Content uploaded by Heather Grab
Author content
All content in this area was uploaded by Heather Grab on Aug 02, 2018
Content may be subject to copyright.
rspb.royalsocietypublishing.org
Research
Cite this article: Grab H, Poveda K, Danforth
B, Loeb G. 2018 Landscape context shifts the
balance of costs and benefits from wildflower
borders on multiple ecosystem services.
Proc. R. Soc. B 285: 20181102.
http://dx.doi.org/10.1098/rspb.2018.1102
Received: 17 May 2018
Accepted: 6 July 2018
Subject Category:
Global change and conservation
Subject Areas:
ecology
Keywords:
ecological intensification, wildflower strips,
landscape, pollination, biological control,
crop yield
Author for correspondence:
Heather Grab
e-mail: hlc66@cornell.edu
Electronic supplementary material is available
online at https://dx.doi.org/10.6084/m9.
figshare.c.4163000.
Landscape context shifts the balance
of costs and benefits from wildflower
borders on multiple ecosystem services
Heather Grab1, Katja Poveda1, Bryan Danforth1and Greg Loeb2
1
Department of Entomology, Cornell University, Ithaca, NY 14853, USA
2
Department of Entomology, New York State Agricultural Experiment Station, Cornell University, Geneva,
NY 14456, USA
HG, 0000-0002-1073-8805
In the face of global biodiversity declines driven by agricultural intensification,
local diversification practices are broadly promoted to support farmland
biodiversityand multiple ecosystem services. The creation of flower-rich habi-
tats on farmland has been subsidized in both the USA and EU to support
biodiversity and promote delivery of ecosystem services. Yet, theory suggests
that the landscape context in which local diversification strategies are
implemented will influence their success. However, few studies have
empirically evaluated this theory or assessed the ability to support multiple
ecosystem services simultaneously. Here, we evaluate the impact of creating
flower-rich habitats in field margins on pollination, pest control, and crop
yield over 3 years using a paired design across a landscape gradient. We
find general positive effects of natural habitat cover on fruit weight and that
flowering borders increase yields by promoting bee visitation to adjacent
crops only in landscapes with intermediate natural habitat cover. Flowering
borders had little impact on biological control regardless of landscape context.
Thus, knowledge of landscape context can be used to target wildflower border
placement in areas where they will have the greatest likelihood for success and
least potential for increasing pest populations or yield loss in nearby crops.
1. Introduction
Presently, 40% of the earth’s terrestrial surface is used for agricultural production
[1] and the continued transition of natural habitat to agricultural use is one of the
primary drivers of biodiversity loss worldwide [2]. Balancing the demand for
agricultural productivity with biodiversity conservation is one of the greatest
challenges facing global humanity. However, agricultural intensification can
undermine the very biodiversity and ecosystem services that would otherwise
benefit crop production [3– 5]. Diverse biological communities support many eco-
system services to agriculture, provide resilience to disturbances, and maintain
the capacity to adapt to future changing environments [6,7]. Agricultural intensi-
fication at both local and landscape scales reduces the spatial and temporal
availability of resources required by beneficial organisms, such as pollinators
and natural enemies [8], while crop pests often benefit from a concentration of
host plants [9].
To increase agricultural sustainability, strategies are needed that reduce
conflicts between biodiversity conservation and crop production. Ecological inten-
sification capitalizes on the biodiversity within agroecosystems to achieve
sustainable increases in crop yields by actively managing communities of ecosys-
tem service providers [10,11]. Yet, a major hurdle to the widespread adoption of
ecological intensification strategies is a framework for predicting the contexts in
which they will be successful. Variable effectiveness of these practices may be
due to the landscape context in which they are implemented [12– 14]. The inter-
mediate landscape complexity hypothesis predicts that local management
strategies will be most effective at improving biodiversity and ecosystem services
&2018 The Author(s) Published by the Royal Society. All rights reserved.
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
when established in landscapes that are dominated by agri-
culture but with at least some natural habitat remaining
[9,12]. In landscapes with high natural habitat cover, beneficial
organisms continuously colonize agricultural habitats. Alterna-
tively, in landscapes with very little natural habitat remaining,
source populations of beneficials are too depauperate to recruit
from. However, in landscapes with intermediate amounts of
remaining natural habitat, regional source populations are pre-
sent, but agricultural habitats are not continuously colonized.
Therefore, ecological intensification in these intermediate land-
scapes is expected to produce the greatest effects and early
findings from Europe support these patterns with respect to
enhancing biodiversity [15– 19]. Whether these findings are
also reflected in the delivery of multiple ecosystem services
and crop yield remains unresolved.
Multiple ecosystem services are expected to benefit from
increases in local habitat diversity. For example, local manage-
ment with flowering strips has been shown to increase the
abundance of pollinators and natural enemies of pests in adja-
cent cropland [19–21]. However, few studies have evaluated
the effect of local habitat management on multiple services
simultaneously [22–24], and only one has evaluated their
combined effects on crop yield [25]. Consequently, our under-
standing of the potential interactions between yield-supporting
ecosystem services and ecological intensification strategies that
can simultaneously support them is limited.
Pests can also benefit from natural habitats at the local
and landscape scale [24,26,27], thus management strategies
aimed at increasing ecosystem services may fail to improve
pest control or crop yield [28]. In these cases, although bio-
diversity may be locally improved, yield gaps may trigger
the transition of natural lands to agriculture elsewhere leading
to a net loss for both biodiversity and ecosystem services.
The planting of flower-rich crop borders is subsidized by pol-
icies in both the USA and EU and many governments and
intergovernmental agencies have recently called for agricul-
tural management practices that support biodiversity and
ecosystem services on farms (White House Pollinator Protec-
tion Task Force 2016, IPBES 2017) making the need to ensure
efficient placement and effectiveness more critical than ever.
Here, we evaluate the benefits and potential costs of a
commonly implemented ecological intensification strategy,
the planting of native perennial wildflowers in field margins.
We explore the effect of wildflower borders on crop visitation
by bees, biological control, pest abundance, crop damage,
and crop yield using a paired design in strawberry plantings
with and without a wildflower border on 12 farms across a
landscape gradient. Following the predictions of the inter-
mediate landscape hypothesis, we expect that wildflower
margins will improve ecosystem services and crop production
to a greater extent when implemented in landscapes with
intermediate amounts of natural habitat cover.
2. Methods
(a) Experimental design
We identified 12 farms within the Finger Lakes region of central
New York State that varied in landscape composition (18 – 61%
natural land cover; electronic supplementary material, figure
S1). On each farm, we established two 10 15 m plots consisting
of five rows of strawberry (var. ‘Jewel’) in the spring of 2012.
Plots were separated by a minimum of 200 m and were randomly
assigned to either a control border or a native perennial wild-
flower border. The distance separating plots represents a
compromise between the relatively small foraging ranges of the
insect communities relevant to strawberry [29] and ensuring
that plot pairs within a farm were within the same landscape
contexts. Composition and management of control borders
were representative of field edge management practices in the
region. Control borders consisted primarily of orchard grass
and were regularly mown over the growing season. Wildflower
borders were approximately 4 m wide by 10 m long and ran
parallel with the crop border consistent with standard imple-
mentations of this management strategy in terms of size and
orientation. Plantings consisted of the following nine US native
perennial species Zizia aurea,Penstemon digitalis,Coreopsis
lanceolata,Potentilla fruticosa,Veronicastrum virginicum,Agastache
neptoides,Silphium perfoliatum,Lobelia siphilitica, and Solidago
canadensis. These species were selected based on their attractive-
ness to bees and natural enemies [13,20,30– 32] and provide
overlapping bloom periods, so that flowers are present through-
out the growing season. When possible, every effort was made to
grow plants from local ecotypes. Both border types were estab-
lished in the autumn of 2012. Plots were managed organically
or involved limited use of pesticides for weed or fungal pathogen
management. Each year, straw mulch was applied to all plots
in the autumn and raked into the row middles in the spring con-
sistent with standard horticultural practices for strawberry in
the northeast. In 2015, one wildflower strip was accidentally
destroyed leaving only 11 site replicates in that year.
At four plots, it was necessary to prevent damage from large
mammalian herbivores by erecting temporary plastic fencing.
Fence gaps were wide enough (3 3 cm) to allow access to
even the largest pollinators (H Grab 2014, personal observation).
In each case, both the control and wildflower treatment plots on
the same farm were fenced.
(b) Landscape
Landscape complexity was characterized using the National
Agricultural Statistics Service Cropland Data Layer for
New York [33] for each year of the study (2013– 2015) in ArcGIS
10.1. The region is characterized by a mix of row crops, fruits
and vegetables, orchard, dairy, old-field habitats, and forest. We
quantified the cover of natural and semi-natural habitats at four
radii from the centre of each plot (500, 750, 1 000, and 1 250 m).
Land cover values were averaged between paired plots in each
farm to generate a single landscape value for each farm in each
year. We fitted separate nonlinear models for each response vari-
able and scale and determined that 750 m was the scale at which
the cover of natural area provided the best fit to the data (based
on AICc values, see electronic supplementary material, table S1).
Previous studies of the pollinators, parasitoids, and pests in this
system have found strong responses to this landscape metric at
similar scales [29,34,35].
(c) Pollinator surveys
The community of pollinators visiting strawberry is dominated
by a diverse fauna of wild bees with honeybees comprising
only 7% of the pollinator community [29]. In the 3 years follow-
ing plot establishment (2013 – 2015), the visitation rate of bees to
strawberry flowers was estimated by conducting visual surveys
on four dates per plot spanning the duration of crop bloom.
Surveys were carried out between 10:00 and 15:30 on sunny
days with temperatures above 168C. Visitation rate was assessed
using standardized 10 min transects through each plot recording
each bee visit to a strawberry flower. The number of open straw-
berry flowers per square foot was estimated for each plot by
averaging counts of flowers in 1 ft
2
quadrats in each of the five
rows. Visitation rates per plot were calculated by dividing the
rspb.royalsocietypublishing.org Proc. R. Soc. B 285: 20181102
2
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
total number of visits recorded during the 10 min transects by
the average number of open flowers per square foot.
To better understand the relative importance of the planted
wildflower species, we monitored pollinator visitation rates to
each plant species as well as visits to flowering weeds within
the borders throughout the season in 2015. All flowering plants
within the border were observed for 10 min and total number
of visits per plant species was recorded.
(d) Pest surveys
The primary pest of strawberry in the region is Lygus lineolaris
(Hemiptera: Miridae), a generalist herbivore that feeds on the
seeds of developing strawberry fruits. From 2013 to 2015, the
abundance of L. lineolaris was estimated in each plot immediately
following strawberry flowering by tapping individual strawberry
inflorescences until a total of 24 nymphs were collected or all
inflorescences in the plot were sampled. We chose a target of
24 nymphs per sample because this number allowed us to
accurately estimate parasitism rates using the protocols described
below. Nymph densities were calculated by dividing the number
of nymphs collected by the total number of inflorescences
tapped.
Because wildflower borders may harbour pest populations
that can spill over into the crop, we estimated the abundance of
L. lineolaris in the wildflower borders compared to the control bor-
ders for an entire growing season in 2015. The abundance of
L. lineolaris adults and nymphs was estimated for each flowering
species present in the wildflower borders by vacuuming (Echo
ES 230 Shred ‘n Vac, Lake Zurich, IL, USA) 25 inflorescences of
each plant species once a week from May to October. Plants
were sampled at the bud, flowering, and seed head phases, so
that our estimate for each species accurately reflected the broad
feeding preferences of L. lineolaris. After sampling a particular
species, all L. lineolaris were returned to the host plant they were
collected from to ensure that the effects of sampling in one
week had little impact on samples in the subsequent weeks.
Sampling also included any weedy flowering species that had
invaded the perennial borders. As some plant species had fewer
than 25 inflorescences on any particular sampling date, the total
number of L. lineolaris collected was divided by the number of
inflorescences vacuumed for each sample. The order of sampling
was randomized for species blooming on a given date. An equiv-
alent number of vacuum samples were obtained from the grassy
margins of control borders for each wildflower species sampled
from its paired wildflower treatment plot.
(e) Parasitism rates
In the study region, the primary natural enemies of L. lineolaris
include a complex of native and introduced parasitoid wasps
in the genus Peristenus [36]. Three species, Peristenus digoneutis
(introduced), Peristenus pallipes (native), and rarely Peristenus
relictus (introduced), are known to attack L. lineolaris; however,
parasitism rates are reduced in landscapes with a high pro-
portion of agricultural land cover [35].Diagnostic PCR assays
were used to simultaneously estimate parasitism rates and para-
sitoid species identity, as they are faster and more accurate than
rearing or dissection [37,38]. Random samples of 24 nymphs
from each sampling period at each site were selected for parasitism
assays. In some cases, fewer than 24 nymphs were collected in a
sampling period. In these cases, all collected nymphs for the
period were processed. In three instances, no nymphs were collected
on a farm in a particular year; therefore, these instances resulted in
only 32 site by year replicates. DNA from nymphs was extracted
using an abbreviated chloroform: isoamyl alcohol protocol devel-
oped by Tilmon & Hoffmann [39]. DNA extractions along with
negative controls were amplified using Peristenus species-specific
primers as in [40]. Using this method, species-specific forward
primers are combined with a genus-specific reverse primer to
amplify a region including ITS1 and ITS2.
(f) Fruit damage and yield
A typical strawberry inflorescence is comprised of a single pri-
mary fruit (king berry), a pair of secondary fruit, and four
tertiary fruit. Strawberries are an aggregate accessory fruit com-
prised of as many as 300 achenes on a primary fruit and 200 on
a secondary fruit [41]. Each achene must be fertilized for the sur-
rounding tissue to develop and an average of four visits per flower
is required to achieve full pollination [42]. Lygus lineolaris nymphs
and adults feed on developing achenes leading to developmental
failure of the surrounding tissues. Fruit weight is highly correlated
with the number of developed undamaged achenes [41]. Fruits
with a high percentage of damaged achenes, either from poorpol-
lination or L. lineolaris feeding, develop with major malformations
that reduce overall yield and marketability [43].
To measure the impact of wildflower borders on crop yield at
each site, 30 flowers were marked and the resulting fruits from
each plot were harvested when ripe and weighed. Owing to
sample processing errors, fruit data are unavailable for one site
in 2013 and one site in 2014. The percentage of poorly pollinated
and damaged achenes was estimated for each fruit. Secondary
fruits were used, as they are less prone to frost damage than
primary fruit and due to their later development are highly
susceptible to damage from L. lineolaris nymph feeding.
(g) Statistical analysis
To evaluate whether wildflower borders had differential effects
across the landscape gradient, we first pooled individual measures
for each variable (ex. pollinator visitation, L. lineolaris, parasitism,
malformations, or weights per individual fruit) by plot and year
averaging over all surveys within a plot in each year. Because the
primary objective of the study was to determine the effectiveness
of the plantings under varying landscape contexts, we calculated
an index of wildflower border effectiveness for each variable. In
this way, we are able to control for the overall variation that
occurs across the landscape gradient and isolate the variation
due to the wildflower border. Absolute values for each variable
on control and wildflower planting across the landscape gradient
are presented in electronic supplementary material, figure S2.
The effectiveness index was calculated by subtracting the average
value observed on the plots with a wildflower border minus the
control divided by the control ((Wildflower 2Control)/Control)
of each farm in each year. The effectiveness index therefore rep-
resents a quantitative measure of the relative effectiveness of the
wildflower planting. Positive values indicate an increase in the
variable of interest on the plot with a wildflower planting com-
pared to the control and negative values indicate the measured
variable was higher on control plantings. We then constructed
linear and nonlinear mixed effects models for each index
(GLMER, R package lme4 [40,44]) with Gaussian error structures.
Fixed effects in each model included year and proportion of com-
bined natural and semi-natural habitat cover as well their
interaction. We constructed linear, logistic, and polynomial
models for each variable and selected the best fit model based on
AICc values. Farm was included as a random effect in each
model to account for the paired experimental design and repeated
measures across years. Additionally, we tested whether site-level
covariates including the distance between the wildflower and con-
trol plot, the total number of flower plant species, and the total
number of native perennial species that established had an effect
on each of the index variables. Each of these site-level covariates
was evaluated in a mixed effects model that included the linear
and polynomial natural habitat cover terms with site as a
random effect. The majority of covariates did not explain a
rspb.royalsocietypublishing.org Proc. R. Soc. B 285: 20181102
3
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
significant portion of the variation in any index and were not
included in the final models.
Differences in L. lineolaris numbers and bee visits to wild-
flower and weedy flowering species within the plot margins
were assessed with generalized linear mixed effects models
with Poisson error distributions. For both variables, plant species
was included as a fixed effect and farm was included as a
random effect. For overall L. lineolaris abundance in wildflower
compared to control borders, an index was computed similar
to those described above. Fixed effects included year and pro-
portion of combined natural and semi-natural habitat cover as
well their interaction.
The contribution of L. lineolaris feeding versus poor pollination
to fruit damage was assessed using a generalized linear mixed
effects model with a Poisson error structure. Fixed effects include
year, average bee visitation, and average L. lineolaris abundance,
as well as, the two-way interactions between year and bee visits,
and year and L. lineolaris abundance. In all models, p-values and
degrees of freedom are calculated based on the Satterthwaite
approximation as implemented in the package lmerTest [45].
3. Results
Wildflowers bloomed from April to November each year begin-
ning in 2013. On average, seven of the nine wildflower species
became established at each site, but no site had fewer than six
species (electronic supplementary material, table S3). In the 3
years following establishment (2013–2015), a total of 5 684
bee visits to strawberry were recorded and 1 307 bee specimens
were collected. Wildbees were the dominant visitors, represent-
ing 95.8% of the community while managed bees (honeybees)
made up only 4.2% of recorded visits. In total, 99 species
were recorded based on net collected specimens.
Following the expectations of the intermediate landscape
hypothesis, the effect of wildflower borders on bee visitation
to the strawberry crop across the landscape gradient was best
described by a second-order polynomial function (AICc
poly
¼
63.86, AICc
log
¼73.86, AICc
linear
¼73.89; Poly: F
1,21
¼7.33,
p¼0.01). Wildflower borders increased bee visitation to
strawberry relative to controls only in landscapes with inter-
mediate amounts of natural habitat (figure 1a). On average,
wildflower borders had little effect on bee visitation in the
first 2 years after establishment, but had positive effects in
2015 (t
1,21
¼2.48, p¼0.02; figure 1b).
A total of 3 197 L. lineolaris nymphs were collected from
the strawberry plantings over the 3 years of the study. The
effect of wildflower borders on L. lineolaris abundance was
also influenced by the landscape according to a second-
order polynomial function (AICc
poly
¼127.7, AICc
log
¼
136.3, AICc
linear
¼136.1; Poly: F
1,18
¼3.71, p¼0.06). Pest
abundances on plots with a wildflower border were greater
than controls in the landscapes with the least and most natu-
ral habitat cover (figure 2a). In intermediate landscapes,
wildflower borders decreased pest pressure below the levels
of control plots. The abundance of L. lineolaris on plots with
a wildflower border differed across the years (F
2,18
¼7.88,
p¼0.003) and was greatest in 2014 (t
1,21
¼4.67, p,0.001;
electronic supplementary material, figure S3).
Parasitism assays revealed an overall parasitism rate of
18%. Three parasitoid species were detected with the intro-
duced P. digoneutis being the dominant natural enemy (96.7%
of parasitism events) and the other two species, P. pallipes
(native, 2.8%) and P. relictus (introduced, 0.05%), represented
at low levels. The effectiveness of wildflower borders across
the landscape gradient on parasitism largely mirrored the pat-
tern observed for pest abundances (figure 2b). Again a
polynomial function best fit the data (AICc
poly
¼117.7,
AICc
log
¼126.5, AICc
linear
¼126.4; Poly: F
1,16
¼4.06, p¼
0.06). However, parasitism rates followed a pattern across
years similar to bee visitation; achieving the highest values
on wildflower plots relative to controls in 2015 (t
1,18
¼2.48,
p¼0.02; electronic supplementary material, figure S4).
Sampling L. lineolaris within the plot margins themselves
revealed that densities of L. lineolaris were higher in wildflower
borders compared to control borders throughout the season
(WF: F
1,10
¼30.47, p¼0.0003). Although there were no differ-
ences in the number of L. lineolaris collected in control margins
between the landscape types, wildflower margins in land-
scapes with intermediate natural habitat cover supported
greater numbers of L. lineolaris relative to landscapes with
either low or high proportions of natural habitat (figure 3a,
–0.2
–0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
2013 2014 2015
y
ear
*
–0.8
–0.6
–0.4
–0.2
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
0.2 0.3 0.4 0.5 0.6
p
ro
p
ortion natural habitat
effect on bee visitation to crop
2013
2014
2015
year
(a)(b)
Figure 1. Effectiveness of wildflower (WF) borders relative to control (C) plots ((WF 2C)/C) for bee visitation to strawberry flowers (a) in relation to the proportion
of natural land cover in a 750 m radius around each site across all 3 years of the study and (b) in each of the study years following wildflower establishment in 2012.
In (a), shaded areas represent 95% confidence intervals. Asterisk in (b) indicates values different from 0 at p,0.05 based on post hoc contrast tests.
rspb.royalsocietypublishing.org Proc. R. Soc. B 285: 20181102
4
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
F
1,10
¼5.42, p¼0.052). In 2015, each wildflower species in the
wildflower border was surveyed for its attractiveness to both
bees and L. lineolaris. The number of L. lineolaris supported by
different species of wildflowers varied (F
12,238
¼1.94, p¼0.03)
as did the number of bee visitors to each species (F
14,86
¼8.12,
p¼0.0001; figure 3b). While some species supported
moderate numbers of both L. lineolaris and pollinators
(ex. Penstemon), the most attractive species were different for
pests (ex. Erigeron) and pollinators (ex. Silphium).
Both lack of pollination by bees and feeding by
L. lineolaris cause malformations to developing strawberry
fruit resulting in yield loss. The relative importance of
L. lineolaris abundance versus bee visitation in predicting
malformations varied across study years (L. lineolaris
year: F
1,11
¼36.03, p,0.001; bee year: F
1,11
¼33.26, p,
0.001). In both 2013 and 2014, L. lineolaris abundance was
the only significant predictor of fruit malformations and
increasing nymph abundance was associated with greater
malformations (2013 L. lineolaris:z¼2.98, p¼0.002, bee:
z¼0.22, p¼0.823; 2014 L. lineolaris:z¼2.17, p¼0.029,
bee: z¼20.21, p¼0.829). In 2015, both groups predicted
malformations; although, bee visitation had a stronger
effect in reducing malformations (bee: z¼22.74, p¼0.006;
L. lineolaris:z¼2.24, p¼0.025; electronic supplementary
material, figure S5) consistent with increasing positive effect
of wildflowers on bees over the 3-year study.
The difference in fruit malformations on plots with a
wildflower border compared to controls was best explained
by a polynomial response to landscape (AICc
poly
¼65.83,
AICc
log
¼72.35, AICc
linear
¼72.58; Poly: F
1,19
¼3.48, p¼
0.07). Malformations caused by both poor pollination and
L. lineolaris feeding were greatest on plots with a wildflower
border relative to control plots in landscapes with the
least natural land cover (figure 4a). Landscapes with inter-
mediate cover of natural habitat had the greatest reduction
in malformations relative to control plots.
For fruit weight, a polynomial function also best described
the relationship between wildflower border effectiveness and
landscape (AICc
poly
¼20.33, AICc
log
¼7.42, AICc
linear
¼
8.52; Poly: F
1,19
¼8.68, p¼0.008). In the landscapes with the
0
2
4
6
effect on L. lineolaris
–1
0
1
2
3
4
5
0.2 0.3 0.4 0.5 0.6
proportion natural habitat
0.2 0.3 0.4 0.5 0.6
proportion natural habitat
effect on parasitism rate
2013
2014
2015
(b)(a)
year
Figure 2. Effectiveness of wildflower (WF) borders relative to control (C) plots ((WF 2C)/C) for (a) the number of L. lineolaris nymphs and (b) the parasitism rate
of nymphs. Shaded areas represent 95% confidence intervals.
02040
Silphium*
Symphyotrichum
Penstemon*
Solidago*
Zizea*
Lotus
Veronicastrum*
Agastache*
Cirsium
Hieracium
Lobelia*
Coreopsis*
Trifolium
Erigeron
bee visits
00.10.20.30.40.50.6
Erigeron
Solidago*
Symphyotrichum
Penstemon*
Silphium*
Achillium
Trifolium
Coreopsis*
Zizea*
Daucus
Veronicastrum*
Agastache*
Lobelia*
Hieracium
L. lineolaris per sample
–5
0
5
0.2 0.3 0.4 0.5 0.6
effect on L. lineolaris in wildflowers
proportion natural habitat
(a)(b)
Figure 3. The average number of L. lineolaris nymphs collected within the wildflower borders (a) relative to control borders and (b) on various wildflower species
(planted species with *) relative to the number of bees visiting each wildflower species.
rspb.royalsocietypublishing.org Proc. R. Soc. B 285: 20181102
5
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
least natural cover, plots with a wildflower border had lower
yields than control plots. By contrast, plots with a wildflower
border had higher yields than controls in landscapes with
intermediate amounts of natural land cover (figure 4b). This
difference between wildflower and control borders decreased
in landscapes with the most natural cover.
4. Discussion
Ecological intensification strategies including the creation of
flower-rich habitats on agricultural lands have been promoted
as a practice to support farmland biodiversity and encourage
the delivery of ecosystem services (White House Initiative,
EU Initiative, IPBES report 2016 [12,13,46– 48]). In the USA,
these practices are subsidized at a rate of $57 million per year
through the Conservation Reserve Program (CRP)and Environ-
mental Quality Incentives Program (EQIP) [49]. Yet, few studies
have evaluated the effectiveness of these practices across a
gradient of landscape contexts or on multiple ecosystem ser-
vices simultaneously, impeding our ability to effectively
implement these practices. Here, we evaluate the impact of
wildflower borders on pollination, pest control, and crop
yield across a landscape gradient and find that while flowering
crop borders can be successful in some contexts, landscapes that
are the most agriculturally intensified will require larger scale
more coordinated conservation approaches.
In terms of supporting crop pollination services, our find-
ings support the prediction of the intermediate landscape
hypothesis [9,12] with bee visitation to crop flowers increas-
ing with the addition of local wildflower borders according
to a polynomial function which peaked in landscapes with
intermediate cover of natural habitats. Interestingly, the inter-
mediate values of land cover that correspond with success of
the wildflower borders in supporting ecosystem services are
shifted strongly towards higher values of natural habitat
compared to those originally proposed by Tscharntke et al.
[12] for supporting biodiversity in European landscapes.
Tscharntke et al. proposed that wildflower borders would
have the strongest effects on biodiversity in landscapes with
1–20% non-crop habitat. In our study, wildflower habitats
were the most successful at increasing the delivery of
ecosystem services in landscapes with 25–55% natural habitat
cover. These differences in threshold values may reflect the
differences in the composition of the current dominant natural
habitat covers (grasslands in Europe, forest in the northeastern
USA) or differences in the history of large-scale agricultural
land use between the regions (thousands of years in Europe,
hundreds in the northeastern USA). Alternatively, the shift
in response curves may represent fundamentally different
landscape optima for supporting ecosystem services compared
to biodiversity with local management practices. One mechan-
ism that may lead to this shift is that flowering crop borders
continue to support increased abundance of functionally
important species past the optima at which species richness
is maximized. Delivery of ecosystem services has been shown
to be driven by the abundance of functionally important
taxa rather than community diversity [50,51]. Indeed, the
effectiveness of supplementing floral resources for enhancing
parasitism rates in California vineyards was greatest when
landscapes contained 20–60% natural habitat [52], supporting
the idea that a higher threshold of natural habitat is required for
benefits to ecosystem services. These results imply that policies
attempting to prioritize areas for either conservation or ecosys-
tem services management need to be tailored, as the response
curves may differ.
For pest pressure, the shape of the relationship between
landscape and effectiveness of flowering crop borders is also
predicted by the intermediate landscape hypothesis, yet the
curve is shifted strongly above zero. This shift indicates a cost
of wildflower borders not predicted by the intermediate land-
scape hypothesis. In landscapes with both the least and
greatest natural habitat cover, plots with a wildflower border
had greater pest abundances than those with a control border.
Although flowering borders are intended to target beneficial
insects, generalist pests like L. lineolaris are also able to take
advantage ofthese additional resources [20,24,26]. We observed
greater numbers of L. lineolaris in wildflower borders in moder-
ately agricultural landscapes compared with more complex
landscapes. This result likely reflects the lower propensity for
L. lineolaris to disperse in agriculturally dominated landscapes
[53] and may lead to increased spillover of pests from the
wildflowers to the crop in the following spring.
–0.8
0.2
1.2
2.2
effect on fruit damage
–0.7
–0.6
–0.5
–0.4
–0.3
–0.2
–0.1
0
0.1
0.2
0.3
0.2 0.3 0.4 0.5 0.6
p
ro
p
ortion natural habitat
0.2 0.3 0.4 0.5 0.6
p
ro
p
ortion natural habitat
effect on fruit weight
year
2013
2014
2015
(a)(b)
Figure 4. Effectiveness of wildflower (WF) borders relative to control (C) plots ((WF 2C)/C) for (a) the malformations to and (b) the weight of strawberry fruits.
Shaded areas represent 95% confidence intervals.
rspb.royalsocietypublishing.org Proc. R. Soc. B 285: 20181102
6
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
The relationship between landscape and effect ofwildflower
borders on parasitism was the opposite of our predictions based
on the intermediate landscape hypotheses. Rather, wildflower
plots with the greatest increases in parasitism relative to con-
trols were in the same landscape contexts that also had the
greatest increases in pest abundances. In this case, parasitoid
responses to wildflower borders may be obscured by density-
dependent responses to host abundance [54]. However, other
studies have found positive effects of wildflower borders on
biological control of pests [20,55], particularly when the pest
was not able to use the flowering crop border as alternative
hosts. Effects of wildflower strips on parasitism rates may
also have lagged behind effects on herbivores as L. lineolaris
had the greatest increase in plots with a wildflower border in
2014 while parasitism increased most strongly in 2015.
The lag in time between the establishment of wildflower
borders and the response of the beneficial insect community
can influence the cost–benefit ratio for farmers implementing
these borders with the goal of enhancing ecosystem services
[31]. These lags are particularly important for annual crops or
short-term perennial crops like strawberry, which are grown
in the same field for only 2– 5 years. In our study, increases in
bee visitation and parasitism rates occurred in the third year fol-
lowing establishment. Although a number of studies report
responses within the first year following establishment
[19,20,32,56], the majority of these studies report on commu-
nities within wildflower plantings rather than in adjacent crop
habitats [19,32] while others use annual plants in their borders
[56]. These results suggest that other border types may be more
appropriate for annual or rotating crops or that growers should
establish flowering borders before the crop.
Ultimately, the benefits of ecological intensification prac-
tices like flowering crop borders can be measured in terms
of increases in crop yields. Yet, while many studies evaluate
the effects of wildflower borders on bee visitation or natural
enemy communities, few assess the impact on crop damage
and the final effect on yield (but see [25,31]). Regardless of
crop border treatment, natural habitat cover had a generally
positive effect on fruit weight. When comparing border treat-
ments using the effectiveness index, wildflower borders
reduced fruit damage and increased yield most strongly in
landscapes with intermediate natural habitat cover, showing
that ecological intensification practices can successfully
improve crop productivity. However, in landscapes with
either high or low natural habitat cover, wildflower borders
tended to increase fruit damage and reduce yield. In these
same landscapes, wildflower borders had little effect on bee
visitation and increased pest abundances, suggesting that
the success of ecological intensification practices will be
dependent on the context in which they are implemented.
Although flowering crop borders had positive yield effects
at intermediate levels of landscape complexity, our study indi-
cates that wildflower border management is not without costs
imposed by increased herbivore pressure when implemented
outside of the optimal landscape window. Yet, increases in her-
bivore pressure were only observed in landscapes where
wildflower borders had the least success in improving bee
visitation. In all landscape contexts, efforts should focus on
selecting wildflower species that are not preferred by crop
pests [21,57] and on managing weedy species that support
high numbers of crop pests. Management practices that
reduce these weedy species canalso increase the establishment
rates of planted species [58]. In simple landscapes where wild-
flower borders have few benefits, efforts should focus on the
conservation of the remaining natural habitat and restoration
of larger areas of natural habitat rather than on field-scale
diversification strategies.
Because of the importance of landscape in mediating the
success of ecological intensification practices like wildflower
crop borders, we propose that landscape context should be
explicitly considered in large policy initiatives that subsidize
the creation of flowering habitats on farmlands. Ranking cri-
teria that incorporate landscape along with other site-level
criteria including slope, proximity to waterways or wetlands,
and other factors will allow land managers and conservation
practitioners to select appropriate conservation measures for
a given site. By implementing these metrics, limited resources
for establishing habitat for beneficial insect conservation can
be targeted to areas where they will have the greatest likeli-
hood for success with the least potential for increasing pest
populations or yield loss in nearby crops.
Data accessibility. Data associated with this manuscript have been depos-
ited in the Dryad Digital Repository: http://dx.doi.org/10.5061/
dryad.425kd01 [59].
Authors’ contributions. H.G., K.P., and G.L. designed experiments. B.D.
provided materials for laboratory assays. H.G. collected data, con-
ducted analyses, and wrote the first draft of the manuscript. All
authors contributed substantially to revisions.
Competing interests. We declare we have no competing interests.
Funding. This work was supported in part by a Northeast SARE
graduate student grant to H.G. (GNE12-036).
Acknowledgements. The authors thank David Kleijn and Matthias
Albrecht for helpful comments on an earlier version of the manu-
script. Thanks to past undergraduate field assistants and to Ellie
McCabe, Alison Wentworth, and Steve Hesler for their assistance
in the field.
References
1. Foley JA et al. 2011 Solutions for a cultivated
planet. Nature 478, 337– 342. (doi:10.1038/
nature10452)
2. Newbold T et al. 2015 Global effects of land use on
local terrestrial biodiversity. Nature 520, 45–50.
(doi:10.1038/nature14324)
3. Kremen C, Williams NM, Thorp RW. 2002 Crop
pollination from native bees at risk from agricultural
intensification. Proc. Natl Acad. Sci. USA 99,
16 812–16 816. (doi:10.1073/pnas.262413599)
4. Thies C, Tscharntke T. 1999 Landscape structure and
biological control in agroecosystems. Science 285,
893–895. (doi:10.1126/science.285.5429.893)
5. Rusch A et al. 2016 Agricultural landscape
simplification reduces natural pest control: a
quantitative synthesis. Agric. Ecosyst. Environ. 221,
198–204. (doi:10.1016/j.agee.2016.01.039)
6. Cardinale BJ et al. 2012 Biodiversity loss and its
impact on humanity. Nature 489, 326. (doi:10.
1038/nature11373)
7. Hooper DU et al. 2012 A global synthesis reveals
biodiversity loss as a major driver of ecosystem
change. Nature 486, 105– 108. (doi:10.1038/
nature11118)
8. Schellhorn NA, Gagic V, Bommarco R. 2015 Time
will tell: resource continuity bolsters ecosystem
services. Trends Ecol. Evol. 30, 524– 530. (doi:10.
1016/j.tree.2015.06.007)
9. Tscharntke T et al. 2012 Landscape moderation of
biodiversity patterns and processes—eight
rspb.royalsocietypublishing.org Proc. R. Soc. B 285: 20181102
7
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
hypotheses. Biol. Rev. 87, 661– 685. (doi:10.1111/j.
1469-185X.2011.00216.x)
10. Bommarco R, Kleijn D, Potts SG. 2013 Ecological
intensification: harnessing ecosystem services for
food security. Trends Ecol. Evol. 28, 230– 238.
(doi:10.1016/j.tree.2012.10.012)
11. Pywell RF, Heard MS, Woodcock BA, Hinsley S,
Ridding L, Nowakowski M, Bullock JM. 2015
Wildlife-friendly farming increases crop yield:
evidence for ecological intensification. Proc. R. Soc.
B282, 20151740. (doi:10.1098/rspb.2015.1740)
12. Tscharntke T, Klein AM, Kruess A, Steffan-Dewenter
I, Thies C. 2005 Landscape perspectives on
agricultural intensification and biodiversity—
ecosystem service management. Ecol. Lett. 8,
857–874. (doi:10.1111/j.1461-0248.2005.00782.x)
13. Isaacs R, Tuell J, Fiedler A, Gardiner M, Landis D.
2009 Maximizing arthropod-mediated ecosystem
services in agricultural landscapes: the role of native
plants. Front. Ecol. Environ. 7, 196– 203. (doi:10.
1890/080035)
14. Kleijn D, Rundlo
¨f M, Scheper J, Smith HG,
Tscharntke T. 2011 Does conservation on farmland
contribute to halting the biodiversity decline?
Trends Ecol. Evol. 26, 474– 481. (doi:10.1016/j.tree.
2011.05.009)
15. Concepcio
´n ED, Dı
´az M, Baquero RA. 2008 Effects of
landscape complexity on the ecological effectiveness
of agri-environment schemes. Landsc. Ecol. 23,
135–148. (doi:10.1007/s10980-007-9150-2)
16. Concepcio
´nEDet al. 2012 Interactive effects of
landscape context constrain the effectiveness of
local agri-environmental management. J. Appl. Ecol.
49, 695–705.
17. Batary P, Baldi A, Kleijn D, Tscharntke T. 2011
Landscape-moderated biodiversity effects of agri-
environmental management: a meta-analysis.
Proc. R. Soc. B 278, 1894– 1902. (doi:10.1098/rspb.
2010.1923)
18. Scheper J, Holzschuh A, Kuussaari M, Potts SG,
Rundlo
¨f M, Smith HG, Kleijn D. 2013 Environmental
factors driving the effectiveness of European agri-
environmental measures in mitigating pollinator
loss—a meta-analysis. Ecol. Lett. 16, 912–920.
(doi:10.1111/ele.12128)
19. Scheper J et al. 2015 Local and landscape-level
floral resources explain effects of wildflower strips
on wild bees across four European countries. J. Appl.
Ecol. 52, 1165–1175. (doi:10.1111/1365-2664.
12479)
20. Blaauw BR, Isaacs R. 2015 Wildflower plantings
enhance the abundance of natural enemies and
their services in adjacent blueberry fields. Biol.
Control. 91, 94– 103. (doi:10.1016/j.biocontrol.
2015.08.003)
21. Tschumi M, Albrecht M, Collatz J, Dubsky V, Entling
MH, Najar-Rodriguez AJ, Jacot K. 2016 Tailored
flower strips promote natural enemy biodiversity
and pest control in potato crops. J. Appl. Ecol. 53,
1169–1176. (doi:10.1111/1365-2664.12653)
22. Otieno M, Woodcock BA, Wilby A, Vogiatzakis IN,
Mauchline AL, Gikungu MW, Potts SG. 2011 Local
management and landscape drivers of pollination
and biological control services in a Kenyan agro-
ecosystem. Biol. Conserv. 144, 2424– 2431. (doi:10.
1016/j.biocon.2011.06.013)
23. Balzan MV, Bocci G, Moonen AC. 2016 Utilisation of
plant functional diversity in wildflower strips for the
delivery of multiple agroecosystem services.
Entomol. Exp. Appl. 158, 304–319. (doi:10.1111/
eea.12403)
24. Mccabe E, Loeb G, Grab H. 2017 Responses of crop
pests and natural enemies to wildflower borders
depends on functional group. Insects 8,1–8.
(doi:10.3390/insects8030073)
25. Sutter L, Albrecht M, Jeanneret P. 2017 Landscape
greening and local creation of wildflower strips and
hedgerows promote multiple ecosystem services.
J. Appl. Ecol. 55, 612 –620. (doi:10.1111/1365-
2664.12977)
26. Balzan MV, Bocci G, Moonen AC. 2014 Augmenting
flower trait diversity in wildflower strips to optimise
the conservation of arthropod functional groups for
multiple agroecosystem services. J. Insect Conserv.
18, 713–728. (doi:10.1007/s10841-014-9680-2)
27. Midega CAO, Jonsson M, Khan ZR, Ekbom B. 2014
Effects of landscape complexity and habitat
management on stemborer colonization, parasitism
and damage to maize. Agric. Ecosyst. Environ. 188,
289–293. (doi:10.1016/j.agee.2014.02.028)
28. Tscharntke T et al. 2016 When natural habitat fails
to enhance biological pest control—five
hypotheses. Biol. Conserv. 204, 449–458. (doi:10.
1016/j.biocon.2016.10.001)
29. Connelly H, Poveda K, Loeb G. 2015 Landscape
simplification decreases wild bee pollination services
to strawberry. Agric. Ecosyst. Environ. 211, 51– 56.
(doi:10.1016/j.agee.2015.05.004)
30. Tuell JK, Fiedler AK, Landis D, Isaacs R, Tuell JK,
Fiedler AK, Landis D. 2008 Visitation by wild and
managed bees (Hymenoptera: Apoidea) to Eastern
U.S. Native plants for use in conservation programs.
Environ. Entomol. 37, 707–718. (doi:10.1603/0046-
225X(2008)37[707:VBWAMB]2.0.CO;2)
31. Blaauw BR, Isaacs R. 2014 Flower plantings increase
wild bee abundance and the pollination services
provided to a pollination-dependent crop. J. Appl.
Ecol. 51, 890–898. (doi:10.1111/1365-2664.12257)
32. Williams NM et al. 2015 Native wildflower plantings
support wild bee abundance and diversity in
agricultural landscapes across the United States.
Ecol. Appl. 25, 2119–2131. (doi:10.1890/14-
1748.1)
33. USDA-NASS. 2014 2013 USDA National Agricultural
Statistics Service Cropland Data Layer.
34. Renauld M, Hutchinson A, Loeb G, Poveda K,
Connelly H. 2016 Landscape simplification constrains
adult size in a native ground-nesting bee. PLoS ONE
11, e0150946.
35. Grab H, Danforth B, Poveda K, Loeb G. 2018
Landscape simplification reduces classical biological
control and crop yield. Ecol. Appl. 28, 348– 355.
(doi:10.1002/eap.1651)
36. Day WH. 1996 Evaluation of biological control of the
tarnished plant bug (Hemiptera: Miridae) in alfalfa
by the introduced parasite Peristenus digoneutis
(Hymenoptera: Braconidae). Environ. Entomol. 25,
512–518. (doi:10.1093/ee/25.2.512)
37. Tilmon KJ, Danforth BN, Day WH, Hoffmann MP.
2000 Determining parasitoid species composition in
a host population: a molecular approach.
Ann. Entomol. Soc. Am. 93, 640 –647.
(doi:10.1603/0013-8746(2000)093[0640:DPSCIA]2.
0.CO;2)
38. Ashfaq M, Braun L, Hegedus D, Erlandson M. 2004
Estimating parasitism levels in Lygus spp.
(Hemiptera: Miridae) field populations using
standard and molecular techniques. Biocontrol. Sci.
Technol. 14, 731– 735. (doi:10.1080/
09583150410001683592)
39. Tilmon KJ, Hoffmann MP. 2003 Biological control of
Lygus lineolaris by Peristenus spp. in strawberry.
Biol. Control. 26, 287– 292. (doi:10.1016/S1049-
9644(02)00159-7)
40. Gariepy TD, Broadbent AB, Ethier S, Kuhlmann U,
Gillott C, Erlandson M. 2008 Detection of European
Peristenus digoneutis loan in mirid populations in
southern Ontario. Biocontrol. Sci. Technol. 18,
583–590. (doi:10.1080/09583150802100197)
41. Webb RA, Purves JV, White BA. 1974 The
components of fruit size in strawberry. Sci. Hortic.
(Amsterdam) 2, 165– 174. (doi:10.1016/0304-
4238(74)90006-5)
42. Chagnon M, Gingras J, de Oliveira D. 1989 Effect of
honey bee (Hymenoptera: Apidae) visits on the
pollination rate of strawberries. J. Econ. Entomol.
82, 1350–1353. (doi:10.1093/jee/82.5.1350)
43. Schaefers GA. 1980 Yield effects of tarnished plant
bug feeding on June-bearing strawberry varieties in
New York State. J. Econ. Entomol. 73, 721– 725.
(doi:10.1093/jee/73.5.721)
44. Bates D, Ma
¨chler M, Bolker B, Walker S. 2014 Fitting
linear mixed-effects models using lme4. J. Stat.
Softw.,eprint arXiv:1406.5823 67, 51. (doi:10.
18637/jss.v067.i01)
45. Kuznetsova A, Brockhoff PB, Christensen RHB. 2016
lmerTest: tests in linear mixed effects models.
46. Dicks LV et al. 2016 Ten policies for pollinators.
Science 354, 975–976. (doi:10.1126/science.
aai9226)
47. Fiedler AK, Landis DA, Wratten SD. 2008
Maximizing ecosystem services from conservation
biological control: the role of habitat management.
Biol. Control 45, 254– 271. (doi:10.1016/j.
biocontrol.2007.12.009)
48. Wratten SD, Gillespie M, Decourtye A, Mader E,
Desneux N. 2012 Pollinator habitat enhancement:
benefits to other ecosystem services. Agric. Ecosyst.
Environ. 159, 112– 122. (doi:10.1016/j.agee.2012.
06.020)
49. USDA-FSA. 2017 Conservation Reserve Program
Monthly Summary—June 2017.
50. Kleijn D et al. 2015 Delivery of crop pollination
services is an insufficient argument for wild
pollinator conservation. Nat. Commun. 6, 7414.
(doi:10.1038/ncomms8414)
51. Winfree R, Fox JW, Williams NM, Reilly JR, Cariveau
DP. 2015 Abundance of common species, not
species richness, drives delivery of a real-world
rspb.royalsocietypublishing.org Proc. R. Soc. B 285: 20181102
8
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
ecosystem service. Ecol. Lett. 18, 626– 635. (doi:10.
1111/ele.12424)
52. Wilson H, Miles AF, Daane KM, Altieri MA. 2017
Landscape diversity and crop vigor outweigh influence
of local diversification on biologicalcontrol of avineyard
pest. Ecosphere 8, e01736. (doi:10.1002/ecs2.1736)
53. Mazzi D, Dorn S. 2012 Movement of insect
pests in agricultural landscapes. Ann. Appl. Biol. 160,
97– 113. (doi:10.1111/j.1744-7348.2012.00533.x)
54. May RM, Hassell MP, Anderson RM, Tonkyn DW.
1981 Density dependence in host-parasitoid models.
J. Anim. Ecol. 50, 532 –543.
55. Jonsson M, Straub CS, Didham RK, Buckley HL, Case
BS, Hale RJ, Gratton C, Wratten SD. 2015
Experimental evidence that the effectiveness of
conservation biological control depends on
landscape complexity. J. Appl. Ecol. 52, 1274– 1282.
(doi:10.1111/1365-2664.12489)
56. Feltham H, Park K, Minderman J, Goulson D. 2015
Experimental evidence that wildflower strips
increase pollinator visits to crops. Ecol. Evol. 5,
3523–3530. (doi:10.1002/ece3.1444)
57. Tschumi M, Albrecht M, Entling MH, Jacot K. 2015 High
effectiveness of tailored flower strips in reducing pests
and crop plant damage. Proc. R. Soc. B 282, 20151369.
(doi:10.1098/rspb.2015.1369)
58. Benvenuti S, Bretzel F. 2017 Agro-biodiversity
restoration using wildflowers: what is the
appropriate weed management for their long-term
sustainability? Ecol. Eng. 102, 519–526. (doi:10.
1016/j.ecoleng.2017.02.062)
59. Grab H, Poveda K, Danforth B, Loeb G. 2018 Data
from: Landscape context shifts the balance of costs
and benefits from wildflower borders on multiple
ecosystem services Dryad Digital Repository (doi:10.
5061/dryad.425kd01)
rspb.royalsocietypublishing.org Proc. R. Soc. B 285: 20181102
9
on August 2, 2018http://rspb.royalsocietypublishing.org/Downloaded from
- A preview of this full-text is provided by The Royal Society.
- Learn more
Preview content only
Content available from Proceedings of the Royal Society B
This content is subject to copyright.
