You are on page 1of 6

Articles

https://doi.org/10.1038/s41893-018-0062-8

There are amendments to this paper

Climate-smart sustainable agriculture in low-to-


intermediate shade agroforests
W. J. Blaser 1
*, J. Oppong2, S. P. Hart 3
, J. Landolt1, E. Yeboah2 and J. Six1

Meeting demands for agricultural production while maintaining ecosystem services, mitigating and adapting to climate change
and conserving biodiversity will be a defining challenge of this century. Crop production in agroforests is being widely imple-
mented with the expectation that it can simultaneously meet each of these goals. But trade-offs are inherent to agroforestry
and so unless implemented with levels of canopy cover that optimize these trade-offs, this effort in climate-smart, sustainable
intensification may simply compromise both production and ecosystem services. By combining simultaneous measurements
of production, soil fertility, disease, climate variables, carbon storage and species diversity along a shade-tree cover gradient,
here we show that low-to-intermediate shade cocoa agroforests in West Africa do not compromise production, while creating
benefits for climate adaptation, climate mitigation and biodiversity. As shade-tree cover increases above approximately 30%,
agroforests become increasingly less likely to generate win–win scenarios. Our results demonstrate that agroforests cannot
simultaneously maximize production, climate and sustainability goals but might optimise the trade-off between these goals at
low-to-intermediate levels of cover.

H
ow agriculture is implemented will be one of the strongest primary forests20,22,23. Trade-offs are, therefore, inherent to agrofor-
determinants of global human and environmental health estry. Unless agroforests are implemented in a way that optimizes
over the coming century, and the challenges are formidable. these trade-offs, this effort in climate-smart, sustainable intensifica-
Meeting food security needs for the world’s population requires tion may simply compromise our ability to meet production, cli-
sustained production, which requires healthy soils, effective control mate, sustainability and conservation goals.
of pests and disease and adaptation to new climates1–3. Meanwhile, The contribution of agroforests to the future of agriculture there-
given agriculture’s particular vulnerability to climate change4,5, and fore requires understanding the balance of their costs and benefits.
an already vast global footprint2,6, best-practice agriculture should First, this requires simultaneous measurements of the multiple
mitigate against further, avoidable climate change, while also lim- effects of agroforests to identify the trade-offs inherent to their
iting impacts on biodiversity and other ecosystem services. These implementation. And second, it requires understanding how the
multi-dimensional requirements for the future of agriculture have costs and benefits accrue with changes in shade-tree cover so that
recently been summarized by two complementary concepts: sus- they can be implemented with shade levels that maximize benefits
tainable intensification and climate-smart agriculture2,6,7. However, while minimizing costs. Here, we quantify the effectiveness of cocoa
while the concepts and requirements for the future of agriculture agroforests along gradients of shade-tree cover to determine if—and
are clear, easily implemented integrative solutions are rare. at what levels of shade-tree cover—agroforests can be both climate-
Agroforests—the deliberate inclusion of trees in cropping sys- smart and sustainable (Fig. 1). Over a period of 2 years, including
tems in some form of spatial arrangement or temporal sequence8— 1 normal and 1 exceptionally dry year, we measured cocoa produc-
are an ideal candidate for climate-smart, sustainable agricultural tion, losses to disease, soil fertility, temperature, relative humidity,
intensification2,7,9. Agroforests are widely expected to buffer crops soil moisture, carbon stocks and biodiversity of multiple taxa in 20
from climate change, improve soil fertility and regulate pests and cocoa agroforest plots selected systematically along a gradient of
pathogens10–15. Moreover, agroforests are believed to contribute shade-tree cover in a globally significant cocoa growing region in
to climate-change mitigation through increased carbon seques- Ghana, West Africa (see Methods for further details).
tration9,16–18, while also maintaining diversity of associated spe- To isolate the effects of shade trees from confounding factors
cies13,19,20. However, although they are widely advocated, and are such as differences in farm management, soil type and local micro-
currently being implemented across large parts of Africa, Asia and climate, all plots were compared with paired monoculture plots on
South America, agroforests generate costs and many of their pre- the same farms (Methods). The shade-tree cover and the species
sumed benefits, while intuitive, are often untested. For example, composition of cocoa agroforests often varies over the lifetime of
climate-change adaptation has been a major reason to encourage cocoa, with higher shade requirements in the establishment phase
agroforests but shade trees are likely to have multiple interacting (0–3 years), which can be provided by food crops in addition to
effects on different climate variables such as temperature and water shade trees, and then a reduction or removal of shade as cocoa trees
availability so that the net effects on crops under climate change mature13. In this context, our study focuses on the effects of shade
may not always be positive. Furthermore, shade trees reduce crop trees in mature cocoa farms (15–25 years since establishment) in
production through competition21, and at the same time agrofor- which shade levels are less likely to vary over time. We focused our
ests are unlikely to match the diversity and service provisioning of assessment on cocoa agroforests in Ghana because agriculture in this

Sustainable Agroecosystems Group, Department of Environmental Systems Science, ETH Zurich, Zurich, Switzerland. 2Council for Scientific and Industrial
1

Research – Soil Research Institute, Kwadaso, Kumasi, Ghana. 3Institute of Integrative Biology, Department of Environmental Systems Science, ETH Zurich,
Zurich, Switzerland. *e-mail: wilma.blaser@usys.ethz.ch

234 Nature Sustainability | VOL 1 | MAY 2018 | 234–239 | www.nature.com/natsustain

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NAture SustAinABility Articles
10% 30% 50% 70% consistent with the smaller size of cocoa trees found under higher
levels of cover (Supplementary Table 1). Yields assessed per cocoa
tree gave similar results (Supplementary Table 1), indicating that
lower yields at higher shade-tree cover were not the effect of lower
1. Agricultural production planting densities of cocoa trees. Notably, our empirical results take
Crop yields –3.6% –73%
a similar functional form to predictions based on mechanistic phys-
* iological models30, but are more optimistic than those models about
Soil fertility +0.1% +0.1%
the ability of cocoa to sustain yields in agroforests.
Pest control * +59% +32% It is commonly believed that the negative effects of shade trees on
2. Climate adaptation
short-term productivity will be partially offset by positive effects of
Temperature * –0.1% –2.4% shade trees on soil fertility13 and disease prevalence14,15, which would
Water availability * –16% –58% then enable the maintenance of production over longer periods of
3. Climate mitigation time10,13. With respect to disease prevalence, our results support this
Aboveground C * +70% +530% contention but only at intermediate levels of cover (Figs. 1 and 2b).
Soil C –0.8% –0.8% While in low- and high-shade agroforests cocoa suffered from dis-
ease more than in monocultures, at around 50% shade-tree cover the
4. Biodiversity conservation * +91% +698% proportion of diseased pods and infestation of cocoa trees by para-
sitic plants (mistletoe) was reduced relative to monocultures (Fig. 2b,
10 20 30 40 50 60 70 80
Supplementary Table 2). Our results corroborate emerging evidence
Canopy cover (%)
that intermediate levels of shade may improve long-term production
Fig. 1 | The simultaneous costs and benefits of agroforests on production, by minimizing losses to disease14, although these benefits were insuf-
climate adaptation, climate mitigation and biodiversity along a gradient ficient to increase production to levels above monocultures in our
of shade-tree cover (10–80%). The four top panels give a schematic aerial study (Fig. 2a). Our results, along with others14,28,31–33, do not support
view of four representative plots along the gradient of increasing shade- the contention that agroforestry will facilitate long-term production
tree cover. Each bar in the lower panels shows the effect of shade trees through increases in soil fertility, at least as currently implemented in
on a particular response variable, with blue indicating shade trees confer our study area. Shade trees did not increase soil carbon stocks (Fig. 2c)
a benefit, red indicating shade trees confer a cost and white indicating or other soil fertility parameters, including total soil nitrogen (Fig. 2d)
no difference, relative to cocoa monocultures. The colour gradients are and phosphorus, resin extractable phosphorus, exchangeable cations
quantitative, matching the predicted values for the percentage difference (potassium, calcium, magnesium and sodium) and effective soil cat-
in each response variable relative to monocultures, and scaled to the ion exchange capacity (Supplementary Table 3). Whether shade trees
ranges shown in Fig. 2. The actual predicted range over the gradient for can increase the productive lifetime of cocoa plantations by reduc-
each variable is indicated with minimum and maximum predicted values of ing premature aging compared to full-sun monocultures remains
percentage difference (%) at the ends of each bar. Predicted values were to be studied. In sum, agroforests with approximately 30% cover
based on significant regression lines (marked with a red asterisk) shown in (and sometimes higher) can have similar production levels to cocoa
Fig. 2, for crop yields (mean annual cocoa yields 2015–2017), pest control monocultures and can reduce pressure from pest and disease (Fig. 1).
(mean annual total losses to disease 2015–2017), temperature (dry season Notably, these levels of cover are likely to be higher than is most
maximum temperature 2016), water availability (dry season soil moisture commonly implemented in this region34,35.
2016), aboveground C (aboveground carbon stocks in tree biomass)
and biodiversity (bird richness). Predicted values for non-significant Do agroforests allow adaptation to climate change? The idea that
relationships (soil fertility (total soil nitrogen stocks) and soil C (total soil agroforests can buffer crops from the worst effects of climate change
carbon stocks)) are averages across the shade-tree cover gradient. is a major reason for their implementation in climate-vulnerable
agricultural sectors and regions11,12. In general, it is most important
for shade trees to buffer climate extremes during the dry season,
when cocoa is most vulnerable to high temperatures and drought
region is critically vulnerable to each of the threats that agroforestry stress30. However, we found that agroforests had both positive and
promises to mitigate. For example, while global demand for cocoa negative effects on climate-related variables during the dry season.
is increasing, current productivity in West Africa (which accounts Increasing shade-tree cover buffered cocoa against temperature
for more than 65% of global production) is declining because of extremes by decreasing daily maximum and daily mean tempera-
degraded soils and pathogens, and future production is threatened tures in the dry season (Figs. 1 and 2e, and Supplementary Table 4).
by climate change24–26. At the same time, agricultural expansion has However, while increasing shade-tree cover increased relative
led to extensive deforestation in this region, with little remaining humidity in the wet season (Supplementary Table 4), evidence that
primary forest habitat27, which makes the biodiversity and carbon trees could do so in the dry season was less convincing (Fig. 2f).
storage potential of agroforests critically important. Moreover, increasing shade-tree cover also decreased soil moisture
during both the short and long dry seasons (Figs. 1 and 2g, and
Results and discussion Supplementary Table 4; see also36).
Can agricultural production be maintained in agroforests? Our Taken together, our results suggest that the effectiveness of
results suggest that shade trees are unlikely to compromise mean agroforests for climate adaptation will depend on the nature of the
annual production of cocoa up to levels of ~30% cover (Figs. 1 principal climate-related threats to production in a given area (high
and 2a). This result provides a pathway to sustainable agricultural temperatures and/or low water availability), and on how agrofor-
intensification. As shade-tree cover increases beyond ~30%, cocoa ests are implemented in relation to these threats. On the one hand,
production declines (Fig. 2a, see also19,28,29). These declines can be a decrease in water availability with increasing shade-tree cover is
significant in magnitude, although shade trees may not necessar- expected to exacerbate drought stress, which has long been consid-
ily compromise production up to ~50% cover (95% confidence ered a major climate-related threat to cocoa production in West Afr
interval at 50% cover: +​  1.2 to −​33.3). The decrease in yield as cover ica24,25,36,37. However, rising temperatures may be an equally or more
increases is likely to occur as a consequence of competition for light important threat to cocoa in this region25,38, suggesting that future cli-
(Supplementary Fig. 1), water (Fig. 2g) and nutrients21,30, which is matic suitability for cocoa may depend on the temperature-buffering

Nature Sustainability | VOL 1 | MAY 2018 | 234–239 | www.nature.com/natsustain 235


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles NAture SustAinABility

Agricultural production
a Yield b Losses to disease c Soil C d Soil N
100 100 100 100
2 2 2
R = 0.49** R = 0.40* R = 0.08 R 2 = 0.02

Difference relative to
Difference relative to

Difference relative to
Difference relative to

monoculture (%)

monoculture (%)
monoculture (%)
50 50 50 50
monoculture (%)

0 0 0 0

–50 –50 −50 −50

–100 –100 −100 −100


0 30 60 90 0 30 60 90 0 30 60 90 0 30 60 90
Shade-tree cover (%) Shade-tree cover (%) Shade-tree cover (%) Shade-tree cover (%)

Climate adaptation and mitigation


e Temperature f Humidity g Soil moisture h Aboveground C
5.0 2 10 2 100 2 1,000
R = 0.23* R = 0.05 R = 0.28° R 2 = 0.49***

Difference relative to
Difference relative to

Difference relative to
Difference relative to

750

monoculture (%)
monoculture (%)

monoculture (%)
monoculture (%)

2.5 5 50

500
0.0 0 0

250
−2.5 −5 −50
0
−5.0 −10 −100
0 30 60 90 0 30 60 90 0 30 60 90 0 30 60 90
Shade-tree cover (%) Shade-tree cover (%) Shade-tree cover (%) Shade-tree cover (%)

Biodiversity conservation
i Tree richness j Bird richness k Ant richness l Frog richness
1,200 2
1,000 2
100 100
R = 0.37** R = 0.47*
Difference relative to

Difference relative to
Difference relative to

Difference relative to

900 750
monoculture (%)

monoculture (%)
monoculture (%)

monoculture (%)

50 50

600 500
0 0

300 250
−50 −50
0 R 2 = 0.06 R 2 = 0.17
0
−100 −100
0 30 60 90 0 30 60 90 0 30 60 90 0 30 60 90
Shade tree cover (%) Shade tree cover (%) Shade tree cover (%) Shade tree cover (%)

Fig. 2 | The effects of agroforests on agricultural production, climate adaptation, climate mitigation and biodiversity along a gradient of shade-tree
cover. a, Mean annual cocoa production (period 2015–2017) was maintained at low-to-intermediate levels of shade, decreasing relative to monocultures
at levels above ~30% cover. b–d, Shade trees reduced the proportion of cocoa pods lost to disease (period 2015–2017) at medium cover (b) but did not
increase soil fertility (soil carbon (C) and soil nitrogen (N)) relative to monocultures (c,d). e–g, Shade trees buffered maximum air temperatures (e) and
minimum relative air humidity (f) but decreased soil moisture (g) during the long dry season (2016). h, Shade trees significantly increased aboveground
carbon stocks. i–l, The diversity of trees (i) and birds (j) increased with increasing shade-tree cover, but the diversity of ants (k) and frogs (l), while
significantly higher in agroforests than in monocultures, was not related to shade-tree cover. Plots show the percentage difference in each response
variable between agroforests and monocultures along a gradient of increasing shade-tree cover (10–80% cover), where positive values indicate higher
values in agroforests. Solid lines show significant relationships and broken lines indicate non-significant relationships with significance levels: °P =​ 0.06;
*P <​ 0.05; **P <​ 0.01; ***P <​0.001. The dotted horizontal line represents no difference between agroforest and monoculture plots. See Supplementary
Tables 1–5 for analyses, including analyses of additional response variables and in additional seasons.

effects of shade trees that we demonstrate. Importantly, benefits the heavily shaded agroforests represents only 34–39% of the carbon
of shade trees for temperature emerged at lower levels of shade- stored in semideciduous natural forests in Ghana39.
tree cover than did costs for soil moisture (Figs. 1 and 2e–g), It is also clear that the biggest potential for carbon sequestra-
providing a potentially important opportunity to optimize the tion in cocoa agroforests lies in aboveground biomass and not in
trade-offs between climate adaptation benefits and costs in agro- soils, where we found no increases in carbon stocks relative to
forestry systems. monocultures (Fig. 2c and Supplementary Table 4). Soil carbon
sequestration has recently been promoted as a major potential
Can agroforests mitigate future climate change? Increasing shade- climate-change mitigation strategy—especially in agroforestry
tree cover has strong positive effects on aboveground carbon seques- systems40. However, our results and others18,28,31,33 suggest that
tration (Figs. 1 and 2h and Supplementary Table 4). Low-shade in perennial cropping systems, such as cocoa and coffee, add-
( <​ 30%) cocoa agroforests had up to 210% more carbon stored in ing shade trees may not have the same potential for soil car-
aboveground standing biomass than paired monocultures, while bon sequestration as in annual cropping systems. This may be
some of the most heavily shaded plots had more than 660% more car- because the litter produced by shade trees in perennial crop-
bon (62.0 Mg C ha-1 versus 8.1 Mg C ha-1, Fig. 2h and Supplementary ping systems might not significantly increase carbon inputs to
Table 4). While the increase relative to monocultures is substantial, it levels above those of the perennial crops alone, as was the case
is also important to note that the aboveground stored carbon in even in our study (mean difference in litter between agroforests and

236 Nature Sustainability | VOL 1 | MAY 2018 | 234–239 | www.nature.com/natsustain

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NAture SustAinABility Articles
monocultures −​8.8%, t-test: t =​  −​1.72, P =​  0.95; Supplementary species in monocultures may be increased by ‘spillover’ from nearby
Methods and Supplementary Table 4). agroforest patches49. Equally, the performance of agroforests may
Taken together, our results suggest that the net carbon sequestra- be affected by conditions in nearby monocultures, and these effects
tion potential of agroforests will depend strongly on land-use history could also extend to some disease and climate adaptation variables.
before their establishment. For example, the incorporation of shade Such effects would be expected to reduce the differences in perfor-
trees into annual or degraded systems will most likely result in a net mance between monocultures and agroforests in our study, suggest-
increase in below- and aboveground carbon stocks18,40,41, whereas ing our conclusions are likely to be conservative.
the establishment of new agroforests through the conversion of While we chose farms that were representative of how agrofor-
natural forests will lead to net carbon losses from both aboveground ests are most commonly managed in the study region, the extent to
biomass (~75% loss) and soils (30–50% loss)22,32,40. Nevertheless, it is which our results apply to other regions, or to farms that are better
clear that the potential for climate mitigation through aboveground maintained will, at least partly, depend on local conditions. Yields in
carbon storage in agroforests is a major advantage of agroforests our study system were relatively low (mean 551.2 ±​ 25.8 kg ha-1) as
over monocultures, and is likely to be particularly important for cli- is typical across large parts of West Africa26,35, and improvements in
mate mitigation in regions where limited primary forest remains16. farm management have been shown to improve yields independent
Moreover, the strong positive relationship between shade and car- of shade-tree cover in agroforests elsewhere19. This may suggest that
bon sequestration (Figs. 1 and 2h) suggests that conversion from yields could be improved further in low-input systems such as those
high-shade to low-shade agroforests will come at the cost of a sub- that we studied, while maintaining many of the benefits of agro-
stantial loss of carbon storage capacity, particularly in landscapes forests that we document. Focussing these management efforts in
such as those in large parts of West Africa where agroforests have agroforests with higher cover (30–50%)—which our data suggests
already replaced large areas of natural forest (see also16). may still have the potential for yields similar to monocultures in
some farms (Fig. 2a)—would lead to much higher realized benefits
How effective are agroforests for the conservation of biodiver- of agroforests for climate (Fig. 2h) and biodiversity (Figs. 2i,j)17,19.
sity? We found that agroforests support higher levels of species While there were no clear patterns in the distribution of par-
diversity than monocultures (Figs. 1 and 2i–l, and Supplementary ticular shade-tree species in our plots (Supplementary Fig. 2 and
Table 5), a result that is consistent with findings from other cocoa Supplementary Table 6), there may also be potential to improve the
and oil palm agroforests42–44. However, one of the major questions effectiveness of agroforests through careful (rather than ad hoc)
surrounding the effectiveness of agroforests is their ability to sup- selection of shade trees with particular functional traits. Soil fertility
port biodiversity without compromising yields19. In this context, benefits, for example, may be more likely to arise with the deliberate
our results suggest that, relative to monocultures, agroforests can inclusion of nitrogen-fixing species, and deep-rooting trees might
have large benefits for biodiversity that do not strongly compro- limit competition for water10,13. The overall economic performance of
mise yields up to a cover of ~30% (Fig. 1). Beyond these levels, agroforests can also be improved (beyond benefits to the yield of the
the positive relationship between shade-tree cover and bird and primary crop) by including species that can be harvested for timber,
tree diversity (Figs. 2i,j) combined with the negative relationship fruit or other products, something not explicitly considered in our
between cover and yield (Fig. 2a) causes an increasingly strong study (but see Supplementary Table 6). Finally, because shade trees
trade-off between diversity and yield as cover increases. This result differ in their crown architecture and size, optimal levels of cover can
is contrary to recent studies of cocoa agroforests in Indonesia19, but either be achieved with a high density of small shade trees or a low
consistent with negative biodiversity–yield relationships in temper- density of large shade trees50 (see Supplementary Fig. 3 for details).
ate systems45,46. In contrast, ants and amphibians showed a posi- Including tall trees that offer small amounts of shade relative to their
tive effect of shade trees (t-test amphibians: t =​  2.27, P =​  0.02; ants: size, for example, could potentially increase biodiversity and carbon
t =​  3.7, P =​ 0.003) but no significant relationship with shade-tree storage relative to monocultures, while also maximizing yields16,17,19.
cover (Figs. 2k,l, and Supplementary Table 5), suggesting that in our
study these taxa benefit from the simple inclusion of shade trees, Agroforestry and the future of climate-smart sustainable agri-
irrespective of the level of shade-tree cover. culture. Our study suggests that agroforests have large potential—
Are the benefits of agroforests for biodiversity worth the cost to across a wide range of production, climate, ecosystem services
yields at higher levels of shade-tree cover, particularly given that bio- and diversity metrics—as a strategy for climate-smart, sustainable
diversity in cocoa agroforests is already strongly reduced compared agriculture. Nevertheless, shade levels are critical such that ad hoc
to natural forests? Studies in Ghana, Indonesia and Uganda that have inclusion of trees in cropping systems is likely to generate subopti-
compared land-sharing practices, such as agroforestry, with land-spar- mal outcomes without careful consideration of which production,
ing strategies have found that more tree and bird species were con- climate and sustainability goals will be compromised. While agri-
served under land-sparing approaches27,47. Our observations suggest a cultural production is highest in monocultures and agroforests up
similar pattern because we recorded no forest interior bird species in to 30% shade-tree cover, the provision of regulating ecosystem ser-
any of our plots, regardless of the shade-tree cover, as well as relatively vices and the value for biodiversity conservation is maximized in
low diversity of trees compared with intact forests. Nevertheless, the heavily shaded systems (Fig. 1). Moreover, the inclusion of shade
conservation value of agroforests can be considerable and might be trees has conflicting effects on the ability of cocoa production to
one of few options for conserving biodiversity in human-dominated adapt to climate change, suggesting trees need to be included at lev-
landscapes that have little remaining primary habitat48. els that minimize costs by addressing the most important climate-
related threats. Despite these difficulties, our results suggest that
Limitations and the potential to maximize agroforest perfor- cocoa agroforests up to 30% cover are far superior to monocultures
mance. To control for confounding factors, we compared mono- because they do not strongly compromise production, while at the
cultures and agroforests on the same farm, which meant that the same time they provide benefits for disease management, climate
size of our study plots was dictated by the approximate size at which mitigation and adaptation and biodiversity conservation (Fig. 1).
monocultures and agroforests are most commonly implemented in
the study area. One potential implication is that the performance
Methods
of these often relatively small monoculture and agroforest patches Study area and sampling design. We did our study in the moist semideciduous
may be affected by other nearby patches that may have higher tropical zone of the Ashanti Region in Ghana, West Africa (06°40’ N and 01°57’ W).
or lower levels of cover. For example, the diversity of associated Closed semideciduous tropical forest in this region has been replaced by a mosaic

Nature Sustainability | VOL 1 | MAY 2018 | 234–239 | www.nature.com/natsustain 237


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles NAture SustAinABility
of rotational fallows, slash-and-burn agricultural systems and perennial cropping were equipped with one logger in the centre of each plot, after we verified that
systems. Cocoa is the dominant perennial crop in this region and is grown this design adequately captured the variability in full-sun monocultures (data not
under variable shade-tree cover, including full-sun monocultures. Mean annual shown). We calculated average daily mean temperature and relative humidity, daily
precipitation in the study area ranges between 1,700 and 1,850 mm and occurs maximum temperatures and daily minimum relative humidity for each month.
across two separate wet seasons (March to mid-July and September to October). We present monthly averages for October to represent the conditions at the end of
The months between the rains are marked by a short dry season in August and a the wet season when shade trees have their full foliage and monthly averages for
long dry season from December to February. Variation in the yield of cocoa trees is February, which represents the height of the long dry period.
affected more by rainfall than by any other climate variable30. The first year of this Volumetric water content in topsoils (0–20 cm) was measured at every metre
study was exceptionally dry, with only 30 mm of rain during the two driest months along six regular transects in each plot (126 measurements per plot) with a time
(January and February 2016) compared to an average of 110 ±​ 16 mm over the past domain reflectometry sensor (HydroSense II, Campbell Scientific Ltd.), during
10 years (Kwadaso Agricultural Station, Ghana). This exceptional dry period is the peak of the short dry season (August 2015) and during the long dry season
believed to have been caused by a particularly strong El Niño event, and these are, (February 2016).
more generally, known to cause droughts in West Africa36. Because El Niño events
are predicted to increase in intensity as a consequence of climate change51, such dry Biodiversity. Tree species richness was recorded in all plots with surveys of all trees
periods may become more common in this region in the future. with DBH >​ 10 cm. In 10 out of 20 paired plots, we recorded the richness of birds,
In 2015, we selected 20 plots in cocoa agroforests along a gradient of shade-tree ants and litter frogs between October and December 2015. Birds were recorded using
cover ranging from low to high shade (14–76% shade-tree cover, see Supplementary visual and acoustic surveys. Surveys were done on three separate mornings (between
Table 6). Because low-shade agroforests are more common in the study region, 06:00 and 10:00) in each plot, for 20 min each time. Surveys were done from the
random selection of plots would have resulted in comparisons of the simple presence centre of each plot, using the fixed-radius point-count method52. All surveys were
versus absence of shade trees, which would not have allowed a comparison of how done by the same observer (J.L.). During each survey, bird calls were also recorded
the effects of agroforests change at different levels of shade-tree cover. Therefore, we continuously using a Tascam DR100 hand-held recorder equipped with a Sennheiser
selected plots systematically to ensure our design could be used to meet our aim of ME66/K6 microphone, and the time-points of individual observations were recorded
quantifying changes in the effectiveness of shade trees to meet production, climate to aid in a posteriori confirmation of bird identification when necessary53. Over
and conservation goals across a gradient of shade-tree cover. This design required the course of 1 month prior to the survey period, the bird calls of the study area
that other systematic differences between plots be controlled for. We therefore paired were systematically recorded to build a reference database53. Species observed only
each agroforest plot with a nearby monoculture plot on the same farm. The paired flying above the canopy were excluded. Litter frogs were surveyed in both the wet
sampling was aimed at controlling for variability between plots due to different soils, and dry seasons. We conducted an exhaustive search for frogs for 30 min between
local climate, farm age, cocoa variety and management practices between farms in 09:00 and 17:00 along 30 parallel, 1-m wide transects. After 30 min, all individuals
order to isolate the effects of shade trees on measured variables. We selected all plots were identified, photographed and returned to the plot. When necessary, species
from relatively well-maintained, low-fertilizer-input, mature cocoa farms (15–25 identification was confirmed using the photographs. Ground-dwelling ants were
years since establishment) with no commercial fruit or timber production, and in sampled using four pitfall traps (diameter 14.5 cm) in each plot. After installation the
which we could locate cocoa growing in monoculture and in agroforests. While traps were closed for 10 days to avoid a temporary increase in the capture rate due
there was a possibility that high-shade plots may have occurred only in inherently to the disturbance caused by the installation54. The traps were subsequently opened
unproductive locations, this possibility was not supported by our soil fertility data and filled with ethanol. Traps were left open for 14 days, and insects were collected
(Fig. 2c,d, and Supplementary Table 3). Plots measured 30 m ×​ 30 m. This plot size every 7 days. All ants were identified to species or morphospecies level. We note
was chosen because it is representative of the size of existing monocultures and that our plots appear to be slightly smaller (for birds and trees) or similar (for frogs
high-shade agroforest patches in the region. Our plots contained between 63 and 195 and ants) in size compared to those used previously in assessments of biodiversity
cocoa trees, with a mean of 108 ±​ 5 cocoa trees per plot. in agroforests and other forest landscapes19,29,32,55,56. We further note that they were
The percentage of shade-tree canopy cover of each plot was estimated using of a sufficiently large size and number to detect differences in biodiversity between
high-resolution aerial photographs captured with a quadcopter drone (Green monocultures and agroforests in all taxa, and significant positive relationships
Copter Tropic-Quadro, Rucon Engineering) equipped with a digital camera between cover and diversity in two of four taxa (trees and birds).
(Canon PowerShot S110). Photographs were analysed using image-processing
software (Agisoft PhotoScan 1.2.3 and Arc GIS 10.4). In each plot, we counted Data analysis. We use the relative percentage difference (d) between monoculture
cocoa and shade trees and calculated their density. We identified each species (m) and agroforest (a) plots according to the following equation for all measured
of shade tree (Supplementary Table 6). We further recorded diameter at breast parameters: d=(a−m)/m×​100. This metric incorporates information from our
height (DBH) of all cocoa and shade trees (Supplementary Table 6), and the height paired design, isolating the effects of shade trees from other confounding factors.
of all shade trees. Based on these measurements, we estimated carbon stocks in Moreover, this metric implicitly uses performance in monoculture as a baseline
aboveground biomass using allometric relationships (Supplementary Methods). to quantify the added advantage/disadvantage (expressed as percent change) of
We note that shade-tree density and DBH are strong determinants of shade-tree the inclusion of shade trees on each of our response variables. With this metric,
cover (multiple regression: F3,16 =​  9.85, R2 =​  0.65, P <​ 0.001), and that similar shade- we used polynomial regression to quantify the relationship between each of our
tree cover could be achieved with a high density of small trees or a low density of response variables and shade-tree cover. Models were simplified using partial F
large trees (see Supplementary Figure 3 for analysis, and for calculations of cover as tests, retaining quadratic terms at a significance level of P <​ 0.20 to guard against
a function of different tree-density and tree-size combinations). type II errors57. Analyses for yield and losses to disease were based on values
averaged over the 2 years. Analyses of microclimate variables were done separately
Agricultural production, pests and disease. Ripe cocoa pods were harvested for both wet and dry seasons in both years. Analyses of soil fertility variables,
within each plot at fortnightly intervals during the main and minor harvesting carbon stocks and biodiversity were done using the single estimates of each of these
seasons from September 2015 to July 2017. We recorded the number of healthy variables collected during the study.
and diseased cocoa pods and categorised diseased pods according to whether they Variables for resin-extractable phosphorus and bulk density were
had black pod (Phytophthora sp.) or other diseases (mostly mirid bugs Sahlbergella log(x +​ constant)-transformed to meet model assumptions. When we detected no
singularis and Distantiella theobroma). Wet-bean cocoa weight was measured for effect of increasing shade-tree cover on a response variable, we carried out t-tests to
healthy pods. Extracted beans from healthy pods on ten farms were fermented and determine whether the simple presence or absence of trees (that is, agroforest versus
dried to determine the wet/dry bean conversion factor. Yields were expressed as dry monoculture) caused differences in our response variable, independent of shade-tree
bean yield per area, as well as yield per cocoa tree. Disease levels were expressed as cover. All analyses were performed in R 3.2.4 (R Development Core Team 2016).
the percentage of harvested pods lost to disease. Infestation with parasitic plants
(mistletoe) was assessed once by counting individual plants in each plot in January Data availability. The data that support the findings of this study are openly
2016. It is also important to note that a visual assessment by an expert in cocoa available in figshare at https://doi.org/10.6084/m9.figshare.6118562
swollen shoot virus did not reveal any signs of the disease in our plots.
Received: 25 November 2017; Accepted: 10 April 2018;
Soil fertility. We collected 36 soil samples with a gouge auger (diameter 3 cm) in Published online: 15 May 2018
each plot (0–15 cm depth) in May 2015. Samples were composited at the plot level
and analysed for pH, total carbon, nitrogen and phosphorus content, content of resin
extractable phosphorus and exchangeable cations (potassium, calcium, magnesium References
and sodium), effective soil cation exchange capacity and soil texture. Soil bulk density 1. Palm, C. A. et al. Identifying potential synergies and trade-offs for meeting
was determined based on nine additional core samples and used to calculate soil food security and climate change objectives in sub-Saharan Africa. Proc. Natl
carbon, nitrogen and phosphorus pools. For details, see Supplementary Methods. Acad. Sci. USA 107, 19661–19666 (2010).
2. Lipper, L. et al. Climate-smart agriculture for food security. Nat. Clim.
Microclimate. Temperature and relative humidity were measured continuously Change 4, 1068–1072 (2014).
with five HOBO data loggers (HOBO Pro v2 U23-001) per agroforest plot, 3. Godfray, H. C. J. et al. Food security: The challenge of feeding 9 billion
installed 0.5 m above the cocoa canopy in September 2015. Monoculture plots people. Science 327, 812–818 (2010).

238 Nature Sustainability | VOL 1 | MAY 2018 | 234–239 | www.nature.com/natsustain

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NAture SustAinABility Articles
4. Parry, M., Evans, A., Rosegrant, M. W. & Wheeler, T. Climate Change and 36. Abdulai, I. et al. Cocoa agroforestry is less resilient to sub-optimal and extreme
Hunger: Responding to the Challenge (World Food Programme, Rome, 2009). climate than cocoa in full sun. Glob. Change Biol. 24, 273–286 (2018).
5. Morton, J. F. The impact of climate change on smallholder and subsistence 37. Carr, M. K. V. & Lockwood, G. The water relations and irrigation
agriculture. Proc. Natl Acad. Sci. USA 104, 19680–19685 (2007). requirements of cocoa (Theobroma cacao L.): a review. Exp. Agric 47,
6. Garnett, T. et al. Sustainable intensification in agriculture: premises and 653–676 (2011).
policies. Science 341, 33–34 (2013). 38. Schroth, G., Läderach, P., Martinez-Valle, A. I. & Bunn, C. From site-level to
7. Campbell, B. M., Thornton, P., Zougmoré, R., van Asten, P. & Lipper, L. regional adaptation planning for tropical commodities: cocoa in West Africa.
Sustainable intensification: What is its role in climate smart agriculture? Mitig. Adapt. Strat. Glob. Change 22, 903–927 (2017).
Curr. Opin. Environ. Sustain. 8, 39–43 (2014). 39. Asase, A., Asitoakor, B. K. & Ekpe, P. K. Linkages between tree diversity and
8. Nair, P. R. An Introduction to Agroforestry (Kluwer Academic Publishers, carbon stocks in unlogged and logged West African tropical forests.
Dordrecht, 1993). Int. J. Biodivers. Sci. Ecosyst. Serv. Manag. 8, 217–230 (2012).
9. Harvey, C. A. et al. Climate-smart landscapes: Opportunities and challenges 40. Paustian, K. et al. Climate-smart soils. Nature 532, 49–57 (2016).
for integrating adaptation and mitigation in tropical agriculture. Conserv. 41. Jagoret, P., Michel-Dounias, I., Snoeck, D., Ngnogue, H. T. & Malezieux, E.
Lett. 7, 77–90 (2014). Afforestation of savannah with cocoa agroforestry systems: A small-farmer
10. Jose, S. Agroforestry for ecosystem services and environmental benefits: an innovation in central Cameroon. Agrofor. Syst. 86, 493–504 (2012).
overview. Agrofor. Syst. 76, 1–10 (2009). 42. Clough, Y., Putra, D. D., Pitopang, R. & Tscharntke, T. Local and landscape
11. Vaast, P., Harmand, J.-M., Rapidel, B., Jagoret, P. & Deheuvels, O. in Climate factors determine functional bird diversity in Indonesian cacao agroforestry.
Change and Agriculture Worldwide (ed. Torquebiau, E.) 209–224 (Springer Biol. Conserv. 142, 1032–1041 (2009).
Netherlands, Dordrecht, 2016). 43. Abrahamczyk, S., Kessler, M., Dwi Putra, D., Waltert, M. & Tscharntke, T.
12. Lin, B. B. Agroforestry management as an adaptive strategy against potential The value of differently managed cacao plantations for forest bird
microclimate extremes in coffee agriculture. Agric. For. Meteorol. 144, conservation in Sulawesi, Indonesia. Bird. Conserv. Int. 18, 349–362 (2008).
85–94 (2007). 44. Teuscher, M. et al. Trade-offs between bird diversity and abundance, yields
13. Tscharntke, T. et al. Multifunctional shade-tree management in tropical and revenue in smallholder oil palm plantations in Sumatra, Indonesia. Biol.
agroforestry landscapes—a review. J. Appl. Ecol 48, 619–629 (2011). Conserv. 186, 306–318 (2015).
14. Andres, C. et al. Agroforestry systems can mitigate the severity of cocoa 45. Kleijn, D. et al. On the relationship between farmland biodiversity
swollen shoot virus disease. Agric. Ecosyst. Environ. 252, 83–92 (2018). and land-use intensity in Europe. Proc. R. Soc. B Biol. Sci. 276,
15. Schroth, G., Krauss, U., Gasparotto, L., Duarte Aguilar, J. A. & Vohland, K. 903–909 (2009).
Pests and diseases in agroforestry systems of the humid tropics. Agrofor. Syst. 46. Nilsson, F. O. L. Biodiversity on Swedish pastures: Estimating biodiversity
50, 199–241 (2000). production costs. J. Environ. Manag. 90, 131–143 (2009).
16. Schroth, G. et al. Contribution of agroforests to landscape carbon storage. 47. Hulme, M. F. et al. Conserving the birds of Uganda’s banana-coffee arc: land
Mitig. Adapt. Strat. Glob. Change 20, 1175–1190 (2015). sparing and land sharing compared. PloS ONE 8, e54597 (2013).
17. Schroth, G. et al. Climate friendliness of cocoa agroforests is compatible with 48. Bhagwat, S. A., Willis, K. J., Birks, H. J. B. & Whittaker, R. J. Agroforestry:
productivity increase. Mitig. Adapt. Strat. Glob. Change 21, 67–80 (2016). a refuge for tropical biodiversity? Trends Ecol. Evol. 23, 261–267 (2008).
18. Feliciano, D., Ledo, A., Hillier, J. & Nayak, D. R. Which agroforestry options 49. Lucey, J. M. & Hill, J. K. Spillover of insects from rain forest into adjacent oil
give the greatest soil and above ground carbon benefits in different world palm plantations. Biotropica 44, 368–377 (2012).
regions? Agric. Ecosyst. Environ. 254, 117–129 (2018). 50. Asare, R. & Ræbild, A. Tree diversity and canopy cover in cocoa systems in
19. Clough, Y. et al. Combining high biodiversity with high yields in tropical Ghana. New For. 47, 287–302 (2016).
agroforests. Proc. Natl Acad. Sci. USA 108, 8311–8316 (2011). 51. Cai, W. et al. Increasing frequency of extreme El Niño eve’nts due to
20. De Beenhouwer, M., Aerts, R. & Honnay, O. A global meta-analysis of the greenhouse warming. Nat. Clim. Change 4, 111 (2014).
biodiversity and ecosystem service benefits of coffee and cacao agroforestry. 52. Hutto, R. L., Pletschet, S. M. & Hendricks, P. A fixed-radius point
Agric. Ecosyst. Environ 175, 1–7 (2013). count method for nonbreeding and breeding season use. Auk 103,
21. Sanchez, P. A. Science in agroforestry. Agrofor. Syst. 30, 5–55 (1995). 593–602 (1986).
22. Guo, L. B. & Gifford, R. M. Soil carbon stocks and land use change: A meta 53. Parker, T. A. III On the use of tape recorders in avifaunal surveys. Auk 108,
analysis. Glob. Change Biol. 8, 345–360 (2002). 443–444 (1991).
23. Gibson, L. et al. Primary forests are irreplaceable for sustaining tropical 54. Leather, S. R. Insect Sampling in Forest Ecosystems (Blackwell, Oxford, 2005).
biodiversity. Nature 478, 378–381 (2011). 55. Ferreira, R. B., Beard, K. H. & Crump, M. L. Breeding guild determines frog
24. Läderach, P., Martinez-Valle, A., Schroth, G. & Castro, N. Predicting the distributions in response to edge effects and habitat conversion in the Brazil’s
future climatic suitability for cocoa farming of the world’s leading producer Atlantic Forest. PLoS ONE 11, e0156781 (2016).
countries, Ghana and Côte d’Ivoire. Climatic Change 119, 841–854 (2013). 56. Kraft, N. J. B. et al. Disentangling the drivers of β​diversity along latitudinal
25. Schroth, G., Laderach, P., Martinez-Valle, A. I., Bunn, C. & Jassogne, L. and elevational gradients. Science 333, 1755–1758 (2011).
Vulnerability to climate change of cocoa in West Africa: patterns, opportunities 57. Quinn, G. P. & Keough, M. J. Experimental Design and Data Analysis for
and limits to adaptation. Sci. Total Environ. 556, 231–241 (2016). Biologists (Cambridge Univ. Press, Cambridge, 2002).
26. Franzen, M. & Borgerhoff Mulder, M. Ecological, economic and social
perspectives on cocoa production worldwide. Biodivers. Conserv. 16,
3835–3849 (2007).
Acknowledgements
We thank K. Oppong and D. Oppong for assistance with fieldwork, and farmers in
27. Phalan, B., Onial, M., Balmford, A. & Green, R. E. Reconciling food
the study region for allowing access to their farms. We further acknowledge J. O. Fening
production and biodiversity conservation: Land sharing and land sparing
for institutional support provided by the Soil Research Institute of Ghana. We thank
compared. Science 333, 1289–1291 (2011).
B. Jahn-Humphrey, G. Asamoah, B. Studer, G. Quansah, T. Afreh and M. Amponsah for
28. Blaser, W. J., Oppong, J., Yeboah, E. & Six, J. Shade trees have limited benefits
laboratory assistance. C. Ofori kindly helped us with the identification of frogs. This study was
for soil fertility in cocoa agroforests. Agric. Ecosyst. Environ. 243, 83–91 (2017).
funded by the Sustainable Agroecosystems Group at ETH Zurich and a grant from the Swiss-
29. Bisseleua, D. H. B., Missoup, A. D. & Vidal, S. Biodiversity conservation,
African Research Cooperation (SARECO) funded by the University of Basel, Switzerland.
ecosystem functioning, and economic incentives under cocoa agroforestry
intensification. Conserv. Biol. 23, 1176–1184 (2009).
30. Zuidema, P. A., Leffelaar, P. A., Gerritsma, W., Mommer, L. & Anten, N. P. R. Author contributions
A physiological production model for cocoa (Theobroma cacao): model W.J.B., S.P.H. and J.S. designed the research. W.J.B., J.O., J.L. and E.Y. performed the
presentation, validation and application. Agric. Syst. 84, 195–225 (2005). research. W.J.B., S.P.H. and J.S. analysed and interpreted the data. W.J.B. wrote the first
31. Jacobi, J. et al. Carbon stocks, tree diversity, and the role of organic draft and S.P.H. and J.S. contributed to revisions.
certification in different cocoa production systems in Alto Beni, Bolivia.
Agrofor. Syst. 88, 1117–1132 (2014). Competing interests
32. Steffan-Dewenter, I. et al. Tradeoffs between income, biodiversity, and The authors declare no competing interests.
ecosystem functioning during tropical rainforest conversion and agroforestry
intensification. Proc. Natl Acad. Sci. USA 104, 4973–4978 (2007).
33. Wartenberg, A. C. et al. Does shade tree diversity increase soil fertility in Additional information
cocoa plantations? Agric. Ecosyst. Environ. 248, 190–199 (2017). Supplementary information is available for this paper at https://doi.org/10.1038/
34. Ruf, F. O. The myth of complex cocoa agroforests: The case of Ghana. Hum. s41893-018-0062-8.
Ecol. 39, 373–388 (2011). Reprints and permissions information is available at www.nature.com/reprints.
35. Gockowski, J., Afari-Sefa, V., Sarpong, D. B., Osei-Asare, Y. B. & Agyeman, N. F.
Improving the productivity and income of Ghanaian cocoa farmers while Correspondence and requests for materials should be addressed to W.J.B.
maintaining environmental services: What role for certification? Int. J. Agric. Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
Sustain. 11, 331–346 (2013). published maps and institutional affiliations.

Nature Sustainability | VOL 1 | MAY 2018 | 234–239 | www.nature.com/natsustain 239


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

You might also like