SC I ENCE ADVANCES | R E S EARCH ART I C L EAGR ICULTUREDainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019Copyright © 2019
The Authors, some
rights reserved;
exclusive licensee
American Association
for the Advancement
of Science. No claim to
originalU.S. Government
Works. Distributed
under a Creative
Commons Attribution
NonCommercial
License 4.0 (CC BY-NC).Downloaded from http://advances.sciencemag.org/ on October 17, 2019A global synthesis reveals biodiversity-mediated
benefits for crop production
Matteo Dainese1,2*, Emily A. Martin2, Marcelo A. Aizen3, Matthias Albrecht4, Ignasi Bartomeus5,
Riccardo Bommarco6, Luisa G. Carvalheiro7,8, Rebecca Chaplin-Kramer9, Vesna Gagic10,
Lucas A. Garibaldi11, Jaboury Ghazoul12, Heather Grab13, Mattias Jonsson6, Daniel S. Karp14,
Christina M. Kennedy15, David Kleijn16, Claire Kremen17, Douglas A. Landis18,
Deborah K. Letourneau19, Lorenzo Marini20, Katja Poveda13, Romina Rader21, Henrik G. Smith22,23,
Teja Tscharntke24, Georg K. S. Andersson22, Isabelle Badenhausser25,26, Svenja Baensch24,27,
Antonio Diego M. Bezerra28, Felix J. J. A. Bianchi29, Virginie Boreux12,30, Vincent Bretagnolle31,
Berta Caballero-Lopez32, Pablo Cavigliasso33, Aleksandar Ćetković34, Natacha P. Chacoff35,
Alice Classen2, Sarah Cusser36, Felipe D. da Silva e Silva37,38, G. Arjen de Groot39,
Jan H. Dudenhöffer40, Johan Ekroos22, Thijs Fijen16, Pierre Franck41, Breno M. Freitas28,
Michael P. D. Garratt42, Claudio Gratton43, Juliana Hipólito11,44, Andrea Holzschuh2,
Lauren Hunt45, Aaron L. Iverson13, Shalene Jha46, Tamar Keasar47, Tania N. Kim48,
Miriam Kishinevsky49, Björn K. Klatt23,24, Alexandra-Maria Klein30, Kristin M. Krewenka50,
Smitha Krishnan12,51,52, Ashley E. Larsen53, Claire Lavigne41, Heidi Liere54, Bea Maas55,
Rachel E. Mallinger56, Eliana Martinez Pachon57, Alejandra Martínez-Salinas58,
Timothy D. Meehan59, Matthew G. E. Mitchell60, Gonzalo A. R. Molina61, Maike Nesper12,
Lovisa Nilsson22, Megan E. O'Rourke62, Marcell K. Peters2, Milan Plećaš34, Simon G. Potts43,
Davi de L. Ramos63, Jay A. Rosenheim64, Maj Rundlöf23, Adrien Rusch65, Agustín Sáez66,
Jeroen Scheper16,39, Matthias Schleuning67, Julia M. Schmack68, Amber R. Sciligo69,
Colleen Seymour70, Dara A. Stanley71, Rebecca Stewart22, Jane C. Stout72, Louis Sutter4,
Mayura B. Takada73, Hisatomo Taki74, Giovanni Tamburini30, Matthias Tschumi4,
Blandina F. Viana75, Catrin Westphal27, Bryony K. Willcox21, Stephen D. Wratten76,
Akira Yoshioka77, Carlos Zaragoza-Trello5, Wei Zhang78, Yi Zou79, Ingolf Steffan-Dewenter2
Human land use threatens global biodiversity and compromises multiple ecosystem functions critical to
food production. Whether crop yield–related ecosystem services can be maintained by a few dominant
species or rely on high richness remains unclear. Using a global database from 89 studies (with 1475 loca-
tions), we partition the relative importance of species richness, abundance, and dominance for pollination;
biological pest control; and final yields in the context of ongoing land-use change. Pollinator and enemy
richness directly supported ecosystem services in addition to and independent of abundance and domi-
nance. Up to 50% of the negative effects of landscape simplification on ecosystem services was due to
richness losses of service-providing organisms, with negative consequences for crop yields. Maintaining
the biodiversity of ecosystem service providers is therefore vital to sustain the flow of key agroecosystem
benefits to society.INTRODUCTION
Natural and modified ecosystems contribute a multitude of functions
and services that support human well-being (1, 2). It has long been re-
cognized that biodiversity plays an important role in the functioning of
ecosystems (3, 4), but the dependence of ecosystem services on bio-
diversity is under debate.An early synthesis revealed inconsistent results
(5), whereas subsequent studies suggest that a few dominant species
may supply the majority of ecosystem services (6, 7). It thus remains
unclear whether a few dominant or many complementary species are
needed to supply ecosystem services.
The interpretation of earlier studies has been controversial because
multiplemechanisms underlying changes in ecosystem service response
to biodiversity can operate in combination (8, 9). On one hand, com-
munities with many species are likely to include species responsible for
large community-wide effects due to statistical selection. On the other
hand, such diverse communities may contain a particular combinationof species that complement each other in service provisioning. While
these mechanisms imply positive effects of species richness on eco-
system service supply, total organism abundance or dominance of cer-
tain species may also drive the number of interactions benefiting
ecosystem service supply. Depending on the relative importance of spe-
cies complementarity, community abundance, and the role of dominant
species, different relationships between species richness and ecosystem
services can be expected (10).
In real-world ecosystems, natural communities consist of a fewhigh-
ly abundant (dominant species) andmany rare ones. The importance of
richness, abundance, and dominance is likely to be influenced by the
extent to which relative abundance changes with species richness (11)
and by differences in the effectiveness and degree of specialization of
service-providing communities. However, these three aspects of diver-
sity have typically been tested in isolation and mainly in small-scale ex-
perimental settings (12, 13), while a synthetic study contrasting their1 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L Erelative importance in real-world ecosystems is still lacking. A major
limitation to resolving these relationships is a lack of evidence from
real-world human-driven biodiversity changes (14, 15), particularly
for ecosystem services in agroecosystems. For instance, changes in rich-
ness and total or relative abundance of service-providing organisms in
response to land-clearing for agriculture (16, 17) could alter the flow of
benefits to people in different ways compared to experimental random
loss of biodiversity.
Over the past half-century, the need to feed a growing world popu-
lation has led to markedly expanded and intensified agricultural pro-
duction, transforming many regions into simplified landscapes (18).
This transformation not only has contributed to enhanced agricultural
production but also has led to the degradation of the global
environment. The loss of biodiversity can disrupt key intermediate
services to agriculture, such as crop pollination (19) and biological pest
control (20), which underpin the final provisioning service of crop1Institute for Alpine Environment, Eurac Research, Viale Druso 1, 39100 Bozen/Bolzan
of Würzburg, Am Hubland, 97074 Würzburg, Germany. 3Grupo de Ecología de la Pol
Rio Negro, Argentina. 4Agroecology and Environment, Agroscope, Reckenholzstrass
grative Ecology, E-41092 Sevilla, Spain. 6Swedish University of Agricultural Sciences
Universidade Federal de Goias (UFG), Goiânia, Brazil. 8Faculdade de Ciencias, Centre f
Lisboa, Portugal. 9Natural Capital Project, Stanford University, Stanford, CA 94618, U
gaciones en Recursos Naturales, Agroecología y Desarrollo Rural (IRNAD), Sede Andin
Carlos de Bariloche, Río Negro, Argentina. 12Department of Environmental Systems Sc
Entomology, Cornell University, Ithaca, NY 14853, USA. 14Department of Wildlife, Fis
15Global Lands Program, The Nature Conservancy, 117 E. Mountain Avenue, Fort Coll
University, Droevendaalsesteeg 3a, Wageningen 6708 PB, Netherlands. 17IRES and Bi
18Department of Entomology and Great Lakes Bioenergy Research Center, Michiga
partment of Environmental Studies, University of California, Santa Cruz, Santa Cruz
Legnaro, Padova, Italy. 21School of Environment and Rural Science, University of New
Research, Lund University, S-223 62 Lund, Sweden. 23Department of Biology, Lund
University of Göttingen, D-37077 Göttingen, Germany. 25USC1339 INRA-CNRS, CEBC
26INRA, Unité de Recherche Pluridisciplinaire Prairies et Plantes Fourragères (URP3
Sciences, University of Göttingen, Germany. 28Departamento de Zootecnia–CCA, Un
Ecology, Wageningen University and Research, P.O. Box 430, 6700 AK Wageningen, N
Freiburg, Tennenbacher Straße 4, 79106 Freiburg, Germany. 31LTSER Zone Atelier
Beauvoir sur Niort 79360, France. 32Arthropods Department, Natural Sciences Museu
pecuaria, Estación Experimental Concordia, Estacion Yuqueri y vias del Ferrocarril s/n,
trg 16, 11000 Belgrade, Serbia. 35Instituto de Ecología Regional (IER), Universidad
Kellogg Biological Station, Michigan State University, Hickory Corners, MI 49060, USA.
of Barra do Garças/MT, 78600-000, Brazil. 38Center of Sustainable Development, Univ
70910-900, Brazil. 39Wageningen Environmental Research, Wageningen University an
Institute, University of Greenwich, Central Avenue, Chatham Maritime, Kent ME44TB
France. 42Centre for Agri-Environmental Research, School of Agriculture, Policy and
mology, University of Wisconsin, Madison, WI 53705, USA. 44Instituto Nacional de Pe
Environment Systems, Ecology, Evolution, and Behavior, Department of Biological S
Biology, University of Texas at Austin, 205 W 24th Street, 401 Biological Laboratories
Haifa, Oranim, Tivon 36006, Israel. 48Department of Entomology, Kansas State Univers
Environmental Biology, University of Haifa, 3498838 Haifa, Israel. 50Institute for Plant S
International, Bangalore 560 065, India. 52Ashoka Trust for Research in Ecology and th
and Management, University of California, Santa Barbara, Santa Barbara, CA 93106
Avenue, Seattle, WA 9812, USA. 55Department of Botany and Biodiversity Research
University of Vienna, Rennweg 14, 1030 Vienna, Austria. 56Department of Entomol
FL 32601, USA. 57Agrosavia, Centro de Investigación Obonuco, Km 5 vía Obonuco, Pa
Agricultural Research and Higher Education Center (CATIE), Cartago, Turrialba 3050
Resources, Environment and Sustainability, University of British Columbia, Vancouver, B
Universidad de Buenos Aires, CABA C1417DSE, Argentina. 62School of Plant and Environ
UnB—Campus Universitário Darcy Ribeiro, Brasília-DF 70910-900, Brazil. 64Department o
65INRA, UMR 1065 Santé et Agroécologie du Vignoble, ISVV, Université de Bordeaux, B
versidad Nacional del Comahue, CONICET, Quintral 1250, 8400 Bariloche, Rio Neg
Senckenberganlage 25, 60325 Frankfurt am Main, Germany. 68Centre for Biodiversit
of Environmental Science, Policy, and Management, University of California Berkele
Research Centre, Private Bag X7, Claremont 7735, South Africa. 71School of Agricultu
4, Ireland. 72School of Natural Sciences, Trinity College Dublin, Dublin 2, Ireland. 73
Sciences, The University of Tokyo, 188-0002 Tokyo, Japan. 74Forestry and Forest Produ
de Biologia, Universidade Federal da Bahia, 40170-210 Salvador, Brazil. 76Bio-Protectio
National Institute for Environmental Studies, 963-770 Fukushima, Japan. 78Environm
tute, Washington, DC 20005, USA. 79Department of Health and Environmental Scien
*Corresponding author. Email: matteo.dainese@eurac.edu
Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019production (21). The recent stagnation or even decline of crop yields
with ongoing intensification (22) indicates that alternative pathways
are necessary tomaintain future stable and sustainable crop production
(23–25). An improved understanding of global biodiversity-driven eco-
system services in agroecosystems and their cascading effects on crop
production is urgently needed to forecast future supplies of ecosystem
services and to pursue strategies for sustainable management (15).
We compiled an extensive database comprising 89 studies that
measured richness and abundance of pollinators, pest natural enemies,
and associated ecosystem services at 1475 sampling locations around
theworld (Fig. 1A).We focused on the ecosystem services of pollination
andbiological pest control because these services are essential to croppro-
duction and have been the focus ofmuch research in recent decades (26).
We quantified pollinator and pest natural enemy richness as the number
of unique taxa sampled fromeach location (field), abundance as thenum-
ber of observed individuals, and evenness (or the complementary term,o, Italy. 2Department of Animal Ecology and Tropical Biology, Biocenter, University
inización, INIBIOMA, Universidad Nacional del Comahue, CONICET, 8400 Bariloche
e 191, 8046 Zurich, Switzerland. 5Estación Biológica de Doñana (EBD-CSIC), Inte
, Department of Ecology, 750 07 Uppsala, Sweden. 7Departamento de Ecologia
or Ecology, Evolution and Environmental Changes (CE3C), Universidade de Lisboa
SA. 10CSIRO, GPO Box 2583, Brisbane, QLD 4001, Australia. 11Instituto de Investi
a, Universidad Nacional de Río Negro (UNRN) y CONICET, Mitre 630, CP 8400 San
ience, ETH Zurich, Universitätstrasse 16, 8092 Zurich, Switzerland. 13Department o
h and Conservation Biology, University of California Davis, Davis, CA 95616, USA
ins, CO 80524, USA. 16Plant Ecology and Nature Conservation Group, Wageningen
odiversity Research Centre, University of British Columbia, Vancouver, BC, Canada
n State University, 204 CIPS, 578 Wilson Ave, East Lansing, MI 48824, USA. 19De
, CA 95064, USA. 20DAFNAE, University of Padova, viale dell’Università 16, 35020
England, Armidale, NSW 2350, Australia. 22Centre for Environmental and Climate
University, S-223 62 Lund, Sweden. 24Agroecology, Department of Crop Sciences
UMR 7372, CNRS and Université de La Rochelle, Beauvoir sur Niort 79360, France
F), Lusignan 86600, France. 27Functional Agrobiodiversity, Department of Crop
iversidade Federal do Ceará, 60.356-000 Fortaleza, CE, Brazil. 29Farming System
etherlands. 30Chair of Nature Conservation and Landscape Ecology, University o
Plaine and Val de Sevre, CEBC UMR 7372, CNRS and Université de La Rochelle
m of Barcelona, 08003 Barcelona, Spain. 33Instituto Nacional de Tecnología Agro
3200 Entre Rios, Argentina. 34Faculty of Biology, University of Belgrade, Studentsk
Nacional de Tucumán, CONICET, 4107 Yerba Buena, Tucumán, Argentina. 36W.K
37Federal Institute of Education, Science and Technology of Mato Grosso, Campu
ersity of Brasília (UnB)—Campus Universitário Darcy Ribeiro, Asa Norte, Brasília-DF
d Research, P.O. Box 47, 6700 AA Wageningen, Netherlands. 40Natural Resource
, UK. 41INRA, UR 1115, Plantes et Systèmes de culture Horticoles, 84000 Avignon
Development, Reading University, Reading RG6 6AR, UK. 43Department of Ento
squisas da Amazônia (INPA), CEP 69.067-375 Manaus, Amazonas, Brazil. 45Human
ciences, Boise State University, Boise, ID 83725, USA. 46Department of Integrative
, Austin, TX 78712, USA. 47Department of Biology and Environment, University o
ity, 125 Waters Hall, Manhattan, KS 66503, USA. 49Department of Evolutionary and
cience and Microbiology, University of Hamburg, Hamburg, Germany. 51Bioversity
e Environment (ATREE), Bangalore, India. 53Bren School of Environmental Science
-5131, USA. 54Department of Environmental Studies, Seattle University, 901 12th
, Division of Conservation Biology, Vegetation Ecology and Landscape Ecology
ogy and Nematology, University of Florida, 1881 Natural Area Drive, Gainesville
sto, Nariño, Colombia. 58Agriculture, Livestock and Agroforestry Program, Tropica
1, Costa Rica. 59National Audubon Society, Boulder, CO 80305, USA. 60Institute fo
C, Canada. 61Cátedra de Avicultura, Cunicultura y Apicultura, Facultad de Agronomía
mental Sciences, Virginia Tech, Blacksburg, VA 24061, USA. 63Department of Ecology
f Entomology and Nematology, University of California, Davis, Davis, CA 95616, USA
ordeaux Sciences Agro, F-33883 Villenave d’Ornon Cedex, France. 66INIBIOMA, Uni
ro, Argentina. 67Senckenberg Biodiversity and Climate Research Centre (SBiK-F)
y and Biosecurity, University of Auckland, Auckland, New Zealand. 69Departmen
y, Berkeley, CA, USA. 70South African National Biodiversity Institute, Kirstenbosch
re and Food Science and Earth Institute, University College Dublin, Belfield, Dublin
Institute for Sustainable Agro-ecosystem Services, School of Agriculture and Life
cts Research Institute, 1 Matsunosato, Tsukuba, Ibaraki 305-8687, Japan. 75Instituto
n Research Centre, Lincoln University, Lincoln, New Zealand. 77Fukushima Branch
ent and Production Technology Division, International Food Policy Research Insti
ces, Xi’an Jiaotong–Liverpool University, 215123, Suzhou, China.
2 of 13
Downloaded from http://advances.sciencemag.org/ on October 17, 2019,
-
,
,
-
f
.
.
-
,
.
s
f
,
-
i
.
s
s
,
-
-
f
,
,
l
r
,
,
.
-
,
t
,
-
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019dominance) as the Evar index (27) that reflects the relative abundance
of those taxa. We derived a standardized index of pollination services
using measures of pollination success (fruit set or seed set) and plant
reproduction (pollen limitation) and of pest control services using
measures of natural enemy activity (predation or parasitism rates)
and crop damage (pest activity) (see Materials and Methods). We
z-transformed each measure separately to remove the effect of differ-
ences in measurement scale between indices and inverted values for
those measures where low values indicate positive contribution to eco-
system service (i.e., pest activity).We also characterized the 1-km-radius
landscape surrounding each field by measuring the percentage of
cropland from high-resolution land-use maps. This landscape metric
has been used as a relevant proxy for characterizing landscape simpli-
fication (28, 29) and is often correlated with other indicators of land-
scape complexity (30, 31).
Using a Bayesian multilevel modeling approach, we addressed three
fundamental yet unresolved questions in the biodiversity-ecosystem
function framework: (i) Do richness, abundance, and dominance sus-
tain ecosystem services in real-world ecosystems? (ii) Does landscape
simplification indirectly affect ecosystem services mediated by a lossDainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019of local community diversity? (iii) How strong are the cascading effects
of landscape simplification on final crop production?RESULTS AND DISCUSSION
We found clear evidence that richness of service-providing organisms
positively influenced ecosystem service delivery. This was detected for
both pollination (Fig. 1B and table S2) and pest control (Fig. 1C and
table S2) and in almost all studies (figs. S1 and S2). As differentmethods
were used in different studies to quantify richness and ecosystem
services, we tested the sensitivity of our results tomethodological differ-
ences. The bivariate relationships between richness and ecosystem
services were robust to the taxonomic resolution to which each orga-
nism was identified, the sampling methods used to collect pollinators
and natural enemies, and the pollination and pest control service mea-
sures used (see Materials and Methods).
The positive contribution of richness to service supply was also
confirmed in a series of path analyses where we partitioned the relative
importance of richness, total abundance, and evenness in driving bio-
diversity–ecosystem services relationships (Fig. 2). Although richnessFig. 1. Distribution of analyzed studies and effects of richness on ecosystem services provisioning. (A) Map showing the size (number of crop fields sampled) and
location of the 89 studies (further details of studies are given in table S1). (B) Global effect of pollinator richness on pollination (n = 821 fields of 52 studies). (C) Global
effect of natural enemy richness on pest control (n = 654 fields of 37 studies). The thick line in each plot represents the median of the posterior distribution of the
model. Light gray lines represent 1000 random draws from the posterior. The lines are included to depict uncertainty of the modeled relationship.3 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019showed the larger contribution in both models, we also found a direct
effect of total abundance (Fig. 2, A andB) and evenness (Fig. 2, C andD)
on ecosystem service delivery.While both richness and total abundance
showed a positive effect, evenness had a negative effect, suggesting that
dominant species contribute more strongly to service supply. This inte-
grative assessment of different aspects of community structure revealed
a more multilayered relationship between biodiversity and ecosystem
services than has been previously acknowledged. Our results
complement previous findings for pollination (7, 21, 32) and pest con-
trol (33) and indicate that richness and total and relative abundance are
notmutually exclusive but concurrently contribute to support these two
key ecosystem services to agriculture. Hence, we find strong support for
the role of species-rich communities in supporting pollination and pest
control services.Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019Furthermore, we found that landscape simplification indirectly af-
fected ecosystem services by reducing the richness of service-providing
organisms. Roughly a third of the negative effects of landscape simpli-
fication on pollination were due to a loss in pollinator richness (Fig. 3A
and table S5). This effect was even greater for pest control where natural
enemy richnessmediated about 50%of the total effect of landscape sim-
plification (Fig. 3B and table S5). A similar pattern was also found
considering abundance in addition to richness. In this case, landscape
simplification indirectly affected ecosystem services by reducing both
richness and abundance of service-providing organisms (fig. S4, A
andB, and table S6). A consistent richness-mediated effectwas also con-
firmed when we tested the direct and indirect effects of landscape sim-
plification on ecosystem services via changes in both richness and
evenness (fig. S4, C and D, and table S7). However, contrary to our ex-Fig. 2. Direct and indirect effects of richness, total abundance, and evenness on ecosystem services. (A) Path model of pollinator richness as a predictor of pollination,
mediated by pollinator abundance. (B) Pathmodel of natural enemy richness as a predictor of pest control, mediated by natural enemy abundance. (C) Path model of pollinator
richness as a predictor of pollination, mediated by pollinator evenness. (D) Path model of natural enemy richness as a predictor of pest control, mediated by natural enemy
evenness. Pollination model, n = 821 fields of 52 studies; pest control model, n = 654 fields of 37 studies. Path coefficients are effect sizes estimated from the median of the
posterior distribution of the model. Black and red arrows represent positive or negative effects, respectively. Arrow widths are proportional to highest density intervals (HDIs).4 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019pectation, landscape simplification led to higher pollinator evenness. It
is likely that landscape simplification has stronger effects on more
specialized, rare species and shifts assemblages toward more evenly
abundant,mobile, generalist specieswith a higher ability to use scattered
resources in the landscape (34, 35). These findings, combined with the
larger contribution of richness when tested in combination with even-
ness, suggest that not only dominant species but also relatively rare spe-Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019cies contributed to pollination services (32, 36). The effect of landscape
simplification on ecosystem services was minimized when not consid-
ering the mediated effect of richness, especially for pest control. We did
not find a direct landscape simplification effect on pest control [all high-
est density intervals (HDIs) overlapped zero; Fig. 3B and table S5].
Together, these results demonstrate strong negative indirect effects of
landscape simplification on biodiversity-driven ecosystem services in
agroecosystems and the importance of the richness and abundance of
service-providing organisms in mediating these effects.
Last, for a subset of the data that had crop production information
(674 fields of 42 studies), we found that the cascading effects of land-
scape simplification, mediated through richness and associated eco-
system services, led to lower crop production. This was detected for
both pollination (Fig. 4A and table S8) and pest control (Fig. 4B
and table S8). Specifically, landscape simplification reduced both pol-
linator and natural enemy richness, which had indirect consequences
for pollination and pest control and, in turn, decreased crop produc-
tion. For pest control, a positive link with crop production was detected
in fields where the study area was not sprayed with insecticides during
the course of the experiment (Fig. 4B) but not when considering all
sites combined (with and without insecticide use; fig. S4). In sprayed
areas, we did not find a pest control effect (all HDIs overlapped zero),
probably because effects were masked by insecticide use (37, 38), in-
dicating that insecticide use undermines the full potential of natural
pest control. A positive link with crop production was detected, al-
though measures used to estimate pest control (natural enemy and
pest activity) were not direct components of crop production, as
was the case of pollination measures (fruit or seed set). We also found
an indirect influence of dominance (Fig. 4 and table S8) but not of
abundance (fig. S4 and table S9) on crop production. However, the
indirect dominance pathway was weaker than that via richness. An
abundance effect may not have been detectable because of the lower
sample size in these submodels with crop production. Although only
available from a subset of the data, this result supports the hypothesis
that the effects of landscape simplification can cascade up to reducing
the final provisioning service of crop production.
Our findings suggest that some previously inconsistent responses of
natural enemy abundance and activity to surrounding landscape
composition (39) can be reconciled by considering richness in addition
to abundance. Although richness and abundance are often correlated,
their response to environmental variation can differ. This was evident in
the path analysis showing a strong effect of landscape simplification on
richness but only a marginal effect on abundance (fig. S3B). Moreover,
effect sizes for natural enemy abundances in individual studies (fig. S5)
showed similarly inconsistent responses to the previous synthesis (39).
Likewise, results for natural enemy activity showed inconsistent re-
sponses to surrounding landscape composition in both syntheses
[Fig. 3B and (39)].
Using an integrative model to assess key ecological theory, we dem-
onstrate that the negative effects of landscape simplification on service
supply and final crop production are primarily mediated by loss of spe-
cies. We found strong evidence for positive biodiversity–ecosystem ser-
vice relationships, highlighting that managing landscapes to enhance
the richness of service-providing organisms (40, 41) is a promising
pathway toward a more sustainable food production globally. In an
era of rapid environmental changes, preserving biodiversity-driven
services will consistently confer greater resilience to agroecosystems,
such that we could expect improved crop production under a broader
range of potential future conditions.Fig. 3. Direct and indirect effects of landscape simplification on richness of
service-providing organisms and associated ecosystem services. (A) Path
model of landscape simplification as a predictor of pollination, mediated by pollina-
tor richness (n = 821 fields of 52 studies). (B) Path model of landscape simplification
as a predictor of pest control, mediated by natural enemy richness (n = 654 fields of
37 studies). Path coefficients are effect sizes estimated from the median of the pos-
terior distribution of the model. Black and red arrows represent positive and negative
effects, respectively. Arrow widths are proportional to HDIs. Gray arrows represent
nonsignificant effects (HDIs overlapped zero).5 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019MATERIALS AND METHODS
Database compilation
We compiled data from crop studies where measures of richness and
abundance of service-providing organisms (pollinators or natural
enemies) and associated ecosystem services (pollination and biologicalDainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019pest control) were available for the same sites. If available, we also in-
cluded information on yield. Studies were identified by first searching
the reference lists of recent meta-analyses (7, 21, 39, 42, 43) and then
directly contacting researchers. For pest control, data were mostly
provided from a recent pest control database (39). Of 191 researchersFig. 4. Direct and cascading effects of landscape simplification on final crop production via changes in richness, evenness, and ecosystem services. (A) Path model
representing direct and indirect effects of landscape simplification on final crop production through changes in pollinator richness, evenness, and pollination (n = 438 fields of
27 studies). (B) Pathmodel representing direct and indirect effects of landscape simplification on final crop production through changes in natural enemy richness, evenness,
and pest control [only insecticide-free areas were considered in the model (n = 185 fields of 14 studies)]. Path coefficients are effect sizes estimated from the median of the
posterior distribution of the model. Black and red arrows represent positive and negative effects, respectively. Arrow widths are proportional to HDIs. Gray arrows represent
nonsignificant effects (HDIs overlapped zero).6 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019initially contacted, 86 provided data that met our criteria. As similar
studies were frequently performed in the same area, occasionally in
the same year, and studies with multiple years usually used different
sites each year, we did not nest year within study. Likewise, some studies
collected data in different crops. Thus, we considered each year ofmulti-
year studies (that is, 10 studies) and each crop ofmulticrop studies (that
is, 5 studies) to be an independent dataset and used study-crop-year
combinations as the highest hierarchical unit. Overall, we analyzed data
from 89 studies and 1475 fields in 27 countries around the world (table
S1). Twenty-nine crops were considered, including a wide array of an-
nual and perennial fruit, seed, nut, stimulant, pulse, cereal, and oilseed
crops. Studies represented the spectrum of management practices, in-
cluding conventional, low-input conventional, integrated pest manage-
ment, and organic farming. Few studies compared differentmanagement
practices in similar landscapes, crop types, or sampled year. In 76% of
fields, pest control experiments were performed in insecticide-free areas.
In some fields, this information was not available (7%) or insecticides
were applied (17%).
Pollinator and pest natural enemy richness and abundance
Studies used a broad range of methods, which we categorized as active
or passive (44), to sample pollinators or natural enemies. Active
sampling methods included netting pollinators seen on crop flowers,
hand-collecting individuals on plants, observational counting, sweep-
netting, and vacuum sampling. Passive samplingmethods weremalaise
traps, pan traps, pitfall traps, and sticky cards. Active sampling was per-
formed in 85% of pollinator sampling fields and in 50% of natural en-
emy sampling fields.
Pollinators included representatives from the orders Hymenoptera,
Diptera, Lepidoptera, and Coleoptera. Bees (Hymenoptera: Apoidea)
were the most commonly observed pollinators and included Apis bees
(Apidae:Apis mellifera,Apis cerana,Apis dorsata, andApis florea), sting-
less bees (Apidae: Meliponini), bumblebees (Apidae: Bombus spp.),
carpenter bees (Apidae: Xylocopini), small carpenter bees (Apidae:
Ceratinini), sweat bees (Halictidae), long-horned bees (Apidae: Eucerini),
plasterer bees (Colletidae), mining bees (Andrenidae), and mason bees
(Megachilidae). Non-bee taxa included syrphid flies (Diptera: Syrphidae),
other flies (Diptera: Calliphoridae, Tachinidae, and Muscidae), but-
terflies and moths (Lepidoptera), various beetle families (Coleoptera),
and hymenopterans including ants (Formicidae) and the paraphyletic
group of non-bee aculeate wasps. Natural enemies included ground
beetles (Coleoptera), flies (Diptera), spiders (Aranea), hymenopterans
including ants (Formicidae) and wasps, bugs (Hemiptera), thrips
(Thysanoptera), net-winged insects (Neuroptera), bats, and birds.
We calculated pollinator and natural enemy richness as the number
of unique taxa sampled per study, method, and field. A taxon was
defined as a single biological type (that is, species, morphospecies, ge-
nus, and family) determined at the finest taxonomic resolution towhich
each organism was identified. In almost 70% of cases, taxonomic reso-
lution was to species level (averaged proportion among all studies), but
sometimes, it was based onmorphospecies (15%), genus (8%), or family
levels (7%). Taxon richness per field varied between 1 and 49 for polli-
nators and between 1 and 40 for natural enemies. Abundance reflected
the sum of individuals sampled per study, method, and field. Pollina-
tor richness was calculated either including or excluding honey bees
(A. mellifera). A. mellifera was considered as the only species within
the honey bee group for consistency across all datasets (43). Other
Apis bees (that is,A. cerana,A. dorsata, andA. florea) were not pooled
into the honey bee category as the large majority of observed individ-Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019ualswere derived from feral populations. Feral andmanagedhoney bees
were analyzed as a single group because they cannot be distinguished
during field observations. Feral honey bees were uncommon in most
studies except for those in Africa and South America (A. mellifera is
native in Africa, while it was introduced to the Americas). In studies
with subsamples within a field (that is, plots within fields or multiple
sampling rounds within fields), we calculated the total number of indi-
viduals and unique taxa across these subsamples.
We also calculated evenness using the Evar index (27). This index is
based on the variance of log abundances (centered on the mean of log
abundances) and then appropriately scaled to cover 0 to 1 (0,maximally
uneven and 1, perfectly even).
Pollination and pest control services
As different methods were used to quantify pollination or pest control
services across studies, standardization was necessary to put all the in-
dices on equivalent terms. Therefore, we transformed each index y in
each field i in each study jusing z scores.We preferred the use of z scores
over other transformations (for example, division by themaximum) be-
cause z scores do not constrain the variability found in the raw data, as
do other indices that are bounded between 0 and 1. We used the pro-
portion of flowers that set fruit (that is, fruit set), the average number of
seeds per fruits (that is, seed set), or the estimatedmeasures of pollinator
contribution to plant reproduction (that is, differences in fruit weight
between plants with and without insect pollination, hereafter D fruit
weight) as measures of pollination services. We then converted these
measures into the pollination index. The pest control index was
calculated using measures of natural enemy activity or pest activity.
Natural enemy activity was measured by sentinel pest experiments
where pests were placed in crop fields and predation or parasitism rates
were monitored or by field exclosure experiments where cages were
used to exclude natural enemies to quantify differences in pest abun-
dance or crop damage between plants with and without natural
enemies. Pest activity was measured as the fraction or amount of each
crop consumed, infested, or damaged.We inverted standardized values
of pest activity bymultiplying by−1, as low values indicate positive con-
tributions to the ecosystem service.
Crop production
Depending on the crop type, marketable crop yield is valued by farmers
not only in terms of area-based yield but also in terms of fruit or seed
weight [for example, in coffee, sunflower, or strawberry fields; (45, 46)]
or seed production per plant [for example, in seed production fields;
(36)]. Moreover, area-based yield and within-plant yield are often
correlated (35, 36). Thus, we used both area-based yield and within-
plant yield as measures of final crop production. Within-plant yield
was measured by the total number (or mass) of seeds or fruits per plant
or by fruit or seed weight. In addition, in this case, we standardized
variables (z scores) to put all the indices on equivalent terms.
Landscape simplification
Landscapes were characterized by calculating the percentage of
cropland (annual and perennial) within a radius of 1 km around the
center of each crop field. This landscape metric has been used as a re-
levant proxy for characterizing landscape simplification (28, 29) and is
often correlated with other indicators of landscape complexity (30, 31).
Moreover, we used this metric because cropland data are readily acces-
sible from publicly available land cover data and aremore accurate than
other land use types such as forests and grasslands (47), especially when7 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019detailed maps are not available. The 1-km spatial extent was chosen to
reflect the typical flight and foraging distances ofmany insects including
pollinators (48, 49) and natural enemies (50, 51). For studies where this
information was not supplied by the authors, land uses were digitized
using GlobeLand30 (52), a high-resolution map of Earth’s land cover.
The derived land cover maps were verified and, if necessary, corrected
using a visual inspection of satellite images (Google Earth). We then
calculated the percentage of cropland within the radius using Quan-
tum GIS 2.18 (Open Source Geospatial Foundation Project; http://
qgis.osgeo.org). The average percentage of cropland was 67.5% for
pollination studies and 41.5% for natural enemy studies.
Data analysis
Data standardization
Before performing the analyses, we standardized the predictors
(abundance, richness, and landscape simplification) using z scores
within each study. This standardization was necessary to allow
comparisons between studies with differences in methodology
and landscape ranges (53). Moreover, this allows the separation
of within-study effects from between-study effects. This separation
is important because it prevents the risk of misinterpreting the
results based on studies differing in methodology and landscape
gradients, by erroneously extrapolating from between-study effects
to within-study effects or vice versa (54, 55).
Relationship between richness and ecosystem services
The relationship between richness of service-providing organisms and
related ecosystemservices (Fig. 1, B andC)was estimated fromaBayesian
multilevel (partial pooling)model that allowed the intercept and the slope
to vary among studies (also commonly referred to as random intercepts
and slopes), following the equation
ESi Nðaj½i þ bj½iRICi; sjÞ
e
aj Nðma; saÞ
e
bj Nðmb; sbÞ
e
where ESi is the ecosystem service index (pollination or pest control de-
pending on the model), RICi is richness of service-providing organisms
(pollinator or natural enemy richness depending on the model), and j[i]
represents observation i of study j. This partial-poolingmodel estimated
both study-level responses [yielding an estimate for each study (bj)] and
the distribution from which the study-level estimates were drawn,
yielding a higher-level estimate of the overall response across crop
systems (mb). In addition, it accounted for variation in variance and
sample size across observations (for example, studies). The intercepts
aj and slopes bj varied between studies according to a normal
distribution with mean m and SD s. Independent within-study errors
also followed a normal distribution ei ~ N(0,s). We used weakly in-
formative priors: normal (0, 10) for the population-level parameters
(a, b) and half–Student t (3, 0, 5) for the group-level SD and residual SD.
Direct and indirect effects of richness, abundance,
and evenness on service provisioning
As natural communities vary not only in number of species but also in
relative abundance of each species (evenness) and the total number of
individuals (abundance), it is important to incorporate these attributes
when assessing or modeling biodiversity effects (12, 56). In a Bayesian
multivariate response model, a form of path analysis, we partitionedDainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019the relative importance of richness and total and relative abundance in
driving biodiversity-ecosystem relationships. We hypothesized that
richness drives both abundance and evenness according to a revised ver-
sion of the “more individuals hypothesis” (57) that proposes that more
species can exploit more diverse resources and may therefore maintain
more individuals than species-poor communities. Specifically, we tested
(i) whether richness per se directly influences ecosystem services or is in-
stead mediated by abundance (Eq. 1, 3; Fig. 2, A and B) and (ii) whether
richness per se directly influences ecosystem services or is instead
mediated by evenness (Eq. 2, 4; Fig. 2, C andD). Before analysis, we also
checked for data collinearity among abundance, evenness, and richness
by calculating the variance inflation factor (VIF). No signal of collinear-
ity was detected in eithermodel (VIFs were below 1.8). In a preliminary
analysis, we also tested a possible interaction between richness and
evenness. No significant interaction was found in both pollination
and pest control models. To illustrate, we first show the univariate mul-
tilevel (partial pooling) models following these equations
RICi Nðaj½i þ bj½iABUi; sjÞ ð1Þ
e
RICi Nðaj½i þ bj½iEVEi; sjÞ ð2Þ
e
ESi Nðaj½i þ bj½iRICi þ bj½iABUi; sjÞ ð3Þ
e
ESi Nðaj½i þ bj½iRICi þ bj½iEVEi; sjÞ ð4Þ
e
where RICi is richness of service-providing organisms, ABUi is abun-
dance, EVEi is evenness, ESi is the ecosystem service index, and the
index j[i] represents observation i of study j.We then specified bothmul-
tivariate multilevel models in a matrix-vector notion (53), as follows
Yi NðXiBr½i;SjÞ
e
Br NðMB;SBÞ
e
where Yi is the matrix of response variables with observations i as rows
and variables r as columns,Xi is thematrix of all predictors for response
r, Br are the regression parameters (a and b) for response r,MB repre-
sents themeanof the distribution of the regression parameters, and∑B is
the covariance matrix representing the variation of the regression pa-
rameters in the population groups. We used weakly informative priors:
normal (0, 10) for the population-level parameters (a, b) and half–Student
t (3, 0, 5) for the group-level SD and residual SD. In building themodel,
we ensured that no residual correlation betweenESi andRICi, betweenESi
and EVEi, or between ESi and ABUi was estimated [see the “set_rescor”
function in the package brms; (58)].
Direct and indirect effects of landscape simplification on
ecosystem services
To estimate the direct and indirect effects of landscape simplification on
richness and associated ecosystem services, we used two models. In a
Bayesian multivariate response model with causal mediation effects
(hereafter, mediation model), we tested whether landscape simplifica-
tion directly influences ecosystem services or is mediated by richness.
Mediation analysis is a statistical procedure to test whether the effect
of an independent variable X on a dependent variable Y (X → Y)
is at least partly explained via the inclusion of a third hypothetical8 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019variable, themediator variableM (X→M→ Y) (59). The three causal
paths a, b, and c′ correspond toX’s effect onM,M’s effect onY, andX’s
effect on Y accounting forM, respectively. The three causal paths cor-
respond to parameters from two regressionmodels, one in whichM is
the outcome and X is the predictor and one in which Y is the outcome
and X and M are the simultaneous predictors (fig. S6). From these
parameters, we can compute themediation effect (the product ab, also
known as the indirect effect) and the total effect of X on Y
c ¼ c′þ ab
Thus, the total causal effect ofX, which is captured by the parameter
c, can be decomposed precisely into two components, a direct effect c′
and an indirect (mediation) effect ab (the product of paths a and b). The
model included the ecosystem service index as response, landscape sim-
plification as predictor, and richness as mediator (Fig. 3). The separate
regression models that made up the Bayesian multivariate multilevel
model followed these equations
RICi Nðaj½i þ bj½iLANDi; sjÞ
e
ESi Nðaj½i þ bj½iRICi þ bj½iLANDi; sjÞ
e
We then compiled a multilevel path analysis testing the direct and
indirect effects of landscape simplification on ecosystem services via
changes in both richness and abundance (Eq. 5, 6, 8; fig. S4, A and B)
or richness and evenness (Eq. 5, 7, 9; fig. S4, C and D). The separate
regression models that made up the model followed these equations
RICi Nðaj½i þ bj½iLANDi; sjÞ ð5Þ
e
ABUi Nðaj½i þ bj½iLANDi; sjÞ ð6Þ
e
EVEi Nðaj½i þ bj½iLANDi; sjÞ ð7Þ
e
ESi Nðaj½i þ bj½iRICi þ bj½iABUi þ bj½iLANDi; sjÞ ð8Þ
e
ESi Nðaj½i þ bj½iRICi þ bj½iEVEi þ bj½iLANDi; sjÞ ð9Þ
e
where RICi is richness of service-providing organisms, LANDi is land-
scape simplification measured as the percentage of arable land
surrounding each study site, ABUi is abundance, EVEi is evenness,
ESi is the ecosystem service index, and the index j[i] represents observa-
tion i of study j. We then specified multivariate multilevel models in a
matrix-vector notion, as explained above. The mediation analysis was
implemented using the R package sjstats [v. 0.15.0; (60)].
Cascading effects of landscape simplification on final
crop production
For 42 studies and 675 fields (pollination model, n = 438 fields of
27 studies; pest control model, n = 236 fields of 15 studies; table
S1), the data allowed us to use a multilevel path analysis to examine
cascading effects of landscape simplification on final crop produc-Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019tion via changes in richness, abundance, evenness, and ecosystem
services. We expected that (i) landscape simplification would have
a direct effect on richness, abundance, and dominance of service-
providing organisms and (ii) richness and abundance dominance
would relate positively to intermediate services, which, in turn,
(iii) would increase final crop production (Fig. 4 and fig. S4). The in-
direct effects of richness and abundance (Eq. 10, 11, 13, 15; Fig. 4) or
richness and evenness (Eq. 10, 12, 14, 15; fig. S4) were tested separately
RICi Nðaj½i þ bj½iLANDi; sjÞ ð10Þ
e
ABUi Nðaj½i þ bj½iLANDi; sjÞ ð11Þ
e
EVEi Nðaj½i þ bj½iLANDi; sjÞ ð12Þ
e
ESi Nðaj½i þ bj½iRICi þ bj½iABUi; sjÞ ð13Þ
e
ESi Nðaj½i þ bj½iRICi þ bj½iEVEi; sjÞ ð14Þ
e
PRODi Nðaj½i þ bj½iESi; sjÞ ð15Þ
e
where RICi is richness of service-providing organisms, LANDi is land-
scape simplification measured as the percentage of arable land
surrounding each study site, ABUi is abundance, EVEi is evenness,
ESi is the ecosystem service index, PRODi is crop production, and the
index j[i] represents observation i of study j. We specified a multivariate
multilevel model in a matrix-vector notion, as explained above.
Parameter estimation
All analyses were conducted in the programming language Stan through
R (v. 3.4.3) using the package brms [v. 2.2.0; (58)]. Stan implements
Hamiltonian Monte Carlo and its extension, the No-U-Turn Sampler.
These algorithms convergemuchmore quickly than otherMarkov chain
Monte Carlo algorithms especially for high-dimensional models (61).
Eachmodel was runwith four independentMarkov chains of 5000 itera-
tions, discarding the first 2500 iterations per chain as warm-up and re-
sulting in 10,000 posterior samples overall. Convergence of the four
chains and sufficient sampling of posterior distributions were confirmed
by (i) the visual inspection of parameter traces, (ii) ensuring a scale reduc-
tion factorðR~Þ below 1.01, and (iii) an effective size (neff) of at least 10%of
the number of iterations. For each model, posterior samples were sum-
marized on the basis of theBayesian point estimate (median), SE (median
absolute deviation), and posterior uncertainty intervals byHDIs, a type of
credible interval that contains the requiredmass such that all points with-
in the interval have a higher probability density than points outside the
interval (62). The advantage of the Bayesian approach is the possibility to
estimate not only expected values for each parameter but also the uncer-
tainty associatedwith these estimates (63). Thus,we calculated 80, 90, and
95% HDIs for parameter estimates.
Sensitivity analyses
Given that different methods were used in different studies to quantify
richness, ecosystem services, and final crop production, we measured
the sensitivity of our results to methodological differences.
(1)We verified whether treating each annual dataset frommultiyear
studies separately could incorrectly account for the dependence of the9 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019data. We refitted the model testing the relationship between richness
and ecosystem services including year nestedwithin study (that is, study
defined as study-crop combination). Then, we comparedmodels (year-
independent model versus year-nested model) using leave-one-out
(LOO) cross-validation, a fully Bayesian model selection procedure
for estimating pointwise out-of-sample prediction accuracy (64).We
calculated the expected log pointwise predictive density^ðelpdlooÞ,
using the log-likelihood evaluated at the posterior simulations of
the parameter values. Model comparison was implemented using
R package loo [v. 2.0.0; (65)]. We found that the year-nested model
had a lower average predictive accuracy than the year-independent
model for both pollination ðD^elpdloo ¼ 1:79Þ and pest control
ðD^elpdloo ¼ 1:09Þ and therefore retained the year-independent
model in our analysis.
(2) We verified whether taxonomic resolution influenced the inter-
pretation of results. We recalculated richness considering only orga-
nisms classified at the fine taxonomy level (species or morphospecies
levels) and refitted the model testing the effect of richness on ecosystem
services. We found no evidence that taxonomic resolution influenced
our results. With a fine taxonomic resolution, the effects of richness
on ecosystem services (bpollinators = 0.1535, 90% HDIs = 0.0967 to
0.2141; benemies = 0.2264, 90% HDIs = 0.1475 to 0.3065; table S2) were
nearly identical to the estimates presented in the main text (bpollinators =
0.1532, 90% HDIs = 0.0892 to 0.2058; benemies = 0.2093, 90% HDIs =
0.1451 to 0.2779; table S2).
(3) We verified whether the sampling methods used to collect polli-
nators (active versus passive sampling techniques) influenced the rela-
tionship between pollinator richness and pollination using Bayesian
hypothesis testing (58). Passive methods do not directly capture flower
visitors and may introduce some bias (for example, they may
underestimate flower visitors). However, our estimate was not influ-
enced by samplingmethod (the one-sided 90% credibility interval over-
lapped zero; table S11). In accordance with this finding, the evidence
ratio showed that the hypothesis tested (that is, estimates of studies with
active sampling > estimates of studies with passive sampling) was only
0.78 times more likely than the alternative hypothesis.
(4) We verified whether methodological differences in measuring
pollination and pest control services influenced the relationship be-
tween richness and ecosystem services. Using Bayesian hypothesis
testing, we tested whether the estimates differed among methods.
The two-sided 95% credibility interval overlapped zero in all compar-
isons (estimates did not differ significantly; table S12), indicating
that our estimate was not influenced by methodological differences
in measuring ecosystem services. Furthermore, we tested effects in-
cluding only inverted pest activity as a reflection of pest control.
We found positive effects of natural enemy richness on inverted
pest activity (b = 0.1307, 90% HDIs = 0.0102 to 0.2456), indicating
that results were robust to the type of pest control measure considered.
(5) As honey bees are the most important and abundant flower visi-
tors in some locations, we verified the potential influence of honey bees
on our results by refitting the path models testing direct and indirect
effects of richness, abundance, and evenness onpollination serviceswith
honey bees. A positive direct contribution of richness to pollination was
confirmed even after including honey bees (fig. S7). In these models,
both abundance and evenness showed a larger effect compared to
models without honey bees (Fig. 2).
(6) Insecticide application during the course of the experiment could
mask the effect of pest control on crop production (37, 38). We verified
the potential influence of insecticide application on our results by refit-Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019ting the model considering only fields where the study area was not
sprayed with insecticide during the course of the experiment (n = 85
fields of 14 studies).We found a pest control effect that wasmaskedwhen
considering all sites combined (with and without insecticide; fig. S4). We
therefore show the insecticide-free model in the main text (Fig. 4).
(7)We verified the consistency of our results considering only studies
that measured area-based yield (submodel). Only the cascading effects of
landscape simplification on final crop production via changes in richness
were tested in a simplified model. We found no evident differences be-
tween the submodel (fig. S8) and the fullmodel presented in themain text
(Fig. 4).SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/
content/full/5/10/eaax0121/DC1
Supplementary Text
Fig. S1. Forest plot of the effect of pollinator richness on pollination for individual studies.
Fig. S2. Forest plot of the effect of natural enemy richness on pest control for individual studies.
Fig. S3. Direct and indirect landscape simplification effects on ecosystem services via changes
in richness and abundance or richness and evenness.
Fig. S4. Direct and cascading landscape simplification effects on final crop production via
changes in natural enemy richness, abundance evenness, and pest control (all sites together,
with and without insecticide application).
Fig. S5. Forest plot of the effect of landscape simplification on natural enemy abundance for
individual studies.
Fig. S6. Mediation model.
Fig. S7. Direct and indirect effects of pollinator richness, abundance, and evenness (with honey
bees) on pollination.
Fig. S8. Direct and cascading landscape simplification effects on area-based yield via changes
in richness, and ecosystem services.
Table S1. List of 89 studies considered in our analyses.
Table S2. Model output for richness–ecosystem service relationships.
Table S3. Model output for path models testing direct and indirect effects (mediated by
changes in abundance) of richness on ecosystem services.
Table S4. Model output for path models testing direct and indirect effects (mediated by
changes in evenness) of richness on ecosystem services.
Table S5. Model output for path models testing direct and indirect effects (mediated by
changes in richness) of landscape simplification on ecosystem services.
Table S6. Model output for path models testing the direct and cascading landscape
simplification effects on ecosystem services via changes in richness and abundance.
Table S7. Model output for path models testing the direct and cascading landscape
simplification effects on ecosystem services via changes in richness and evenness.
Table S8. Model output for path models testing the direct and cascading landscape
simplification effects on final crop production via changes in richness, evenness, and
ecosystem services.
Table S9. Model output for path models testing the direct and cascading landscape
simplification effects on final crop production via changes in richness, abundance, and
ecosystem services.
Table S10. Model output for path models testing direct and indirect effects of pollinator
richness, abundance, and evenness (with honey bees) on pollination.
Table S11. Results of pairwise comparison of richness–ecosystem service relationships
according to the methods used to sample pollinators and natural enemies.
Table S12. Results of pairwise comparison of richness–ecosystem service relationships
according to the methods used to quantify pollination and pest control services.
Database S1. Data on pollinator and natural enemy diversity and associated ecosystem
services compiled from 89 studies and 1475 locations around the world.
References (66–111)
View/request a protocol for this paper from Bio-protocol.REFERENCES AND NOTES
1. S. Díaz, U. Pascual, M. Stenseke, B. Martín-López, R. T. Watson, Z. Molnár, R. Hill,
K. M. A. Chan, I. A. Baste, K. A. Brauman, S. Polasky, A. Church, M. Lonsdale,
A. Larigauderie, P. W. Leadley, A. P. E. van Oudenhoven, F. van der Plaat, M. Schröter,
S. Lavorel, Y. Aumeeruddy-Thomas, E. Bukvareva, K. Davies, S. Demissew, G. Erpul,
P. Failler, C. A. Guerra, C. L. Hewitt, H. Keune, S. Lindley, Y. Shirayama, Assessing nature’s
contributions to people. Science 359, 270–272 (2018).10 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 20192. S. Díaz, J. Settele, E. Brondízio, Summary for policymakers of the global
assessment report on biodiversity and ecosystem services of the Intergovernmental
Science-Policy Platform on Biodiversity and Ecosystem Services (United Nations
Paris, Fr., 2019), 1–39.
3. S. Naeem, J. E. Duffy, E. Zavaleta, The functions of biological diversity in an age of
extinction. Science 336, 1401–1406 (2012).
4. D. Tilman, F. Isbell, J. M. Cowles, Biodiversity and ecosystem functioning. Annu. Rev. Ecol.
Evol. Syst. 45, 471–493 (2014).
5. B. J. Cardinale, J. E. Duffy, A. Gonzalez, D. U. Hooper, C. Perrings, P. Venail, A. Narwani,
G. M. Mace, D. Tilman, D. A. Wardle, A. P. Kinzig, G. C. Daily, M. Loreau, J. B. Grace,
A. Larigauderie, D. S. Srivastava, S. Naeem, Biodiversity loss and its impact on humanity.
Nature 486, 59–67 (2012).
6. R. Winfree, J. W. Fox, N. M. Williams, J. R. Reilly, D. P. Cariveau, Abundance of common
species, not species richness, drives delivery of a real-world ecosystem service. Ecol. Lett.
18, 626–635 (2015).
7. D. Kleijn, R. Winfree, I. Bartomeus, L. G. Carvalheiro, M. Henry, R. Isaacs, A.-M. Klein,
C. Kremen, L. K. M'Gonigle, R. Rader, T. H. Ricketts, N. M. Williams, N. Lee Adamson,
J. S. Ascher, A. Báldi, P. Batáry, F. Benjamin, J. C. Biesmeijer, E. J. Blitzer, R. Bommarco,
M. R. Brand, V. Bretagnolle, L. Button, D. P. Cariveau, R. Chifflet, J. F. Colville,
B. N. Danforth, E. Elle, M. P. D. Garratt, F. Herzog, A. Holzschuh, B. G. Howlett, F. Jauker,
S. Jha, E. Knop, K. M. Krewenka, V. le Féon, Y. Mandelik, E. A. May, M. G. Park, G. Pisanty,
M. Reemer, V. Riedinger, O. Rollin, M. Rundlöf, H. S. Sardiñas, J. Scheper, A. R. Sciligo,
H. G. Smith, I. Steffan-Dewenter, R. Thorp, T. Tscharntke, J. Verhulst, B. F. Viana,
B. E. Vaissière, R. Veldtman, K. L. Ward, C. Westphal, S. G. Potts, Delivery of crop
pollination services is an insufficient argument for wild pollinator conservation.
Nat. Commun. 6, 7414 (2015).
8. M. Loreau, A. Hector, Partitioning selection and complementarity in biodiversity
experiments. Nature 412, 72–76 (2001).
9. D. U. Hooper, F. S. Chapin III, J. J. Ewel, A. Hector, P. Inchausti, S. Lavorel, J. H. Lawton,
D. M. Lodge, M. Loreau, S. Naeem, B. Schmid, H. Setälä, A. J. Symstad, J. Vandermeer,
D. A. Wardle, Effects of biodiversity on ecosystem functioning: A consensus of current
knowledge. Ecol. Monogr. 75, 3–35 (2005).
10. M. Schleuning, J. Fründ, D. García, Predicting ecosystem functions from biodiversity and
mutualistic networks: An extension of trait-based concepts to plant-animal interactions.
Ecography 38, 380–392 (2015).
11. T. H. Larsen, N. M. Williams, C. Kremen, Extinction order and altered community
structure rapidly disrupt ecosystem functioning. Ecol. Lett. 8, 538–547 (2005).
12. J. Reiss, J. R. Bridle, J. M. Montoya, G. Woodward, Emerging horizons in biodiversity and
ecosystem functioning research. Trends Ecol. Evol. 24, 505–514 (2009).
13. K. H. Bannar-Martin, C. T. Kremer, S. K. M. Ernest, M. A. Leibold, H. Auge, J. Chase,
S. A. J. Declerck, N. Eisenhauer, S. Harpole, H. Hillebrand, F. Isbell, T. Koffel, S. Larsen,
A. Narwani, J. S. Petermann, C. Roscher, J. S. Cabral, S. R. Supp, Integrating community
assembly and biodiversity to better understand ecosystem function: The community
assembly and the functioning of ecosystems (CAFE) approach. Ecol. Lett. 21, 167–180
(2018).
14. P. Balvanera, I. Siddique, L. Dee, A. Paquette, F. Isbell, A. Gonzalez, J. Byrnes,
M. I. O’Connor, B. A. Hungate, J. N. Griffin, Linking biodiversity and ecosystem services:
Current uncertainties and the necessary next steps. Bioscience 64, 49–57 (2014).
15. F. Isbell, A. Gonzalez, M. Loreau, J. Cowles, S. Díaz, A. Hector, G. M. Mace, D. A. Wardle,
M. I. O'Connor, J. E. Duffy, L. A. Turnbull, P. L. Thompson, A. Larigauderie, Linking the
influence and dependence of people on biodiversity across scales. Nature 546, 65–72
(2017).
16. R. Chaplin-Kramer, M. E. O’Rourke, E. J. Blitzer, C. Kremen, A meta-analysis of crop
pest and natural enemy response to landscape complexity. Ecol. Lett. 14, 922–932
(2011).
17. C. M. Kennedy, E. Lonsdorf, M. C. Neel, N. M. Williams, T. H. Ricketts, R. Winfree,
R. Bommarco, C. Brittain, A. L. Burley, D. Cariveau, L. G. Carvalheiro, N. P. Chacoff,
S. A. Cunningham, B. N. Danforth, J.-H. Dudenhöffer, E. Elle, H. R. Gaines,
L. A. Garibaldi, C. Gratton, A. Holzschuh, R. Isaacs, S. K. Javorek, S. Jha, A. M. Klein,
K. Krewenka, Y. Mandelik, M. M. Mayfield, L. Morandin, L. A. Neame, M. Otieno,
M. Park, S. G. Potts, M. Rundlöf, A. Saez, I. Steffan-Dewenter, H. Taki, B. F. Viana,
C. Westphal, J. K. Wilson, S. S. Greenleaf, C. Kremen, A global quantitative synthesis
of local and landscape effects on wild bee pollinators in agroecosystems. Ecol. Lett.
16, 584–599 (2013).
18. J. A. Foley, R. DeFries, G. P. Asner, C. Barford, G. Bonan, S. R. Carpenter, F. S. Chapin,
M. T. Coe, G. C. Daily, H. K. Gibbs, J. H. Helkowski, T. Holloway, E. A. Howard, C. J. Kucharik,
C. Monfreda, J. A. Patz, I. C. Prentice, N. Ramankutty, P. K. Snyder, Global consequences of
land use. Science 309, 570–574 (2005).
19. T. H. Ricketts, J. Regetz, I. Steffan-Dewenter, S. A. Cunningham, C. Kremen, A. Bogdanski,
B. Gemmill-Herren, S. S. Greenleaf, A. M. Klein, M. M. Mayfield, L. A. Morandin,
A. Ochieng', S. G. Potts, B. F. Viana, Landscape effects on crop pollination services: Are
there general patterns? Ecol. Lett. 11, 499–515 (2008).Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 201920. F. J. J. A. Bianchi, C. J. H. Booij, T. Tscharntke, Sustainable pest regulation in agricultural
landscapes: A review on landscape composition, biodiversity and natural pest control.
Proc. Biol. Sci. 273, 1715–1727 (2006).
21. L. A. Garibaldi, I. Steffan-Dewenter, R. Winfree, M. A. Aizen, R. Bommarco,
S. A. Cunningham, C. Kremen, L. G. Carvalheiro, L. D. Harder, O. Afik, I. Bartomeus,
F. Benjamin, V. Boreux, D. Cariveau, N. P. Chacoff, J. H. Dudenhoffer, B. M. Freitas,
J. Ghazoul, S. Greenleaf, J. Hipolito, A. Holzschuh, B. Howlett, R. Isaacs, S. K. Javorek,
C. M. Kennedy, K. M. Krewenka, S. Krishnan, Y. Mandelik, M. M. Mayfield, I. Motzke,
T. Munyuli, B. A. Nault, M. Otieno, J. Petersen, G. Pisanty, S. G. Potts, R. Rader,
T. H. Ricketts, M. Rundlof, C. L. Seymour, C. Schuepp, H. Szentgyorgyi, H. Taki,
T. Tscharntke, C. H. Vergara, B. F. Viana, T. C. Wanger, C. Westphal, N. Williams, A. M. Klein,
Wild pollinators enhance fruit set of crops regardless of honey bee abundance.
Science 339, 1608–1611 (2013).
22. D. K. Ray, N. Ramankutty, N. D. Mueller, P. C. West, J. A. Foley, Recent patterns of crop
yield growth and stagnation. Nat. Commun. 3, 1293 (2012).
23. J. A. Foley, N. Ramankutty, K. A. Brauman, E. S. Cassidy, J. S. Gerber, M. Johnston,
N. D. Mueller, C. O’Connell, D. K. Ray, P. C. West, C. Balzer, E. M. Bennett, S. R. Carpenter,
J. Hill, C. Monfreda, S. Polasky, J. Rockström, J. Sheehan, S. Siebert, D. Tilman,
D. P. M. Zaks, Solutions for a cultivated planet. Nature 478, 337–342 (2011).
24. R. Bommarco, D. Kleijn, S. G. Potts, Ecological intensification: Harnessing ecosystem
services for food security. Trends Ecol. Evol. 28, 230–238 (2013).
25. J. Pretty, Intensification for redesigned and sustainable agricultural systems.
Science 362, eaav0294 (2018).
26. Millennium Ecosystem Assessment, Ecosystems and Human Well-being: Biodiversity
Synthesis (World Resources Institute, 2005).
27. B. Smith, J. B. Wilson, A consumer’s guide to evenness indices. Oikos 76, 70–82 (1996).
28. T. D. Meehan, B. P. Werling, D. A. Landis, C. Gratton, Agricultural landscape simplification
and insecticide use in the Midwestern United States. Proc. Natl. Acad. Sci. U. S. A. 108,
11500–11505 (2011).
29. M. Dainese, N. J. B. Isaac, G. D. Powney, R. Bommarco, E. Öckinger, M. Kuussaari, J. Pöyry,
T. G. Benton, D. Gabriel, J. A. Hodgson, W. E. Kunin, R. Lindborg, S. M. Sait, L. Marini,
Landscape simplification weakens the association between terrestrial producer and
consumer diversity in Europe. Glob. Chang. Biol. 23, 3040–3051 (2017).
30. T. Tscharntke, A. M. Klein, A. Kruess, I. Steffan-Dewenter, C. Thies, Landscape
perspectives on agricultural intensification and biodiversity—Ecosystem service
management. Ecol. Lett. 8, 857–874 (2005).
31. D. A. Landis, Designing agricultural landscapes for biodiversity-based ecosystem
services. Basic Appl. Ecol. 18, 1–12 (2017).
32. R. Winfree, J. R. Reilly, I. Bartomeus, D. P. Cariveau, N. M. Williams, J. Gibbs, Species
turnover promotes the importance of bee diversity for crop pollination at regional
scales. Science 359, 791–793 (2018).
33. D. K. Letourneau, J. A. Jedlicka, S. G. Bothwell, C. R. Moreno, Effects of natural enemy
biodiversity on the suppression of arthropod herbivores in terrestrial ecosystems.
Annu. Rev. Ecol. Evol. Syst. 40, 573–592 (2009).
34. L. Marini, E. Öckinger, K.-O. Bergman, B. Jauker, J. Krauss, M. Kuussaari, J. Pöyry,
H. G. Smith, I. Steffan-Dewenter, R. Bommarco, Contrasting effects of habitat area and
connectivity on evenness of pollinator communities. Ecography 37, 544–551 (2014).
35. S. Gámez-Virués, D. J. Perović, M. M. Gossner, C. Börschig, N. Blüthgen, H. de Jong,
N. K. Simons, A.-M. Klein, J. Krauss, G. Maier, C. Scherber, J. Steckel, C. Rothenwöhrer,
I. Steffan-Dewenter, C. N. Weiner, W. Weisser, M. Werner, T. Tscharntke, C. Westphal,
Landscape simplification filters species traits and drives biotic homogenization.
Nat. Commun. 6, 8568 (2015).
36. T. P. M. Fijen, J. A. Scheper, T. M. Boom, N. Janssen, I. Raemakers, D. Kleijn, Insect
pollination is at least as important for marketable crop yield as plant quality in a seed
crop. Ecol. Lett. 21, 1704–1713 (2018).
37. H. Liere, T. N. Kim, B. P. Werling, T. D. Meehan, D. A. Landis, C. Gratton, Trophic cascades
in agricultural landscapes: Indirect effects of landscape composition on crop yield.
Ecol. Appl. 25, 652–661 (2015).
38. V. Gagic, D. Kleijn, A. Báldi, G. Boros, H. B. Jørgensen, Z. Elek, M. P. D. Garratt,
G. A. de Groot, K. Hedlund, A. Kovács-Hostyánszki, L. Marini, E. Martin, I. Pevere,
S. G. Potts, S. Redlich, D. Senapathi, I. Steffan-Dewenter, S. Świtek, H. G. Smith, V. Takács,
P. Tryjanowski, W. H. van der Putten, S. van Gils, R. Bommarco, Combined effects of
agrochemicals and ecosystem services on crop yield across Europe. Ecol. Lett. 20,
1427–1436 (2017).
39. D. S. Karp, R. Chaplin-Kramer, T. D. Meehan, E. A. Martin, F. DeClerck, H. Grab, C. Gratton,
L. Hunt, A. E. Larsen, A. Martínez-Salinas, M. E. O'Rourke, A. Rusch, K. Poveda, M. Jonsson,
J. A. Rosenheim, N. A. Schellhorn, T. Tscharntke, S. D. Wratten, W. Zhang, A. L. Iverson,
L. S. Adler, M. Albrecht, A. Alignier, G. M. Angelella, M. Z. Anjum, J. Avelino, P. Batáry,
J. M. Baveco, F. J. J. A. Bianchi, K. Birkhofer, E. W. Bohnenblust, R. Bommarco, M. J. Brewer,
B. Caballero-López, Y. Carrière, L. G. Carvalheiro, L. Cayuela, M. Centrella, A. Ćetković,
D. C. Henri, A. Chabert, A. C. Costamagna, A. De la Mora, J. de Kraker, N. Desneux, E. Diehl,
T. Diekötter, C. F. Dormann, J. O. Eckberg, M. H. Entling, D. Fiedler, P. Franck, F. J. F. van Veen,11 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 2019T. Frank, V. Gagic, M. P. D. Garratt, A. Getachew, D. J. Gonthier, P. B. Goodell, I. Graziosi,
R. L. Groves, G. M. Gurr, Z. Hajian-Forooshani, G. E. Heimpel, J. D. Herrmann, A. S. Huseth,
D. J. Inclán, A. J. Ingrao, P. Iv, K. Jacot, G. A. Johnson, L. Jones, M. Kaiser, J. M. Kaser, T. Keasar,
T. N. Kim, M. Kishinevsky, D. A. Landis, B. Lavandero, C. Lavigne, A. Le Ralec, D. Lemessa,
D. K. Letourneau, H. Liere, Y. Lu, Y. Lubin, T. Luttermoser, B.Maas, K.Mace, F.Madeira, V.Mader,
A. M. Cortesero, L. Marini, E. Martinez, H. M. Martinson, P. Menozzi, M. G. E. Mitchell,
T. Miyashita, G. A. R. Molina, M. A. Molina-Montenegro, M. E. O'Neal, I. Opatovsky,
S. Ortiz-Martinez, M. Nash, Ö. Östman, A. Ouin, D. Pak, D. Paredes, S. Parsa, H. Parry,
R. Perez-Alvarez, D. J. Perović, J. A. Peterson, S. Petit, S. M. Philpott, M. Plantegenest, M. Plećaš,
T. Pluess, X. Pons, S. G. Potts, R. F. Pywell, D. W. Ragsdale, T. A. Rand, L. Raymond, B. Ricci,
C. Sargent, J. P. Sarthou, J. Saulais, J. Schäckermann, N. P. Schmidt, G. Schneider, C. Schüepp,
F. S. Sivakoff, H. G. Smith, K. S.Whitney, S. Stutz, Z. Szendrei, M. B. Takada,H. Taki, G. Tamburini,
L. J. Thomson, Y. Tricault, N. Tsafack, M. Tschumi, M. Valantin-Morison, M. Van Trinh,
W. van der Werf, K. T. Vierling, B. P. Werling, J. B. Wickens, V. J. Wickens, B. A. Woodcock,
K. Wyckhuys, H. Xiao, M. Yasuda, A. Yoshioka, Y. Zou, Crop pests and predators exhibit
inconsistent responses to surrounding landscape composition. Proc. Natl. Acad. Sci. U. S. A.15,
E7863–E7870 (2018).
40. C. Kremen, A. M. Merenlender, Landscapes that work for biodiversity and people.
Science 362, eaau6020 (2018).
41. E. A. Martin, M. Dainese, Y. Clough, A. Báldi, R. Bommarco, V. Gagic, M. P. D. Garratt,
A. Holzschuh, D. Kleijn, A. Kovács-Hostyánszki, L. Marini, S. G. Potts, H. G. Smith,
D. Al Hassan, M. Albrecht, G. K. S. Andersson, J. D. Asís, S. Aviron, M. V. Balzan,
L. Baños-Picón, I. Bartomeus, P. Batáry, F. Burel, B. Caballero-López, E. D. Concepción,
V. Coudrain, J. Dänhardt, M. Diaz, T. Diekötter, C. F. Dormann, R. Duflot, M. H. Entling,
N. Farwig, C. Fischer, T. Frank, L. A. Garibaldi, J. Hermann, F. Herzog, D. Inclán, K. Jacot,
F. Jauker, P. Jeanneret, M. Kaiser, J. Krauss, V. Le Féon, J. Marshall, A.-C. Moonen,
G. Moreno, V. Riedinger, M. Rundlöf, A. Rusch, J. Scheper, G. Schneider, C. Schüepp,
S. Stutz, L. Sutter, G. Tamburini, C. Thies, J. Tormos, T. Tscharntke, M. Tschumi, D. Uzman,
C. Wagner, M. Zubair-Anjum, I. Steffan-Dewenter, The interplay of landscape
composition and configuration: New pathways to manage functional biodiversity and
agroecosystem services across Europe. Ecol. Lett. 22, 1083–1094 (2019).
42. L. A. Garibaldi, L. G. Carvalheiro, B. E. Vaissiere, B. Gemmill-Herren, J. Hipolito,
B. M. Freitas, H. T. Ngo, N. Azzu, A. Saez, J. Astrom, J. An, B. Blochtein, D. Buchori,
F. J. C. Garcia, F. Oliveira da Silva, K. Devkota, M. de Fátima Ribeiro, L. Freitas,
M. C. Gaglianone, M. Goss, M. Irshad, M. Kasina, A. J. S. P. Filho, L. H. P. Kiill, P. Kwapong,
G. N. Parra, C. Pires, V. Pires, R. S. Rawal, A. Rizali, A. M. Saraiva, R. Veldtman, B. F. Viana,
S. Witter, H. Zhang, Mutually beneficial pollinator diversity and crop yield outcomes
in small and large farms. Science 351, 388–391 (2016).
43. R. Rader, I. Bartomeus, L. A. Garibaldi, M. P. D. Garratt, B. G. Howlett, R. Winfree,
S. A. Cunningham, M. M. Mayfield, A. D. Arthur, G. K. S. Andersson, R. Bommarco,
C. Brittain, L. G. Carvalheiro, N. P. Chacoff, M. H. Entling, B. Foully, B. M. Freitas,
B. Gemmill-Herren, J. Ghazoul, S. R. Griffin, C. L. Gross, L. Herbertsson, F. Herzog,
J. Hipólito, S. Jaggar, F. Jauker, A. M. Klein, D. Kleijn, S. Krishnan, C. Q. Lemos,
S. A. M. Lindström, Y. Mandelik, V. M. Monteiro, W. Nelson, L. Nilsson, D. E. Pattemore,
N. de O. Pereira, G. Pisanty, S. G. Potts, M. Reemer, M. Rundlöf, C. S. Sheffield, J. Scheper,
C. Schüepp, H. G. Smith, D. A. Stanley, J. C. Stout, H. Szentgyörgyi, H. Taki, C. H. Vergara,
B. F. Viana, M. Woyciechowski, Non-bee insects are important contributors to global
crop pollination. Proc. Natl. Acad. Sci. U. S. A. 113, 146–151 (2016).
44. E. M. Lichtenberg, C. M. Kennedy, C. Kremen, P. Batáry, F. Berendse, R. Bommarco,
N. A. Bosque-Pérez, L. G. Carvalheiro, W. E. Snyder, N. M. Williams, R. Winfree, B. K. Klatt,
S. Åström, F. Benjamin, C. Brittain, R. Chaplin-Kramer, Y. Clough, B. Danforth, T. Diekötter,
S. D. Eigenbrode, J. Ekroos, E. Elle, B. M. Freitas, Y. Fukuda, H. R. Gaines-Day, H. Grab,
C. Gratton, A. Holzschuh, R. Isaacs, M. Isaia, S. Jha, D. Jonason, V. P. Jones, A. M. Klein,
J. Krauss, D. K. Letourneau, S. Macfadyen, R. E. Mallinger, E. A. Martin, E. Martinez,
J. Memmott, L. Morandin, L. Neame, M. Otieno, M. G. Park, L. Pfiffner, M. J. O. Pocock,
C. Ponce, S. G. Potts, K. Poveda, M. Ramos, J. A. Rosenheim, M. Rundlöf, H. Sardiñas,
M. E. Saunders, N. L. Schon, A. R. Sciligo, C. S. Sidhu, I. Steffan-Dewenter, T. Tscharntke,
M. Veselý, W. W. Weisser, J. K. Wilson, D. W. Crowder, A global synthesis of the effects
of diversified farming systems on arthropod diversity within fields and across
agricultural landscapes. Glob. Chang. Biol. 23, 4946–4957 (2017).
45. L. G. Carvalheiro, R. Veldtman, A. G. Shenkute, G. B. Tesfay, C. W. W. Pirk, J. S. Donaldson,
S. W. Nicolson, Natural and within-farmland biodiversity enhances crop productivity.
Ecol. Lett. 14, 251–259 (2011).
46. V. Boreux, C. G. Kushalappa, P. Vaast, J. Ghazoul, Interactive effects among ecosystem
services and management practices on crop production: Pollination in coffee
agroforestry systems. Proc. Natl. Acad. Sci. U.S.A 110, 8387–8392 (2013).
47. R. Tropek, O. Sedláček, J. Beck, P. Keil, Z. Musilova, I. Šímova, D. Storch, Comment on “High-
resolution global maps of 21st-century forest cover change”. Science 344, 981 (2014).
48. M. Rundlöf, J. Bengtsson, H. G. Smith, Local and landscape effects of organic farming on
butterfly species richness and abundance. J. Appl. Ecol. 45, 813–820 (2008).
49. A. Holzschuh, M. Dainese, J. P. González-Varo, S. Mudri-Stojnić, V. Riedinger, M. Rundlöf,
J. Scheper, J. B. Wickens, V. J. Wickens, R. Bommarco, D. Kleijn, S. G. Potts, S. P. M. Roberts,Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019H. G. Smith, M. Vilà, A. Vujić, I. Steffan-Dewenter, Mass-flowering crops dilute pollinator
abundance in agricultural landscapes across Europe. Ecol. Lett. 19, 1228–1236 (2016).
50. C. Thies, T. Tscharntke, Landscape structure and biological control in agroecosystems.
Science 285, 893–895 (1999).
51. A. Rusch, R. Chaplin-Kramer, M. M. Gardiner, V. Hawro, J. Holland, D. Landis, C. Thies,
T. Tscharntke, W. W. Weisser, C. Winqvist, M. Woltz, R. Bommarco, Agricultural landscape
simplification reduces natural pest control: A quantitative synthesis. Agric. Ecosyst. Environ.
221, 198–204 (2016).
52. J. Chen, Y. Ban, S. Li, China: Open access to Earth land-cover map. Nature 514, 434
(2014).
53. A. Gelman, J. Hill, Data Analysis Using Regression and Multilevel Hierarchical Models
(Cambridge Univ. Press, 2007).
54. M. van de Pol, J. Wright, A simple method for distinguishing within- versus between-subject
effects using mixed models. Anim. Behav. 77, 753–758 (2009).
55. H. Schielzeth, Simple means to improve the interpretability of regression coefficients.
Methods Ecol. Evol. 1, 103–113 (2010).
56. B. J. Cardinale, D. S. Srivastava, J. Emmett Duffy, J. P. Wright, A. L. Downing, M. Sankaran,
C. Jouseau, Effects of biodiversity on the functioning of trophic groups and ecosystems.
Nature. 443, 989–992 (2006).
57. P.-C. Bürkner, brms: An R package for Bayesian multilevel models using Stan. J. Stat.
Softw. 80, 1–27 (2017).
58. D. P. MacKinnon, A. J. Fairchild, M. S. Fritz, Mediation analysis. Annu. Rev. Psychol. 58,
593–614 (2007).
59. D. Lüdecke, sjstats: Statistical functions for regression models. 10.5281/
ZENODO.1308979 (2018).
60. B. Carpenter, A. Gelman, M. D. Hoffman, D. Lee, B. Goodrich, M. Betancourt, M. Brubaker,
J. Guo, P. Li, A. Riddell, Stan: A probabilistic programming language. J. Stat. Softw. 76,
1–37 (2017).
61. J. Kruschke, Doing Bayesian Data Analysis: A Tutorial with R, JAGS, and Stan (Academic
Press, ed. 2, 2015).
62. R. McElreath, Statistical Rethinking: A Bayesian Course with Examples in R and Stan (CRC
Press, 2016).
63. A. Vehtari, A. Gelman, J. Gabry, Practical Bayesian model evaluation using leave-one-out
cross-validation and WAIC. Stat. Comput. 27, 1413–1432 (2017).
64. A. Vehtari, A. Gelman, J. Gabry, Y. Yao, A. Gelman, loo: Efficient leave-one-out cross-
validation and WAIC for Bayesian models. R package version, 2.0.0 2018;
https://cran.r-project.org/package=loo.
65. U. Samnegård, G. Alins, V. Boreux, J. Bosch, D. García, A. K. Happe, A. M. Klein, M. Miñarro,
K. Mody, M. Porcel, A. Rodrigo, L. Roquer-Beni, M. Tasin, P. A. Hambäck, Management trade-
offs on ecosystem services in apple orchards across Europe: Direct and indirect effects of
organic production. J. Appl. Ecol. 56, 802–811 (2019).
66. M. P. D. Garratt, C. L. Truslove, D. J. Coston, R. L. Evans, E. D. Moss, C. Dodson, N. Jenner,
J. C. Biesmeijer, S. G. Potts, Pollination deficits in UK apple orchards. J. Pollinat. Ecol. 12,
9–14 (2014).
67. G. A. de Groot, R. van Kats, M. Reemer, D. van der Sterren, J.C. Biesmeijer, D. Kleijn,
“De bijdrage van (wilde) bestuivers aan de opbrengst van appels en blauwe bessen:
Kwantificering van ecosysteemdiensten in Nederland,” Alterra, Wageningen-UR, 2015;
http://edepot.wur.nl/353774.
68. R. E. Mallinger, C. Gratton, Species richness of wild bees, but not the use of managed
honeybees, increases fruit set of a pollinator-dependent crop. J. Appl. Ecol. 52, 323–330
(2015).
69. M. P. D. Garratt, D. J. Coston, C. L. Truslove, M. G. Lappage, C. Polce, R. Dean,
J. C. Biesmeijer, S. G. Potts, The identity of crop pollinators helps target conservation for
improved ecosystem services. Biol. Conserv. 169, 128–135 (2014).
70. D. d. L. Ramos, M. M. C. Bustamante, F. D. D. S. E. Silva, L. G. Carvalheiro, Crop fertilization
affects pollination service provision—Common bean as a case study. PLOS ONE. 13,
e0204460 (2018).
71. H. Taki, K. Okabe, S. Makino, Y. Yamaura, M. Sueyoshi, Contribution of small insects to
pollination of common buckwheat, a distylous crop. Ann. Appl. Biol. 155, 121–129
(2009).
72. H. Taki, K. Okabe, Y. Yamaura, T. Matsuura, M. Sueyoshi, S.’. Makino, K. Maeto, Effects of
landscape metrics on Apis and non-Apis pollinators and seed set in common
buckwheat. Basic Appl. Ecol. 11, 594–602 (2010).
73. A. Holzschuh, J.-H. Dudenhöffer, T. Tscharntke, Landscapes with wild bee habitats
enhance pollination, fruit set and yield of sweet cherry. Biol. Conserv. 153, 101–107
(2012).
74. V. Boreux, S. Krishnan, K. G. Cheppudira, J. Ghazoul, Impact of forest fragments on bee
visits and fruit set in rain-fed and irrigated coffee agro-forests. Agric. Ecosyst. Environ.
172, 42–48 (2013).
75. V. Boreux, P. Vaast, L. P. Madappa, K. G. Cheppudira, C. Garcia, J. Ghazoul, Agroforestry
coffee production increased by native shade trees, irrigation, and liming. Agron. Sustain. Dev.
36, 10.1007/s13593-016-0377-7, (2016).12 of 13
SC I ENCE ADVANCES | R E S EARCH ART I C L E
Downloaded from http://advances.sciencemag.org/ on October 17, 201976. A. Classen, M. K. Peters, S. W. Ferger, M. Helbig-Bonitz, J. M. Schmack, G. Maassen,
M. Schleuning, E. K. V. Kalko, K. Böhning-Gaese, I. Steffan-Dewenter, Complementary
ecosystem services provided by pest predators and pollinators increase quantity and
quality of coffee yields. Proc. Biol. Sci. 281, 20133148 (2014).
77. J. Hipólito, D. Boscolo, B. F. Viana, Landscape and crop management strategies to
conserve pollination services and increase yields in tropical coffee farms. Agric. Ecosyst.
Environ. 256, 218–225 (2018).
78. S. Krishnan, C. G. Kushalappa, R. U. Shaanker, J. Ghazoul, Status of pollinators and
their efficiency in coffee fruit set in a fragmented landscape mosaic in South India.
Basic Appl. Ecol. 13, 277–285 (2012).
79. S. Krishnan, K. G. Cheppudira, J. Ghazoul, Pollinator services in coffee agroforests of the
western Ghats, in Agroforestry, J. Dagar, V. Tewari, Eds. (Springer Singapore, 2017),
pp. 771–795.
80. M. Nesper, C. Kueffer, S. Krishnan, C. G. Kushalappa, J. Ghazoul, Shade tree diversity
enhances coffee production and quality in agroforestry systems in the Western Ghats.
Agric. Ecosyst. Environ. 247, 172–181 (2017).
81. S. Cusser, J. L. Neff, S. Jha, Natural land cover drives pollinator abundance and richness,
leading to reductions in pollen limitation in cotton agroecosystems. Agric. Ecosyst.
Environ. 226, 33–42 (2016).
82. N. P. Chacoff, M. A. Aizen, Edge effects on flower-visiting insects in grapefruit
plantations bordering premontane subtropical forest. J. Appl. Ecol. 43, 18–27 (2006).
83. N. P. Chacoff, M. A. Aizen, V. Aschero, Proximity to forest edge does not affect crop
production despite pollen limitation. Proc. R. Soc. B Biol. Sci. 275, 907–913 (2008).
84. L. G. Carvalheiro, C. L. Seymour, R. Veldtman, S. W. Nicolson, Pollination services decline
with distance from natural habitat even in biodiversity-rich areas. J. Appl. Ecol. 47,
810–820 (2010).
85. L. G. Carvalheiro, C. L. Seymour, S. W. Nicolson, Creating patches of native flowers
facilitates crop pollination in large agricultural fields: Mango as a case study. J. Appl. Ecol.
49, 1373–1383 (2012).
86. I. Bartomeus, V. Gagic, R. Bommarco, Pollinators, pests and soil properties interactively
shape oilseed rape yield. Basic Appl. Ecol. 16, 737–745 (2015).
87. D. A. Stanley, D. Gunning, J. C. Stout, Pollinators and pollination of oilseed rape crops
(Brassica napus L.) in Ireland: Ecological and economic incentives for pollinator
conservation. J. Insect Conserv. 17, 1181–1189 (2013).
88. D. A. Stanley, J. C. Stout, Pollinator sharing between mass-flowering oilseed rape and
co-flowering wild plants: Implications for wild plant pollination. Plant Ecol. 215, 315–325
(2014).
89. Y. Zou, F. J. J. A. Bianchi, F. Jauker, S. Luo, W. van der Werf, Wild pollinators enhance
oilseed rape yield in small-holder farming systems in China. BMC Ecol. 17, 6 (2017).
90. H. Connelly, K. Poveda, G. Loeb, Landscape simplification decreases wild bee pollination
services to strawberry. Agric. Ecosyst. Environ. 211, 51–56 (2015).
91. H. Grab, K. Poveda, B. Danforth, G. Loeb, Landscape context shifts the balance of costs
and benefits from wildflower borders on multiple ecosystem services. Proc. R. Soc. B Biol.
Sci. 285, 20181102 (2018).
92. R. I. A. Stewart, G. K. S. Andersson, C. Brönmark, B. K. Klatt, L.-A. Hansson, V. Zülsdorff,
H. G. Smith, Ecosystem services across the aquatic-terrestrial boundary: Linking ponds
to pollination. Basic Appl. Ecol. 18, 13–20 (2016).
93. B. Ricci, P. Franck, J.-F. Toubon, J.-C. Bouvier, B. Sauphanor, C. Lavigne, The influence of
landscape on insect pest dynamics: A case study in southeastern France. Landsc. Ecol.
24, 337–349 (2009).
94. M. Maalouly, P. Franck, J.-C. Bouvier, J.-F. Toubon, C. Lavigne, Codling moth parasitism is
affected by semi-natural habitats and agricultural practices at orchard and landscape
levels. Agric. Ecosyst. Environ. 169, 33–42 (2013).
95. B. Caballero-López, R. Bommarco, J. M. Blanco-Moreno, F. X. Sans, J. Pujade-Villar,
M. Rundlöf, H. G. Smith, Aphids and their natural enemies are differently affected by
habitat features at local and landscape scales. Biol. Control. 63, 222–229 (2012).
96. G. Tamburini, S. De Simone, M. Sigura, F. Boscutti, L. Marini, Conservation tillage
mitigates the negative effect of landscape simplification onbiological control. J. Appl. Ecol. 53,
233–241 (2016).
97. G. Tamburini, S. De Simone, M. Sigura, F. Boscutti, L. Marini, Soil management shapes
ecosystem service provision and trade-offs in agricultural landscapes. Proc. R. Soc. B Biol. Sci.
283, 20161369 (2016).
98. G. Tamburini, I. Pevere, N. Fornasini, S. de Simone, M. Sigura, F. Boscutti, L. Marini,
Conservation tillage reduces the negative impact of urbanisation on carabid
communities. Insect Conserv. Divers. 9, 438–445 (2016).
99. H. Taki, K. Maeto, K. Okabe, N. Haruyama, Influences of the seminatural and natural
matrix surrounding crop fields on aphid presence and aphid predator abundance
within a complex landscape. Agric. Ecosyst. Environ. 179, 87–93 (2013).
100. D. K. Letourneau, S. G. Bothwell Allen, J. O. Stireman III, Perennial habitat fragments,
parasitoid diversity and parasitism in ephemeral crops. J. Appl. Ecol. 49, 1405–1416
(2012).Dainese et al., Sci. Adv. 2019;5 : eaax0121 16 October 2019101. B. Maas, Y. Clough, T. Tscharntke, Bats and birds increase crop yield in tropical
agroforestry landscapes. Ecol. Lett. 16, 1480–1487 (2013).
102. M. E. O. O’Rourke, K. Rienzo-Stack, A. G. Power, A multi-scale, landscape approach
to predicting insect populations in agroecosystems. Ecol. Appl. 21, 1782–1791 (2011).
103. M. Jonsson, H. L. Buckley, B. S. Case, S. D. Wratten, R. J. Hale, R. K. Didham, Agricultural
intensification drives landscape-context effects on host-parasitoid interactions in
agroecosystems. J. Appl. Ecol. 49, 706–714 (2012).
104. M. Kishinevsky, T. Keasar, A. R. Harari, E. Chiel, A comparison of naturally growing
vegetation vs. border-planted companion plants for sustaining parasitoids in
pomegranate orchards. Agric. Ecosyst. Environ. 246, 117–123 (2017).
105. E. A. Martin, B. Seo, C.-R. Park, B. Reineking, I. Steffan-Dewenter, Scale-dependent
effects of landscape composition and configuration on natural enemy diversity, crop
herbivory, and yields. Ecol. Appl. 26, 448–462 (2016).
106. K. Poveda, E. Martínez, M. F. Kersch-Becker, M. A. Bonilla, T. Tscharntke, Landscape
simplification and altitude affect biodiversity, herbivory and Andean potato yield.
J. Appl. Ecol. 49, 513–522 (2012).
107. M. B. Takada, A. Yoshioka, S. Takagi, S. Iwabuchi, I. Washitani, Multiple spatial scale
factors affecting mirid bug abundance and damage level in organic rice paddies.
Biol. Control. 60, 169–174 (2012).
108. M. G. E. Mitchell, E. M. Bennett, A. Gonzalez, Agricultural landscape structure affects
arthropod diversity and arthropod-derived ecosystem services. Agric. Ecosyst. Environ.
192, 144–151 (2014).
109. F. Geiger, J. Bengtsson, F. Berendse, W. W. Weisser, M. Emmerson, M. B. Morales,
P. Ceryngier, J. Liira, T. Tscharntke, C. Winqvist, S. Eggers, R. Bommarco, T. Pärt,
V. Bretagnolle, M. Plantegenest, L. W. Clement, C. Dennis, C. Palmer, J. J. Oñate,
I. Guerrero, V. Hawro, T. Aavik, C. Thies, A. Flohre, S. Hänke, C. Fischer, P. W. Goedhart,
P. Inchausti, Persistent negative effects of pesticides on biodiversity and biological
control potential on European farmland. Basic Appl. Ecol. 11, 97–105 (2010).
110. M. Plećaš, V. Gagić, M. Janković, O. Petrović-Obradović, N. G. Kavallieratos, Ž. Tomanović,
C. Thies, T. Tscharntke, A. Ćetković, Landscape composition and configuration influence
cereal aphid-parasitoid-hyperparasitoid interactions and biological control
differentially across years. Agric. Ecosyst. Environ. 183, 1–10 (2014).
111. M. Tschumi, M. Albrecht, M. H. Entling, K. Jacot, High effectiveness of tailored flower
strips in reducing pests and crop plant damage. Proc. R. Soc. Biol. Sci. 282, 20151369
(2015).
Acknowledgments: We thank R. Carloni for producing icons in Figs. 1 to 4. We also thank the
Department of Innovation, Research and University of the Autonomous Province of Bozen/
Bolzano for covering the Open Access publication costs. For all further acknowledgements,
see the Supplementary Materials. Funding: This work was funded by EU-FP7 LIBERATION
(311781) and Biodiversa-FACCE ECODEAL (PCIN-2014–048). Author contributions: M.D., E.A.M.,
and I.S.-D. conceived the study. M.D. performed statistical analyses and wrote the manuscript
draft. The authors named from E.A.M. to T.T. and I.S.-D. discussed and revised earlier
versions of the manuscript. The authors named from G.K.S.A. to Y.Z. are listed alphabetically,
as they contributed equally in gathering field data and providing several important corrections
to subsequent manuscript drafts. Competing interests: The authors declare that they have
no competing interests. Data and materials availability: All data needed to evaluate the
conclusions in the paper are present in the paper and/or the Supplementary Materials.
Additional data related to this paper may be requested from the authors.
Submitted 14 February 2019
Accepted 22 September 2019
Published 16 October 2019
10.1126/sciadv.aax0121
Citation: M. Dainese, E. A. Martin, M. A. Aizen, M. Albrecht, I. Bartomeus, R. Bommarco,
L. G. Carvalheiro, R. Chaplin-Kramer, V. Gagic, L. A. Garibaldi, J. Ghazoul, H. Grab, M. Jonsson,
D. S. Karp, C. M. Kennedy, D. Kleijn, C. Kremen, D. A. Landis, D. K. Letourneau, L. Marini, K. Poveda,
R. Rader, H. G. Smith, T. Tscharntke, G. K. S. Andersson, I. Badenhausser, S. Baensch, A. D. M. Bezerra,
F. J. J. A. Bianchi, V. Boreux, V. Bretagnolle, B. Caballero-Lopez, P. Cavigliasso, A. Ćetković,
N. P. Chacoff, A. Classen, S. Cusser, F.D. da Silva e Silva, G. A. deGroot, J.H. Dudenhöffer, J. Ekroos,
T. Fijen, P. Franck, B. M. Freitas, M. P. D. Garratt, C. Gratton, J. Hipólito, A. Holzschuh, L. Hunt,
A. L. Iverson, S. Jha, T. Keasar, T. N. Kim, M. Kishinevsky, B. K. Klatt, A.-M. Klein, K. M. Krewenka,
S. Krishnan, A. E. Larsen, C. Lavigne, H. Liere, B. Maas, R. E. Mallinger, E. Martinez Pachon,
A. Martínez-Salinas, T. D. Meehan, M. G. E. Mitchell, G. A. R. Molina, M. Nesper, L. Nilsson,
M. E. O’Rourke, M. K. Peters, M. Plećaš, S. G. Potts, D. d. L. Ramos, J. A. Rosenheim, M. Rundlöf,
A. Rusch, A. Sáez, J. Scheper, M. Schleuning, J. M. Schmack, A. R. Sciligo, C. Seymour, D. A. Stanley,
R. Stewart, J. C. Stout, L. Sutter,M. B. Takada, H. Taki, G. Tamburini, M. Tschumi, B. F. Viana,
C. Westphal, B. K. Willcox, S. D. Wratten, A. Yoshioka, C. Zaragoza-Trello, W. Zhang, Y. Zou,
I. Steffan-Dewenter, A global synthesis reveals biodiversity-mediated benefits for crop
production. Sci. Adv. 5, eaax0121 (2019).13 of 13
A global synthesis reveals biodiversity-mediated benefits for crop production
Matteo Dainese, Emily A. Martin, Marcelo A. Aizen, Matthias Albrecht, Ignasi Bartomeus, Riccardo Bommarco, Luisa G.
Carvalheiro, Rebecca Chaplin-Kramer, Vesna Gagic, Lucas A. Garibaldi, Jaboury Ghazoul, Heather Grab, Mattias Jonsson, 
Daniel S. Karp, Christina M. Kennedy, David Kleijn, Claire Kremen, Douglas A. Landis, Deborah K. Letourneau, Lorenzo
Marini, Katja Poveda, Romina Rader, Henrik G. Smith, Teja Tscharntke, Georg K. S. Andersson, Isabelle Badenhausser, 
Svenja Baensch, Antonio Diego M. Bezerra, Felix J. J. A. Bianchi, Virginie Boreux, Vincent Bretagnolle, Berta
Caballero-Lopez, Pablo Cavigliasso, Aleksandar Cetkovic, Natacha P. Chacoff, Alice Classen, Sarah Cusser, Felipe D. da 
Silva e Silva, G. Arjen de Groot, Jan H. Dudenhöffer, Johan Ekroos, Thijs Fijen, Pierre Franck, Breno M. Freitas, Michael P. D.
Garratt, Claudio Gratton, Juliana Hipólito, Andrea Holzschuh, Lauren Hunt, Aaron L. Iverson, Shalene Jha, Tamar Keasar, 
Tania N. Kim, Miriam Kishinevsky, Björn K. Klatt, Alexandra-Maria Klein, Kristin M. Krewenka, Smitha Krishnan, Ashley E.
Larsen, Claire Lavigne, Heidi Liere, Bea Maas, Rachel E. Mallinger, Eliana Martinez Pachon, Alejandra Martínez-Salinas, 
Timothy D. Meehan, Matthew G. E. Mitchell, Gonzalo A. R. Molina, Maike Nesper, Lovisa Nilsson, Megan E. O'Rourke, Marcell
K. Peters, Milan Plecas, Simon G. Potts, Davi de L. Ramos, Jay A. Rosenheim, Maj Rundlöf, Adrien Rusch, Agustín Sáez, 
Jeroen Scheper, Matthias Schleuning, Julia M. Schmack, Amber R. Sciligo, Colleen Seymour, Dara A. Stanley, Rebecca
Stewart, Jane C. Stout, Louis Sutter, Mayura B. Takada, Hisatomo Taki, Giovanni Tamburini, Matthias Tschumi, Blandina F. 
Viana, Catrin Westphal, Bryony K. Willcox, Stephen D. Wratten, Akira Yoshioka, Carlos Zaragoza-Trello, Wei Zhang, Yi Zou
and Ingolf Steffan-Dewenter
Sci Adv 5 (10), eaax0121.
DOI: 10.1126/sciadv.aax0121
ARTICLE TOOLS http://advances.sciencemag.org/content/5/10/eaax0121
SUPPLEMENTARY http://advances.sciencemag.org/content/suppl/2019/10/10/5.10.eaax0121.DC1
MATERIALS
REFERENCES This article cites 101 articles, 13 of which you can access for free
http://advances.sciencemag.org/content/5/10/eaax0121#BIBL
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Use of this article is subject to the Terms of Service
Science Advances (ISSN 2375-2548) is published by the American Association for the Advancement of Science, 1200 New
York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive licensee American 
Association for the Advancement of Science. No claim to original U.S. Government Works. The title Science Advances is a
registered trademark of AAAS.
Downloaded from http://advances.sciencemag.org/ on October 17, 2019