Content uploaded by Andrea De Stefano
Author content
All content in this area was uploaded by Andrea De Stefano on Nov 13, 2017
Content may be subject to copyright.
Soil carbon sequestration in agroforestry systems:
a meta-analysis
Andrea De Stefano .Michael G. Jacobson
Received: 12 October 2016 / Accepted: 22 October 2017
ÓSpringer Science+Business Media B.V. 2017
Abstract Agroforestry systems may play an impor-
tant role in mitigating climate change, having the
ability to sequester atmospheric carbon dioxide (CO
2
)
in plant parts and soil. A meta-analysis was carried out
to investigate changes in soil organic carbon (SOC)
stocks at 0–15, 0–30, 0–60, 0–100, and 0 C100 cm,
after land conversion to agroforestry. Data was
collected from 53 published studies. Results revealed
a significant decrease in SOC stocks of 26 and 24% in
the land-use change from forest to agroforestry at 0–15
and 0–30 cm respectively. The transition from agri-
culture to agroforestry significantly increased SOC
stock of 26, 40, and 34% at 0–15, 0–30, and 0–100 cm
respectively. The conversion from pasture/grassland
to agroforestry produced significant SOC stock
increases at 0–30 cm (9%) and 0–30 cm (10%).
Switching from uncultivated/other land-uses to agro-
forestry increased SOC by 25% at 0–30 cm, while a
decrease was observed at 0–60 cm (23%). Among
agroforestry systems, significant SOC stocks increases
were reported at various soil horizons and depths in the
land-use change from agriculture to agrisilviculture
and to silvopasture, pasture/grassland to agrosilvopas-
toral systems, forest to silvopasture, forest plantation
to silvopasture, and uncultivated/other to agrisilvicul-
ture. On the other hand, significant decreases were
observed in the transition from forest to agrisilvicul-
ture, agrosilvopastoral and silvopasture systems, and
uncultivated/other to silvopasture. Overall, SOC
stocks increased when land-use changed from less
complex systems, such as agricultural systems. How-
ever, heterogeneity, inconsistencies in study design,
lack of standardized sampling procedures, failure to
report variance estimators, and lack of important
explanatory variables, may have influenced the
outcomes.
Keywords Agroforestry Carbon sequestration
Soil organic carbon Climate change Meta-analysis
Introduction
Anthropogenic carbon dioxide emission as cause
of climate change
One of the recognized causes of climate change is the
increasing concentration of atmospheric carbon
Electronic supplementary material The online version of
this article (https://doi.org/10.1007/s10457-017-0147-9) con-
tains supplementary material, which is available to authorized
users.
A. De Stefano
School of Renewable Natural Resources, Louisiana State
University, Baton Rouge, LA 70808, USA
M. G. Jacobson (&)
Department of Ecosystem Science and Management, The
Pennsylvania State University, University Park,
PA 16802, USA
e-mail: mgj2@psu.edu
123
Agroforest Syst
DOI 10.1007/s10457-017-0147-9
dioxide (Nair et al. 2009a; Jose and Bardhan 2012).
Since the industrial revolution, atmospheric carbon
dioxide (CO
2
) concentration has increased more than
40%, rising from 280 ppm in 1750 to about 392 ppm
in 2012, it was expected to exceed 400 ppm by 2015
(Hutchinson et al. 2007; Jose and Bardhan 2012). The
most recent measurements confirm the expectations
that atmospheric CO
2
concentration exceed the
400 ppm threshold in 2016 (http://www.esrl.noaa.
gov/gmd/ccgg/trends/weekly.html). Agriculture
accounts for about 25% of the CO
2
, 50% of the CH
4
,
and 70% of the N
2
O emitted on a global scale through
anthropogenic sources (Hutchinson et al. 2007). One
of the feasible strategies to reduce CO
2
atmospheric
concentration is carbon (C) sequestration, defined as
the process of removing C from the atmosphere and
depositing it in a reservoir (http://unfccc.int/essential_
background/glossary/items/3666.php#C) (Nair et al.
2009a).
Climate change and land uses
C stock modifications associated with land-use change
can occur naturally or as a result of human activities
(Guo and Gifford 2002). Within human driven land-
use change, agricultural and forestry practices have the
potential to mitigate CO
2
concentration through C
sequestration, allowing the land acts as a ‘‘sink’’ for C
(Hutchinson et al. 2007). In 1992, during the Kyoto
Protocol, afforestation and reforestation were recog-
nized as a form of GHG-offset activities (Nair et al.
2009a). Subsequently, different land management
strategies, such as forest-, crop-, and grazing-land
management, and revegetation, were added to the of
land-use, land-use change and forestry (LULUCF)
activities in 2001, including agroforestry practices
(Nair et al. 2009a). With regard to those activities, C
sequestration implies the removal of CO
2
from
atmosphere and its storage in long-lived pools of C,
such as aboveground plant biomass, belowground
biomass (roots, soil microorganisms), stable forms of
organic and inorganic C in the soil, and long-lived
products (i.e. timber products) (Sanchez 2000;
Roshetko et al. 2002; Kirby and Potvin 2007; Nair
et al. 2009a).
Soil as carbon sinks
Soil plays an important role in C sequestration, being
able to store 1.5–3 times more C than in vegetation
(Young 1997). The amount of C sequestered in the soil
depends on a large number of factors, including the
region, site quality, current land-use, previous land-
use, and the portion of soil profile in case of land-use
changes, (Nair et al. 2010). Generally, soil accounts
for 60% of the total C stored in tree-based land-use
systems (Lal 2004d; Lorenz and Lal 2005; Lal 2007;
Nair et al. 2010). Soil C sequestration occurs in two
different ways: (1) direct fixation of atmospheric CO
2
,
which transforms CO
2
into soil inorganic C com-
pounds, (2) and indirect fixation of atmospheric CO
2
,
in which atmospheric CO
2
is incorporated in plant
tissue through the photosynthetic process, and subse-
quently, part of plant biomass is indirectly sequestered
as SOC during decomposition processes (Burras et al.
2001). On a global scale, the total soil C pool was
estimated in the rage of 2157–2296 Pg, of which
1462–1548 Pg is soil organic C (SOC), and 659–748
Pg is soil inorganic C (SIC) (Batjes 1996). The total
soil C pool is about three times the estimated
atmospheric C pool, and 3.8 times the vegetation C
pool (Lal 2004b; Nair et al. 2010); therefore, any
variation in the soil carbon pool would have a
significant impact on the global C budget. Among
land uses, agricultural and degraded soil have a
promising C sequestration potential: those soils were
depleted of a significant part of their original organic C
pool, and the adoption of specific management
practices, such as the implementation of trees and
permanent vegetation, could significantly boost their
C sequestration potential (Albrecht and Kandji 2003;
Lal 2004c).
The role of agroforestry and other land uses
in mitigating climate change
Agroforestry could offer a viable opportunity to deal
with climate change issues, having the potential to
sequester and store atmospheric CO
2
over long periods
(Albrecht and Kandji 2003; Lorenz and Lal 2014). In
sustainable-managed agroforestry systems, a large
portion of organic C returns to the soil in the form of
crop residues and tree litter (Oelbermann et al. 2004).
Those inputs can help to stabilize soil organic matter
(SOM) and decrease biomass decomposition rate and
Agroforest Syst
123
SOM destabilization, improving SOC stocks (Young
1997; Oelbermann et al. 2004; Lal 2004a; Sollins et al.
2007). Commonly, most agricultural soils act as a
major source of greenhouse gasses (CO
2
,CH
4
, and
N
2
O), having lost an important portion of their original
soil organic carbon (Stavi and Lal 2013). Agroforestry
can help to recover up to 35% of the original forest C
stock lost due to slash and burn agriculture (Sanchez
2000). Previous individual studies and secondary
quantitative reviews have investigated the overall
effect of land-use change and management practices
on SOC stocks, showing how the loss of soil coverage
related to deforestation, forest clearing, land-use
change, and other disturbance factors can negatively
affect SOC stocks (Allen 1985; Detwiler 1986; Mann
1986; Schlesinger 1986; Neill et al. 1997; Fearnside
and Imbrozio Barbosa 1998; Conant et al. 2001). Guo
and Gifford (2002) found decreased SOC stocks after
the conversion from pasture to plantation, native forest
to crops, and pasture to crops. Afforestation, defined as
the conversion of non-forested lands to forest planta-
tions, decreased soil C due when Pinus species where
used (Berthrong et al. 2009). Laganiere et al. (2010)
investigated on the factors that contribute to restoring
SOC stocks after afforestation, finding out that previ-
ous land-use, tree species planted, soil clay content,
pre-planting disturbance, and climatic zone had a
significant effect. Specifically, the positive impact of
afforestation on SOC stocks was more evident in
cropland soils than in pastures or natural grasslands.
Don et al. (2011) found SOC losses from conversion of
primary forest into cropland and perennial crops,
forest into grassland. Furthermore, SOC losses were
partly reversible with afforestation of agricultural
land, and with cropland converted to grassland.
Similar patterns were shown by Li et al. (2012):
carbon stock increased following afforestation on
cropland and pasture. The introduction of the cover
crop in agricultural systems had a significant positive
effect on SOC stock (Poeplau and Don 2015). Those
results suggest that the introduction of trees may
increase SOC stocks, supporting the potential pf
agroforestry. However, despite the availability of
important individual studies, there are few compre-
hensive quantitative reviews on the effect of agro-
forestry on SOC stocks during land-use change
process. The potential of agroforestry in mitigating
climate change is widely recognized, but appears to be
based on opinions rather than quantifiable compre-
hensive data.
Objectives
The main objective of this paper was to investigate the
effects of agroforestry on soil carbon stocks by
summarizing the data from literature using meta-
analytical techniques. Meta-analysis is defined as
statistical procedure that allows one to compare results
from different studies, in order to accomplish a
statistical synthesis, finding common patterns, dis-
crepancies or other interesting relationships that may
come to light in the context of multiple studies
(Borenstein et al. 2009). The core of meta-analysis is
to identify a common measure, reflecting the magni-
tude of the strength of a phenomenon, called the effect
size (Borenstein et al. 2009). Cohen (1992) defined an
effect as a statistical measure that depicts the degree to
which a given event is present in a sample, while
Rosenberg et al. (2000) considered it as the different
measures of a specific effect and their magnitude.
More details about the effect size and its measure will
be provided in Methods section. Specific objectives
were to investigate the effects of: (1) overall agro-
forestry, (2) and specific agroforestry practices on
SOC stocks as consequence of land-use change from
non-agroforestry land uses to agroforestry.
Methods
Data collection
A literature survey of peer-reviewed publications was
carried out using ISI-Web of Science, CAB Direct,
and Google Scholar (Google Inc., Mountain View,
CA, USA). The first step was the identification of the
potential studies to be included in the analysis through
the examination of the abstract, using agroforestry,C
sequestration,soil C sequestration,soil C stock,soil C
pool, and their combination their combination, as
keywords for the search. The result yielded a total 250
observations from 52 publications (see Supplemental
Material), encompassing over 20 countries, mostly
located in Northern, Central, and Southern America,
Africa, and Asia. To be included, studies had to
contain information about soil C concentration or
stocks per unit land area (i.e. Mg of C ha
-1
or in other
Agroforest Syst
123
appropriate equivalent), for both non-agroforestry and
agroforestry land uses (see Supplemental Material).
When C concentration was reported as (Mg C Mg
-1
),
SOC stocks (Mg C ha
-1
) were calculated as follows
(Don et al. 2011):
SOC Stock ¼X
n
i¼1
SOCðMg C Mg1ÞBD SV
ð1Þ
where nis the number of soil layers, BD is soil bulk
density (Mg cm
-3
), and SV is soil volume (m
3
ha
-1
).
Studies considered different soil depths, ranging form
7.5 to 250 cm. SOC content increases with sampling
depth, and its vertical distribution is function of
several features, such as climate, soil texture, clay
content, and vegetation type (Jobba
´gy and Jackson
2000,2001). To reduce the variability due to the
different sampling depths, the dataset was divided into
the following sampling depths: (1) 0–15 cm, (2)
0–30 cm, (3) 0–60 cm, (4) 0–100 cm, and (5) 0
C100 cm. Studies that reported information for
different sampling depths were included in more than
one category. Details about the studies and the number
of observations relative to each sampling depth are
listed in the Supplemental Material.
Variance estimators and weighting function
One of the challenges in conduction of a systematic
review deals with incomplete reporting of outcomes,
which often results in the omission of estimates of
variations—usually standard deviation (SD) or stan-
dard error (SE)—in the included studies (Furukawa
et al. 2006; Wiebe et al. 2006). In meta-analysis, effect
size estimates and related inferences are dependent on
the weighting method (Hungate et al. 2009). The
weighting function conventionally used in meta-
analyses is based on the inverse of pooled variance
(van Groenigen et al. 2011), and omitting studies that
fail to report estimates of variance can be detrimental
for the analysis itself, and may induce bias in the
outcomes (Wiebe et al. 2006). Different approaches
can be used to impute SD, to mitigate any loss in
statistical power and to avoid bias. Authors were
contacted to obtain non-reported or missing data.
When no variance data were obtained, means and
variation estimates from the all studies included in the
meta-analysis were to calculate the coefficient of
variation (CV) as follows (Bracken 1992):
CV ¼SDmean
Meanmean
100 ð2Þ
where SD
mean
and Mean
mean
are the mean of the SD
and the mean of the mean of all the studies included in
the meta-analysis. Missing SD are computed as
follows:
SDmissing ¼CV Meanmissing SD
100 ð3Þ
where Mean
missing SD
is the mean of the study with
missing SD.
This approach allowed to include of all experimen-
tal comparisons contained in the data set.
Categories
Due to the high variability of agroforestry systems and
land uses in general, categories were created. Agro-
forestry systems were grouped according to the
composition and arrangement of the system’s ele-
ments, functions of the system, its socioeconomic
scale, level of management, and its ecological spread
(Nair 1987). Three main agroforestry groups (treat-
ment) were considered: (1) agrisilviculture (crop-
s?trees); (2) silvopastoral (pasture/
animals ?trees); (3) agrosilvopastoral (crops ?pas-
ture/animals ?trees) (Table 1). Non-agroforestry
land uses (control) were grouped into five categories,
according to the information provided by the authors:
(1) agriculture; (2) forest; (3) forest plantation; (4)
pasture/grassland; (5) uncultivated land/other. The last
category includes uncultivated land, marginal or
degraded land, and control (no tree), because of the
lack of an adequate number of studies/observations
needed for using those land uses as standalone
categories. For a similar reason grassland and herbland
(natural ecosystem) were merged with pasture (hu-
man-induced ecosystems) into one single category
(pasture/grassland).
Data analysis
An effect size estimator widely used in meta-analysis
is the magnitude of the experimental treatment mean
(X
e
) relative to the control treatment mean (X
c
). A
common effect size metric is the response ratio
Agroforest Syst
123
R=X
e
/X
c
. To be statistically useful, Ris often log
transformed such that ln R =ln X
e
-ln X
c
. When the
distribution of X
e
and X
c
is normal, and X
c
is unlikely
to be negative, Rwill be approximately normally
distributed with a mean approximately equal to the
true response log ratio (Gurevitch and Hedges 2001).
The effect of agroforestry on SOC stock was quanti-
fied by calculating the natural log of the response ratio
using MetaWin 2.0 (Rosenberg et al. 2000). A random-
effects model was used for analysis of all data sets,
based on the assumption that random variation in SOC
stocks occurred between observations. Mean effect
size for each observation was calculated with 95%
confidence intervals (CIs) generated by a bootstrap-
ping procedures (4999 interactions) (Efron 1979).
Positive effect size indicates a positive impact of
agroforestry. Likewise, the effect of agroforestry is
considered negative for negative values of effect size.
To facilitate the interpretation of effect size, it was
transformed as percentage change ([ln R -1] 9100).
Categorical meta-analysis
One of the specific objectives was to determine the
effect of specific agroforestry practices on SOC
stocks. For agroforestry practices, the difference in
mean response among various categories of agro-
forestry and land-use changes was statistically tested,
following a procedure similar to analysis of variance
(ANOVA). In random-effects models, the heterogene-
ity of true effect size among studies is supposed to be
due to random variation around the mean effect size of
the population of studies, allowing different study-
specific effect sizes (Borenstein et al. 2009). The total
heterogeneity for a group of comparisons (Q
T
)is
partitioned into within-class heterogeneity (Q
w
) and
between class heterogeneity (Q
b
), such that Q
T
=-
Q
w
?Q
b
. To assess whether there is significant
heterogeneity in a sample, Q is tested against a Chi
square distribution with k-1 degrees of freedom,
where kis the number of categories. The null
hypothesis for this test is that all effect sizes are equal
(Gurevitch and Hedges 2001; Borenstein et al. 2009).
A significant Q indicates that the variance among
effect sizes is greater than expected by sampling error
(Cooper 1998). Between-group heterogeneity (Q
b
)
was analyzed across the data and the overall mean and
confidence intervals (CIs) were calculated. Means
were considered significant when the CIs did not
overlap with zero (Gurevitch and Hedges 2001).
Datasets
Natural forests and vegetation have the ability to
increase the SOM and consequently SOC, due to their
organic matter inputs (litter, decomposing material)
and soil protection provided by trees, shrubs, and
perennial plants reduces C losses due to precipitation
leaching and other disturbances factors (Guo and
Gifford 2002; Don et al. 2011; Poeplau and Don
2015). Therefore, outcomes also included results
deriving from the exclusion of observations relative
to conversions from forest (but not forest plantations)
and uncultivated land/other. For each sampling depth,
two meta-analyses were performed: (1) a meta-anal-
ysis investigating the overall effect of agroforestry, (2)
Table 1 Agroforestry grouping according to Nair (1987)
Category Components Systems
Agrisilvicultural
systems
Woody perennials, agricultural
species
Improved fallow species in shifting cultivation, alley cropping, multispecies
tree gardens, multipurpose trees/shrubs farmlands, plantation crops and
other crops, mixtures of plantation crops, shade trees for commercial
plantation crops, agroforestry for fuelwood production, shelterbelts,
windbreaks, riparian buffers, and soil conservation edges
Silvopastoral
systems
Woody perennials, pasture/
animals
Protein bank (multipurpose fodder trees on or around farmlands, living
fences or fodder hedges and shrubs, trees and shrubs on pastures, and
integrated production of animals and wood products
Agrosilvopastoral
systems
Woody perennials, agricultural
species, pasture/animals
Tree-livestock-crop mix around the homestead (home gardens), crop-
animal-wood integrated production, woody hedgerows for browse, green
manure, and soil conservation
Agroforest Syst
123
a meta-analysis investigating the effect of each
agroforestry systems listed in Table 1. Publications
used in the meta-analysis presented various experi-
mental designs such as paired sites, pseudo-replica-
tion, chronosequence, and repeated measures. An
explanation of experimental designs was provided by
Laganiere et al. (2010). In paired site design, a
treatment site is compared to an adjacent control site,
allowing comparisons among different treatments.
Paired site design assumes certain variables to be fixed
between fixed and control sites, and the sampling
constitutes a single measurement in time. A chronose-
quence is defined as the combination of a series of
paired sites, supposedly having similar characteristics,
spread out over time to simulate a succession. The
basic assumption of this design is that each site in the
sequence differs only in age, and that each has the
same history.
Results
0–15 cm
Analysis of the effect size revealed no significant
difference in the transition to agroforestry (Fig. 1a).
However, the conversion from agriculture to agro-
forestry showed significant differences (Fig. 1a),
increasing SOC stocks of 26% (Fig. 1b). The conver-
sion from forest to agroforestry indicated a negative
and significant effect (Fig. 1a), with a decrease of 26%
(Fig. 1b). No significant differences in SOC stocks
were found in the shift from pasture/grassland and
uncultivated/other to agroforestry (Fig. 1a). Remov-
ing natural forest and uncultivated/other from the
analysis the effect size produced a significant effect in
the land-use change towards agroforestry (Fig. 1a),
with a 13% increase in SOC stocks (Fig. 1b). Change
from agriculture to agroforestry increased the SOC
stocks by 25%, while there was no difference in the
conversion from agroforestry to pasture/grassland
(Fig. 1a, b). Test for heterogeneity was not significant
for both full (Q=30.917, df =32, pvalue 0.521) and
reduced datasets (Q=9.509, df =14, p-value 0.797)
(See Supplemental Material).
Looking at specific agroforestry systems, signifi-
cant differences on SOC stocks were detected in the
transition from forest to agrisilviculture (12%
decrease), forest to silvopasture (44% decrease), and
agriculture to agrisilviculture (25% increase), while
the overall agroforestry effect produced no significant
differences (Fig. 2a, b). Land-use change towards
agroforestry showed a significant effect on SOC stocks
(9% increase), when natural forest and uncultivated/
other were removed from the analysis (Fig. 2a, b). No
noticeable variations were observed in the transition
from pasture/grassland to agrisilviculture, pas-
ture/grassland to silvopasture, and agriculture to
agrisilviculture (Fig. 2a, b). Test for heterogeneity
was significant for the full dataset (Q=62.949,
df =32, p-value 0.001), while was not significant
for the reduced dataset (Q=9.199, df =14, p-value
0.818) (See Supplemental Material).
Fig. 1 Effects of agroforestry (AF) on SOC stocks (0–15 cm
sampling depth). On the left avalues are effect size and 95%
bootstrap confidence intervals (CIs). On the right bvalues are
percentage change of SOC stocks. Effect size is considered
significant when confidence intervals did not overlap with zero.
Numbers in parentheses indicate number of observations, and *
denotes results generated when land-use changes from forest
and uncultivated/other to agroforestry were excluded from the
analysis
Agroforest Syst
123
0–30 cm
Land-use change to agroforestry denoted a positive
and significant effect size on SOC stocks, with an
increase of 12% (Fig. 3a, b). Other significant effect
sizes on SOC stocks were observed in the transition
from forest to agroforestry (22% decrease), pas-
ture/grassland to agroforestry (9% increase), agricul-
ture to agroforestry (40% increase), and uncultivated/
other to agroforestry (25% increase), (Fig. 3a, b). The
removal of forest and natural/uncultivated categories
confirmed the positive trend, increasing the magnitude
of effect size for the conversion to overall agroforestry
(26% increase), while the results for pasture/grassland
and agriculture to agroforestry were identical (Fig. 3a,
b). Test for heterogeneity was not significant for both
the full (Q=30.986, df =44, p-value 0.931) and the
reduced dataset (Q=11.229, df =27, p-value 0.997)
(See Supplemental Material).
Significant differences on SOC stocks were
detected in the transition from forest to agrisilviculture
(24% decrease), agriculture to agrisilviculture (40%
increase), pasture/grassland to agrosilvopastoral sys-
tems (13% increase), uncultivated/other to agrisilvi-
culture (55% increase), uncultivated/other to
agrosilvopastoral systems (7% increase), and overall
agroforestry effect (12%) (Fig. 4a, b). Removing
forest and uncultivated/other form the analysis dou-
bled the SOC stocks % change due to overall
agroforestry effect (from 12% to 25%) (Fig. 4a, b).
Fig. 2 Effects of specific agroforestry systems on SOC stocks
(0–15 cm sampling depth). On the left avalues are effect size
and 95% bootstrap confidence intervals (CIs). On the right
bvalues are percentage change of SOC stocks. Effect size is
considered significant when confidence intervals did not overlap
with zero. Numbers in parentheses indicate number of
observations, and * denotes results generated when land-use
changes from forest and uncultivated/other to agroforestry were
excluded from the analysis
Fig. 3 Effects of agroforestry (AF) on SOC stocks (0–30 cm
sampling depth). On the left avalues are effect size and 95%
bootstrap confidence intervals (CIs). On the right bvalues are
percentage change of SOC stocks. Effect size is considered
significant when confidence intervals did not overlap with zero.
Numbers in parentheses indicate number of observations, and *
denotes results generated when land-use changes from forest
and uncultivated/other to agroforestry were excluded from the
analysis
Agroforest Syst
123
Test for heterogeneity was not significant for both the
full (Q=29.230, df =42, p-value 0.918) and the
reduced dataset (Q=10.916, df =26, p-value 0.996)
(See Supplemental Material).
0–60 cm
The conversion of pasture/grassland to agroforestry
significantly increased SOC stocks by 10% (Fig. 5a,
b). On the other hand, the conversion of uncultivated/
other to agroforestry decreased SOC stocks by (23%).
No significant effect sizes on SOC stocks were found
in the conversion from forest to agroforestry, forest
plantation to agroforestry, agriculture to agroforestry,
and in the overall agroforestry effect (Fig. 5a, b). The
removal of forest and uncultivated/other did not
noticeably affected the previous outcomes, but
changed the overall agroforestry effect from not
significant to significant, increasing SOC stocks by
10% (Fig. 5a, b). Test for heterogeneity was signifi-
cant for both the full (Q=101.479, df =50, p-value
0.000) and the reduced dataset (Q=59.030, df =35,
p-value 0.006) (See Supplemental Material).
Significant changes in SOC stocks were reported in
the conversion from forest plantation to silvopasture
(17% increase), forest to agrosilvopastoral systems
(27% decrease), agriculture to agrosilvopastoral sys-
tems (21% increase), agriculture to silvopasture (66%
increase), pasture/grassland to agrisilviculture (8%
increase) (Fig. 6a, b). The removal of forest and
uncultivated/other changed the effect size of overall
agroforestry from not significant to significant, with an
11% increase in SOC stocks (Fig. 6a, b). Test for
heterogeneity was significant for both the full
Fig. 4 Effects of specific agroforestry systems on SOC stocks
(0–30 cm sampling depth). On the left avalues are effect size
and 95% bootstrap confidence intervals (CIs). On the right
bvalues are percentage change of SOC stocks. Effect size is
considered significant when confidence intervals did not overlap
with zero. Numbers in parentheses indicate number of
observations, and * denotes results generated when land-use
changes from forest and uncultivated/other to agroforestry were
excluded from the analysis
Fig. 5 Effects of agroforestry (AF) on SOC stocks (0–60 cm
sampling depth). On the left avalues are effect size and 95%
bootstrap confidence intervals (CIs). On the right bvalues are
percentage change of SOC stocks. Effect size is considered
significant when confidence intervals did not overlap with zero.
Numbers in parentheses indicate number of observations, and *
denotes results generated when land-use changes from forest
and uncultivated/other to agroforestry were excluded from the
analysis
Agroforest Syst
123
(Q=113.950, df =47, p-value 0.000) and the
reduced dataset (Q=65.062, df =34, p-value
0.001) (See Supplemental Material).
0–100 cm
The land-use change from agriculture to agroforestry
significantly increased the SOC stocks by 34%, and
the overall effect of agroforestry on SOC stocks was
found significant (8% increase) (Fig. 7a, b). On the
other hand, there was no significant difference in the
land-use change from forest, pasture/grassland, forest
plantation, and uncultivated/other to agroforestry
(Fig. 7a, b). Removing forest and uncultivated/other
from the analysis increased the SOC stocks due the
overall agroforestry effect by 5 percentage points
(from 8 to 13%) (Fig. 7a, b). Test for heterogeneity
was significant for both the full dataset (Q=80.108,
df =80, p-value 0.476), while was significant for the
reduced dataset (Q=84.383, df =51, p-value 0.002)
(See Supplemental Material).
The effect of agroforestry was found significant and
positive, showing an increase SOC stock of 9%
(Fig. 8a, b). Significant effect on land-use change on
SOC stocks were found in the conversion from
agriculture to agrisilviculture (10% increase), agricul-
ture to agrosilvopastoral systems (30% increase),
pasture to agrisilviculture (11% decrease), forest to
agrosilvopastoral systems (36% decrease), forest to
silvopasture (32% increase), agriculture to silvopas-
ture (31% increase), uncultivated/other to silvopasture
(47% decrease), and forest plantation to silvopasture
Fig. 6 Effects of specific agroforestry systems on SOC stocks
(0–60 cm sampling depth). On the left avalues are effect size
and 95% bootstrap confidence intervals (CIs). On the right
bvalues are percentage change of SOC stocks. Effect size is
considered significant when confidence intervals did not overlap
with zero. Numbers in parentheses indicate number of
observations, and * denotes results generated when land-use
changes from forest and uncultivated/other to agroforestry were
excluded from the analysis
Fig. 7 Effects of agroforestry (AF) on SOC stocks (0–100 cm
sampling depth). On the left avalues are effect size and 95%
bootstrap confidence intervals (CIs). On the right bvalues are
percentage change of SOC stocks. Effect size is considered
significant when confidence intervals did not overlap with zero.
Numbers in parentheses indicate number of observations, and *
denotes results generated when land-use changes from forest
and uncultivated/other to agroforestry were excluded from the
analysis
Agroforest Syst
123
(3% increase) (Fig. 8a, b). Removing forest and
uncultivated/other had a positive influence on overall
agroforestry effect, increasing the SOC stocks from 9
to 14% (Fig. 8a, b). Test for heterogeneity was
significant for both the full (Q=113.303, df =79,
p-value 0.006) and the reduced dataset (Q=86.623,
df =50, p-value 0.001) (See Supplemental Material).
0C100 cm
No significant influence on SOC stocks was detected
in land-use conversions to agroforestry, nor in the
overall effect of agroforestry (Fig. 9a, b). The removal
of forest and uncultivated/other from the analysis did
not affect the previous outcome (Fig. 9a, b). Test for
heterogeneity was significant for both the full
(Q=110.457, df =65, p-value 0.000) and the
reduced dataset (Q=47.838, df =30, p-value
0.021) (See Supplemental Material).
The same non-significant results were observed in
the effect of specific agroforestry systems on SOC
stocks with or without the removal of forest and
uncultivated/other from the analysis (Fig. 10a, b). Test
for heterogeneity was significant for both the full
(Q=105.846, df =64, p-value 0.001) and the
reduced dataset (Q=42.744, df =29, p-value
0.048) (See Supplemental Material).
Fig. 8 Effects of specific agroforestry systems on SOC stocks
(0–100 cm sampling depth). On the left avalues are effect size
and 95% bootstrap confidence intervals (CIs). On the right
bvalues are percentage change of SOC stocks. Effect size is
considered significant when confidence intervals did not overlap
with zero. Numbers in parentheses indicate number of
observations, and * denotes results generated when land-use
changes from forest and uncultivated/other to agroforestry were
excluded from the analysis
Fig. 9 Effects of agroforestry (AF) on SOC stocks
(0 C100 cm sampling depth). On the left avalues are effect
size and 95% bootstrap confidence intervals (CIs). On the right
bvalues are percentage change of SOC stocks. Effect size is
considered significant when confidence intervals did not overlap
with zero. Numbers in parentheses indicate number of
observations, and * denotes results generated when land-use
changes from forest and uncultivated/other to agroforestry were
excluded from the analysis
Agroforest Syst
123
Discussion
Overall effect of agroforestry
According to our results and considering a full dataset
with forest and uncultivated/other land-uses included,
agroforestry revealed a significant and positive effect
on SOC stocks at 0–30 and 0–100 cm depths. In the
reduced dataset, the significant positive effect of
agroforestry was observed at all depths, except for
0C100 cm. Overall, and confirming other studies,
incorporating trees on land leads to an increase in SOC
stocks (Haile et al. 2008; Nair et al. 2009a). This is
possibly due to changes in quantity and quality of litter
inputs (Jobba
´gy and Jackson 2000), and soil charac-
teristics related to SOC sequestration dynamics and
storage, such as humification, aggregation, transloca-
tion of biomass into subsoil by root system, and
leaching of inorganic C into groundwater (Lal 2001).
Nair et al. (2009a) ranked SOC stocks as follows:
forests [agroforests [tree plantations [arable
crops, and our findings seemed to be in line with the
pattern.
Forest to agroforestry
Land-use conversion from forest to agroforestry
decreased the SOC stocks at 0–15 and 0–30 cm
depths, while no significant effect was detected at the
other investigated depths. The trend was somehow
expected, since forests retain the most part of their
SOC in the topsoil, and generally the conversion from
forest to another land-use, such as agriculture, caused
loss in SOC, especially in those upper layers (Brown
and Lugo 1990; Guo and Gifford 2002; Leuschner
et al. 2013). Specifically, significant decreases in SOC
carbon were observed in the conversion from forest to
agrisilviculture (0–15, 0–30 cm). Despite the presence
of woody plants, agroforestry systems lack diversifi-
cation, density, and structural complexity typical of
natural ecosystems. However, the effect of natural
vegetation on SOC is less evident in deeper layers,
where no significant changes in SOC stocks were
observed in the land-use change from forest to
agrisilviculture (0–60, 0–100, 0 C100 cm), and
forest to agrosilvopasture (0–100 cm). On the other
hand, perennial grasses present in silvopastoral sys-
tems seem to be more efficient than woody plants in
storing C in soil. Generally trees deposit a larger
fraction of OM on the soil surface than grasses; here
the decomposition process is dominant, and might lead
to less formation of SOM and consequently less SOC
(Post and Kwon 2000; Guo and Gifford 2002). Higher
SOC stocks have been found in grasslands than forests
(Brown and Lugo 1990; Jobba
´gy and Jackson 2000;
Conant et al. 2001). This study showed silvopastoral
systems having increased SOC stocks than forest at
0–100 cm of depth.
Agriculture to agroforestry
Findings suggested that the conversion of agricultural
land to agroforestry significantly increased SOC
stocks at 0–15, 0–30, 0–100, but not at 0–60
and [100 cm. Different authors (Brown and Lugo
1990; Guo and Gifford 2002) also indicated that in the
Fig. 10 Effects of specific agroforestry systems on SOC stocks
(0 C100 cm sampling depth). On the left avalues are effect
size and 95% bootstrap confidence intervals (CIs). On the right
bvalues are percentage change of SOC stocks. Effect size is
considered significant when confidence intervals did not overlap
with zero. Numbers in parentheses indicate number of
observations, and * denotes results generated when land-use
changes from forest and uncultivated/other to agroforestry were
excluded from the analysis
Agroforest Syst
123
conversion from cropland to systems with trees, C
sequestration increased. Again, trees appear to have a
positive effect on SOC, thanks to higher inputs, deeper
deposition, and reduced decomposability of OM (Post
and Kwon 2000). Successful examples of SOC stock
recovery after afforestation and reforestation are
common in literature (Detwiler 1986; Houghton
1995; Don et al. 2011). Generally, agroforestry reduce
tillage and soil disturbance regimes, which can help to
maintain or even increase SOC pools (Aslam et al.
1999). Under proper agroforestry management, con-
siderable litter inputs and vegetation residues from
pruning are returned to the soil, which might increase
SOC (Montagnini and Nair 2004). However, the
results for 0–60 and 0 C100 cm deviated from the
general trend, and further investigation with additional
studies included are recommended. Among agro-
forestry systems, positive significant increases of
SOC stocks were observed in the change from
agriculture to agrisilviculture (0–15, 0–30,
0–100 cm), agriculture to agrosilvopastoral systems
(0–60, 0–100 cm), and agriculture to silvopasture
(0–100 cm). The inclusion of perennial grasses in
agrosilvopastoral systems and silvopasture seems to
increase SOC stocks as a consequence of their highly
developed root system, allowing higher belowground
translocation of C (Kuzyakov and Domanski 2000). In
addition, grasses provide a continuous soil coverage,
reducing soil temperatures, and can have higher
belowground productivity and turnover rates that
increase SOM (Brown and Lugo 1990; Conant et al.
2001).
Pasture/grassland to agroforestry
The establishment of agroforestry systems requires a
certain level of inputs, although minimized under
sustainable management, such as site preparation, tree
planting, and other activities, which could reduce
original grasslands and pastures SOC stocks (Guo and
Gifford 2002). Findings seems to support agroforestry:
no significant differences were observed in the con-
version from pasture/grassland to agroforestry (0–15,
0–30, 0–100, 0 C100 cm), while a significant
increase was observed at 0–60 cm. SOC stocks were
found significantly reduced when pasture/grassland
were converted into agrisilvicultural systems
(0–100 cm), probably due to the lack of perennial
grasses. In pastures, the annual turnover of organic
matter from roots is higher than trees, which deposit
more recalcitrant material (Jobba
´gy and Jackson
2001), increasing SOC stocks. Legumes are often
sown to increase forage production, and they can
increase soil nitrogen, improve soil fertility, increase
belowground productivity, and consequently below-
ground C inputs (Watson 1963; Vallis 1972; Boddey
et al. 1997; Conant et al. 2001).
Uncultivated/other to agroforestry
The category encompassed all land-uses that are not
classified as forest by the authors and other without
additional information available, such as bare land or
control plots. Results indicated no significant differ-
ence (0–15, 0–100, 0 C100 cm), significant increase
(0–30 cm), and significant decrease (0–60 cm) in SOC
stocks. The conversion from uncultivated/other to
agrisilviculture and agrosilvopastoral systems in SOC
at 0–30 cm, while a significant reduction was
observed in the transition from uncultivated/other to
agrisilviculture at 0–60 cm. Findings were contrasting
for the top layers, where both significant and not
significant effects were detected. It is possible that the
wide variation of the land-uses included in unculti-
vated/other category was somehow responsible of the
diverging results. Hence, more detailed information
about control groups is needed, in particular avoiding
land-uses such as bare land or no vegetation as only
control group.
Forest plantation to agroforestry
No significant differences in SOC stocks were
detected in the conversion from forest plantation to
agroforestry at 0–60 and 0–100 cm. Similarly, the
land-use change from forest plantation to silvopasture
did not produce significant results at the same depths.
Unfortunately, the database for this category had
observations only from a few studies (see Supple-
mental Material for more information). Guo and
Gifford (2002) observed a decrease in SOC stocks in
the transition from pasture to forest plantation, recog-
nizing the negative impacts of site preparation activ-
ities, which can break soil aggregates, disturb soil
structure, and disrupt physical protection provided by
vegetation coverage. They also found how tree species
may influence C storage: decreases in SOC stocks
were observed in the conversion of pastures with
Agroforest Syst
123
conifers plantations, while little effects were reported
in broadleaf plantations. Although fertilization inputs
can enhance decomposition and reduce belowground
C allocation (Haynes and Gower 1995), N-fixing
species and fertilization were found, in certain case, to
increase SOC sequestration rates, due to the additional
nitrogen inputs (Ewers et al. 1996; Guo and Gifford
2002). Therefore, species and management practices
may play an important role in SOC stocks in
agroforestry systems compared to forest plantations.
Methodological issues
In the last decades studies investigating agroforestry
SOC stocks have pointed out critical methodological
challenges linked to sampling, analysis, computations,
and interpretation of results (Nair 2011). The hetero-
geneous nature of agroforestry in terms of site, soil
type, tree species, and management practices, pro-
duced an inconsistent estimate of agroforestry C
stocks (Montagnini and Nair 2004; Nair
2011,2012). Nair (2011,2012) have largely discussed
about methodological discrepancies and issues related
to (1) soil sampling depth, (2) preparation of samples,
(3) experimental design, and (4) calculation/presenta-
tion of results, pointing out the lack of rigorous
standards. The consistency of effects across studies is
a desirable requisite in meta-analysis (Higgins et al.
2003) and Liberati et al. (2001) advocated for it. The
lack of standardized methodologies certainly influ-
ences findings, in particular heterogeneity, which
might bias outcomes (Borenstein et al. 2009). Incon-
sistency across the studies is noticeable in wide
confidence intervals, such as in Fig. 10a, b. One way
to deal with the problem is to include more precise and
consistent studies: increasing the sample size
increases precision, reduces the variation, and the
point estimate is more likely to represent the true mean
value in the investigated population (Zlowodzki et al.
2007). However, the available database is unfortu-
nately limited. The adoption of consistent standards in
agroforestry research will increase the precision of
future studies, and strongly suggest it includes (1)
variance estimators, (2) detailed information about
previous land-uses, and (3) follow the guidelines
suggested by Nair (2011,2012). Also, studies should
include important explanatory variables, such as age
of the systems and time to have C sequestration rates
(quantity of C per area unit per time) rather than SOC
stock quantities (Kim et al. 2016), adopted manage-
ment practices (fertilization, irrigation, tillage, etc.),
depth-relative BD, soil texture, pH, silt and clay
content values (Brown and Lugo 1990; Laganiere
et al. 2010), climatic factors (Brown and Lugo 1990);
and vegetation species (Guo and Gifford 2002).
Conclusions
The conversion from forest to agroforestry lead to
losses in SOC stocks in the top layers, while no
significant differences were detected when deeper
layers were included. On the other hand, the conver-
sion from agriculture to agroforestry increased SOC
stocks in most of the cases. Significant increases were
also observed in the transition from pasture/grassland
to agroforestry in the top layers, especially with the
inclusion perennial in the systems, such as in sil-
vopasture and agrosilvopastoral systems. Finally, the
conversion from uncultivated/other land-uses to agro-
forestry produced inconsistent results, perhaps due to
the high variability of the category, and the little
available land-use history. Overall, SOC stocks
increased when land-use changed from less complex
systems, such as agricultural systems.
The purpose of the study was to provide an
empirical foundation to support agroforestry systems
as a strategy to reduce atmospheric CO
2
concentration
and mitigate climate change, as proposed by several
authors (Albrecht and Kandji 2003; Montagnini and
Nair 2004; Nair et al. 2009a,b,2010; Mosquera-
Losada et al. 2011; Nair 2012). However, important
methodological issues, lack of information, and
knowledge gaps might bias the outcome of the meta-
analysis. Specific efforts are needed to build a more
robust database for future research, in particular the
adoption of unified and rigorous standards in study
design, sampling collection and preparation, informa-
tion completeness, and data presentation.
References
Albrecht A, Kandji ST (2003) Carbon sequestration in tropical
agroforestry systems. Agric Ecosyst Environ 99:15–27.
https://doi.org/10.1016/S0167-8809(03)00138-5
Allen JC (1985) Soil response to forest clearing in the United
States and the tropics: geological and biological factors.
Biotropica 17:15–27. https://doi.org/10.2307/2388373
Agroforest Syst
123
Aslam T, Choudhary MA, Saggar S (1999) Tillage impacts on
soil microbial biomass C, N and P, earthworms and
agronomy after 2 years of cropping following permanent
pasture in New Zealand. Soil Tillage Res 51:103–111.
https://doi.org/10.1016/S0167-1987(99)00032-X
Batjes NH (1996) Total carbon and nitrogen in the soils of the
world. Eur J Soil Sci 47:151–163. https://doi.org/10.1111/j.
1365-2389.1996.tb01386.x
Berthrong ST, Jobba
´gy EG, Jackson RB (2009) A global meta-
analysis of soil exchangeable cations, pH, carbon, and
nitrogen with afforestation. Ecol Appl 19:2228–2241.
https://doi.org/10.1890/08-1730.1
Boddey RM, De Moraes Sa
´JC, Alves BJR, Urquiaga S (1997)
The contribution of biological nitrogen fixation for sus-
tainable agricultural systems in the tropics. Soil Biol Bio-
chem 29:787–799. https://doi.org/10.1016/S0038-0717
(96)00221-0
Borenstein M, Hedges LV, Higgins JPT, Rothstein HR (2009)
Introduction to meta-analysis. Wiley, Chichester
Bracken MB (1992) Statistical methods for analysis of effects of
treatment in overviews of randomized trials. Effective care
of the newborn infant. Oxford University Press, Oxford,
pp 13–18
Brown S, Lugo AE (1990) Effects of forest clearing and suc-
cession on the carbon and nitrogen content of soils in
Puerto Rico and US Virgin Islands. Plant Soil 124:53–64.
https://doi.org/10.1007/BF00010931
Burras CL, Kimble J, Lal R et al (2001) Carbon sequestration:
position of the soil science society of America. Carbon Seq
Position Soil Sci Soc Am 65:1–4
Cohen J (1992) A power primer. Psychol Bull 112:155
Conant RT, Paustian K, Elliott ET (2001) Grassland manage-
ment and conversion into grassland: effects on soil carbon.
Ecol Appl 11:343–355. 10.1890/1051-
0761(2001)011[0343:GMACIG]2.0.CO;2
Cooper HM (1998) Synthesizing research: a guide for literature
reviews. Sage, New York
Detwiler RP (1986) Land use change and the global carbon
cycle: the role of tropical soils. Biogeochemistry 2:67–93.
https://doi.org/10.1007/BF02186966
Don A, Schumacher J, Freibauer A (2011) Impact of tropical
land-use change on soil organic carbon stocks—a meta-
analysis. Glob Change Biol 17:1658–1670. https://doi.org/
10.1111/j.1365-2486.2010.02336.x
Efron B (1979) Bootstrap methods: another look at the jack-
knife. Ann Stat 7:1–26
Ewers B, Binkley D, Bashkin M (1996) Influence of adjacent
stand on spatial patterns of soil carbon and nitrogen in
Eucalyptus and Albizia plantations. Can J For Res
26:1501–1503. https://doi.org/10.1139/x26-167
Fearnside PM, Imbrozio Barbosa R (1998) Soil carbon changes
from conversion of forest to pasture in Brazilian Amazonia.
For Ecol Manag 108:147–166. https://doi.org/10.1016/
S0378-1127(98)00222-9
Furukawa TA, Barbui C, Cipriani A et al (2006) Imputing
missing standard deviations in meta-analyses can provide
accurate results. J Clin Epidemiol 59:7–10. https://doi.org/
10.1016/j.jclinepi.2005.06.006
Guo LB, Gifford RM (2002) Soil carbon stocks and land use
change: a meta analysis. Glob Change Biol 8:345–360.
https://doi.org/10.1046/j.1354-1013.2002.00486.x
Gurevitch J, Hedges L V. (2001) Meta-analysis: combing the
results of independent experiments. In: Design and analysis
of ecological experiments, 2nd edn. Oxford University
Press, Oxford
Haile SG, Nair PKR, Nair VD (2008) Carbon storage of dif-
ferent soil-size fractions in Florida silvopastoral systems.
J Environ Qual 37:1789. https://doi.org/10.2134/jeq2007.
0509
Haynes BE, Gower ST (1995) Belowground carbon allocation
in unfertilized and fertilized red pine plantations in north-
ern Wisconsin. Tree Physiol 15:317–325
Higgins JPT, Thompson SG, Deeks JJ, Altman DG (2003)
Measuring inconsistency in meta-analyses. BMJ Br Med J
327:557–560
Houghton RA (1995) Changes in the storage of terrestrial car-
bon since 1850. Soils Glob Change 5:45–65
Hungate BA, van Groenigen KJ, Six J et al (2009) Assessing the
effectof elevated carbon dioxide on soilcarbon: a comparison
of four meta-analyses. Glob Chang Biol 15:2020–2034.
https://doi.org/10.1111/j.1365-2486.2009.01866.x
Hutchinson JJ, Campbell CA, Desjardins RL (2007) Some
perspectives on carbon sequestration in agriculture. Agric
For Meteorol 142:288–302. https://doi.org/10.1016/j.
agrformet.2006.03.030
Jobba
´gy EG, Jackson RB (2000) The vertical distribution of soil
organic carbon and its relation to climate and vegetation.
Ecol Appl 10:423–436. 10.1890/1051-0761(2000)010
[0423:TVDOSO]2.0.CO;2
Jobba
´gy EG, Jackson RB (2001) The distribution of soil nutri-
ents with depth: global patterns and the imprint of plants.
Biogeochemistry 53:51–77. https://doi.org/10.1023/A:
1010760720215
Jose S, Bardhan S (2012) Agroforestry for biomass production
and carbon sequestration: an overview. Agrofor Syst
86:105–111. https://doi.org/10.1007/s10457-012-9573-x
Kim D-G, Kirschbaum MUF, Beedy TL (2016) Carbon
sequestration and net emissions of CH4 and N2O under
agroforestry: synthesizing available data and suggestions
for future studies. Agric Ecosyst Environ 226:65–78.
https://doi.org/10.1016/j.agee.2016.04.011
Kirby KR, Potvin C (2007) Variation in carbon storage among
tree species: implications for the management of a small-
scale carbon sink project. For Ecol Manag 246:208–221.
https://doi.org/10.1016/j.foreco.2007.03.072
Kuzyakov Y, Domanski G (2000) Carbon input by plants into
the soil. Review. J Plant Nutr Soil Sci 163:421–431. https://
doi.org/10.1002/1522-2624(200008)163:4\421:AID-
JPLN421[3.0.CO;2-R
Laganiere J, Angers DA, Pare D (2010) Carbon accumulation in
agricultural soils after afforestation: a meta-analysis. Glob
Chang Biol 16:439–453. https://doi.org/10.1111/j.1365-
2486.2009.01930.x
Lal R (2001) Potential of desertification control to sequester
carbon and mitigate the greenhouse effect. Clim Change
51:35–72. https://doi.org/10.1023/A:1017529816140
Lal R (2004a) Soil carbon sequestration in natural and managed
tropical forest ecosystems. J Sustain For 21:1–30. https://
doi.org/10.1300/J091v21n01_01
Lal R (2004b) Soil carbon sequestration impacts on global cli-
mate change and food security. Science 304:1623–1627.
https://doi.org/10.1126/science.1097396
Agroforest Syst
123
Lal R (2004c) Soil carbon sequestration to mitigate climate
change. Geoderma 123:1–22. https://doi.org/10.1016/j.
geoderma.2004.01.032
Lal R (2004d) Soil carbon sequestration in natural and managed
tropical forest ecosystems. J Sustain For 21:1–30. https://
doi.org/10.1300/J091v21n01_01
Lal R (2007) Soil carbon stocks under present and future climate
with specific reference to European ecoregions. Nutr Cycl
Agroecosyst 81:113–127. https://doi.org/10.1007/s10705-
007-9147-x
Leuschner C, Moser G, Hertel D et al (2013) Conversion of
tropical moist forest into cacao agroforest: consequences
for carbon pools and annual C sequestration. Agrofor Syst
87:1173–1187. https://doi.org/10.1007/s10457-013-9628-7
Li DJ, Niu SL, Luo YQ (2012) Global patterns of the dynamics
of soil carbon and nitrogen stocks following afforestation:
a meta-analysis. New Phytol 195:172–181. https://doi.org/
10.1111/j.1469-8137.2012.04150.x
Liberati A, Buzzetti R, Grilli R et al (2001) Which guidelines
can we trust?: assessing strength of evidence behind rec-
ommendations for clinical practice. West J Med
174:262–265
Lorenz K, Lal R (2005) The depth distribution of soil organic
carbon in relation to land use and management and the
potential of carbon sequestration in subsoil horizons. In:
Sparks DL (ed) Advances in agronomy, vol 88. Academic
Press, New York, pp 35–66
Lorenz K, Lal R (2014) Soil organic carbon sequestration in
agroforestry systems. A review. Agron Sustain Dev
34:443–454. https://doi.org/10.1007/s13593-014-0212-y
Mann LK (1986) Changes in soil carbon storage after cultiva-
tion. Soil Sci 142:288–289. https://doi.org/10.1097/
00010694-198611000-00006
Montagnini F, Nair PKR (2004) Carbon sequestration: an
underexploited environmental benefit of agroforestry sys-
tems. Agrofor Syst 61–2:281–295. https://doi.org/10.1023/
b:agfo.0000029005.92691.79
Mosquera-Losada MR, Freese D, Rigueiro-Rodrı
´guez A (2011)
Carbon sequestration in European Agroforestry Systems B.
In: Kumar MB, Nair RPK (eds) carbon sequestration
potential of agroforestry systems: opportunities and chal-
lenges. Springer, Dordrecht, pp 43–59
Nair PKR (1987) Agroforestry systems inventory. Agrofor Syst
5:301–317. https://doi.org/10.1007/BF00119128
Nair PKR (2011) Methodological challenges in estimating
carbon sequestration potential of agroforestry systems. In:
Kumar BM, Nair PKR (eds) Carbon sequestration potential
of agroforestry systems. Springer, Dordrecht, pp 3–16
Nair PKR (2012) Carbon sequestration studies in agroforestry
systems: a reality-check. Agrofor Syst 86:243–253. https://
doi.org/10.1007/s10457-011-9434-z
Nair PKR, Kumar BM, Nair VD (2009a) Agroforestry as a
strategy for carbon sequestration. J Plant Nutr Soil Sci Fur
Pflanzenernahrung Und Bodenkd 172:10–23. https://doi.
org/10.1002/jpln.200800030
Nair PKR, Nair VD, Kumar BM, Haile SG (2009b) Soil carbon
sequestration in tropical agroforestry systems: a feasibility
appraisal. Environ Sci Policy 12:1099–1111. https://doi.
org/10.1016/j.envsci.2009.01.010
Nair PKR, Nair VD, Kumar BM, Showalter JM (2010) Carbon
sequestration in agroforestry systems. In: Sparks DL (ed)
Advances in agronomy, vol 108. Academic Press, New
York, pp 237–307
Neill C, Melillo JM, Steudler PA et al (1997) Soil carbon and
nitrogen stocks following forest clearing for pasture in the
southwestern Brazilian Amazon. Ecol Appl 7:1216–1225.
https://doi.org/10.2307/2641209
Oelbermann M, Paul Voroney R, Gordon AM (2004) Carbon
sequestration in tropical and temperate agroforestry sys-
tems: a review with examples from Costa Rica and
southern Canada. Agric Ecosyst Environ 104:359–377.
https://doi.org/10.1016/j.agee.2004.04.001
Poeplau C, Don A (2015) Carbon sequestration in agricultural
soils via cultivation of cover crops—a meta-analysis. Agric
Ecosyst Environ 200:33–41. https://doi.org/10.1016/j.
agee.2014.10.024
Post WM, Kwon KC (2000) Soil carbon sequestration and land-
use change: processes and potential. Glob Chang Biol
6:317–327
Rosenberg MS, Adams DC, Gurevitch J (2000) MetaWin: sta-
tistical software for meta-analysis. Sinauer Associates
Sunderland, Sunderland
Roshetko J, Delaney M, Hairiah K, Purnomosidhi P (2002)
Carbon stocks in Indonesian homegarden systems: can
smallholder systems be targeted for increased carbon
storage? Am J Altern Agric 17:138–148. https://doi.org/10.
1079/AJAA200116
Sanchez PA (2000) Linking climate change research with food
security and poverty reduction in the tropics. Agric Ecosyst
Environ 82:371–383. https://doi.org/10.1016/S0167-
8809(00)00238-3
Schlesinger WH (1986) Changes in soil carbon storage and
associated properties with disturbance and recovery. In:
Trabalka JR, Reichle DE (eds) The changing carbon cycle:
a global analysis. Springer, New York, pp 194–220
Sollins P, Swanston C, Kramer M (2007) Stabilization and
destabilization of soil organic matter—a new focus. Bio-
geochemistry 85:1–7. https://doi.org/10.1007/s10533-007-
9099-x
Stavi I, Lal R (2013) Agroforestry and biochar to offset climate
change: a review. Agron Sustain Dev 33:81–96. https://doi.
org/10.1007/s13593-012-0081-1
Vallis I (1972) Soil nitrogen changes under continuously grazed
legume-grass pastures in subtropical coastal Queensland.
Aust J Exp Agric 12:495–501
van Groenigen KJ, Osenberg CW, Hungate BA et al (2011)
Increased soil emissions of potent greenhouse gases under
increased atmospheric CO
2
. Nature 475:214–216. https://
doi.org/10.1038/nature10176
Watson ER (1963) The influence of subterranean clover pas-
trues on soil fertility. I. Short-term effects. Aust J Agric Res
14:796–807
Wiebe N, Vandermeer B, Platt RW et al (2006) A systematic
review identifies a lack of standardization in methods for
handling missing variance data. J Clin Epidemiol
59:342–353. https://doi.org/10.1016/j.jclinepi.2005.08.017
Young A (1997) Agroforestry for soil conservation, 2nd edn.
CAB International, Wallingford
Zlowodzki M, Poolman RW, Kerkhoffs GM et al (2007) How to
interpret a meta-analysis and judge its value as a guide for
clinical practice. Acta Orthop 78:598–609
Agroforest Syst
123
A preview of this full-text is provided by Springer Nature.
Content available from Agroforestry Systems
This content is subject to copyright. Terms and conditions apply.