How ants, birds and bats affect crop yield along shade gradients in tropical cacao agroforestry

Abstract

Tropical agroforests are diverse systems where several predator groups shape animal communities and plant–arthropod interactions. Ants, birds and bats in particular can reduce herbivore numbers and thereby increase crop yield. However, the relative importance of these groups, whether they interact, and how this interaction is affected by management and landscape context, is poorly understood. 2.We jointly manipulated access of ants, birds and bats in Indonesian smallholder cacao agroforestry across gradients of shade and distance to natural forest. We quantified arthropod abundance, pest damage and yield. 3.In control treatments, yield was highest under 30–40% canopy cover. Ant exclusion strongly reduced yield (from 600 to 300 kg ha−1 year−1) at 15% canopy cover. Bird exclusion impaired yield (from 400 to 250 kg ha−1 year−1) at 60% and enhanced yield (from 600 to 900 kg ha−1 year−1) at 15% canopy cover, while bats had no effect. Yield increased with forest proximity, a pattern not related to predator access. 4.No interactive effects among predator exclusions on yield, pest damage and arthropod communities were found. Ant exclusion increased numbers of herbivores below 30% canopy cover, without reducing spider abundances. Bird exclusion reduced herbivore and increased spider abundances. 5.Synthesis and applications. Using exclusion studies, we estimated that ants and birds cause cacao yield to vary between 100 to 800 kg ha−1 year−1, depending on shade-tree management. In all but the most shaded agroforests, ants were pivotal in supporting yields. Yields under low-canopy cover were strongly dependent on access by predator groups, with birds reducing rather than increasing yield. Hence, cacao farmers should refrain from disturbing ant communities and maintain 30–40% shade-tree canopy cover not only for ecophysiological reasons but also to buffer variability in predator communities. This article is protected by copyright. All rights reserved.

Full text

How ants, birds and bats affect crop yield along shade
gradients in tropical cacao agroforestry
Pierre Gras
1,
*, Teja Tscharntke
1
, Bea Maas
1,2
, Aiyen Tjoa
3
, Awal Hafsah
3
and Yann Clough
1,4
1
Agroecology, Georg-August-Universit€
at G€
ottingen, G€
ottingen, Germany;
2
Division of Tropical Ecology and Animal
Biodiversity, Department of Botany and Biodiversity Research, University of Vienna, Vienna, Austria;
3
Fakultas
Pertanian, Universitas Tadulako, Palu, Indonesia; and
4
Centre for Environmental and Climate Research, Lund
University, Lund, Sweden
Summary
1. Tropical agroforests are diverse systems where several predator groups shape animal com-
munities and plant–arthropod interactions. Ants, birds and bats in particular can reduce her-
bivore numbers and thereby increase crop yield. However, the relative importance of these
groups, whether they interact, and how this interaction is affected by management and land-
scape context, is poorly understood.
2. We jointly manipulated access of ants, birds and bats in Indonesian smallholder cacao
agroforestry across gradients of shade and distance to natural forest. We quantified arthro-
pod abundance, pest damage and yield.
3. In control treatments, yield was highest under 30–40% canopy cover. Ant exclusion
strongly reduced yield (from 600 to 300 kg ha
1
year
1
) at 15% canopy cover. Bird exclusion
impaired yield (from 400 to 250 kg ha
1
year
1
) at 60% and enhanced yield (from 600 to
900 kg ha
1
year
1
) at 15% canopy cover, while bats had no effect. Yield increased with for-
est proximity, a pattern not related to predator access.
4. No interactive effects among predator exclusions on yield, pest damage and arthropod
communities were found. Ant exclusion increased numbers of herbivores below 30% canopy
cover, without reducing spider abundances. Bird exclusion reduced herbivore and increased
spider abundances.
5. Synthesis and applications. Using exclusion studies, we estimated that ants and birds cause
cacao yield to vary between 100 and 800 kg ha
1
year
1
, depending on shade-tree manage-
ment. In all but the most shaded agroforests, ants were pivotal in supporting yields. Yields
under low-canopy cover were strongly dependent on access by predator groups, with birds
reducing rather than increasing yield. Hence, cacao farmers should refrain from disturbing
ant communities and maintain 30–40% shade-tree canopy cover not only for ecophysiological
reasons but also to buffer variability in predator communities.
Key-words: agricultural intensification, biocontrol, canopy cover, ecosystem services, forest
distance, mesopredator release, predation, Theobroma cacao, trophic interactions, yield
Introduction
Predators exert top-down control and can positively or
negatively influence plant development through direct and
indirect interactions affecting pests and diseases (Vander-
meer et al. 2002; Vandermeer, Perfecto & Philpott 2010).
Detailed reviews covering effects of ants, birds and bats
on arthropod communities and crop yield, often assessed
using exclusion experiments, underline their importance
for ecosystem service provision (Philpott & Armbrecht
2006; Van Bael et al. 2008; Whelan, Wenny & Marquis
2008; Mooney et al. 2010; Kunz et al. 2011; Wenny et al.
2011; Maas et al. 2015b).
Ants are effective biocontrol agents, especially in tropi-
cal agroforestry. For example, in Mexican coffee arboreal
ants protect trees from colonization by important pests
(Gonthier et al. 2013). However, impacts of ants depend
on the environmental context (e.g. temperature), ant
species involved (Philpott & Armbrecht 2006; Gove 2007;
Wielgoss et al. 2014), potential counterproductive effects
*Correspondence author. Department of Crop Sciences, Agroe-
cology, Georg-August-Universit€
at G€
ottingen, Grisebachstr. 6,
37077 G€
ottingen, Germany. E-mail: pgras@gwdg.de
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society
Journal of Applied Ecology 2016 doi: 10.1111/1365-2664.12625

on mesopredators such as spiders (Eubanks 2001), mutu-
alism and plant-sucking trophobionts (i.e. aphids and
mealbugs) that are harmful to crops. Birds have also been
shown to reduce abundances of pests (Holmes, Schultz &
Nothnagle 1979; Johnson et al. 2009), with yield losses
up to 310 kg ha
1
year
1
prevented in coffee agroforestry,
for example (Karp et al. 2013). Effects of birds on
arthropod communities are not always that strong, can
negatively affect mesopredators and do not necessarily
trickle down to plants (Williams-Guill
en, Perfecto & Van-
dermeer 2008; Maas, Clough & Tscharntke 2013). Bats
have received increased attention as biocontrol agents in
recent years following several seminal reports from natural
forests and agroforests (Kalka, Smith & Kalko 2008; Wil-
liams-Guill
en, Perfecto & Vandermeer 2008). Efforts to dis-
entangle bird and bat effects have partly confirmed the
potential of bats as biocontrol agents in agroforestry
(Maas, Clough & Tscharntke 2013; Maas et al. 2015b) and
partly suggested that negative impacts on spiders can can-
cel out the effects that bats may have on herbivores (Karp
&Daily2014).
Effects of these predator groups have usually been con-
sidered in isolation, using exclusion experiments con-
ducted in only one or a handful of locations. In reality,
density and diversity of predators, and thus probably also
their biocontrol effect, depend on local management or
landscape context (Clough et al. 2009; Karp et al. 2013).
In addition, it is unclear whether impacts of the three
predator groups are complementary or redundant. Few
studies have tested interactive effects (but see Mestre
et al. 2013b; Mooney 2007; Pi~
nol et al. 2010; Philpott
et al. 2004; Spiller & Schoener 2001). Finally, few studies
quantify impacts on crop yield, making economic assess-
ments difficult since plants can often compensate for
damage (but see Mooney et al. 2010; Wielgoss et al.
2014; Maas, Clough & Tscharntke 2013; Karp & Daily
2014).
Here, we simultaneously investigate effects and manage-
ment dependency of ant, bird and bat exclusions in 15
smallholder cacao plantations differing in the percentage
of shade-tree canopy cover and distance to forest margins.
In our study region, Central Sulawesi (Indonesia), results
from separate experiments revealed that both ant and
combined bird–bat exclusions decreased yield by ~30%
(Maas, Clough & Tscharntke 2013; Wielgoss et al. 2014).
These effects may change along canopy cover gradients as
local shade-tree management impacts productivity, fruit
abortion, and arthropod and vertebrate communities
(Bos, Steffan-Dewenter & Tscharntke 2007a; Clough et al.
2009). Increasing distance from natural forest can also
negatively affect the density and diversity of bird and bat
communities, and therefore the degree of biocontrol
(Estrada & Coates-Estrada 2002; Clough et al. 2009). In
the present study, we addressed the following questions:
1. Does exclusion of ants, birds and bats affect cacao
yield, and if so, does this change with shade-tree canopy
cover or distance to forest margin?
2. Are effects of the three manipulated predator groups
synergistic, additive or antagonistic?
3. What are the likely processes explaining impacts
of ant, bird or bat exclusions on cacao crop yield,
pest and disease infestation, and fruit set or fruit
abortion?
4. What are the management implications for cacao farm-
ers in terms of canopy cover, landscape-scale forest
preservation and predator conservation?
Materials and methods
SITE SELECTION AND GRADIENTS
Fifteen smallholder Theobroma cacao L. plantations were selected
at the eastern border of the Lore Lindu National Park in Central
Sulawesi, Indonesia (Fig. 1; 1°230318″S 120°1805755″E, ~1130 m
a.s.l., 1990–3804 mm rainfall) to cover five distances to the forest
margin (Fig. S1, three agroforests in each category: 0–250, 251–
500, 501–1500, 1501–2500, 2501–3000 m) and three canopy cover
types (Fig. S1, five agroforests in each category: <30%, 30–50%,
>50%). Later, both variables were measured in metres to the for-
est margin or percentage of shade-tree canopy cover. The latter
was measured at the experimental treatment level (within agro-
forest), but variability was larger between than within agro-
forests. Local farmers managed all agroforests and we
incorporated pesticide-, herbicide- and fungicide-free areas to
minimize management differences. Nonetheless, each agroforest
provided a unique habitat due to agroforest size, cacao tree age/
abundance, previous/nearby land-use, topography, drainage, fer-
tilizer use (Table S18) or intercropped vegetables and fruits
(Tables S2 and S3). The agroforests were separated by at least
500 m edge-to-edge distance. Two planted legume shade-tree spe-
cies –Erythrina subumbrans Merr. and Gliricidia sepium Kunth –
represented 30–91% of all recorded non-crop trees (Table S2,
herb species in Table S3). We trimmed the herb layers every
2 months using a motor scythe. During the experiment, we
recorded 49 ant and 69 bird species (Tables S4–S6).
EXCLUSION TREATMENTS
Bird and bat exclusions were implemented in 2010 (cf. Maas,
Clough & Tscharntke 2013), and ant exclusions were added in
April 2011; both were maintained until June 2012. Exclusions con-
sisted of eight randomly positioned treatments (two trees each)
per study site: (i) no exclusion, (ii) ant exclusion, (iii) bird exclu-
sion, (iv) bat exclusion, (v) ant–bird exclusion, (vi) ant–bat exclu-
sion, (vii) bird–bat exclusion and (viii) ant–bird–bat exclusion.
We used bamboo scaffolds covered with fishing nets
(35 935 mm mesh size) to exclude birds and/or bats. The bird–
bat exclusion treatments had fixed nets. Bird exclusion nets were
manually opened in the morning (05:30 h) and closed in the eve-
ning (18:30 h), while bat exclusions were opened in the evening
(18:30 h) and closed in the morning (05:30 h). Controls were
always open. Open nets (top and all sides) were bound tightly to
the scaffolds (Fig. S7). To exclude ants, we used cone-shaped
insect glue rings fitted to the tree trunks (Fig. S8), thereby reduc-
ing contact between ring and trunk and preventing trunks from
getting mouldy –a problem encountered while planning a previ-
ous study (Wielgoss et al. 2014). Thus, ant exclusion effects are
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
2P. Gras et al.

similar to conventional insect glue applications. We destroyed ant
nests during establishment of ant exclusions, whenever discovered
and during a monthly ‘search-and-destroy’ survey. Arthropod
exchange through net–tree contact was avoided by regularly
pruning branches to keep at least 30 cm between foliage and
nets/scaffolds. Effectiveness of ant exclusion was quantified by
counting ant individuals from canopy knock-down fogging sam-
ples, showing that ant abundance was strongly reduced (by 60–
90%, see Results) in exclusions compared to control treatments.
SAMPLING AND SURVEYING METHODS
We obtained distances to forest margins and agroforest area
using a Garmin Oregon 550 Global Positioning System device
(Fig. 1). We took treatment-level digital hemispherical canopy
photos and calculated percentage of canopy cover using the
CIMES-FISHEYE software (Walter 2009). We measured several
potentially yield-related covariates: (i) cacao tree crown volume
and (ii) mean d.b.h. to represent tree size, (iii) daily mean temper-
ature and (iv) precipitation to represent abiotic conditions, (v)
non-crop tree richness known to correlate with the bird commu-
nity composition (Clough et al. 2009), (vi) the average branch
perimeter as indicator of fruit carrying capacity and (vii) the pod
groove depth which indicates ‘on-fruit’ hiding opportunities for
arthropods and can differ between agroforests depending on the
cacao phenotype (Tables S9 and S10).
Every 2 weeks from April 2011 to May 2012, we surveyed all
experimental trees counting and classifying cacao fruits by size/
development and presence/absence of pest and disease symptoms
and harvesting of ripe fruits (classification details, Tables S11
and S12). We separately quantified the weight of marketable
and damaged beans (i.e. due to the damage of the cacao pod
borer Conopomorpha cramerella). Damage caused by the most
economically important organisms (i) C. cramerella, (ii) Helopel-
tis sulawesi and (iii) Phytophthora palmivora was assessed by
counting fruit with symptoms of damage.
We recorded other pests damaging leaves, flowers and fruits
(e.g. aphids, herbivorous bugs and caterpillars), as well as meso-
predators (e.g. spiders, earwigs and lacewing larvae) through
canopy knock-down fogging with a mixture of 5% Malathion
(Fumithion 1150 ULV) and diesel fuel (Bos, Steffan-Dewenter &
Tscharntke 2007b; Rizali et al. 2013) at the end of the experi-
ment. We fogged all treatments (covered by plastic tents) for
5 min and left the tents closed for 1 h. Fogging samples were
used to test predator exclusion effects on arthropod orders.
Every 4 weeks from May 2011 to March 2012, ants were baited
on trees without ant exclusions to detect effects of shade/forest
distance on ant community abundance, richness and composition.
We used protein (tinned tuna in brine) and carbohydrate (satu-
rated sugar solution) baits. We identified ant/morphospecies using
taxonomic literature (Bolton 1994; Fisher 2010) and regional ant
collections (Rizali et al. 2013; Wielgoss et al. 2014).
Between September 2010 and June 2011, the bird community
was recorded by repeated mist netting surveys and point count
recordings on all 15 study sites (see Maas et al. 2015a for details).
STATISTICAL ANALYSES
Data from 28 harvests were summed for each agroforest and
treatment. Response variables directly related to yield (total mar-
ketable yield, fruit abundance and fruit weight) were analysed in
three steps. First, we fitted a full model containing the random
factor ‘experimental agroforest’, and the terms of interest: (i)
three binary exclusion variables and their interaction terms, (ii)
agroforest-level design variables (canopy cover, forest distance)
and (iii) their interaction with each exclusion treatment variable,
(iv) seven covariates: crown volume, mean d.b.h., temperature,
precipitation, non-crop tree richness, branch perimeter and pod
groove depth. A second-order polynomial term was included for
canopy cover, to allow for nonlinearity (Tscharntke et al. 2011).
Continuous explanatory variables were z-transformed. We fitted
the model using maximum likelihood and Gaussian (marketable
Fig. 1. Study area in the Napu valley of
Central Sulawesi, Indonesia. Bottom: The
red area indicates Sulawesi. Top left:
Topography of Sulawesi and study area
(black square). Top right: Cycles show
experimental cacao agroforests, filling style
indicates canopy cover (empty <30%, half
=30–50%, filled >50%), white labels
show rainfall in mm, green area indicates
rain forest, reddish areas indicate houses,
and bright areas indicate open land and
lines indicate streets.
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
Shade-dependent predation effects on crop yield 3

yield, yield per fruit, number of harvested fruit, leaf area, leaf
damage), overdispersion-corrected Poisson (arthropod abun-
dances), or binomial (proportions of infested fruits) distribution.
For Gaussian models, root- or log-transformed response variables
were used to meet assumptions of homoscedasticity and (approxi-
mate) normality of the residuals. Models were assessed with vari-
ance inflation factors and diagnostic plots for residual normality,
heteroscedasticity and leverage.
Secondly, retaining all variables of interest, the model was sim-
plified by keeping those covariates that contributed to the model,
as indicated by a multi-model inference procedure (‘MuMIn’ R-
package, AIC based). The tree crown volume differed between
treatments and negatively correlated with canopy cover
(F=2092, P=004). Therefore, we included crown volume as
a permanent covariate during all statistical analyses. Thirdly, the
same procedure was used to identify non-significant interactions
that were removed for model simplification and refitted a model
using restricted maximum likelihood. In the results section, vari-
ables of interest and their interactions were reported when rela-
tive variable importance values (proportion of models including
the variable vs. models excluding the variable out of the best set
of models, identified by delta AIC <2) exceeded 09 for main
effect interactions and 07 for covariates. We used the lme4 pack-
age in R (Bates et al. 2014; R Core Team 2014) and reported sta-
tistical significance from the final model using ANOVA Type II,
Wald chi-square test (Kuznetsova, Brockhoff & Christensen
2014).
To determine the mechanisms leading to differences in mar-
ketable yields, only predictors identified as important for mar-
ketable yield (see Appendix S13) were used on the following
responses: (i) open flowers, (ii) aborted small fruits and (iii) leaf
herbivory. To test patterns of herbivorous arthropod abundances,
incidence of pests and disease symptoms, and beneficial arthro-
pod abundances, we used: (i) fruit infestation of H. sulawesi and
C. cramerella, (ii) Sternorrhyncha, (iii) caterpillars >10 mm, (iv)
Coleoptera >10 mm, (v) Auchenorrhyncha, (vi) Dermaptera, (vii)
Neuroptera larvae, (viii) hunting spiders <4 mm, (ix) hunting spi-
ders 4–10 mm, (x) web spiders <4 mm, (xi) web spiders 4–
10 mm, (xii) ants and (xiii) Diptera <4 mm.
Finally, changes in community composition were tested using
species richness, diversity, evenness and composition of ants and
birds along the gradients of canopy cover and forest distance.
The measurements were rarefied to 737 individuals per site for
ants and 122 individuals per site for birds. We conducted a
redundancy analysis and visualized community composition using
Minimum Convex Polygons (MCP) at the level of the agroforest
(birds) and experimental treatment (ants) and, finally, tested for
the influences of canopy cover, forest distance and treatment
using a permutation test (n=999). Separate analyses were con-
ducted for each variable, with both remaining variables and
crown volume being set as conditions. Crown volume and treat-
ment were not included for bird community analyses, as birds
were recorded at the agroforest level.
Results
EFFECTS OF PREDATOR EXCLUSION, CANOPY COVER
AND DISTANCE TO FOREST ON CACAO YIELD
Yields of control trees varied between 75% and 135% of
the average productivity per hectare of 540 kg ha
1
year
1
for Indonesia in 2012 (Fig. 2, Appendix S13.1; yield
Indonesia 2012: http://faostat3.fao.org). In control trees,
marketable yield, proportion of fruits without pest infesta-
tion and the number of harvested fruits peaked under 30–
40% shade-tree canopy cover (Figs 2 and 3g–l,
Appendix S13.1–S13.3). Yield was affected by ant and bird
exclusion with magnitude and direction of effects
depending on shade-tree canopy cover. Yield was not
affected by bat exclusion (Fig. 2, Appendix S13.1).
Interactions between exclusion treatments were not
significant. Independently of predator exclusions,
marketable yields and numbers of harvested
fruits decreased with increasing forest distance (Appendix
S13.1–S13.3).
ANT EXCLUSION
Ant exclusions reduced ant abundances by 60–90%
(Fig. 5a; Appendix S13.11). Marketable yield, the number
of harvested fruit and fruit weight were reduced under ant
exclusion (Figs 2 and 3j,m; Appendix S13.1–S13.3). Ant
abundances were positively correlated with marketable
yields while controlling for exclusion treatments (v
2
=722,
P<001).
Fig. 2. Effects of predator exclusion on
marketable cacao bean yield. Dry yield (y-
axis) of exclusion treatments (control =
continuous black, no ants =blue, no birds
=dotted red, no bats =dashed green)
dependent on percentage of canopy cover
(x-axis), grey line marks the average cacao
production of Indonesia 2012 (http://
faostat3.fao.org), and yield was standard-
ized to 12 months for plotting only.
v2
Intercept indicates influences of predator
exclusions, while v2
Slope indicates canopy
cover dependency of predator exclusions.
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
4P. Gras et al.

Below 30% canopy cover, ant exclusions were associ-
ated with fewer flowers, and a trend towards fewer
aborted small fruits compared to control treatments
(Figs 3a,d; Appendix S13.4 and S13.5). Ant exclusion did
not affect pest and disease incidence at fruits (Fig. 3g;
Appendix S13.6). Canopy cover above 50% reversed the
(a) (b) (c)
(c) (d) (e)
(f) (g) (h)
(i) (j) (k)
(l) (m) (n)
Fig. 3. Effects of predator exclusion on cacao tree phenology. Small figures show the response of cacao trees to ant (left), bird (mid) or
bat (right) exclusion related to canopy cover (x-axis), each circle/dot corresponds to two trees, line types: control =continuous black, no
ants =blue, no birds =dotted red, no bats =dashed green. v2
Intercept indicates influences of predator exclusions, while v2
Slope indicates
canopy cover dependency of predator exclusions.
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
Shade-dependent predation effects on crop yield 5

differences in aborted small fruits, with higher numbers in
ant exclusions than in controls (Fig. 3d). Losses of flowers
and small fruits were reflected in reduced numbers of har-
vested fruits under ant exclusion (Fig. 3j;
Appendix S13.3). Additionally, fruit weight was reduced
in the absence of ants (Fig. 3m; Appendix S13.2).
Abundances of lepidopteran caterpillars (>10 mm) and
beetles significantly increased in response to ant exclusions
while abundances of caterpillars (0–4 mm) and leafhoppers
remained unaffected (Figs 4a–j; Appendix S13.7–S13.10).
Leaf damage was increased in ant exclusions below 30%
canopy cover (v
2
=654, P=001; Appendix S13.16).
Spiders were largely unaffected by ant exclusion, with
only a trend towards small web-building spiders being
more frequent in ant exclusions below 30% canopy cover.
(Fig. 5d,g; Appendix S13.12 and S13.13). Below 30%
canopy cover, earwigs were absent, but when present,
their abundance was reduced in ant exclusions (Fig. 5j;
Appendix S13.14). Small beetles were more abundant
under ant exclusion and with above 50% canopy cover
(Fig. 5m; Appendix S13.15).
BIRD EXCLUSION
Compared to controls, marketable cacao yields were
slightly reduced in bird exclusions when canopy cover was
above 50%. Canopy cover below 30% lead to higher
yields in bird exclusions (500 kg ha
1
year
1
to
830 kg ha
1
year
1
, Fig. 2; Appendix S13.1). Comparable
patterns were found for flowers, aborted small fruits,
fruits without damage by pests such as H. sulawesi and C.
cramerella, and harvested fruits (Figs 3b–h;
Appendix S13.4–S13.7). Amounts of non-infected and
harvested fruits were significantly correlated (v
2
=3287,
P<001). The fruit weight remained unaffected by bird
exclusion (Fig. 3n; Appendix S13.3).
Caterpillar and beetle abundances remained unaffected
by bird exclusion, but abundances of Auchenorrhyncha
were reduced (Figs. 4b–k, Appendix S13.7–S13.10). Simul-
taneously, web-building spider abundance increased,
although this effect was only significant for large individu-
als (4–10 mm), where it was stronger over 50% canopy
cover (Fig. 5h, Appendix S13.13). Increases in small web
spiders (<4 mm) were marginally significant and tended to
be higher with canopy cover <30% (Fig. 5e,
Appendix S13.12). Small beetles (<4 mm) tended to be less
abundant with more than 30% canopy cover (Fig. 5n,
Appendix S13.15). Earwigs had lower densities due to bird
exclusion, an effect restricted to canopy cover higher than
40% (Fig. 5k, Appendix S13.14). Ants were not signifi-
cantly affected by bird exclusion (Fig. 5b,
Appendix S13.11). Yield effects of bird exclusion persisted
when adding ant abundance as a model covariate. Small
dipterans were less frequent under bird exclusions
(v2
Bird exclusion =659, P=001; Appendix S15.17 in Support-
ing Information).
BAT EXCLUSION
Excluding bats had no significant effect on marketable
yield. No effect could be detected on flowers, aborted
small fruits, fruit infestation by pests and diseases,
harvested fruits or fruit weight (Figs 2 and 3c–o;
Appendix S15.1–S15.6). Bat exclusion was associated with
higher numbers of leaf hoppers and, for canopy cover
over 50%, higher numbers of large caterpillars (4–10 mm)
(Fig. 4f,l, Appendix S15.7/10). Small caterpillars or bee-
tles remained unaffected (Fig. 4a,i, Appendix S15.7/9).
Bat exclusion did not affect mesopredators smaller
than 4 mm in body length, but was associated with higher
spider and earwig abundance (Fig. 5c–o, Appendix
S15.11–15).
ANT AND BIRD DIVERSITY ALONG THE CANOPY COVER
AND FOREST DISTANCE GRADIENT
Ant species composition (redundancy analysis, Fig. S14)
was similar across those treatments where ants were present
(F=076, P=084), but ant species composition strongly
changed with percentage of canopy cover (F=281,
P=0001) and forest distance (F =293, P=0001).
Species composition of the local bird assemblage
(redundancy analysis, Fig. S15) changed due to forest dis-
tance (F =131, P=003), but did not respond to canopy
cover (F =112, P=033). Shannon index, species
evenness and rarefied species richness of ants and birds
were independent of canopy cover and forest distance
(Fig. S16).
Discussion
Access by predators, percentage of shade-tree canopy
cover and distance to forest margin had economically rel-
evant impacts on cacao yield. While the negative effect of
forest distance was independent of predator exclusion,
percentage of canopy cover affected the impact of preda-
tors. Yield peaked at 30–40% canopy cover. Canopy
cover below 15% or above 55% resulted in less than half
of the optimum yield. Ant exclusion reduced yields at
15% canopy cover from 600 to 300 kg ha
1
year
1
. Bird
exclusion decreased yields at 60% canopy cover from 400
to 250 kg ha
1
year
1
. Unexpectedly, bird exclusion
increased yields at 15% canopy cover from 600 to 900 kg
ha
1
year
1
. Bats had no effect on yield.
Canopy cover-dependent effects of ant and bird
exclusions in agroforests have never been shown previ-
ously, despite several recent studies targeting these
groups in tropical agroforestry (Wielgoss et al. 2014;
Maas et al. 2015b). Our results can partly be explained
by patterns in yield formation, as well as impacts on
herbivorous and predatory arthropods that were quanti-
fied by fogging the experimental trees at the end of the
study period.
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
6P. Gras et al.

CANOPY COVER-DEPENDENT EFFECTS OF ANTS
Ants affect plants by tending/defending or predating/dis-
placing herbivores, but also predating/displacing predators
(such as spiders) and pollinators, spreading propagules of
plant pathogens (Wielgoss et al. 2014), pollinating flowers,
predating/dispersing seeds and changing soil conditions
(Bartlett 1961; Way 1963; Mestre et al. 2013b; 2014).
Their ecological dominance in tropical agroforestry sys-
tems likely affects trees in multiple ways (e.g. plant
growth, leaf and fruit development). In our study, impacts
of ant exclusion have their likely cause in the interaction
of arboreal ants and other arboreal organisms such as
herbivores: ants were excluded from trees and not the
ground below, tent-building species able to spread patho-
gens were absent, and ants are not able to pollinate cacao
(Leston 1970). Interference between ants and spiders has
been reported from tree crops (Pi~
nol, Espadaler & Ca~
nel-
las 2012; Mestre et al. 2012; Mestre, Bucher & Entling
2014 but see Mar
ın & Perfecto 2013). Here, and in con-
trast to what was observed in response to bird exclusion,
no significant change in spider abundance occurred under
ant exclusion, suggesting that ants do not limit the abun-
dance of spiders. Instead, ants reduced abundances of her-
bivores such as caterpillars and beetles that damage not
only leaves but also cacao flowers (YC personal observa-
(a) (b) (c)
(d) (e) (f)
(g) (h) (i)
(j) (k) (l)
Fig. 4. Effects of predator exclusion on cacao tree herbivores. Small figures show the response of herbivores to ant (left), bird (mid) or
bat (right) exclusion related to canopy cover (x-axis), each circle/dot corresponds to two trees, line types: control =continuous black, no
ants =blue, no birds =dotted red, no bats =dashed green, v2
Intercept indicates influences of predator exclusions, while v2
Slope indicates
canopy cover dependency of predator exclusions.
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
Shade-dependent predation effects on crop yield 7

tion; Bos, Steffan-Dewenter & Tscharntke 2007a; Maas,
Clough & Tscharntke). Increased herbivore abundance,
parallel increased leaf damage and fewer flowers under
ant exclusion probability resulted in fewer fruits.
Simultaneously, reinforcing negative effects on yield, fruit
weight was reduced when ants were absent, which is asso-
ciated with infestation by fruit-damaging pests such as H.
sulawesi and C. cramerella (Wielgoss et al. 2014). Impacts
(a) (b) (c)
(d) (e) (f)
(g) (h) (i)
(j) (k) (l)
(m) (n) (o)
Fig. 5. Effects of predator exclusion on cacao tree mesopredators. Small figures show the response of mesopredators to ant (left), bird
(mid) or bat (right) exclusion related to canopy cover (x-axis), each circle/dot corresponds to two trees, line types: control =continuous
black, no ants =blue, no birds =dotted red, no bats =dashed green, v2
Intercept indicates influences of predator exclusions, while v2
Slope
indicates canopy cover dependency of predator exclusions.
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
8P. Gras et al.

of ants on crop yield, including their economic impor-
tance, are similar to those reported from sites situated in
the same region at lower altitudes (Wielgoss et al. 2014),
but our results demonstrate that ant predation effects
were only economically relevant for agroforests providing
<40% canopy cover. Generally, low-shade conditions can
be more stressful for cacao trees, both in terms of eco-
physiology, and in terms of herbivory (De Almeida &
Valle 2007; Tscharntke et al. 2011). These findings are in
accordance with our results, showing that small beetles
and caterpillar densities increased under low-shade condi-
tions. Besides this, we could show that there was signifi-
cant ant species turnover across the gradient in canopy
cover. It thus cannot be excluded that ant species more
dominant under low-canopy cover conditions provide a
more effective pest control than species dominant under
high-canopy cover.
BIRD EXCLUSIONS CAN DECREASE OR INCREASE
YIELDS DEPENDING ON CANOPY COVER
Against expectation, excluding birds led to more fruits and
higher proportions of healthy fruit below 30% canopy
cover. Lowered or similar yields under bird exclusion would
have been expected throughout the canopy cover gradient,
but were observed only in relation to canopy cover above
40%. We assume that the effects of excluding birds are
caused by the absence of insectivorous birds, of which 31
species were recorded at cacao canopy level (Maas et al.
2015a). Indeed, none of the bird species recorded feed on
cacao fruits. Woodpecker damage can occasionally be seen
on fruit (YC personal observation), but is likely to only
affect fruit with prior pest infestation. Pollination reduction
due to bird exclusion is unlikely, since cacao flowers are very
small and pollinated by midges, not birds (Leston 1970).
The bird community composition did not change with
canopy cover, suggesting that shade dependency of yields
under bird exclusions may be due to observed differences in
arthropod pest and mesopredator densities along the gradi-
ent. Generally, there was little response in herbivore abun-
dance to bird exclusion. Leafhopper numbers were
depressed and earwig numbers promoted under bird exclu-
sion when canopy cover was below 30%. This, together with
lower numbers of damaged fruits under bird exclusion,
points to indirect positive effects on herbivores by birds
(e.g. mesopredator predation) under little-shaded condi-
tions where herbivory affects cacao trees more severely
(Tscharntke et al. 2011). The exact mechanism is unclear, as
the fogging data show only a trend towards increases in
predator densities at canopy cover below 30% for ants and
small web spiders with increases and decreases at canopy
cover above 50% for larger web spiders and earwigs, respec-
tively. Generally, increased abundances of predators, such
as ants, spiders and earwigs, under bird exclusions, are com-
mon (Gunnarsson 2007; Williams-Guill
en, Perfecto & Van-
dermeer 2008; Pi~
nol et al. 2010; Maas, Clough &
Tscharntke 2013; Mestre et al. 2013a). Overall, increases in
most predatory arthropods and decreases in individual her-
bivores due to bird exclusion suggest mesopredator release
and may at least partly explain the ambivalent effects of
birds on cacao yield in our study, although food web data
will be necessary to confirm this. The impact of lower abun-
dances in small Dipterans (<4 mm), which includes cacao-
pollinating Ceratopogonidae (Leston 1970), is unknown,
yet could conceivably reduce pollination and thereby cause
yield losses (Groeneveld et al. 2010).
BATS
In 2010, at least 16 insectivorous bat species were
recorded in agricultural areas around the Lore Lindu
National Park in two separate studies by Graf and Boon-
man (unpublished data, Table S17). Several of the species
glean insects from leaves or perch on plants to prey on
arthropods associated with plants and might therefore be
relevant to our experimental exclusions. However, bat
exclusion did not affect yield, or yield-related plant vari-
ables, but enhanced abundances of leafhoppers, large
caterpillars, and earwigs and large spiders, that is both
herbivorous and predatory arthropods. Bats have strong
effects on arthropod communities in tropical forests and
agricultural systems (Kalka, Smith & Kalko 2008; Wil-
liams-Guill
en, Perfecto & Vandermeer 2008; Wanger et al.
2014). In coffee, bats can promote herbivorous arthro-
pods by reducing spider densities and, as a result, do not
necessarily provide biocontrol (Karp & Daily 2014). A
previous study from cacao in our study region, however,
showed bats were instrumental in decreasing pests (Maas,
Clough & Tscharntke 2013). This suggests that impacts of
bats could be strongly dependent on the net balance of
arthropod herbivores and predators, with bat impacts on
both groups cancelling out any top-down effect on the
plants in the present study. Leaf-gleaning bats are less
abundant in agricultural systems than in natural forest
(Phommexay et al. 2011) suggesting natural forests are
sources of bats for nearby agroforests, but we did not find
any evidence for forest distance-dependent effects of bats.
STUDY LIMITATIONS AND CAVEATS
The extent to which we can causally underpin the detected
significant effects is limited by the study design in some
respects. First, the impact of forest distance on yield, while
significant, cannot be fully elucidated, as it seems not to
involve the manipulated predators. Secondly, while we tried
to control for variables that could not be standardized,
unwanted exclusion effects may have occurred. For
instance, reduced earwig abundances under ant exclusion
may have been due to the ant exclusion rings partly exclud-
ing crawling earwigs. While this may have led to overesti-
mated impacts of ants under high-canopy cover, impacts of
ants were strongest below 30% canopy cover, where ear-
wigs were scarce, so our conclusions on ant effects are
robust. Moreover, earwig abundances that increased over
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
Shade-dependent predation effects on crop yield 9

40% canopy cover in bat exclusions did not affect cacao
yield. Thirdly, we cannot exclude that the presence of nets
deterred certain birds or bats from foraging inside the agro-
forests, due to the risk of collision. Finally, the drawback of
a long study duration was the impossibility of adequately
surveying arthropods during the whole course of the experi-
ment. Parallels between herbivores and predators recorded
through fogging, and cacao tree variables recorded over the
whole study duration, allow us to draw hypotheses on
mechanisms, but not to formally test them.
MANAGEMENT RECOMMENDATIONS
Impacts of ants and birds on cacao yield are economically
relevant, but depend on shade-tree management. In all but
most shaded agroforests, ants were pivotal in supporting
yields, reducing populations of herbivorous insects rather
than other beneficial organisms such as spiders. This con-
firms the importance of ants for the economic performance
of cacao agroforestry in Central Sulawesi. Farmers should
be aware of changes in ant communities due to shifts in
management or species invasion, and avoid destroying ant
nests, as commonly done for weaver ants, for example
(YC & PG personal observation). Moreover, current cacao
farming methods entail self-shaded or low-shaded (~10%
shade-tree canopy cover) plantations impairing arthropod
and plant diversity (Rice & Greenberg 2000). The observa-
tion that, on average, shade-tree canopy cover of 30–40%
yielded most, suggests that farmers should maintain such
intermediate canopy cover levels not only for ecophysio-
logical reasons (Tscharntke et al. 2011), but also to buffer
variability in predator communities. Our data suggests
that beneficial effects of forest proximity on yield were
unlikely to be related to pest limitation services by ants
and birds. However, positive effects of forest proximity on
predation of dummy caterpillars, mediated by higher
abundances of a locally common white-eye species Zos-
terops chloris, have been demonstrated in our study area
(Maas et al. 2015a), and together these results suggest that
farmers should have an interest in the stability of forest
margins, which are threatened by encroachment.
Acknowledgements
We are grateful to partners at Tadulako University, IPB and LIPI. We
thank Arjan Boonman and Stefan Graf for providing bat survey data of
our study region, Akhmad Rizali (ants), Hardianto Mangopo (trees) and
Firdaus (herbs) for specimen identifications and Arno Wielgoss, Iris
Motzke, Ilfianty Kasmundin and Pak Abdul Rauf for support. Special
thanks we address to all research (Dadang Dwi-Putra, Edi Djismin, Alfi-
anus Rumede), laboratory (Ivon, Abe and Fatma) and field assistants
(Opu, Nimus, Anto, Tia, Papa Ifer, Idi, Soni, Wandi, Anki, Sardin and
Sony), all farmers and project members, especially Wolfram Lorenz. The
project was funded by DFG (ELUC CL-474/1-1 and CRC 990 EFForTS).
Data accessibility
Data are available through EFForTS-Information System
(https://efforts-is.uni-goettingen.de) and Dryad Digital Repository
doi: 10.5061/dryad.90329 (Gras et al. 2016).
References
Bartlett, B.R. (1961) The influence of ants upon parasites, predators, and
scale insects. Annals of the Entomological Society of America,54, 543–
551.
Bates, D., Maechler, M., Bolker, B. & Walker, S. (2014) lme4: linear
mixed-effects models using S4 classes. R package version 1.1-6. R.
Bolton, B. (1994) Identification Guide to the Ant Genera of the World. Har-
vard University Press, Cambridge.
Bos, M.M., Steffan-Dewenter, I. & Tscharntke, T. (2007a) Shade tree
management affects fruit abortion, insect pests and pathogens of cacao.
Agriculture, Ecosystems & Environment,120, 201–205.
Bos, M.M., Steffan-Dewenter, I. & Tscharntke, T. (2007b) The contribution
of cacao agroforests to the conservation of lower canopy ant and beetle
diversity in Indonesia. Biodiversity and Conservation,16, 2429–2444.
Clough, Y., Dwi-Putra, D., Pitopang, R. & Tscharntke, T. (2009) Local
and landscape factors determine functional bird diversity in Indonesian
cacao agroforestry. Biological Conservation,142, 1032–1041.
De Almeida, A.-A.F. & Valle, R.R. (2007) Ecophysiology of the cacao
tree. Brazilian Journal of Plant Physiology,19, 425–448.
Estrada, A. & Coates-Estrada, R. (2002) Bats in continuous forest, forest
fragments and in an agricultural mosaic habitat-island at Los Tuxtlas,
Mexico. Biological Conservation,103, 237–245.
Eubanks, M.D. (2001) Estimates of the direct and indirect effects of red
imported fire ants on biological control in field crops. Biological Con-
trol,21,35–43.
Fisher, B.L. (2010) Ants of Borneo - Guide to the Ant Genera of Borneo
(Ant course 2010).
Gonthier, D.J., Ennis, K.K., Philpott, S.M., Vandermeer, J. & Perfecto, I.
(2013) Ants defend coffee from berry borer colonization. BioControl,
58, 815–820.
Gove, A.D. (2007) Ant biodiversity and the predatory function (A
response to Philpott and Armbrecht, 2006). Ecological Entomology,32,
435–436.
Gras, P., Tscharntke, T., Maas, B., Tjoa, A., Hafsah, A. & Clough, Y.
(2016) Data from: How ants, birds and bats affect crop yield along
shade gradients in tropical cacao agroforestry. Dryad Digital Repository,
http://dx.doi.org/10.5061/dryad.90329.
Groeneveld, J.H., Tscharntke, T., Moser, G. & Clough, Y. (2010) Experi-
mental evidence for stronger cacao yield limitation by pollination than
by plant resources. Perspectives in Plant Ecology, Evolution and System-
atics,12, 183–191.
Gunnarsson, B. (2007) Bird predation on spiders: ecological mechanisms
and evolutionary consequences. Journal of Arachnology,35, 509–529.
Holmes, R.T., Schultz, J.C. & Nothnagle, P. (1979) Bird predation on for-
est insects: an exclosure experiment. Science,206, 462–463.
Johnson, M.D., Levy, N.J., Kellermann, J.L. & Robinson, D.E. (2009)
Effects of shade and bird exclusion on arthropods and leaf damage on
coffee farms in Jamaica’s Blue Mountains. Agroforestry Systems,76,
139–148.
Kalka, M.B., Smith, A.R. & Kalko, E.K.V. (2008) Bats limit arthropods
and herbivory in a tropical forest. Science,320, 71.
Karp, D.S. & Daily, G.C. (2014) Cascading effects of insectivorous birds
and bats in tropical coffee plantations. Ecology,95, 1065–1074.
Karp, D.S., Mendenhall, C.D., Sand

ı, R.F., Chaumont, N., Ehrlich, P.R.,
Hadly, E.A. & Daily, G.C. (2013) Forest bolsters bird abundance, pest
control and coffee yield. Ecology letters,16, 1339–1347.
Kunz, T.H., de Torrez, E.B., Bauer, D., Lobova, T. & Fleming, T.H.
(2011) Ecosystem services provided by bats. Annals of the New York
Academy of Sciences,1223,1–38.
Kuznetsova, A., Brockhoff, P.B. & Christensen, H.B. (2014) lmerTest:
Tests for random and fixed effects for linear mixed effect models (lmer
objects of lme4 package). R package version, R package version 2.0–6.
Leston, D. (1970) Entomology of the cocoa farm. Annual Review of Ento-
mology,15, 273–294.
Maas, B., Clough, Y. & Tscharntke, T. (2013) Bats and birds increase
crop yield in tropical agroforestry landscapes. Ecology Letters,16,
1480–1487.
Maas, B., Tscharntke, T., Saleh, S., Dwi-Putra, D. & Clough, Y. (2015a)
Avian species identity drives predation success in tropical cacao agro-
forestry. Journal of Applied Ecology,52, 735–743.
Maas, B., Karp, D.S., Bumrungsri, S., Darras, K., Gonthier, D., Huang,
J.C.C. et al. (2015b) Bird and bat predation services in tropical
forests and agroforestry landscapes. Biological Reviews, doi:10.1111/
brv.12211.
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
10 P. Gras et al.

Mar
ın, L. & Perfecto, I. (2013) Spider diversity in coffee agroecosystems:
the influence of agricultural intensification and aggressive ants. Environ-
mental Entomology,42, 204–213.
Mestre, L., Bucher, R. & Entling, M.H. (2014) Trait-mediated effects
between predators: ant chemical cues induce spider dispersal. Journal of
Zoology,293, 119–125.
Mestre, L., Pi~
nol, J., Barrientos, J.A., Cama, A. & Espadaler, X. (2012)
Effects of ant competition and bird predation on the spider assemblage
of a citrus grove. Basic and Applied Ecology,13, 355–362.
Mestre, L., Garcia, N., Barrientos, J.A., Espadaler, X. & Pi~
nol, J. (2013a)
Bird predation affects diurnal and nocturnal web-building spiders in a
Mediterranean citrus grove. Acta Oecologica,47,74–80.
Mestre, L., Pi~
nol, J., Barrientos, J.A. & Espadaler, X. (2013b) Ant exclu-
sion in citrus over an 8-year period reveals a pervasive yet changing
effect of ants on a Mediterranean spider assemblage. Oecologia,173,
239–248.
Mooney, K.A. (2007) Tritrophic effects of birds and ants on a canopy
food web, tree growth, and phytochemistry. Ecology,88, 2005–2014.
Mooney, K.A., Gruner, D.S., Barber, N.A., Van Bael, S.A., Philpott,
S.M. & Greenberg, R. (2010) Interactions among predators and the cas-
cading effects of vertebrate insectivores on arthropod communities and
plants. Proceedings of the National Academy of Sciences of the United
States of America,107, 7335–7340.
Philpott, S.M. & Armbrecht, I. (2006) Biodiversity in tropical agroforests
and the ecological role of ants and ant diversity in predatory function.
Ecological Entomology,31, 369–377.
Philpott, S.M., Greenberg, R., Bichier, P. & Perfecto, I. (2004) Impacts of
major predators on tropical agroforest arthropods: comparisons within
and across taxa. Oecologia,140, 140–149.
Phommexay, P., Satasook, C., Bates, P., Pearch, M. & Bumrungsri, S.
(2011) The impact of rubber plantations on the diversity and activity of
understorey insectivorous bats in southern Thailand. Biodiversity and
Conservation,20, 1441–1456.
Pi~
nol, J., Espadaler, X. & Ca~
nellas, N. (2012) Eight years of ant-exclusion
from citrus canopies: effects on the arthropod assemblage and on fruit
yield. Agricultural and Forest Entomology,14,49–57.
Pi~
nol, J., Espadaler, X., Ca~
nellas, N., Mart

ınez-Vilalta, J., Barrientos, J.A.
& Sol, D. (2010) Ant versus bird exclusion effects on the arthropod
assemblage of an organic citrus grove. Ecological Entomology,35, 367–
376.
R Core Team (2014) R: A Language and Environment for Statistical Com-
puting. R Foundation for Statistical Computing, Vienna, Austria, URL
http://www.R–project.org/.
Rice, R.A. & Greenberg, R. (2000) Cacao cultivation and the conservation
of biological diversity. Ambio,29, 167–173.
Rizali, A., Clough, Y., Buchori, D., Hosang, M.L.A., Bos, M.M. &
Tscharntke, T. (2013) Long-term change of ant community structure in
cacao agroforestry landscapes in Indonesia. Insect Conservation and
Diversity,6, 328–338.
Spiller, D.A. & Schoener, T.W. (2001) An experimental test for preda-
tor-mediated interactions among spider species. Ecology,82, 1560–
1570.
Tscharntke, T., Clough, Y., Bhagwat, S.A., Buchori, D., Faust, H., Hertel,
D. et al. (2011) Multifunctional shade-tree management in tropical
agroforestry landscapes–a review. Journal of Applied Ecology,48, 619–
629.
Van Bael, S.A., Philpott, S.M., Greenberg, R., Bichier, P., Barber, N.A.,
Mooney, K.A. & Gruner, D.S. (2008) Birds as predators in tropical
agroforestry systems. Ecology,89, 928–934.
Vandermeer, J., Perfecto, I. & Philpott, S. (2010) Ecological complexity
and pest control in organic coffee production: uncovering an autono-
mous ecosystem service. BioScience,60, 527–537.
Vandermeer, J., Perfecto, I., Nu~
nez, G.I., Phillpott, S. & Ballinas, A.G.
(2002) Ants (Azteca sp.) as potential biological control agents in shade
coffee production in Chiapas, Mexico. Ecology,56, 271–276.
Walter, J.-M. (2009) CIMES-FISHEYE. Hemispherical Photography of
Forest Canopies.
Wanger, T.C., Darras, K., Bumrungsri, S., Tscharntke, T. & Klein, A.-M.
(2014) Bat pest control contributes to food security in Thailand. Biologi-
cal Conservation,171, 220–223.
Way, M.J. (1963) Mutualism between ants and honeydew-producing
Homoptera. Annual Review of Entomology,8, 307–344.
Wenny, D.G., Devault, T.L., Johnson, M.D., Kelly, D., Sekercioglu,
C.H., Tomback, D.F. & Whelan, C.J. (2011) The need to quantify
ecosystem services provided by birds. The Auk,128,1–14.
Whelan, C.J., Wenny, D.G. & Marquis, R.J. (2008) Ecosystem Services
provided by birds. Annals of the New York Academy of Sciences,1134,
25–60.
Wielgoss, A., Tscharntke, T., Rumede, A., Fiala, B., Seidel, H., Shahabud-
din, S. & Clough, Y. (2014) Interaction complexity matters: disentan-
gling services and disservices of ant communities driving yield in
tropical agroecosystems. Proceedings of the Royal Society B,281,
20132144.
Williams-Guill
en, K., Perfecto, I. & Vandermeer, J. (2008) Bats limit
insects in a neotropical agroforestry system. Science,320, 70.
Received 19 September 2015; accepted 4 February 2016
Handling Editor: Peter Manning
Supporting Information
Additional Supporting Information may be found in the online version
of this article.
Fig. S1. Schematic experimental design.
Table S2. Non-crop tree species within chemical application free
area.
Table S3. Herbal plant species.
Table S4. Discovered ant species.
Table S5. Discovered bird species.
Table S6. Ant and bird community.
Fig S7. Open bat exclusion.
Fig. S8. Conical-shaped ant exclusion rings (without insect glue).
Table S9. Potentially yield influencing variables (additional to the
predator exclusion).
Table S10. Characteristics of experimental trees and plantations at
treatment level.
Table S11. Fruit development categories.
Table S12. Pest and disease categories.
Appendix S13. ANOVA Tables.
Fig. S14. RDA –Management, landscape, and predator access
manipulation influences on the ant community composition (mor-
pho species level).
Fig. S15. RDA –Management and landscape effects on the bird
community (species level).
Fig. S16. Visualization of bird and ant community composition
related to shade cover and forest distance.
Table S17. Bat species.
©2016 The Authors. Journal of Applied Ecology ©2016 British Ecological Society, Journal of Applied Ecology
Shade-dependent predation effects on crop yield 11