Seasonal dynamic variation of pollination network is associated with the number of species in flower in an oceanic island communityXiangping Wang , Tong Zeng, Mingsong Wu and Dianxiang Zhang*
Key Laboratory of Plant Resources Conservation and Sustainable Utilization, South China Botanical Garden, Chinese Academy of
Received: 10 March 2020, Revised: 14 July 2020, Accepted: 4 August 2020, Advanced Access publication: 11 August 2020
Citation: Wang X, Zeng T, Wu M, et al. (2020) Seasonal dynamic variation of pollination network is associated with the number of species in flower in an oceanic
island community. J Plant Ecol 13:657–666. https://doi.org/10.1093/jpe/rtaa054
Abstract
Aims Plant–pollinator interaction networks are dynamic entities, and seasonal variation in plant phenology can reshape their structure on both short and long timescales. However, such seasonal dynamics are rarely considered, especially for oceanic island pollination networks. Here, we assess changes in the temporal dynamics of plant–pollinator interactions in response to seasonal variation in floral resource richness in oceanic island communities.
Methods We evaluated seasonal variations of pollination networks in the Yongxing Island community. Four temporal qualitative pollination networks were analyzed using plant–pollinator interaction data of the four seasons. We collected data on plant–pollinator interactions during two consecutive months in each of the four seasons. Four network-level indices were calculated to characterize the overall structure of the networks. Statistical analyses of community dissimilarity were used to compare this community across four seasons to explore the underlying factors driving these patterns. We also evaluated the temporal variation in two species-level indices of plant and pollinator functional groups.
Important Findings Both network-level specialization and modularity showed a significantly opposite trend compared with plant species richness across four seasons. Increased numbers of plant species might promote greater competition among pollinators, leading to increased niche overlap and causing decreased specialization and modularity and vice versa. Further analyses suggested that the season-to-season turnover of interactions was dominated by interaction rewiring. Thus, the seasonal changes in niche overlap among pollinators lead to interaction rewiring, which drives interaction turnover in this community. Hawkmoths had higher values of specialization and Apidae had higher values of species strength compared with other pollinator functional groups. These findings should be considered when exploring plant–pollinator interactions in ecosystems of isolated oceanic islands and in other ecosystems.
Keywords: interaction rewiring, modularity, oceanic island, pollination networks, seasonality, specialization, species strength
(Blüthgen et al. 2006). Species strength quantifies the sum of the
dependencies of each species, i.e. the proportions of interactions
performed by a given species across all of its interaction partners.
Higher values of species strength indicate that more plants depend on
the specific pollinator species, and vice versa (Bascompte et al. 2006).
Calculations of all network-related indices were conducted with the
‘bipartite’ package version 2.05 (Dormann et al. 2008) in R.
Statistical analysis
Following previous studies (CaraDonna et al. 2017; Poisot et al. 2012;
Rabeling et al. 2019), we used Whittaker’s dissimilarity index (Whittaker
1960) (βint
) for calculating seasonal interaction turnover of plant–
pollinator interactions across four seasons to evaluate and compare the
level of temporal variation in the community. Interaction turnover (βint
)
quantifies the dissimilarity of interactions between each pair of seasonal
networks. Values of βint
range from 0 to 1, and higher values indicate
higher turnover, i.e. less links in common between two interaction
networks (CaraDonna et al. 2017; Poisot et al. 2012). As Whittaker’s
index treats interactions as present or absent and is more robust
than other β-dissimilarity indices when dealing with heterogeneous
dataset sizes (Koleff et al. 2003; Poisot et al. 2012), it is appropriate for
qualitative interactions and can meet our goal of exploring the variation
of interactions through time. βint
can be partitioned as βint
= βst + β
rw,
where the link dissimilarity (βint
) is consisting of species turnover (βst)
and interaction rewiring (βrw
) (Poisot et al. 2012). Furthermore, the
contribution of species turnover (βst) to link dissimilarity (β
int) covaries
with species turnover (βs), i.e. the difference in species composition
of seasonal networks (Poisot et al. 2012). We also calculated other
dissimilarity indexes which refer to interactions between species
common to networks in both seasons, species (βs), plant species (β
pl)
and pollinator species (βpo
), respectively. This partitioning of interaction
turnover and the corresponding analyses enabled us to determine
whether the seasonal dynamics in interaction networks were driven
by the changes in species composition (βst) (i.e. turnover in plant
species, pollinator species or both), or due to rewiring of interactions
among species (βrw
) (i.e. establishment of new types of interactions)
or some combination of both (CaraDonna et al. 2017; Rabeling et al.
2019). To determine whether the degree of dissimilarity detected is
driven by similar underlying factors, we then performed correlation
analyses to test the relationships between the different dissimilarity
indexes. For example, a correlation relationship between interaction
turnover (βint
) with species turnover (βs) would indicate that changes in
interactions are mainly driven by changes in species. We performed the
non-parametric Spearman’s rank correction analysis between two β-
dissimilarity indices because these indices are not normally distributed.
We evaluated whether seasons were important determinants of
species-level indices for plants and functional groups of pollinators.
Pollinators were classified as Apidae, non-Apidae Hymenoptera,
Syrphidae, non-Syrphidae Diptera, butterflies and hawkmoths.
As Hemiptera (Triatominae), Passeriformes (Zosterops japonicus)
and Arctiidae (Utetheisa lotrix) each included only one species and
performed few interactions in the networks, they were excluded from
our analysis. When we considered plant and pollinator species that
Dow
nloaded from https://academ
ic.oup.com/jpe/article/13/5/657/5891221 by guest on 20 August 2022
Copyedited by: GB
JOURNAL OF PLANT ECOLOGY RESEARCh ARTiCLE
660 JOURNAL OF PLANT ECOLOGY | VOL 13 | October 2020 | 657–666
occurred in all four seasons, Diptera were excluded from our analysis
because only one species met this criterion. We used linear mixed-
effects models for the species-level data. Season (dry or rainy; spring,
summer, autumn or winter) and functional group were included in
the pollinator model as fixed effects, and species identity nested within
vegetation type was included as a random effect using the ‘lme4’ R
package (Bates et al. 2015; Souza et al. 2018). To test whether season
or functional group had a significant effect on species-level indices, a
likelihood ratio test was used to compare the model with and without
the fixed factor using the ‘car’ R package (Souza et al. 2018). After we
detected that the pollinator functional group was a significant factor,
a general linear hypothesis test was performed in the ‘multcomp’ R
package (Hothorn et al. 2008; Souza et al. 2018). We repeated the
species-level analysis on plant and pollinator species that occurred in
both dry and rainy seasons and on plant and pollinator species that
occurred in all four seasons.
RESULTS
In total, 891 interactions between 64 pollinator species and 63 plant
species from 26 plant families were recorded across all sampling
intervals. The dry season and rainy season had similar network sizes.
The summer season network had a greater number of plant and
pollinator species, followed by the winter, spring and autumn networks
(Table 1). Among the plant families recorded, Asteraceae (7 species)
were the most generalized, receiving 24% of the total interactions,
followed by Euphorbiaceae (9%) and Fabaceae (8%). Hymenoptera
contributed 57% of all interactions, with Apidae accounting for 40%
of these, and Braunapis puangensis was the most generalized pollinator
species, visiting 45 plant species. Among the Hymenoptera species,
Campsomeriella collaris and Micromeriella marginella were categorized
as female and male individuals based on their obvious morphological
differences and their contrasting patterns of plant species visitation.
Diptera and Lepidoptera contributed 23% and 19% of all interactions,
respectively. Syrphidae accounted for 51% of all Diptera interactions,
and Paragus bicolor was the most generalized species, visiting 34 plant
species. Butterflies and hawkmoths contributed 67% and 26% of all
Lepidoptera interactions, respectively. The seasonal variation in the
number of pollinator species in each functional group is shown in
Fig. 1.
Variation of network-level indices across seasons
All seven networks were more specialized and modular than expected
by the null models (Table 1). The rainy season network showed higher
specialization and modularity but lower nestedness and connectance
than the dry season network. Moreover, the spring and autumn
networks showed higher specialization and modularity than the
summer and winter networks. Among the four seasons, the spring
network showed the highest specialization and modularity, whereas
the autumn network showed the highest nestedness. The summer
and winter networks showed similar specialization, nestedness,
connectance and modularity (Table 1; Fig. 2; Supplementary Fig. S1).
The plant and pollinator species number, network size and interaction
number showed a consistent dynamic variation across the four seasons
(i.e. they jointly increased or decreased), while both the network-
level specialization and modularity apparently showed a contrasting
response compared with the plant species number across four seasons
(i.e. values of specialization and modularity increased as the number of
plant species decreased and vice versa) (Fig. 3).
Values of seasonal interaction turnover (βint
) range from 0.218
to 0.256 (Supplementary Table S1). Interaction turnover (βint
) was
positively correlated with interaction rewiring (i.e. the reassembly
of plant–pollinator interactions, βrw
) (Spearman’s coefficient,
r = 0.943, P = 0.005) (Supplementary Table S2; Fig. 4). No significant
relationships were detected between interaction turnover (βint
) and
species turnover (βs) (r = 0.543, P = 0.266), plant species turnover (β
pl)
(r = 0.257, P = 0.623) and pollinator species turnover (βpo
) (r = 0.754,
P = 0.084). There was no correlation between interaction rewiring (βrw
)
and species turnover (βs) (r = 0.371, P = 0.468), suggesting that factors
driving interaction changes are different from those driving species
turnovers. Furthermore, both plant species turnover (βpl) (r = 0.829,
P = 0.042) and pollinator species turnover (r = 0.841, P = 0.036) were
positively correlated with species turnover (βs), indicating that the
main driver of species turnover at Yongxing Island community was
related to changes in the number of plants in flower and the number
of active pollinator species.
Variation in species-level indices across seasons
There were no species-level differences in specialization (plants:
χ2 = 0.48, P = 0.49; pollinators: χ2 = 0.03, P = 0.86) or species strength
(plants: χ2 = 1.26, P = 0.26; pollinators: χ2 = 1.35, P = 0.25) for plants
or pollinators between the dry and rainy seasons (Supplementary Fig.
S2). When we considered only plants and pollinators that occurred in
both the dry and rainy seasons (55 plant species, 49 pollinator species),
no species-level differences in specialization (plants: χ2 = 0.52, P = 0.47;
pollinators: χ2 = 0.01, P = 0.91) or species strength (plants: χ2 = 0.93,
P = 0.33; pollinators: χ2 = 1.00, P = 0.32) were detected between
the two seasons (Supplementary Tables S3–S5; Fig. 5). Hawkmoths
showed higher specialization than the other groups in the dry season.
In the rainy season, hawkmoths showed higher specialization than
non-Apidae Hymenoptera and Syrphidae, and Apidae had a higher
species strength than the other groups (Supplementary Fig. S2).
Similar results were obtained for pollinator species that occurred in
both the dry and rainy seasons (Fig. 5c and d).
Among four seasons, there were some differences in plant
specialization (χ2 = 8.80, P = 0.03), but species strength did not
differ (χ2 = 3.55, P = 0.31) (Supplementary Table S3; Fig. S3a and
b). Hawkmoths were more specialized in autumn than in winter
Table 1: Network metrics of seven plant–pollinator networks in the Yongxing Island community, showing values for the complete sampling period, the dry and rainy seasons, and the spring, summer, autumn and winter seasons
Network
Plant
species number
Pollinator
species number
Interaction
number Network size Specialization H′2 Nestedness Connectance Modulariy Q
Complete 63 64 891 127 0.25* 14.69* 0.22* 0.22*
Dry 57 61 642 118 0.28* 13.36* 0.19* 0.23*
Rainy 61 55 563 116 0.30* 11.83* 0.17* 0.25*
Spring 52 39 279 91 0.39* 11.38* 0.14* 0.33*
Summer 55 57 520 112 0.32* 12.88* 0.17* 0.25*
Autumn 39 43 262 82 0.37* 14.71* 0.16* 0.32*
Winter 52 51 440 103 0.32* 12.39* 0.17* 0.28*
*Statistically significant network metrics that did not overlap null model expectations (95% confidence interval).
Dow
nloaded from https://academ
ic.oup.com/jpe/article/13/5/657/5891221 by guest on 20 August 2022
661JOURNAL OF PLANT ECOLOGY | VOL 13 | October 2020 | 657–666
(χ2 = 13.17, P = 0.004), but no differences were detected for species
strength (χ2 = 1.35, P = 0.25) (Supplementary Fig. S3c and d). Apidae
had a higher value of species strength than other groups in all four
seasons (Supplementary Fig. S3c and d). When we considered plants
and pollinators that occurred in all four seasons (30 plant species, 29
pollinator species), the specialization values for plants showed some
seasonal differences (χ2 = 8.31, P = 0.04) but the species strength values
did not (χ2 = 5.22, P = 0.16) (Fig. 6a and b). Notably, plants showed
higher specialization in spring than in summer (Fig. 6a). For pollinators,
species-level indices did not differ among the seasons (specialization:
χ2 = 7.63, P = 0.054; species strength: χ2 = 3.44, P = 0.33). Hawkmoths
had higher values for specialization, and Apidae had higher values
for species strength than other groups in summer and autumn. In
winter, Apidae had higher values of species strength than butterflies,
hawkmoths and non-Apidae Hymenoptera (Fig. 6c and d), with more
plant species depending on them for pollination.
DISCUSSION
Network-level indices and seasonality
The pollination network of the Yongxing Island community showed
higher specialization and modularity and lower nestedness and
connectance in the rainy season when more floral resources were
available than in the dry season. This result contrasts with that of
Souza et al. (2018), who found higher specialization of the pollination
network in the Central region of Brazil during the dry season when
fewer floral resources were available. Most studies on temporal
networks in temperate regions have suggested that network-level
indices such as nestedness, specialization, modularity and connectance
are highly conserved between successive plant reproductive seasons
(Alarcón et al. 2008; Burkle and Irwin 2009; Dupont et al. 2009; Fang
and Huang 2012, 2016; Olesen et al. 2008; Petanidou et al. 2008).
However, other studies have shown temporal variation in network
properties (Olesen et al. 2008; Petanidou et al. 2008), with seasonal
variations more prominent than interannual variations (Olesen et al.
2008). In the Yongxing Island community, plant–pollinator interactions
occurred throughout the year, but summer and winter appeared to be
the most favorable seasons with the largest network sizes. This pattern
was different from that observed in a temperate forest, where spring
and autumn were the most favorable seasons (Basilio et al. 2006). Low
rainfall and high temperatures coincide to generate a soil water deficit
beginning in June, and this may strongly influence the autumn activity
decline on Yongxing Island. Meanwhile, two additional phenomena
Figure 1: Seasonal variation in the number of pollinator species from different functional groups sampled year-round in the Yongxing Island community.
Figure 2: Seven plant–pollinator interaction networks in the Yongxing Island community (graphs generated with the software Gephi). Plant and pollinator species are represented by green and red circles, respectively.
Dow
nloaded from https://academ
ic.oup.com/jpe/article/13/5/657/5891221 by guest on 20 August 2022
662 JOURNAL OF PLANT ECOLOGY | VOL 13 | October 2020 | 657–666
occurred in the winter season: a specialized pollinator functional group
(Lepidoptera) was more species-rich, and more specialized plants were
blooming. In the Yongxing Island community, all network qualitative
indices showed significant seasonal variations in a single network
throughout its annual cycle that were consistent with previous studies
(Basilio et al. 2006; Jordano et al. 2003; Sajjad et al. 2017). The seasonal
changes in network-level nestedness, connectance, specialization and
modularity suggest that species exhibited more specialized interactions
when the number of species and interactions was at a minimum.
Similar results have also been reported in subtropical areas (Sajjad
et al. 2017). Because specialization is not affected by system size and
sampling intensity (Blüthgen et al. 2006), our findings regarding
seasonal variations in specialization are likely robust.
One of our results was consistent with the expectation of higher
specialization and modularity and lower nestedness and connectance
in the rainy season when a greater richness of plant species was in
full bloom. However, network-level specialization and modularity
have a consistent, opposite trend in seasonal variation compared with
plant species number, while nestedness and connectance showed
no such trend with plant species number. A possible explanation is
that oceanic island networks typically have lower pollinator to plant
ratios and fewer interacting species because isolation from mainland
land masses presents a major barrier to the arrival of many plant and
animal species (Bernardello et al. 2001; Gillespie and Roderick 2002;
Padrón et al. 2009; Trøjelsgaard and Olesen 2013). Thus, some animal
species that successfully colonize isolated islands tend to broaden
their trophic niches (Olesen et al. 2002). Oceanic island networks are
usually smaller and topologically simplified, with a lower interaction
diversity and a higher plant niche overlap than mainland areas
(Traveset et al. 2016), potentially leading to greater seasonal variation
in pollination networks. When larger numbers of plant species are in
full bloom in the oceanic island community, pollinators may continue
to interact with a greater number of species in order to survive in this
low diversity ecosystem, ultimately increasing the generalization of
pollination networks.
The expansion of feeding niches with increasing numbers of plant
species in full bloom may characterize entire communities and reduce
the specialization and modularity of pollination networks in oceanic
island communities. Also, pollinators in such communities behave
as generalists and forage less selectively when resources are rich but
become more selective when resources are rare. Seasonal changes in
the number of plant species in full bloom and contrasting patterns of
variation in network-level metrics between the dry and rainy season
networks and the four seasons networks suggest that different network
timescales may affect the analysis of the pollination system. Our study
strengthens the viewpoint that a proper description of plant–pollinator
interactions in communities with year-round activity should include
finer timescales and not just a few months of dry and rainy seasons.
Our results suggest that the main factor driving seasonal variation
of plant–pollinator interactions was interaction rewiring, even
though the plant and pollinator species also varied between seasons,
suggesting that the expansion of feeding niches with increasing
Figure 4: Results for interaction turnover (βint
) and interaction rewiring (βrw
) rates in the Yongxing Island community. High βint
and βrw
values indicate high dissimilarity between interactions and new types of plant–pollinator interactions.
Figure 3: Seasonal variation of eight properties of the plant–pollinator networks in the Yongxing Island community. For each property, data points represent the value of individual season network.
Dow
nloaded from https://academ
ic.oup.com/jpe/article/13/5/657/5891221 by guest on 20 August 2022
Copyedited by: GB
JOURNAL OF PLANT ECOLOGYRESEARCh ARTiCLE
663JOURNAL OF PLANT ECOLOGY | VOL 13 | October 2020 | 657–666
numbers of plant species lead to interaction rewiring, which may
influence the seasonal dynamic variation of networks structure.
Similarly, CaraDonna et al. (2017) found that week-to-week turnover
of interactions is also dominated by interaction rewiring in a subalpine
ecosystem. Interaction rewiring can play a consistently dominant
role in influencing the structure and dynamics of ecological networks
(CaraDonna et al. 2017). The richness of seasonal co-occurring flowers
and pollinators, and changes in this richness could directly influence
plant–pollinator interactions, and each of them has the potential to
increase interaction rewiring. The temporal dynamic variation in
pollination networks seems to be system-specific and factors such as
seasonality, diversity and composition could influence the interactions
between plant and pollinator and the degree of change over time.
Studies of fine-scale temporal dynamics of plant–pollinator interactions
of communities with different habitats and climate conditions and
their driving factors are particularly important in improving our ability
to estimate the reassembly and resilience of communities with respect
to climate change.
We recognize that conclusions based on the number of plant
species in full bloom as a measure of floral resource availability are
imperfect. The ideal means of assessing resource availability is to
obtain estimates of the sugar and amino acid contents of nectar and
pollen (Zimmerman and Pleasants 1982), but such estimates are not
feasible in many situations. Hegland and Totland (2005) argued for
the use of proxies because the number of flowers was related to the
nectar amount in several studies. However, counting flowers may
also generate imprecise estimates of food availability (Benadi et al.
2014) because pollinators prefer dense patches that reduce search
costs (Hegland and Totland 2005) and may use patches, rather than
individual flowers or inflorescences, as cues to find food resources
(Dauber et al. 2010). On the other hand, the number of plant species
in full bloom is usually positively correlated with pollinator species
richness (Ebeling et al. 2008), and some studies have used only
presence–absence data to predict floral resource availability (Kitahara
et al. 2008). Furthermore, higher species richness is correlated with
higher resource diversity. In our community, the observed plant species
were all in full bloom and distributed all over the island, suggesting that
floral resource availability would increase with increasing numbers of
plant species in full bloom. Therefore, the number of plant species in
full bloom probably reflects the floral resource availability relatively
effectively, suggesting the accuracy of our results. We investigated a
single year of plant–pollinator interaction data, which may not be fully
representative of the Yongxing Island community over the long term.
Although floral resource composition varies considerably among years
(Alarcón et al. 2008), our study nonetheless provides a snapshot of
plant–pollinator interactions in the Yongxing Island community. In
Figure 5: Plants and pollinators that occurred both in dry and rainy seasons: (a) species-level specialization and (b) species strength for plant species in dry and rainy seasons; (c) species-level specialization and (d) species strength for pollinator functional groups (mean ± SE). Diptera and Hymenoptera in (c) and (d) represent non-Syrphidae Diptera and non-Apidae Hymenoptera, respectively. Different letters in (c) and (d) represent significant differences (Tukey post hoc test, P < 0.05).
Dow
nloaded from https://academ
ic.oup.com/jpe/article/13/5/657/5891221 by guest on 20 August 2022
Copyedited by: GB
JOURNAL OF PLANT ECOLOGY RESEARCh ARTiCLE
664 JOURNAL OF PLANT ECOLOGY | VOL 13 | October 2020 | 657–666
the future, the temporal dynamics of plant–pollinator interactions
in oceanic island communities should be further investigated across
months, seasons and years.
Species-level indices and seasonality
Neither all the plant and pollinator species nor the species that occurred
in both dry and rainy seasons showed significant differences in species-
level specialization and species strength between the two seasons, a
finding that was inconsistent with the network-level result. A possible
reason for this inconsistency was that our study included data from
only 1 year, which may have been sufficient to fully characterize the
seasonal dynamics of the network-level metrics. Therefore, successive
multiyear investigations of pollination networks in oceanic island
communities are needed.
Similarly, there were no significant differences among the four
seasons in species-level specialization and the species strength of plants.
For plants that flowered in all four seasons, no significant differences
in species-level species strength were detected, although plants in the
summer showed lower specialization than plants in the spring. It is
possible that higher summer resource availability, coupled with more
active pollinator species, resulted in lower specialization. Among
pollinators, Hawkmoths had higher values of specialization in the
summer and autumn than in the spring and winter. More specialized
plant species were blooming and more hawkmoth species were active
in the winter on Yongxing Island, and this may have generated lower
specialization values. Some plants were mainly visited by hawkmoths
and rarely or never visited by other pollinator functional groups.
Apidae had lower values of specialization and higher values of species
strength than other pollinator functional groups, suggesting that most
plant species visited by Apidae are used only by this functional group.
When we considered pollinators active in all four seasons, Apidae and
Syrphidae had a higher value of species strength than other functional
groups in the winter. One explanation is that some small, green
flowers that occurred in winter were pollinated only by Syrphidae,
causing them to be more dependent on this pollinator functional group
(Ollerton 2017).
In summary, the present study provides a qualitative description
of seasonal changes in plant–pollinator interaction networks
on an oceanic island. Although the rainy season network was
characterized by more plant species in full bloom and showed
higher network-level specialization and modularity, network-
level specialization and modularity have a consistent, opposite
trend in seasonal variation compared with plant species richness.
Moreover, the changes in niche overlap between pollinators lead
to interaction rewiring which drives interaction turnover in this
community. Our results also indicate that data collected during the
dry season, the rainy season, and throughout the year generates
lower estimates of network specialization than data collected during
four individual seasons, perhaps because most pollinator species’
activity spans periods longer than a single season. Depending on the
season, different networks may be found because the aggregation of
temporally extensive data can generate many temporal ‘forbidden
links’ (Carvalheiro et al. 2014). Variation in the number of plant
species across seasons offers the opportunity to learn about the
processes that structure pollination networks, and it is important to
understand these processes in insular, species-poor oceanic island
networks in which flowering occurs year-round. Studies include not
only seasonal information on species interactions but also resource
Figure 6: Plants and pollinators that occurred in all four seasons: (a) species-level specialization and (b) species strength for plant species in four seasons; (c) species-level specialization and (d) species strength for pollinator functional groups (mean ± SE). Hymenoptera in (c) and (d) represent non-Apidae Hymenoptera. Different letters in (a), (c) and (d) represent significant differences (Tukey post hoc test, P < 0.05).
Dow
nloaded from https://academ
ic.oup.com/jpe/article/13/5/657/5891221 by guest on 20 August 2022
Copyedited by: GB
JOURNAL OF PLANT ECOLOGYRESEARCh ARTiCLE
665JOURNAL OF PLANT ECOLOGY | VOL 13 | October 2020 | 657–666
use might be important and necessary for accurate analyses of
oceanic islands communities.
Supplementary MaterialSupplementary material is available at Journal of Plant Ecology online.
Table S1: Season-to-season turnover values for all dissimilarity indices
calculated for the plant–pollinator community in Yongxing Island (Sp:
spring, Su: summer, Au: autumn, Wi: winter).
Table S2: Spearman’s correlation results for the relationships between
dissimilarity measures. βint
: interaction turnover; βrw
: interaction
rewiring; βst: interaction turnover due to species dissimilarity; β
s:
species turnover.
Table S3: Results from mixed-effects models of species-level network
metrics. Significance of the terms was obtained from a likelihood ratio
test in which deviances of the models with and without a specific fixed
variable were compared (P < 0.05 shown in bold).
Table S4: Values of specialization (d′) and species strength for plant
species recorded in dry and rainy seasons during our studied period.
Table S5: Values of specialization (d′) and species strength for pollinator
species recorded in dry and rainy seasons during our studied period
(letters F and M represent female and male individuals, respectively).
