meas.

70

, 56

–

73 (2010).

75. Lajeunesse, M. J. On the meta-analysis of response ratios for studies

with correlated and multi-group designs.

Ecology

92

,

2049

–

2055 (2011).

76. Egger, M., Smith, G. D., Schneider, M. & Minder, C. Bias in meta-

analysis detected by a simple, graphical test.

Bmj

.

315

,

629

–

634 (1997).

77. Duval, S. & Tweedie, R. A nonparametric

“

trim and

ﬁ

ll

”

method of

accounting for publication bias in meta-analysis.

J. Am. Stat. Assoc.

95

, 89

–

98 (2000).

78. Viechtbauer, W. Conducting meta-analyses in

R

with the

metafor

package.

J. Stat. Softw

.

36

, 1

–

48 (2010).

79. Williams, D. R., Rast, P. & Bürkner, P.-C.

Bayesian Meta-Analysis with

Weakly Informative Prior Distributions

.

https://osf.io/7tbrm

(2018)

https://doi.org/10.31234/osf.io/7tbrm

.

80. Bürkner, P.-C.

brms

: An

R

package for bayesian multilevel models

using stan.

J. Stat. Softw

.

80

, 1

–

28 (2017).

81. Vehtari, A., Gelman, A. & Gabry, J. Practical Bayesian model eva-

luation using leave-one-out cross-validation and WAIC.

Stat. Com-

put.

27

, 1413

–

1432 (2017).

82. Beillouin, D. MetaSynthesis: An R package to perform second-order

meta-analysis. (2023).

83. FAO & ITPS. Global Soil Organic Carbon Map (GSOCmap). (2020).

Acknowledgements

Support for this study to M.C. was provided through the CGIAR initiative
Mitigate + : Research for Low-Emission Food Systems, and through the
contributions of all funders to the CGIAR Trust Fund.

Author contributions

D.B.: Data retrieval, formal analysis, draft, review. M.C.: Data retrieval,
draft, review. J.D.: Data retrieval, draft, review. DavidB: Data retrieval,
review. A.B.: Data retrieval. A.F.: Data retrieval, review. F.F.: Data retrie-
val, review. R.C.: Data retrieval, draft, review.

Competing interests

The authors declare no competing interests.

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

9

Additional information

Supplementary information

The online version contains

supplementary material available at

https://doi.org/10.1038/s41467-023-39338-z

.

Correspondence

and requests for materials should be addressed to

Damien Beillouin.

Peer review information

Nature Communications

thanks Nanthi Bolan

and the other, anonymous, reviewer(s) for their contribution to the peer
review of this work. A peer review

ﬁ

le is available.

Reprints and permissions information

is available at

http://www.nature.com/reprints

Publisher

’

s note

Springer Nature remains neutral with regard to jur-

isdictional claims in published maps and institutional af

ﬁ

liations.

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/

.

Article

https://doi.org/10.1038/s41467-023-39338-z

Nature Communications

| (2023) 14:3700

10

Beillouin et al. (2023). A global meta-analysis of soil organic carbon in the Anthropocene

Page 1

Page 2

Page 3

Page 4

Page 5

Page 6

Page 7

Page 8

Page 9

Page 10

Document Outline