Content uploaded by Ingo Grass
Author content
All content in this area was uploaded by Ingo Grass on May 29, 2024
Content may be subject to copyright.
Flower–bee versus pollen–bee
metanetworks in fragmented landscapes
Felipe Librán-Embid1,3, Ingo Grass4,5, Carine Emer6, Viviana Alarcón-Segura1,7,
Hermann Behling2, Siria Biagioni2, Cristina Ganuza8, Celina Herrera-Krings1,
Christina Ani Setyaningsih2 and Teja Tscharntke1
1Agroecology, and 2Department of Palynology and Climate Dynamics, Albrecht‐von‐Haller‐Institute for Plant
Sciences, University of Göttingen, Göttingen 37077, Germany
3Justus Liebig University of Gießen, Institute of Animal Ecology and Systematics, Heinrich-Buff-Ring 26, Gießen
35390, Germany
4Department of Ecology of Tropical Agricultural Systems, and 5Center for Biodiversity and Integrative Taxonomy
(KomBioTa), University of Hohenheim, Stuttgart 70599, Germany
6Instituto de Pesquisas Jardim Botânico do Rio de Janeiro, Rua Pacheco Leão, 915, Jardim Botânico, Rio de Janeiro
CEP22460-030, Brazil
7Animal Ecology, Department of Biology, University of Marburg, Marburg 35037, Germany
8Department of Animal Ecology and Tropical Biology, University of Würzburg, Würzburg 97074, Germany
Understanding the organization of mutualistic networks at multiple spatial
scales is key to ensure biological conservation and functionality in human-
modified ecosystems. Yet, how changing habitat and landscape features
affect pollen–bee interaction networks is still poorly understood. Here, we
analysed how bee–flower visitation and bee–pollen-transport interactions
respond to habitat fragmentation at the local network and regional
metanetwork scales, combining data from 29 fragments of calcareous
grasslands, an endangered biodiversity hotspot in central Europe. We
found that only 37% of the total unique pairwise species interactions
occurred in both pollen-transport and flower visitation networks, whereas
28% and 35% were exclusive to pollen-transport and flower visitation
networks, respectively. At local level, network specialization was higher
in pollen-transport networks, and was negatively related to the diversity
of land cover types in both network types. At metanetwork level, pollen
transport data revealed that the proportion of single-fragment interactions
increased with landscape diversity. Our results show that the specialization
of calcareous grasslands’ plant–pollinator networks decreases with
landscape diversity, but network specialization is underestimated when
only based on flower visitation information. Pollen transport data, more
than flower visitation, and multi-scale analyses of metanetworks are
fundamental for understanding plant–pollinator interactions in human-
dominated landscapes.
1. Introduction
Plant–pollinator interactions are the ecological foundations of animal-medi-
ated pollination, a process on which 88% of flowering plants depend [1,2].
The structural analysis of pollination interaction networks may provide key
information on network stability and robustness under environmental change
[3–5]. Landscape-scale simplifications by habitat loss and habitat fragmenta-
tion have strong negative impacts on the characteristics of plant–pollinator
network structure and functionality [3,6,7], directly through non-random loss
of interactions [8] and indirectly through changes in species richness and
abundance [9,10].
Land-use consequences on ecological networks must be explored at local
and regional spatial scales because of the occurrence of important nonlinear
© 2024 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution
License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original
author and source are credited.
Research
Cite this article: Librán-Embid F, Grass I, Emer C,
Alarcón-Segura V, Behling H, Biagioni S, Ganuza C,
Herrera-Krings C, Setyaningsih CA, Tscharntke T.
2024 Flower–bee versus pollen–bee
metanetworks in fragmented landscapes. Proc. R.
Soc. B 291: 20232604.
https://doi.org/10.1098/rspb.2023.2604
Received: 4 May 2023
Accepted: 15 April 2024
Subject Category:
Ecology
Subject Areas:
ecology
Keywords:
plant–pollinator interactions, calcareous
grasslands, habitat fragmentation, pollen-
transport networks, landscape ecology, life-
history traits
Author for correspondence:
Felipe Librán-Embid
e-mail: felipe.libran-embid@allzool.bio.uni-
giessen.de
Electronic supplementary material is available
online at https://doi.org/10.6084/
m9.figshare.c.7204022
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
emergent properties that demand the use of methods that can deal with multiple levels of ecological complexity [3,11,12].
Metanetworks (i.e. a group of scattered local networks connected by shared interactions) are an emerging approach to study the
consequences of habitat fragmentation on ecological networks at a regional scale [10,12,13].
Plant–pollinator networks have been traditionally constructed using data on flower visitation [14]. However, in order for
successful pollination to occur, viable pollen grains need to be transported from the anthers of a flowering plant to a receptive
stigma of a conspecific [15]. Therefore, the sole flower visitation of an animal is expected to be a poor predictor of its capacity
as a pollinator [16]. For instance, many flower visitors forage exclusively for nectar and do not contact flower anthers; other
species lack morphological traits to carry pollen and thus cannot act as pollinators [17,18]. Two methods have been proposed
to overcome this challenge. First, stigmas and styles can be analysed to identify pollen deposition after an animal visit [19,20].
However, pollen deposition analyses are extremely time-consuming and consequently prohibitive for landscape-scale studies.
Alternatively, pollen loads analyses of flower visitors also provide valuable information regarding an animal’s capacity as a
pollinator and are suitable for large-scale studies given their relative simplicity [21,22]. Pollen-transport networks have been
studied at singular sites and local scales [23–26] and pollen metanetworks across land-use types [27], but, to our knowledge,
there is no study analysing pollen-transport networks and metanetworks over gradients of habitat size, isolation and landscape
diversity.
To compare the structure of different types of networks, such as flower visitation and pollen-transport interaction net-
works—between each other and across environmental gradients—it is essential to quantify network specialization at the
community level [28]. This can be done through qualitative and quantitative indices such as network connectance and Blüth-
gen’s 'H2' [28]. The specialization of pollination networks can be higher than that of visitation networks [23,29] given that the
pollen richness on the bodies of flower visitors is usually a subset of the flowers they visit [30]. However, pollen analyses can
also reveal interactions established with rare and unfrequently visited plant species that could lead to higher pollen-transport
network connectance [22,31] and lower network specialization [32,33]. The specialization of mutualistic networks is affected
by habitat loss and isolation through species turnover [34] and through a shift towards a higher prevalence of opportunistic
interactions among generalists [35].
Species traits can determine the probability that a flower visit is accompanied by pollen transport and therefore that a certain
interaction gets recorded with flower visitation observations, pollen load analyses or both. For example, given that bumblebees
are social, usually hairier, larger and more numerous than other bees, they are generally expected to have a larger carrying
capacity for pollen transport than most solitary bees in Europe [36]. Therefore, interactions established by bumblebees should
have a higher probability of occurrence in both visitation and pollen-transport networks than those established by other bees.
Furthermore, habitat specialist plants, rather than habitat generalists, often have adaptations to maximize visitation and amount
of pollen transferred to pollinators per visit, such as larger floral displays (i.e. increased attractiveness [37]). Hence, habitat
specialist plants should establish interactions with a higher probability of occurrence in both types of networks.
Mutualistic metanetworks in fragmented landscapes are characterized by a large majority of interactions that are unique to
single fragments [11,12]. In a mutualistic metanetwork (sensu [12]), single-fragment interactions are those plant–bee interactions
that are recorded in only one of the natural or seminatural studied habitat fragments. Based on flower visitation data of
bees and butterflies in calcareous grasslands, Librán-Embid et al. [11] found that landscape diversity had a positive effect on
the richness of single-fragment interactions and a negative effect on the proportion of single-fragment interactions. However,
whether this effect is also captured with bee pollen loads analyses is still unknown.
Here, we tested how habitat fragmentation affects plant–pollinator interactions at four distinct but complementary levels of
biological organization, i.e. (i) flower visitation and (ii) pollen-transport interactions at the (iii) local network and (iv) regional
metanetwork scales.
We studied a gradient of habitat change of European calcareous grasslands, a highly threatened biodiversity hotspot
characterized by a vast number of rare and endangered species [38]. Specifically, we asked (i) at the local scale, how landscape
characteristics (fragment area, connectivity and landscape diversity of cover types) would shape the structure (connectance
and specialization) of visitation and pollen-transport networks, and (ii) at the regional scale, which functional traits (i.e. bee
body size, bee group, bee habitat specialization, flower size and plant habitat specialization) are associated with the probability
that a flower visit involves pollen transport and how are single-fragment interactions affected by fragmentation and landscape
diversity in both metanetworks.
We hypothesized that (i) the specialization of light-microscopy local pollen-transport networks will be higher than that of
local visitation networks and their connectance lower. This is expected if the flower visitors’ morphological and behavioural
constraints to pollen transport prevail over the emergence of interactions established with rare and unfrequently visited plant
species; (ii) network specialization will decrease in larger and less isolated calcareous grasslands that are surrounded by more
diverse land cover types, as the presence of more species should increase the probability of multiple interacting partners; (iii)
owing to rare and ineffective interactions (i.e. interactions that contribute little to the plant reproductive success (sensu [15])), a
high number of interactions unique to the pollen-transport and visitation metanetworks, respectively, is expected; (iv) interac-
tion occurrence in both network types simultaneously depends on bee and flower size, bee and plant habitat specialization and
bee group [11]; and finally, (v) the richness and proportion of single-fragment interactions increases with landscape diversity
and fragment area and decreases with fragment connectivity in both types of networks given the central-place foraging
behaviour of bees.
2
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
2. Methods
(a) Study system
Calcareous grasslands in central Europe mainly result from low-intensity grazing activities associated with human livestock
since the Anthropocene and are therefore considered seminatural habitats [38]. In the absence of extensive grazing, bush
encroachment leads to the degradation of this formerly common and extended habitat [38]. In recent decades, agricultural
intensification and land-use change have largely determined the reduction and fragmentation of this biodiverse habitat that
harbours the highest richness of vascular plants, butterflies and grasshoppers in central Europe [39–42]. Due to these charac-
teristics, calcareous grasslands are considered core habitats and conservation priorities in Europe and are therefore legally
protected in the European Union [43,44].
(b) Study area
Data were collected from April until August 2018 on 29 calcareous grasslands in the surroundings of the city of Göttingen,
Germany (electronic supplementary material, figure S1). These grasslands were selected in a previous study [45], from a larger
regional pool (~300 fragments), to vary along independent gradients of size and isolation from other calcareous grasslands.
Arable land and European beech (Fagus sylvatica) forests are the two main land use types in the region, with 31% and 38% land
cover, respectively [45].
(c) Flower visitation interactions
We performed three rounds of sampling throughout the season in each calcareous grassland to capture the succession of flower
visitors (hereafter, pollinators) and wildflower species. Seven fixed-point observation plots of 10 min were established at each
site. We followed a protocol established by van Swaay et al. [46] to carry out our surveys. We collected data from 9.00 to 17.00
on days with a minimum temperature of 15°C and at least 50% clear sky, or with a minimum temperature of 18°C in any sky
condition. To avoid any confounding effect of daytime, sites were surveyed at different times of the day.
Our observational plots were established in flower-rich areas and were circular (3 m radius, 28.3 m2). Within these, all
interactions between bees (Hymenoptera: Apiformes) and flowering plants were recorded. We focused on bees because they are
(jointly with lepidopterans) the most abundant pollinators in grasslands [47] and because they carry significantly larger pollen
loads than butterflies [23]. A bee visit was considered an interaction once the insect contacted the plant’s reproductive organs.
Bees that were not easily recognizable from a distance were sweep netted or collected for later identification by taxonomists
and the timer was stopped while handling insects. We excluded interactions involving Apis mellifera because the presence of this
species in our region relates exclusively to the occurrence of bee keepers in the vicinity [11]. Apis mellifera interactions accounted
for 334 from a total of 1499 interactions registered and were present in all sites (range 1–75 A. mellifera interactions per site).
Bees were classified as solitary bees or bumblebees (hereafter, bee group). All bumblebees are eusocial and belong to the genus
Bombus spp. Within the group of ‘solitary bees,’ seven species presented some degree of sociality but were grouped within the
solitary bees because of the morphological and genetic similarities with these. The seven species are Andrena scotica (communal),
Halictus confusus, Halictus rubicundus, Halictus tumulorum, Lasioglossum calceatum, Lasioglossum morio and Lasioglossum pauxillum.
(d) Pollen-transport interactions
Pollen was taken from bees’ bodies, head and antennae by bathing bees in Eppendorf tubes filled with distilled water, using
a modified protocol from Dafni [48]. As some interactions were very abundant, we collected pollen from all bees that visited
flowers in our observation plots, with an upper limit of collecting six pollen samples from the same interaction in each site and
sampling round following Zhao et al. [25]. Pollen baskets were considered following de Manincor et al. [49], who found that
there is no significant difference in the number of observed links between analyses based on pollen passively transported on the
body and that collected in specialized structures such as the corbiculae. Samples were later acetolysed [50] using a protocol lab
technique and analysed using light microscopy at 40× magnification (one slide represented one bee). We also created a reference
collection of pollen from the flowering plants of the region to aid sample pollen identification. We did not consider slides
with fewer than 30 pollen grains. From all the others, we counted 200 pollen grains on each slide, except for five slides that
had 50–200 pollen grains. To ensure that pollen diversity was captured, the slides were scanned systematically in consecutive
horizontal lines starting from the left upper corner of each slide and up to the count of 200 pollen grains. Following Bosch et
al. [22], we considered the presence of at least 10 pollen grains in our samples as proof of true visitation to the corresponding
flowering species (i.e. threshold for pollen–bee interactions).
(e) Plant–pollinator traits
Plants and pollinators were classified according to their habitat specialization, following Poschlod et al. [51] and Brückmann et
al. [52] for plants and Jauker et al. [53] and Hopfenmüller et al. [54] for bees. All body length values for bees were taken from
Westrich [55]. We consider Cirsium sp. (cluster of four species mostly represented by the habitat specialist Cirsium acaule) and
3
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
Ononis sp. (cluster of two hybridizing species including the specialist Ononis repens) as habitat specialists (i.e. species mostly
restricted to available calcareous grassland habitat fragments).
(f) Landscape metrics
We tested the effects of fragment size, fragment connectivity and landscape diversity of cover types on the structure of local
fragment networks in terms of specialization (i.e. network connectance and H2') and also on the richness and proportion of
single-fragment interactions (i.e. interactions that were only recorded in a single fragment). Fragment area was calculated with
ArcGis 10.5.1 [56] and ranged from 82 to 52 557 m², excluding zones dominated by shrubs. Fragment spatial connectivity and
the Shannon diversity of land cover types (as a measure of landscape diversity) were calculated using the ‘landscapemetrics’
package [57]. For fragments’ spatial connectivity, we used a connectivity index developed by Hanski et al. [58] and considered
all calcareous grasslands in a radius of 2 km around the study grasslands (see electronic supplementary material for details).
Larger values of this index indicate higher spatial connectivity (electronic supplementary material, table S1). The mapped cover
types were: oilseed rape, grainfield, maize, other crops, forest open, forest closed, field margin, hedgerow, pasture, calcareous
grassland, orchard, settlements, water bodies, streets, grass-roads and bare soil (see figure S1 in the electronic supplementary
material). The basemap for habitat classification was provided by ‘the rural development and agricultural promotion service
centre’ (Servicezentrum Landentwicklung und Agrarförderung). Shapefiles of land use were constructed using ArcGis 10.5.1
and all statistics were performed in R 4.1.0 [59].
(g) Network and statistical analysis
Our study involved two levels of biological organization: (i) local scale, in which each of the 29 calcareous grasslands corre-
sponds to a local network of both flower-visitation and pollen transport interactions, totalling 58 networks; and (ii) regional
scale, in which we scaled-up from local fragment networks to regional metanetworks of both flower-visitation and pollen–trans-
port interactions. Below, we describe how we analysed that complexity in the light of our hypotheses.
We constructed local quantitative bipartite networks (one for each calcareous grassland fragment) and regional metanet-
works using data on flower visitation (hereafter, visitation networks) and pollen loads (hereafter, pollen-transport networks),
respectively. Local networks were constructed as Aij adjacency matrices in which i are the plant species, j the pollinator species
and the aij element represents the frequency of interactions between i and j. At the landscape level, metanetworks were built by
pooling the 29 calcareous grasslands into Akl adjacency matrices in which k are the studied sites and l either the plant visitation
or the pollen-transport interactions. To make visitation and pollen-transport networks comparable we did not consider pollen
from trees (e.g. Picea spp. or Pinus spp.), crops (e.g. Vicia faba), grasses (e.g. Poaceae) or ornamental plants (e.g. Astrantia major)
because observations were done exclusively on herbaceous plants of calcareous grasslands.
(h) Local networks
To test whether pollen-transport networks were more specialized than visitation networks at the local level (hypothesis 1)
and whether they were affected by habitat fragmentation and landscape diversity (hypothesis 2), we calculated network
connectance, defined as ‘the realized proportion of possible links’ and the H2' index, which is based on ‘the deviation of
a species’ realized number of interactions expected from each species’ total number of interactions’ [28,60,61]. We used a
linear mixed model with network type, (log) fragment area, (log) connectivity index and landscape diversity at 350 m as
explanatory variables, and fragment identity as random intercept. To choose the spatial scale surrounding the focal calcareous
grassland habitats at which landscape diversity effects were stronger, we compared models fitted at all scales from 100 to 500
m in 50 m intervals and compared them using the corrected Akaike information criterion (AICc) for small samples. Because
almost all indices of network structure are at least partially affected by network size, we standardized our response variables,
network connectance (more affected by network size) and H2' (less affected by network size) relative to a null model to allow
for meaningful comparisons among networks of different fragments [61,62]. We followed Grass et al. [3] by creating null
distributions based on 1000 replicates of Patefield’s algorithm. All network metrics were calculated using the ‘bipartite’ package
[63] and model diagnostics was done using the ‘DHARMa’ package [64].
(i) Metanetworks
We modelled the simultaneous presence of interactions in both metanetworks (hypothesis 4, i.e. flower visitation interactions
resulting in pollen transport) using a generalized linear mixed model with binomial distribution and pollinator and plant
species identity as crossed random intercepts. Our explanatory variables in the full additive model above were the plant and
pollinator habitat specialization, flower size (area), pollinator size and group (i.e. bumblebee or solitary bee). Finally, to study
the effects of landscape diversity and habitat fragmentation (i.e. fragment area and connectivity) on the richness and proportion
of single-fragment interactions (hypothesis 5), we used generalized linear models with negative binomial distribution and linear
models (i.e. normal distribution), respectively. The explanatory variables tested in the full models were the Shannon diversity
index of land cover types at 150 m (for single-fragment interaction richness) and 500 m (for the proportion of single-fragment
interactions), (log) fragment area and (log) connectivity index. The minimum adequate models were found with backwards
4
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
model selection using likelihood ratio tests. All non-significant explanatory variables (p > 0.05) were sequentially removed.
Models were created using the ‘lme4’ package [65]. All network and statistical analyses were performed in R 4.1.0 [59] .
3. Results
We observed 1165 flower–bee interaction events among 71 plant species and 67 bee species, resulting in 250 unique pairwise
interactions. Of those, 31 (43.7%) plant species were only visited by bumblebees and 19 (26.8%) plant species were only visited
by solitary bees, while 23 (32.4%) plant species were visited by both, totalling 71 plant species visited (electronic supplementary
material, figure S2a, table S2). Some examples include Fragaria vesca, which was only visited by solitary bees, and Trifolium
pratense, Salvia pratensis, Prunella grandiflora, Carlina vulgaris and Anthyllis vulneraria, which were only visited by bumblebees
(electronic supplementary material, table S2). Furthermore, we analysed pollen samples of 830 bee individuals and found 474
individuals carrying 0–30 pollen grains, 5 carrying 50–200 pollen grains and 351 carrying ≥200 pollen grains. We identified 44
bee species transporting pollen from 64 plant species, resulting in 222 unique pollen–bee pairwise interactions from a total of
626 pollen interaction events. Pollen of 20 (31.3%) plant species was only transported by bumblebees and pollen of 12 (18.8%)
plant species was exclusively transported by solitary bees (electronic supplementary material, figure S2b, table S3), while pollen
of 32 (50%) plant species was transported by both groups. For example, pollen of Knautia arvensis was only transported by
bumblebees and pollen of Potentilla sp. was only transported by solitary bees (electronic supplementary material, table S3).
At the local network level, our results show that pollen-transport networks were significantly more specialized than
visitation networks (F1,27 = 11.33, p = 0.002, figure 1). We also found a negative effect of landscape diversity at the 350 m
scale on H2' specialization of both visitation and pollen-transport networks (F1,26 = 13.56, p = 0.001, figure 1). On the other hand,
connectance did not differ between the visitation and pollen-transport networks (F1,27 = 1.03, p = 0.32) and was also not affected
by landscape diversity (F1,26 = 1.97, p = 0.17). Fragment area and fragment connectivity had no significant effects neither on H2'
network specialization nor connectance (electronic supplementary material, table S4).
At the regional level, we found a total of 345 unique combinations of plant–pollinator interactions considering both visitation
and pollen-transport networks, from which 127 (36.8%) were found in both types (figure 2, electronic supplementary material,
table S5). From a total of 222 unique pairwise interactions detected in the pollen-transport metanetwork, 95 (42.8%) were
exclusive to it (i.e. they were not registered in the visitation metanetwork, electronic supplementary material, table S6) and 123
out of 250 (49.2%) were recorded only in the visitation metanetwork (electronic supplementary material, table S7). Furthermore,
we identified important differences in the number of interactions established by some plant species in both metanetworks
(electronic supplementary material, tables S8 and S9). The most outstanding case was K. arvensis, which was visited by 19
different bees but only four of them (all bumblebees) transported its pollen.
Additionally, we found that the presence of an interaction in both metanetworks (i.e. visitation and pollen transport) was
affected by the plant habitat specialization and the pollinator group. Specifically, interactions involving habitat specialist plants
(χ2 = 6.47, d.f. = 1, p = 0.011) and bumblebees (χ2 = 17.24, d.f. = 1, p = 0.0071) had a higher occurrence in both metanetworks
than those involving habitat generalist plants and solitary bees (electronic supplementary material, figure S3). Flower size, bee
size and bee habitat specialization had no significant effects on the presence of an interaction in both metanetworks (electronic
supplementary material, table S10).
In the visitation metanetwork, we found 76 (30.4%) unique pairwise interactions that occurred in at least two calcareous
grassland fragments but these accounted for 913 (78.4%) interaction events. In the pollen transport metanetwork (figure 3), we
found 70 (31.5%) unique pairwise interactions occurring in at least two fragments but summing up to 452 (72.2%) interaction
0
–1
–2
–3
–4
1.00 1.25 1.50
Landscape diversity (350 m)
Network type
Pollen transport
Visitation
Network specialization (H2′)
1.75 2.00
Figure 1. Relationship between standardized network specialization (h2'), network type and landscape diversity (i.e. Shannon diversity of land cover types). Each
network type includes 28 local networks (fragments) in each dataset (pollen transport and visitation). Bands represent 95% confidence intervals of the model fit.
5
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
events. Finally, we found a significant positive effect of landscape diversity on the number of single-fragment interactions for
both network types (electronic supplementary material, figure S4). However, the spatial scale at which this effect was stronger
differed for the visitation and pollen-transport networks. Specifically, the number of single-fragment interactions increased
with landscape diversity at the 150 m scale for the visitation data (χ2 = 4.59, d.f. = 1, p = 0.032, electronic supplementary
material, figure S4a) and at the 500 m scale for the pollen transport data (χ2 = 5.96, d.f. = 1, p = 0.015, electronic supplementary
material, figure S4b). Moreover, landscape diversity at the 500 m scale significantly increased the proportion of single-fragment
interactions (F1,27 = 5.26, p = 0.030), but this effect was only found for the pollen-transport networks (electronic supplementary
material, figure S5). Fragment area and fragment connectivity had no significant effect on the number of single-fragment
interactions or the proportion of single-fragment interactions (electronic supplementary material, table S11).
4. Discussion
In this study, we analysed multiple levels of ecological complexity of plant–pollinator networks constructed from bee–flower
visitation and bee–pollen-transport interactions across a gradient of habitat fragmentation of a biodiversity hotspot. Of all
interactions found, 63.2% were exclusive to either the visitation or pollen-transport networks, highlighting both the numerous
low-frequency interactions that are not captured by observations of flower visits (27.5%) and the high number of interactions
(35.7%) that do not translate into pollen transport. Pollen-transport networks were more specialized than visitation networks,
and a higher diversity of land cover types in the surroundings of a habitat fragment decreased network specialization.
(a) Network type effects on network specialization
Ecological and methodological aspects can affect the relationship between the specialization of pollen-transport and visitation
networks. Regarding ecology, an important aspect is whether studies consider all flower-visiting taxa [31–33,68] or rather focus
on bees [29,49]. Different from other flower-visiting arthropods, female bees have specialized structures for pollen transport
(i.e. scopae) and actively search for pollen to feed their larvae. As bees carry more pollen than other flower visitors [25,69], the
chance of detecting rare interactions with pollen analyses is higher for bees. These bee traits would favour the higher presence
of transport-only interactions and the lower presence of visitation-only interactions compared to other pollinator taxa. On the
other hand, as central-place foragers, bees are bound to the areas surrounding their nest location, and they therefore can have a
lower probability of carrying pollen from plants not belonging to the local habitat community compared to other flower visitors,
such as butterflies that can move more freely through the landscape. This implies that the pollen that bees carry has a higher
chance of being a subset of the flowers observed in a focal study site than is the case in other taxa. This difference is expected
to be especially important in bee communities with many small- and medium-sized solitary bees, but even larger bees like Apis
spp. and Bombus spp. are known to forage mostly in proximity to their nests when enough flowering resources are available,
which is usually the case in calcareous grasslands [70–72]. In line with our findings, previous bee studies using microscopy
have found higher specialization of pollen-transport networks compared to visitation networks [29], or no significant difference
between them [49], suggesting a prevalence of the later phenomenon compared to the former.
Regarding methodology, several aspects can affect specialization indices. First, if plant species recorded through pollen
but missing in the study sites are not excluded from the analyses, there would be an overrepresentation of interactions in
the pollen-transport networks and an underrepresentation of detected interactions in the visitation networks. This would
increase the probability of finding pollen-transport networks to be more generalized than visitation ones. In addition, the lower
resolution of plant taxa identification by light microscopy or metabarcoding, compared to field observations of flower visits,
Total interactions by method
Unique pairwise interactions
Pollen transportVisitation
Interactions exclusive to the
pollen transport metanetwork
(e.g. Osmia bicolor–Trifolium
repens)
Interactions ocurring in both metanetworks
(e.g. Bombus terrestris–Centaurea scabiosa)
Interactions exclusive to the
visitation metanetwork
(e.g. Bombus terrestris–
Genista tinctoria)
28%37%35%
Figure 2. Bipartite diagram representation of the overlap between the visitation and pollen-transport metanetworks. The rectangles in the upper part represent the
total interactions detected by each method (1165 and 626 total interactions in the visitation and pollen-transport metanetworks, respectively). The rectangles at the
bottom represent the unique flower–bee (sky blue) and pollen–bee (pink) pairwise interactions. The size of the squares and thickness of the bars are proportional to
the frequency of each unique pairwise interaction. Those interactions exclusive to the pollen transport dataset occur to the right (in pink) and those exclusive to the
visitation dataset are shown to the left (in sky blue). Unique pairwise interactions occurring in both datasets are in the middle and highlighted in violet. The percentage
of unique pairwise interactions occurring in each dataset is also indicated.
6
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
needs to be taken into account. This can be done, for example, by lowering the resolution of the plant visitation data from
species to genus level for those species that cannot be discriminated based on pollen morphology or DNA analyses.
The location on the bee's body from which pollen is removed to construct pollen-transport networks can also affect the
results. Previous studies disagree on removing pollen located in the corbiculae from the family Apidae [29,31,33,68], pollen
located in the scopae [23,25], or considering all carried pollen [32,49]. The exclusion of pollen located in specialized bee
structures can considerably reduce the number of detected interactions. Further, most studies based on microscopy report
network specialization to be higher in pollen-transport networks compared to visitation networks [23,25,29,68], while studies
based on metabarcoding report the opposite [32,33,73]. A possible reason for this effect is that metabarcoding results can vary
depending on the plant reference database used, eventually inflating false detections [73]. Light microscopy also has limitations,
including the expertise of the observer and the possibility of lower taxonomic resolution than metabarcoding [74]. Lastly,
the use (or not) of a threshold to distinguish pollen contamination from legitimate pollen transport [25,33,49] also affects the
number of interactions found and can therefore impact specialization estimations. The methodological approach used in our
study was designed to minimize pitfalls related to spurious interactions detection and to maximize the reliability of results.
A higher specialization of pollen-transport networks in calcareous grasslands indicates that these pollination networks
might be more vulnerable to collapse following disturbance because increased specialization can make pollination networks
less robust and more prone to co-extinction cascades [75]. The vast majority of plant–pollinator network studies are based on
visitation data and conclusions regarding biodiversity conservation are derived mostly from them. In light of our results and
previous studies [23,25,29,68], we call for attention to the risk of an overestimation of plant–pollinator networks' stability and
robustness in past studies based solely on flower visitation data. Still, whether the higher specialization of pollen transport with
respect to visitation networks is the rule across different habitats and geographical locations and how much is it affected by
methodological artefacts remain to be investigated.
(b) Landscape diversity effects on network specialization
Plant–pollinator network specialization is affected by species richness and species behaviour [35,75,76]. Habitat fragmentation
and landscape simplification may decrease the availability of interacting partners as a consequence of reduced population sizes
or local extinctions [77]. The absence of interacting partners can have opposite effects on species specialization. On the one
hand, pollinators may visit more plant species to compensate for missing resources, therefore increasing their generalization
[78]. However, in case of low behavioural plasticity or high plant fidelity, specialization could increase after disturbance (i.e.
due to the loss of a plant partner) as pollinators would be unable to establish new interactions. For plants, losing a pollinator
may directly increase plant specialization by reducing the number of interacting partners. Nonetheless, reduced competition
for resources among pollinators could facilitate visitation of opportunistic (and usually less effective) pollinators, therefore
increasing plant generalization [78].
In protected natural and seminatural habitats, mutualistic networks present higher nestedness and specialization than
expected by chance, reducing interspecific competition through niche partitioning and allowing coexistence [8,79,80]. Habitat
fragmentation and extreme climatic events reduce mutualistic network specialization through the loss of specialized and rare
Centaurea scabiosa/
Bombus terrestris
Trifolium pratense/
Bombus pascuorum
Figure 3. Pollen-transport metanetwork structure among calcareous grassland fragments and unique pairwise pollen–bee interactions (n = 29 and n = 222,
respectively). Circles indicate pairwise pollen–bee interactions and squares represent grassland fragments. Interactions occurring in at least two fragments form
links between sites. Thickness of links (grey lines) is proportional to interaction abundance. Colours represent metanetwork modules based on the Walktrap
community-finding algorithm (igraph package) [66]. This algorithm indicates the presence of subgraphs that constitute a distinctive community. Nodes with greater
centrality occur in the central positions of the graph [67].
7
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
interactions [8,76], thereby increasing the relative proportion of interactions involving generalist species [6]. Jauker et al. [35]
also reported this pattern when analysing plant–pollinator networks in calcareous grasslands. They found that habitat loss
decreased network specialization through the loss of species and interactions, resulting in small and tightly connected networks.
Given the pronounced process of fragmentation of European calcareous grasslands in the past century [38,81], it is expected that
the remaining, commonly small and isolated fragments have already lost many specialized and rare interactions, as shown for
butterflies on the same calcareous grasslands [82].
Habitat generalist bees can use floral and nesting resources of several land cover types and may establish opportunistic
interactions with multiple (habitat generalist and specialist) plant partners. In our study, landscape diversification stands for
high variety of land cover types in the matrix, dominated by oilseed rape, grainfields, maize, pastures, grasslands, field margin
strips and forest stands. Any increase in landscape diversification in the surrounding of a calcareous grassland habitat fragment
(e.g. matrix diversification) may increase the influx (i.e. spillover) of habitat generalist species establishing new generalized
interactions, further decreasing grassland network specialization. Hence, specialization would be reduced by increasing the
number of potential interaction partners available, closing the gap between the potential and the realized number of interacting
partners of each species.
Accordingly, habitat loss and landscape simplification may have opposite effects on European calcareous grasslands network
specialization by affecting groups of species with different traits. Habitat loss may cause a selective reduction in richness or
abundance of species with low behavioural plasticity and high plant fidelity (i.e. specialists), resulting in small and homogene-
ous networks with an overrepresentation of generalist species. In contrast, landscape simplification may result in a reduction
of species with high behavioural plasticity and low plant fidelity (i.e. generalists), resulting in fragile and specialized networks.
The most plausible mechanism that could explain the patterns found in this study is that pollinators did not compensate for the
missing plant partners and that plants did not get extra visits once a pollinator was lost following landscape homogenization,
resulting in more specialized interaction networks.
(c) Visitation and pollen-transport unique interactions
The large proportion of flower visitor species transporting few or no pollen, due to morphological or behavioural constraints,
is an intrinsic characteristic of pollination systems [23,25,29,49]. However, it is important to note that this proportion is usually
reported to be larger in light microscopy studies than metabarcoding studies [74]. The selection of a threshold of pollen
grains carried by a bee as proof of legitimate pollen transport (in opposition to accidental transport due to contamination) is
fundamental to avoid false positives, however, it is also non-trivial [22]. Popic et al. [29] used a 10 grains threshold and reported
only 38% of visitation links resulting in pollen transport (51% in our study). Manincor et al. [49], also studying bee interactions
in calcareous grasslands of Europe (threshold five pollen grains), reported doubling the total number of interactions found (40%
increase in our study) when looking at pollen transport (i.e. pollen-transport only interactions). Our approach was conservative;
the choice of a lower threshold would have most likely augmented the differences found. Despite the more conservative
threshold used in our study, the structural differences found between bee visitation and pollen-transport networks are in
accordance with the literature and challenge the assumption that visitation data are sufficient surrogates of animal-mediated
pollen transport [29,49].
The large presence of flower visitors with a relatively small capacity for pollen transport raises many questions regarding
their importance for pollination [83]. In theory, deposition of a single conspecific pollen grain could be enough for pollination to
occur, but pollen deposition thresholds are common given that not all pollen deposited by pollinators is viable [84]. Therefore,
a relatively high amount of conspecific pollen deposition is usually needed for a meaningful pollination success [84]. The
concomitant deposition of heterospecific pollen is also an important factor considering its negative effects on pollination
[85,86]. Actually, from a plant species perspective, a strategy based on maximizing pollinators’ visits might come at the cost
of high heterospecific pollen deposition on their stigmas. Contrastingly, a strategy based on the attraction of a small number
of specialized pollinators (and therefore larger potential for conspecific pollen deposition) comes at the cost of a higher
dependence on a small group of pollinators, with a higher risk of local extinction and a lower probability of visitation. Habitat
fragmentation and landscape homogenization may impose a reduced set of pollinator partners to interact with. Consequently,
higher plant specialization could arise as an indirect result of the lack of alternative partners and not as part of an ecological
strategy to increase reproductive success.
The pollen-transport networks revealed a high amount of rare interactions. This implies that plant–pollinator networks based
only on flower visitation data are not just biased by the inclusion of interactions with no potential for pollination, but also by
missing many rare interactions. Consequently, pollen loads analysis represents an important complementary approach to study
pollination systems because the actual pollen dispersal across the plant community can be quantified. Visitation data, on the
other hand, appear fundamental to understand plant–pollinator interactions from the pollinator perspective, as competition
among pollinators and the different foraging strategies that pollinators use to maximize their fitness can be analysed. For
example, bees visit some flowers exclusively to collect nectar and eventually may not enter into contact with flower anthers.
These interactions may not be important from the plant perspective because pollen transport may not occur, but they are still
fundamental for bees as they will influence their diet. The detection of interactions involving rare habitat specialist plants,
such as Scabiosa columbaria [87,88], indicates that pollen load analyses can contribute to improve conservation strategies by
identifying remaining small populations of these rare species. For example, restoration efforts targeting these small populations
could be undertaken in places where the plants were thought to be locally extinct.
8
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
(d) Plant habitat specialization and pollinator group
Bumblebees in central Europe have a high capacity for pollen transport given that they are abundant, larger than most solitary
bees and have dense hair [36,89,90]. Our results support this by demonstrating that the probability of a bumblebee carrying
pollen after a flower visit is higher than that of solitary bees. However, studies on pollen transfer (i.e. pollen deposition on
a conspecific stigma) after flower visits would be necessary to test whether bumblebees are also able to deposit more pollen
on stigmas than solitary bees, since pollen transport does not always translate into pollen deposition [19]. For example, the
pollen located in specialized structures, such as the corbiculae, is generally assumed not to be available for pollination [91].
This may not make a difference in pollen transport analyses [49], but it could be important for pollen transfer to stigmas.
Although bumblebees are hairier than other bee groups, which is associated with a higher pollination effectiveness [36], plants'
pollination also depends on pollinator behaviour, flower morphology and on the ratio between conspecific and heterospecific
pollen deposition [85,92]. Therefore, solitary bees may be irreplaceable for the reproductive success of some plant species.
In fact, solitary bees were found to be fundamental for the pollen transport of many plant species. In particular, they
transported pollen from 16 habitat specialist plants and were the only pollen vector for three of them. Considering that many
solitary bee species are vulnerable and threatened with extinction [53,93], these results signal to the importance of their role in
calcareous grasslands and to the potential risk of their absence for habitat specialist plants' reproductive success. Our findings
reveal that both bumblebees and solitary bees are complementary for pollen transport of calcareous grasslands plant species.
We found a smaller representation of habitat generalist than habitat specialist plants on pollen-transport networks in
calcareous grasslands. Hence, flower visits to habitat specialist plants have a higher probability to translate into pollen transport
than visits to habitat generalist plants. The habitat generalist K. arvensis, for example, was visited by many species of both
bee guilds (bumblebees and solitary bees), being involved in a total of 105 interaction events. However, we found only 10
interactions with K. arvensis in the pollen-transport networks, involving only four bumblebee species and no solitary bees. In
contrast, pollen from the habitat specialist Onobrychis viciifolia was transported by all of its seven species of visitors from both
bee guilds.
A higher representation of habitat specialist plants in pollen-transport networks cannot be solely related to a higher
attractiveness of habitat specialist flowers or pollen, as interactions involving habitat specialist plants in the pollen-transport
dataset were less than half of the total interactions found (46.8%). This result is rather a consequence of different mechanisms
that allow habitat specialist plants to allocate their pollen more frequently on flower visitors than habitat generalists. Habitat
specialist plants are expected to have a long history of evolutionary adaptations to the local pollinator pool and, therefore,
to have developed mechanisms for efficient pollen transport through those pollinators [37]. Conversely, generalist plants
should lack such adaptations, as they would exhibit more opportunistic strategies to quickly adapt to different environments.
The adaptations of plants to increase pollination success can occur at many levels, including pollen vector attraction, pollen
presentation, pollen transport and pollen germination [94]. At the visitation level, traits such as flower size, flower abundance
and the quantity and quality of offered flower rewards (i.e. pollen and nectar) may increase visitation rates [95]. At the pollen
transport level, plants may possess mechanisms to place larger amounts of pollen at specific places on the flower visitors’ body
[94]. At the pollen transfer level, plant traits such as the stigma type (i.e. wet or dry), pollen morphological traits or behavioural
characteristics of pollinators may affect the quantity and quality of pollen deposition [19,94,96]. Even after pollen deposition on
stigmas, plants may exhibit mechanisms to regulate receptiveness depending on the characteristics of the flower visitor [97].
(e) Landscape diversity effect on single-fragment interactions
The spatial scale at which landscape diversity most strongly affected the number of single-fragment interactions was larger for
the pollen-transport networks compared to the visitation networks. This suggests that landscape-scale conservation measures
to protect plant–pollinator networks might be undertaken at the wrong spatial scales when solely based on flower visitation
data. The increased number of single-fragment interactions with landscape diversity is not solely related to a general increase in
the total number of interactions with landscape diversification but rather that landscape diversification has a disproportionately
positive effect on the occurrence of single-fragment interactions compared to the total amount of interactions. Importantly, this
effect was only captured with the pollen transport data and highlights that landscape structure effects can remain undetected in
plant–pollinator studies solely based on visitation data. In contrast to a previous study in the area [11], the effect of landscape
diversity on the proportion of single-fragment interactions could be explained by the central-place foraging behaviour of bees.
Differently from mobile butterflies, bees may determine that the positive effects of landscape diversification at small spatial
scales do not spread through the metanetwork but rather increase the local diversity of plant–pollinator interactions in single
fragments.
5. Conclusion
By analysing plant–pollinator networks across a gradient of habitat fragmentation, we found that pollen-transport networks
were more specialized than visitation networks, indicating that plant–pollinator networks could be more vulnerable than
previously believed. Only 36.8% of the total number of registered plant–pollinator interactions occurred in both flower
visitation and pollen-transport networks. Our landscape analysis of a pollen-transport metanetwork also revealed that the
properties of pollination networks are affected by landscape diversity at scales that differ from those informed by visitation
networks, which may increase the accuracy and effectiveness of landscape-level measures for the conservation of plant–
9
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
pollinator networks. Interactions involving habitat specialist plants and bumblebees had a higher probability of simultaneously
occurring in the visitation and pollen-transport networks than interactions involving habitat generalist plants and solitary bees.
Nonetheless, the pollen of several plant species was found to be transported only by solitary bees. Our study shows that
conservation of pollination systems and related pollination services needs finer data on the biological processes underlying
plant–pollinator interaction networks, such as pollen load analyses. Our metanetwork approach allowed us to identify rarity
of plant–pollinator interactions and local uniqueness, which can be further used by local authorities to design tailored conserva-
tion strategies. Our results have important consequences for the understanding of the responses of plant–pollinator networks to
habitat fragmentation and contribute to unveiling important processes underpinning the dynamics of these networks.
Ethics. This work did not require ethical approval from a human subject or animal welfare committee.
Data accessibility. All data associated to this manuscript can be accessed at [98].
Electronic supplementary material is available online [99].
Declaration of AI use. We have not used AI-assisted technologies in creating this article.
Authors’ contributions. F.L.E.: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing—
original draft, writing—review and editing; I.G.: conceptualization, formal analysis, methodology, supervision, writing—review and editing;
C.E.: conceptualization, formal analysis, methodology, supervision, writing—review and editing; V.A.S: investigation, methodology, writing—
review and editing; H.B.: project administration, supervision, writing—review and editing; S.B.: data curation, investigation, methodology,
resources, software, supervision, validation, writing—review and editing; C.G.: investigation, writing—review and editing; C.H.K: data
curation, investigation, writing—review and editing; C.A.S.: data curation, investigation, T.T.: conceptualization, funding acquisition,
methodology, project administration, resources, supervision, writing—review and editing.
All authors gave final approval for publication and agreed to be held accountable for the work performed therein.
Conflict of interest declaration. We declare we have no competing interests.
Funding. This research was supported by the DFG RTG 1644 "Scaling Problems in Statistics", grant no. 152112243.
Acknowledgements. We thank the anonymous editor and reviewers of this manuscript for their contributions. Our gratitude goes also to Julia
Morley, Guillermo Gallardo Quilacán and all assistants who contributed to data collection and pollen analyses. C.E. was supported by FAPESP
(2015/15172-7) and FAPERJ (E-26/200.610/2022).
References
1. Giannini TC, Cordeiro GD, Freitas BM, Saraiva AM, Imperatriz-Fonseca VL. 2015 The dependence of crops for pollinators and the economic value of pollination in Brazil. J. Econ.
Entomol. 108, 849–857. (doi:10.1093/jee/tov093)
2. Ollerton J, Winfree R, Tarrant S. 2011 How many flowering plants are pollinated by animals? Oikos 120, 321–326. (doi:10.1111/j.1600-0706.2010.18644.x)
3. Grass I, Jauker B, Steffan-Dewenter I, Tscharntke T, Jauker F. 2018 Past and potential future effects of habitat fragmentation on structure and stability of plant–pollinator and host–
parasitoid networks. Nat. Ecol. Evol. 2, 1408–1417. (doi:10.1038/s41559-018-0631-2)
4. Memmott J, Waser NM, Price MV. 2004 Tolerance of pollination networks to species extinctions. Proc. R. Soc. B 271, 2605–2611. (doi:10.1098/rspb.2004.2909)
5. Spiesman BJ, Inouye BD. 2013 Habitat loss alters the architecture of plant–pollinator interaction networks. Ecology 94, 2688–2696. (doi:10.1890/13-0977.1)
6. Xiao Y, Li X, Cao Y, Dong M. 2016 The diverse effects of habitat fragmentation on plant–pollinator interactions. Plant Ecol. 217, 857–868. (doi:10.1007/s11258-016-0608-7)
7. Ferreira PA, Boscolo D, Viana BF. 2013 What do we know about the effects of landscape changes on plant–pollinator interaction networks? Ecol. Indic. 31, 35–40. (doi:10.1016/j.
ecolind.2012.07.025)
8. Aizen MA, Sabatino M, Tylianakis JM. 2012 Specialization and rarity predict nonrandom loss of interactions from mutualist networks. Science 335, 1486–1489. (doi:10.1126/
science.1215320)
9. Tylianakis JM, Morris RJ. 2017 Ecological networks across environmental gradients. Annu. Rev. Ecol. Evol. Syst 48, 25–48. (doi:10.1146/annurev-ecolsys-110316-022821)
10. Hagen M, etal. 2012 Biodiversity, species interactions and ecological networks in a fragmented world. Adv. Ecol. Res 46, 89–210. (doi:10.1016/B978-0-12-396992-7.00002-2)
11. Librán-Embid F, Grass I, Emer C, Ganuza C, Tscharntke T. 2021 A plant–pollinator metanetwork along a habitat fragmentation gradient. Ecol. Lett. 24, 2700–2712. (doi:10.1111/ele.
13892)
12. Emer C, Galetti M, Pizo MA, Guimarães PR, Moraes S, Piratelli A, Jordano P. 2018 Seed-dispersal interactions in fragmented landscapes – a metanetwork approach. Ecol. Lett. 21,
484–493. (doi:10.1111/ele.12909)
13. González AMM, Ornelas JF, Dalsgaard B, Márquez-Luna U, Lara C. 2018 Meta-networks for the study of biogeographical traits in ecological networks: the Mexican hummingbird-
plant assemblage. Sci. Nat. 105, 54. (doi:10.1007/s00114-018-1578-5)
14. Ballantyne G, Baldock KCR, Willmer PG. 2015 Constructing more informative plant–pollinator networks: visitation and pollen deposition networks in a heathland plant community.
Proc. R. Soc. B 282, 20151130. (doi:10.1098/rspb.2015.1130)
15. Freitas L. 2013 Concepts of pollinator performance: is a simple approach necessary to achieve a standardized terminology? Braz. J. Bot. 36, 3–8. (doi:10.1007/s40415-013-0005-6)
16. King C, Ballantyne G, Willmer PG. 2013 Why flower visitation is a poor proxy for pollination: measuring single‐visit pollen deposition, with implications for pollination networks and
conservation. Methods Ecol. Evol. 4, 811–818. (doi:10.1111/2041-210X.12074)
17. Genini J, Morellato LPC, Guimarães PR, Olesen JM. 2010 Cheaters in mutualism networks. Biol. Lett. 6, 494–497. (doi:10.1098/rsbl.2009.1021)
18. Stavert JR, Liñán-Cembrano G, Beggs JR, Howlett BG, Pattemore DE, Bartomeus I. 2016 Hairiness: the missing link between pollinators and pollination. PeerJ 4, e2779. (doi:10.
7717/peerj.2779)
19. Emer C, Vaughan IP, Hiscock S, Memmott J. 2015 The Impact of the invasive alien plant, Impatiens glandulifera, on pollen transfer networks. PLoS One 10, e0143532. (doi:10.1371/
journal.pone.0143532)
20. Fang Q, Huang SQ. 2013 A directed network analysis of heterospecific pollen transfer in a biodiverse community. Ecology 94, 1176–1185. (doi:10.1890/12-1634.1)
21. Lopezaraiza-Mikel ME, Hayes RB, Whalley MR, Memmott J . 2007 The impact of an alien plant on a native plant–pollinator network: an experimental approach. Ecol. Lett. 10, 539–
550. (doi:10.1111/j.1461-0248.2007.01055.x)
10
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
22. Bosch J, González AMM, Rodrigo A, Navarro D. 2009 Plant–pollinator networks: adding the pollinator’s perspective. Ecol. Lett. 12, 409–419. (doi:10.1111/j.1461-0248.2009.01296.
x)
23. Alarcón R. 2010 Congruence between visitation and pollen-transport networks in a California plant–pollinator community. Oikos 119, 35–44. (doi:10.1111/j.1600-0706.2009.
17694.x)
24. Gresty CEA, Clare E, Devey DS, Cowan RS, Csiba L, Malakasi P, Lewis OT, Willis KJ. 2018 Flower preferences and pollen transport networks for cavity-nesting solitary bees:
Implications for the design of agri-environment schemes. Ecol. Evol. 8, 7574–7587. (doi:10.1002/ece3.4234)
25. Zhao Y etal. 2019 The topological differences between visitation and pollen transport networks: a comparison in species rich communities of the Himalaya–Hengduan Mountains.
Oikos 128, 551–562. (doi:10.1111/oik.05262)
26. Macgregor CJ, Kitson JJN, Fox R, Hahn C, Lunt DH, Pocock MJO, Evans DM. 2018 Construction, validation and application of nocturnal pollen transport networks in an agro-
Ecosystem: a comparison using microscopy and DNA metabarcoding. Ecol. Entomol. 282, 538. (doi:10.1101/325084)
27. Hall MA, Stavert JR, Saunders ME, Barr S, Haberle SG, Rader R. 2022 Pollen–insect interaction meta-networks identify key relationships for conservation in mosaic agricultural
landscapes. Ecol. Appl. 32, e2537. (doi:10.1002/eap.2537)
28. Blüthgen N, Menzel F, Blüthgen N. 2006 Measuring specialization in species interaction networks. BMC Ecol. 6, 9. (doi:10.1186/1472-6785-6-9)
29. Popic TJ, Wardle GM, Davila YC. 2013 Flower‐visitor networks only partially predict the function of pollen transport by bees. Austral Ecology 38, 76–86. (doi:10.1111/j.1442-9993.
2012.02377.x)
30. de Santiago‐Hernández MH, Martén‐Rodríguez S, Lopezaraiza‐Mikel M, O yama K, González‐Rodríguez A, Quesada M. 2019 The role of pollination effectiveness on the attributes of
interaction networks: from floral visitation to plant fitness. Ecology 100, e02803. (doi:10.1002/ecy.2803)
31. Barker DA, Arceo-Gomez G. 2021 Pollen transport networks reveal highly diverse and temporally stable plant–pollinator interactions in an Appalachian floral community. AoB
Plants 13, plab062. (doi:10.1093/aobpla/plab062)
32. Encinas-Viso F etal. 2023 Pollen DNA metabarcoding reveals cryptic diversity and high spatial turnover in alpine plant–pollinator networks. Mol. Ecol. 32, 6377–6393. (doi:10.1111/
mec.16682)
33. Pornon A, Andalo C, Burrus M, Escaravage N. 2017 DNA metabarcoding data unveils invisible pollination networks. Sci. Rep. 7, 16828. (doi:10.1038/s41598-017-16785-5)
34. Zhang X et al. 2023 Habitat fragmentation increases specialization of multi-trophic interactions by high species turnover. Proc. R. Soc. B 290, 20231372. (doi:10.1098/rspb.2023.
1372)
35. Jauker F, Jauker B, Grass I, Steffan-Dewenter I, Wolters V. 2019 Partitioning wild bee and hoverfly contributions to plant–pollinator network structure in fragmented habitats.
Ecology 100, e02569. (doi:10.1002/ecy.2569)
36. Roquer-Beni L, Rodrigo A, Arnan X, Klein AM, Fornoff F, Boreux V, Bosch J. 2020 A novel method to measure hairiness in bees and other insect pollinators. Ecol. Evol. 10, 2979–2990.
(doi:10.1002/ece3.6112)
37. Miller-Struttmann NE. 2013 Rarity and reproductive biology: habitat specialists reveal a complex relationship. Botany 91, 349–359. (doi:10.1139/cjb-2012-0274)
38. Habel JC, Dengler J , Janišová M, Török P, Wellstein C, Wiezik M. 2013 European grassland ecosystems: threatened hotspots of biodiversity. Biodivers. Conserv. 22, 2131–2138. (doi:
10.1007/s10531-013-0537-x)
39. Cremene C, Groza G, Rakosy L, Schileyko AA, Baur A, Erhardt A, Baur B. 2005 Alterations of steppe‐like grasslands in Eastern Europe: a threat to regional biodiversity hotspots. Biol.
Conserv. 19, 1606–1618. (doi:10.1111/j.1523-1739.2005.00084.x)
40. WallisDeVries MF, Poschlod P, Willems JH. 2002 Challenges for the conservation of calcareous grasslands in northwestern Europe: integrating the requirements of flora and fauna.
Biol. Conserv. 104, 265–273. (doi:10.1016/S0006-3207(01)00191-4)
41. Poschlod P, WallisDeVries MF. 2002 The historical and socioeconomic perspective of calcareous grasslands—lessons from the distant and recent past. Biol. Conserv. 104, 361–376.
(doi:10.1016/S0006-3207(01)00201-4)
42. Steffan-Dewenter I, Tscharntke T. 2002 Insect communities and biotic interactions on fragmented calcareous grasslands—a mini review. Biol. Conserv. 104, 275–284. (doi:10.
1016/S0006-3207(01)00192-6)
43. Filz KJ, Engler JO, Stoffels J, Weitzel M, Schmitt T. 2013 Missing the target? A critical view on butterfly conservation efforts on calcareous grasslands in south-western Germany.
Biodivers. Conserv. 22, 2223–2241. (doi:10.1007/s10531-012-0413-0)
44. van Swaay C etal. 2011 Applying IUCN criteria to invertebrates: how red is the red list of European butterflies? Biol. Conserv. 144, 470–478. (doi:10.1016/j.biocon.2010.09.034)
45. Krauss J, Steffan‐Dewenter I, Tscharntke T. 2003 How does landscape context contribute to effects of habitat fragmentation on diversity and population density of butterflies? J.
Biogeogr. 30, 889–900. (doi:10.1046/j.1365-2699.2003.00878.x)
46. van Swaay C, Brereton T, Kirkland P, Warren M. 2012 Manual for butterfly monitoring. Report Vs2012. De Vlinderstichting/Dutch Butterfly Conservation, Butterfly Conservation UK &
Butterfly Conservation Europe. See http://www.bc-europe.eu/upload/Manual_Butterfly_Monitoring.pdf.
47. Sexton AN, Emery SM. 2020 Grassland restorations improve pollinator communities: a meta-analysis. J. Insect Conserv. 24, 719–726. (doi:10.1007/s10841-020-00247-x)
48. Dafni A. 1992 Pollination ecology: a practical approach. Oxford, UK: Oxford University Press.
49. de Manincor N, Hautekèete N, Mazoyer C, Moreau P, Piquot Y, Schatz B, Schmitt E, Zélazny M, Massol F. 2020 How biased is our perception of plant-pollinator networks? A
comparison of visit- and pollen-based representations of the same networks. Acta Oecologica 105, 103551. (doi:10.1016/j.actao.2020.103551)
50. Jones GD. 2014 Pollen analyses for pollination research, acetolysis. J. Poll. Ecol. 13, 203–217. (doi:10.26786/1920-7603(2014)19)
51. Poschlod P, Kiefer S, Tränkle U, Fischer S, Bonn S. 1998 Plant species richness in calcareous grasslands as affected by dispersability in space and time. Appl. Veg. Sci. 1, 75–91. (doi:
10.2307/1479087)
52. Brückmann SV, Krauss J, Steffan‐Dewenter I. 2010 Butterfly and plant specialists suffer from reduced connectivity in fragmented landscapes. J. Appl. Ecol. 47, 799–809. (doi:10.
1111/j.1365-2664.2010.01828.x)
53. Jauker B, Krauss J, Jauker F, Steffan-Dewenter I. 2013 Linking life history traits to pollinator loss in fragmented calcareous grasslands. Landscape Ecol. 28, 107–120. (doi:10.1007/
s10980-012-9820-6)
54. Hopfenmüller S, Steffan-Dewenter I, Holzschuh A. 2014 Trait-specific responses of wild bee communities to landscape composition, configuration and local factors. PLoS One 9,
e104439. (doi:10.1371/journal.pone.0104439)
55. Westrich P. 2018 Die Wildbienen Deutschlands. Stuttgart, Germany: Verlag Eugen Ulmer.
56. ESRI. 2018 ArcGIS Desktop Version 10.5.1. Environmental Systems Research Institute
57. Hesselbarth MHK, Sciaini M, With KA, Wiegand K, Nowosad J. 2019 landscapemetrics: an open‐source R tool to calculate landscape metrics. Ecography 42, 1648–1657. (doi:10.
1111/ecog.04617)
11
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
58. Hanski I, Kuussaari M, Nieminen M. 1994 Metapopulation structure and migration in the butterfly Melitaea cinxia. Ecology 75, 747–762. (doi:10.2307/1941732)
59. R Development Core Team (2020). R: a language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. http://www.Rproject.org
60. Dunne JA, Williams RJ, Martinez ND. 2002 Food-web structure and network theory: the role of connectance and size. Proc. Natl Acad. Sci. USA 99, 12917–12922. (doi:10.1073/pnas.
192407699)
61. Dormann CF, Frund J, Bluthgen N, Gruber B. 2009 Indices, graphs and null models: analyzing bipartite ecological networks. Open Ecol. J 2, 7–24. (doi:(10.2174/
1874213000902010007)
62. Dormann CF, Strauss R. 2014 A method for detecting modules in quantitative bipartite networks.. Methods Ecol. Evol. 5, 90–98. (doi:10.1111/2041-210X.12139)
63. Dormann CF, Gruber B, Fründ J. 2008 Introducing the bipartite package: analysing ecological networks. R News 8, 8–11.
64. Hartig F. 2022 Dharma: residual diagnostics for Hierarchical (multi-level/mixed) regression models. R Package v. 0.4.6. See https://cran.r-project.org/web/packages/DHARMa/
vignettes/DHARMa.html.
65. Bates D, Sarkar D, Bates MD, Matrix L. 2006 The lme4 package. See http://cran. r-project. org/740src/contrib/Descriptions/lme4. html.
66. Csárdi G, Nepusz T, Traag V, Horvát S, Zanini F, Noom D, Müller K. 2024 Igraph: Network Analysis and Visualization in R. Zenodo (doi:10.5281/zenodo.7682609)
67. Bannister MJ, Eppstein D, Goodrich MT, Trott L. 2013 Force-directed graph drawing using social gravity and Scaling. In Graph drawing. 20th International symposium, GD 2012,
Redmond, WA, USA, September 19-21, 2012: revised selected papers (eds W Didimo, M Patrignani), pp. 414–425. Berlin, Germany; New York, NY: Springer. (doi:10.1007/978-3-642-
36763-2)
68. Souza CS, Maruyama PK, Santos K, Varassin IG, Gross CL, Araujo AC. 2021 Plant-centred sampling estimates higher beta diversity of interactions than pollinator-based sampling
across habitats. New Phytol. 230, 2501–2512. (doi:10.1111/nph.17334)
69. Barrios B, Pena SR, Salas A, Koptur S. 2016 Butterflies visit more frequently, but bees are better pollinators: the importance of mouthpart dimensions in effective pollen removal and
deposition. AoB Plants 8, plw001. (doi:10.1093/aobpla/plw001)
70. Abou-Shaara HF. 2014 The foraging behaviour of honey bees, Apis mellifera: a review. Vet. Med. 59, 1–10. (doi:10.17221/7240-VETMED)
71. Grüter C, Hayes L. 2022 Sociality is a key driver of foraging ranges in bees. Curr. Biol. 32, 5390–5397. (doi:10.1016/j.cub.2022.10.064)
72. Robinson SVJ, Hoover SE, Pernal SF, Cartar RV. 2022 Optimal distributions of central-place foragers: honey bee foraging in a mass flowering crop. Behav. Ecol. 33, 386–397. (doi:10.
1093/beheco/arab143)
73. Arstingstall KA, DeBano SJ, Li X, Wooster DE, Rowland MM, Burrows S, Frost K. 2021 Capabilities and limitations of using DNA metabarcoding to study plant–pollinator interactions.
Mol. Ecol. 30, 5266–5297. (doi:10.1111/mec.16112)
74. Milla L, Sniderman K, Lines R, Mousavi-Derazmahalleh M, Encinas-Viso F. 2021 Pollen DNA metabarcoding identifies regional provenance and high plant diversity in Australian
honey. Ecol. Evol. 11, 8683–8698. (doi:10.1002/ece3.7679)
75. Classen A, Eardley CD, Hemp A, Peters MK, Peters RS, Ssymank A, Steffan-Dewenter I. 2020 Specialization of plant–pollinator interactions increases with temperature at Mt.
Kilimanjaro. Ecol. Evol. 10, 2182–2195. (doi:10.1002/ece3.6056)
76. Hoiss B, Krauss J, Steffan-Dewenter I. 2015 Interactive effects of elevation, species richness and extreme climatic events on plant–pollinator networks. Glob. Chang. Biol. 21, 4086–
4097. (doi:10.1111/gcb.12968)
77. Gámez-Virués S etal. 2015 Landscape simplification filters species traits and drives biotic homogenization. Nat. Commun. 6, 8568. (doi:10.1038/ncomms9568)
78. Brosi BJ, Briggs HM. 2013 Single pollinator species losses reduce floral fidelity and plant reproductive function. Proc. Natl Acad. Sci. USA 110, 13044–13048. (doi:10.1073/pnas.
1307438110)
79. Bastolla U, Fortuna MA, Pascual-García A, Ferrera A, Luque B, Bascompte J. 2009 The architecture of mutualistic networks minimizes competition and increases biodiversity. Nature
458, 1018–1020. (doi:10.1038/nature07950)
80. Benadi G, Blüthgen N, Hovestadt T, Poethke HJ. 2013 Contrasting specialization–stability relationships in plant–animal mutualistic systems. Ecol. Model. 258, 65–73. (doi:10.1016/
j.ecolmodel.2013.03.002)
81. Schils RLM et al. 2022 Permanent grasslands in Europe: land use change and intensification decrease their multifunctionality. Agric. Ecosyst. Environ. 330, 107891. (doi:10.1016/j.
agee.2022.107891)
82. Steffan‐Dewenter I, Tscharntke T. 2000 Butterfly community structure in fragmented habitats. Ecol. Lett. 3, 449–456. (doi:10.1111/j.1461-0248.2000.00175.x)
83. Moquet L, Bruyère L, Pirard B, Jacquemart AL. 2017 Nectar foragers contribute to the pollination of buzz-pollinated plant species. Am. J. Bot. 104, 1451–1463. (doi:10.3732/ajb.
1700090)
84. Li K, Tscharntke T, Saintes B, Buchori D, Grass I. 2019 Critical factors limiting pollination success in oil palm: a systematic review. Agric. Ecosyst. Environ. 280, 152–160. (doi:10.1016/
j.agee.2019.05.001)
85. Arceo-Gómez G, Ashman TL. 2011 Heterospecific pollen deposition: does diversity alter the consequences? New Phytol. 192, 738–746. (doi:10.1111/j.1469-8137.2011.03831.x)
86. Brown BJ, Mitchell RJ. 2001 Competition for pollination: effects of pollen of an invasive plant on seed set of a native congener. Oecologia 129, 43–49. (doi:10.1007/s004420100700)
87. Angeloni F, Vergeer P, Wagemaker CAM, Ouborg NJ. 2014 Within and between population variation in inbreeding depression in the locally threatened perennial Scabiosa
columbaria. Conserv. Genet. 15, 331–342. (doi:10.1007/s10592-013-0541-4)
88. Bijlsma R, Ouborg NJ, Treuren R. 1994 On genetic erosion and population extinction in plants: a case study in Scabiosa Columbaria and Salvia Pratensis. In Conservation Genetics (eds
V Loeschcke, SK Jain, J Tomiuk), pp. 255–271. Basel, Switzerland: Springer. (doi:10.1007/978-3-0348-8510-2)
89. Velthuis HHW, van Doorn A. 2006 A century of advances in bumblebee domestication and the economic and environmental aspects of its commercialization for pollination.
Apidologie 37, 421–451. (doi:10.1051/apido:2006019)
90. Willmer PG, Bataw AAM, Hughes JP. 1994 The superiority of bumblebees to honeybees as pollinators: insect visits to raspberry flowers. Ecol. Entomol. 19, 271–284. (doi:10.1111/j.
1365-2311.1994.tb00419.x)
91. Horskins K, Turner VB. 1999 Resource use and foraging patterns of honeybees, Apis mellifera, and native insects on flowers of Eucalyptus costata. Austral. J. Ecol. 24, 221–227. (doi:
10.1046/j.1442-9993.1999.00965.x)
92. Morales CL, Traveset A. 2008 Interspecific pollen transfer: magnitude, prevalence and consequences for plant fitness. Crit. Rev. Plant Sci. 27, 221–238. (doi:10.1080/
07352680802205631)
93. Nieto A, etal. 2014 European red list of bees. Luxembourg: Publications Office of the European Union.
94. Minnaar C, Anderson B, de Jager ML, Karron JD. 2019 Plant–pollinator interactions along the pathway to paternity. Ann. Bot. 123, 225–245. (doi:10.1093/aob/mcy167)
95. Conner JK, Rush S. 1996 Effects of flower size and number on pollinator visitation to wild radish, Raphanus raphanistrum. Oecologia 105, 509–516. (doi:10.1007/BF00330014)
12
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
96. Konzmann S, Koethe S, Lunau K. 2019 Pollen grain morphology is not exclusively responsible for pollen collectability in bumble bees. Sci. Rep. 9, 4705. (doi:10.1038/s41598-019-
41262-6)
97. Betts MG, Hadley AS, Kress WJ. 2015 Pollinator recognition by a keystone tropical plant. Proc. Natl Acad. Sci. USA 112, 3433–3438. (doi:10.1073/pnas.1419522112)
98. Librán-Embid F. 2024 Flower-bee vs pollen-bee metanetworks in fragmented landscapes (doi:10.17605/OSF.IO/TMHU5)
99. Librán-Embid F, Grass I, Emer C, Alarcón-Segura V, Behling H, Biagioni S, et al. 2024 Supplementary material from: Flower–bee vs pollen–bee metanetworks in fragmented
landscapes. Figshare. (doi:10.6084/m9.figshare.c.7204022)
13
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20232604
Downloaded from https://royalsocietypublishing.org/ on 29 May 2024
