Introduction
Network analysis tools are important and have become widely used in pollination studies (Vázquez et al. Reference Vázquez, Blüthgen, Cagnolo and Chacoff2009). It is widely known that the generation of plant–pollinator interaction networks involves the summation of flower visitation data collected over periods of days, weeks, months, or even years. Studies on plant–pollinator interaction networks have addressed temporal structure over both the seasonal and annual timescale (Burkle et al. Reference Burkle, Marlin and Knight2013, Dupont et al. Reference Dupont, Padrón, Olesen and Petanidou2009, Fang & Huang Reference Fang and Huang2012, Reference Fang and Huang2016; Funamoto Reference Funamoto2019, Kaiser-Bunbury et al. Reference Kaiser-Bunbury, Muff, Memmott and Müller2010, Medan et al. Reference Medan, Basilio, Devoto, Bartoloni, Torretta and Petanidou2006, Petanidou et al. Reference Petanidou, Kallimanis, Tzanopoulos, Sgardelis and Pantis2008, Sajjad et al. Reference Sajjad, Saeed, Ali, Khan, Kwon and Devoto2017, Souza et al. Reference Souza, Maruyama, Aoki, Sigrist, Raizer, Gross and de Araujo2018). However, not all of the detectable links between plants and pollinators occur at the same time. The impact of certain plant–pollinator interactions depends critically on their relative timing (Baldock et al. Reference Baldock, Memmott, Ruiz-Guajardo, Roze and Stone2011). For example, plant–pollinator interactions over field seasons or years generate many temporal ‘forbidden links’ (Carvalheiro et al. Reference Carvalheiro, Biesmeijer, Benadi, Fründ, Stang, Bartomeus and Baldock2014, Jordano et al. Reference Jordano, Bascompte and Olesen2003), in which specific plant–pollinator interactions cannot be discovered due to differences in plant and pollinator seasonal phenology. Nonetheless, the finer-scale temporal dynamics of plant–pollinator networks, such as variations in the network structure within a single day, have received limited attention so far (Baldock et al. Reference Baldock, Memmott, Ruiz-Guajardo, Roze and Stone2011, Fründ et al. Reference Fründ, Dormann and Tscharntke2011, Olesen et al. Reference Olesen, Bascompte, Elberling and Jordano2008). Changes in the flower opening patterns of plant species within a single day may affect the whole pollination network through the dynamics of competition for pollinators. Although optimal flower opening times could be predicted by models (Miyake & Yahara Reference Miyake and Yahara1999), studies focused on the effect of flower opening and closure on network dynamics are lacking.
Flowers of many plant species open for a limited time period during the day, and the resulting flowering pattern is influenced by the endogenous rhythms of plants induced by changes in light, temperature and humidity (van Doorn & van Meeteren Reference van Doorn and van Meeteren2003, von Hase et al. Reference Von Hase, Cowling and Ellis2006). Therefore, flowers of such species often close in the morning, around midday, or early afternoon (van Doorn & Kamdee Reference van Doorn and Kamdee2014). The daily pattern of flower opening and closure, however, has rarely been connected to plant–pollinator interactions (Fründ et al. Reference Fründ, Dormann and Tscharntke2011), which can provide information on pollination network structure, and can help predict community dynamics (Kaiser-Bunbury & Blüthgen Reference Kaiser-Bunbury and Blüthgen2015, Vázquez et al. Reference Vázquez, Blüthgen, Cagnolo and Chacoff2009).
Oceanic islands are particularly well-suited for studying the effects of flower opening and closure on plant–pollinator interactions and community structure over the one-day timescale. On the one hand, global ecosystems markedly differ in floral longevity, with plants on islands tending to have the shortest-lived flowers due to lower elevation, higher temperature and lower water availability than those inhabiting mainlands (Jorgensen & Arathi Reference Jorgensen and Arathi2013, Vesprini & Pacini Reference Vesprini and Pacini2005). On Yongxing Island (Paracel Islands, South China Sea), a high proportion of plant species opens their flowers in the morning and closes them around noon, while other species only open them during the afternoon (flower opening and closure times for 37 species are presented in Table S1). Oceanic islands harbour plants flowering for several hours a day, thus providing an ideal community to study the daily temporal characteristics of pollination networks. On the other hand, pollinators are a limiting resource on such islands due to their isolation and the relatively poor dispersal ability of insects (Bernardello et al. Reference Bernardello, Anderson, Stuessy and Crawford2001, Gillespie & Roderick Reference Gillespie and Roderick2002, Traveset et al. Reference Traveset, Tur, Trøjelsgaard, Heleno, Castro-Urgal and Olesen2016, Whittaker & Fernández-Palacios Reference Whittaker and Fernández-Palacios2007). The reproductive success of co-flowering plant species influence each other through the sharing of pollinators, as well as competition for pollination services (Mitchell et al. Reference Mitchell, Flanagan, Brown, Waser and Karron2009, van der Kooi et al. Reference van Der Kooi, Pen, Staal, Stavenga and Elzenga2016). Highly generalized species easily settle in island communities (Kaiser-Bunbury et al. Reference Kaiser-Bunbury, Memmott and Müller2009), although most islands are inhabited by a limited number of insect species, thus plants may be subject to strong interspecific competition for pollinators (Traveset et al. Reference Traveset, Heleno, Chamorro, Vargas, McMullen, Castro-Urgal and Olesen2013). Competition for pollinator visits can result in the structuring of communities by competitive exclusion, thus leading to minimized competition between plant species for pollinators in communities structured in this manner (Pauw Reference Pauw2013). One adaptation to minimize this competition may be a shift in flowering time (Devaux & Lande Reference Devaux and Lande2010). The opening and closure of flowers may be regulated by light and temperature signals from the environment (Maguvu et al. Reference Maguvu2018), although they may also be influenced by biotic factors, such as pollinators. One of the most challenging questions is how plant species coexist when they share a limited pool of pollinators, as pollinator sharing may result in the deleterious effect of interspecific pollination that hinders plant reproductive success for both interacting plant species (Flanagan et al. Reference Flanagan, Mitchell, Knutowski and Karron2009). Competition for pollination services may also cause the ecological sorting of flowering phenology (Aizen & Vázquez Reference Aizen and Vázquez2006). The temporal separation of blooming time of co-flowering species may reduce competition for pollination and permit coexistence. However, the question of whether flower opening and closure time could affect plant–pollinator interactions, and the mechanism of sharing a limited pollinator pool among co-flowering plants remain largely unexplored.
In the current study, the effects of early flower closure and late flower opening on plant–pollinator interaction networks are analysed. The following questions are addressed: (1) Is there temporal variation between morning and afternoon in plant–pollinator interaction networks and species-level specialization? (2) Do the pollinator species of early-closing plants only visit them during the morning and switch to other plants flowering in the afternoon? Is there a difference between morning and afternoon in the visitation rate of shared pollinators of whole-day flowering plants?
Materials and Methods
Study site and periods
The Xisha Islands (Paracel Islands) are a series of coral islets, formed about 7000 years ago through coral growth and crust uplift (Taylor & Hayes Reference Taylor and Hayes1980), located in the South China Sea (15°46′–17°08′N, 110°11′–112°54′E). Yongxing Island (16°49′N, 112°20′E), with a total area of 2.6 km2, is the largest islet of this archipelago. The data on flowering plants and potential pollinators were collected from 1 July to 31 August 2018 in Yongxing Island community. Sampling was conducted in 30 quadrats sized 5 × 5 m2, at least 10 m away from each other. Plant species in blooming had not changed across the two-monthly period, and all pollinator species were present over the full two months. To maximize the possibility of detecting different floral visitors and reduce the effects of geographic distribution, we collected data at quadrats rather than species.
Pollination data collection
Pollinators were investigated on sunny days without wind over two consecutive months, between 8:00 h and 18:00 h (the webs have data before 12:00 h and after 13:00 h). Flowers or flower heads were counted for each plant species of 30 quadrats. In each quadrat, visitation was quantified for each plant species with at least 15 floral units. Animal visits to flowers of each plant species were recorded during 30-min observations in each sampling interval (08:00–12:00 and 13:00–18:00). Each quadrat was observed for two consecutive days. For each observation interval, each visit was recorded if a visitor contacted with the anther and/or stigma for more than 1 s of a floral unit. Return back visits of the same individual in its foraging bout to the same floral unit were counted as a second visit. Therefore, all flower visiting insects that feed on flowers were recorded, regardless of the efficacy of their visit. The sampling times in AM and PM were long enough (nearly 200 hours) for collecting the pollination data. All flower visitors were firstly morphotyped and noted to family in the field and captured for further identification to the lowest possible taxonomic level in the laboratory by entomologists (see Acknowledgements).
To verify if the flower visitors are the potential pollinators for each plant species, the pollen loads from five collected insect individuals of each species were examined for pollen analysis. Pollen grains on the body of visitors were taken off with gelatin cubes and then transferred to a clean slide (Tong & Huang Reference Tong and Huang2018). The slides were warmed gently to melt the jelly, and pollen samples were viewed under electron microscope. Pollen grains were identified by comparing with a reference library of pollen based on those removed from field-collected flowers (Fang & Huang Reference Fang and Huang2013). If four or five specimens of a plant-flower visitor pair carried the host plant pollen, we presumed that the visitor was a potential pollinator (hereafter, named pollinator), although we did not evaluate the role in the subsequent production of fruits to visited plant species. Vouchers for all plant species were collected, identified and deposited in the SCBG herbarium (international acronym: IBSC). Samples of collected insects were deposited at the South China Botanical Garden, Chinese Academy of Sciences.
Data analysis
We conducted two quantitative bipartite plant–pollinator networks from the two-monthly data for Yongxing Island community, using visitation frequencies of pollinator species to each plant species as a surrogate of interaction strength. To analyse the temporal dynamics of plant–pollinator interactions within a single day, we separated the network into morning (AM) and afternoon (PM) subwebs. For this purpose, ‘noon’ was defined as the time of solar zenith (13:00 h) (Fründ et al. Reference Fründ, Dormann and Tscharntke2011). We calculated metrics illustrating distinct structural properties of the network which were previously shown to be less sensitive to sampling effort (Fründ et al. Reference Fründ, McCann and Williams2016, Vizentin-Bugoni et al. Reference Vizentin-Bugoni, Maruyama, Debastiani, Duarte, Dalsgaard and Sazima2016). Connectance (and weighted connectance) which is the fraction of interactions occurring in the network, increasing with network generalization. Nestedness (and weighted nestedness) quantifies the degree of interactions of specialized species in the network. Network-wide specialization (H2′) describes if species restrict their interactions from those randomly expected based on a partner’s availability (Blüthgen et al. Reference Blüthgen, Menzel and Blüthgen2006), ranging from 0 (low specialization) to 1 (high specialization). Network modularity (Q) quantifies the prevalence of interactions within subsets of species in the community. Q was calculated using the QuaBiMo algorithm based on a hierarchical random graph approach adapted for quantitative bipartite networks (Dormann & Strauss Reference Dormann and Strauss2014). We ran the QuaBiMo algorithm 10 times and retained the iteration with highest Q value, as the algorithm is a stochastic process. NODF is the nestedness measure, correcting for matrix fill and matrix dimensions, ranging from 0 (non-nestedness) to 100 (perfect nesting). Network metrics can be influenced by intrinsic characteristics such as the number of interacting species and sampling effort (Blüthgen et al. Reference Blüthgen, Menzel and Blüthgen2006, Fründ et al. Reference Fründ, McCann and Williams2016, Vizentin-Bugoni et al. Reference Vizentin-Bugoni, Maruyama, Debastiani, Duarte, Dalsgaard and Sazima2016); hence the significance of metrics is assessed by comparison with null model networks. We used the Patefield null model, which fixes the network size and the marginal totals while shuffling interactions randomly (Patefield Reference Patefield1981). We estimated the 95% confidence interval for each metric from the 1000 simulated values, and a metric value was considered significant if it did not overlap with the confidence interval. The quantitative bipartite plant–pollinator networks were compared between two subwebs using Bray–Curtis dissimilarity of visitation frequencies.
To check the role of species within networks and how it varies between morning and afternoon, we calculated two species-level indices that present distinct topological properties of a species: (1) species-level specialization index d′, which quantifies the level of specialization of each species based on its discrimination from random selection of interacting partners, with higher values indicating higher specialization (Blüthgen et al. Reference Blüthgen, Menzel and Blüthgen2006), ranging from 0 (no specialization) to 1 (perfect specialists); and (2) species strength, which is the sum of the proportions of interactions performed by a given species across all its interaction partners, with higher values indicating that more pollinators depend on a specific plant species, and vice versa (Bascompte et al. Reference Bascompte, Jordano and Olesen2006). Web plots and all network-related indices calculation were created with the ‘bipartite’ package version 2.05 (Dormann et al. Reference Dormann, Gruber and Fründ2008) in R3.4.4 (R Development Core Team Reference Development Core Team2016). We compared species-level specialization and species strength of whole-day flowering plant species and different functional pollinator groups between morning and afternoon subwebs by one-way analysis (ANOVA) of variance using Tukey HSD’s method in IBM SPSS 19.0, and data were presented as mean ± SE.
To assess the effect of half-day blooming flowers on the competition of pollinators among pollinator-sharing plants in the community, we calculated separately for morning and afternoon visitation rate to five whole-day flowering plant species (Morinda citrifolia, Sesuvium portulacastrum, Scaevola taccada, Tridax procumbens and Wedelia trilobata) which were mainly specialized visited by the most common pollinator Apis cerana in the community. Since A. cerana was the main pollinator species of the six half-day flowering plant species, we compared the difference of visitation rate by A. cerana among four morning flowering plant species and two afternoon flowering plant species. We also calculated separately for morning and afternoon visitation rate of six whole-day flowering plant species by another two Apidae visitor species in the community, because these two visitor species were also shared by half-day flowering plants and whole-day flowering plants: Ceratina lieftincki visiting Phyla nodiflora, Tridax procumbens and Wedelia trilobata, and Braunapis puangensis visiting Messerschmidia argentea, Scaevola taccada and Tridax procumbens. Finally, we calculated separately for morning and afternoon visitation rate of Canavalia maritima and Cordia subcordata which were high specialized species and attracted different pollinator spectrum. Visitation rate was calculated as the number of visits from each pollinator species per hour per flower for each plant species. We used generalized linear mixed-effects models (negative binomial) for the visitation rate data. Morning or afternoon was included in the model as fixed effect, and species identity was included as a random effect using the ‘lme4’ R package. Visitation rate of the plant species by the same pollinator species were compared between morning and afternoon by one-way analysis (ANOVA) of variance using Tukey HSD’s method in IBM SPSS 19.0 and data were presented as mean ± SE.
Results
The present study yielded a total of 33,864 interactions between 32 pollinator species and 37 plant species distributed across 21 families for all of the evaluated periods. In the AM subweb, 22,735 visits between 34 plant species and 28 pollinator species were recorded, compared with 11,129 visits between 25 plant species and 29 pollinator species in the PM subweb. More than a third (34.0%) of all observed interactions in the AM subweb were associated with 10 morning flowering plant species, while the PM subweb was characterized by three afternoon flowering plant species accounting for 20.7% of all observed interactions (Figure 1). The flower opening and closure times in a single day for 37 plant species are shown in Table S1. Plant–pollinator interactions involving plants blooming in the morning or in the afternoon occurred mainly between 9 am and 10 am, or 4 pm and 5 pm, respectively (Figure 1). Pollinator specimens observed were representatives of Apidae, non-Apidae Hymenoptera, Syrphidae, non-Syrphidae Diptera, butterflies, hawkmoths, Arctiidae and Passeriformes. The dominant pollinator taxon was Apidae representing 81.3% and 81.6% of all recorded visits in the morning and afternoon networks, respectively, accounting for more visits in total than the other functional groups combined (Figure 1). In addition, visits by A. cerana made up 61.8% and 88.9% of all Apidae visits in the morning and afternoon networks, respectively.
Figure 1. Circadian pattern of flower visitation to (a) all plants in the Yongxing Island community and (b) plant species blooming only in the morning or afternoon. (a) and (b) suggest that Apidae was the most active pollinators through the whole day, and plants flowering during the morning or afternoon were competition for their pollination service.
Clear temporal variation in the plant–pollinator interactions was observed upon visual comparisons of morning and afternoon networks (Figure 2). The Bray-Curtis dissimilarities of plant–pollinator interaction between morning and afternoon subsets was quantified as 0.69, and this value dropped to 0.58 when only whole-day flowering plants were considered. Meanwhile, most network-level indices presented higher values for the morning when compared with the afternoon network (Table 1).
Figure 2. Quantitative plant–pollinator interaction networks change between morning and afternoon in Yongxing Island community. (a) and (c) Morning subweb (all interactions observed before 12:00 h), (b) and (d) Afternoon subweb (after 13:00 h). The rectangles represent pollinator species (top row) and plant species (bottom row), and the connecting lines represent links between plant and pollinator species. The width of rectangles and connecting lines are proportional to the visitation frequency per nodes. Pollinator group are colour-coded as follows: red, Apidae; yellow, non-Apidae hymenoptera; blue, Syrphidae; brown, non-Syrphidae diptera; pink, butterflies; purple, Arctiidae; cyan, Passeriformes; orange, hawkmoths. Plant species are colour-coded as follows: green, whole-day flowering plants; black, morning flowering plants; grey, afternoon flowering plants. All networks are drawn to the same scale. Full species names of plants, and flowering open and closure times are available in Table S1. Full species names of pollinators and the original links between plants and pollinators in the whole day are available in Table S2.
Table 1. Network metrics for the quantitative plant–pollinator networks from Yongxing Island community, showing the values for morning, afternoon and whole-day networks. * represents network metrics which are significant (do not overlap null model expectations) (95% confidence interval).
Neither species-level differences in specialization (F = 0.394, P = 0.53) or species strength (F = 0.222, P = 0.64) were detected for whole-day flowering plant species, nor did species-level indices differ between morning and afternoon networks for the different functional groups (Table 2). Furthermore, certain plant and pollinator species presented higher specialization and species strength in morning networks compared with afternoon ones, while others showed the opposite trend (Figures 3 and 4).
Table 2. Comparison of species-level species strength and specialization (d′) (means ± SE) of each functional pollinator group between morning and afternoon networks in the Yongxing Island community.
Figure 3. Changes of species-level specialization (a) and species strength (b) of each plant species between morning and afternoon in the Yongxing Island community.
Figure 4. Changes of species-level specialization (a) and species strength (b) of each pollinator species between morning and afternoon in the Yongxing Island community. Functional pollinator groups are colour-coded as follow: red, Apidae; yellow, non-Apidae hymenoptera; green, Syrphidae; orange, non-Syrphidae diptera; blue, butterflies; purple, Arctiidae; pink, Passeriformes; grey, hawkmoths.
For whole-day flowering plants, the rates of flower visitation by A. cerana were not significantly different between the morning and the afternoon (z = 1.580, P = 0.11): Morinda citrifolia (F = 1.887, P = 0.18), Sesuvium portulacastrum (F = 0.328, P = 0.58), Scaevola taccada (F = 0.614, P = 0.44) and Tridax procumbens (F = 0.376, P = 0.54); however, Wedelia trilobata flowers were visited less frequently during the afternoon than the morning (F = 8.065, P = 0.007) (Figure 5). Nonetheless, plant species blooming only in the morning (four species) or the afternoon (two species) showed some differences in the rates of visitation by A. cerana (df = 5, F = 7.547, P < 0.001). Notably, the visitation rate of Sesbania cannabina, which flowers in the afternoon, was significantly higher than that of plants flowering during the morning (Figure 5). The rates of visitation by Ceratina lieftincki of whole-day flowering plant species, specifically of Phyla nodiflora (F = 1.592, P = 0.22), T. procumbens (F = 0.550, P = 0.46) and W. trilobata (F = 3.540, P = 0.07) did not significantly differ between morning and afternoon (z = 0.836, P = 0.40). Meanwhile, the rates of visitation by Braunapis puangensis of whole-day flowering plant species showed certain differences between morning and afternoon (z = 3.163, P = 0.002): S. taccada (F = 2.219, P = 0.15) and T. procumbens (F = 1.628, P = 0.12), however, the visitation rate of Messerschmidia argentea (F = 13.211, P = 0.001) was reduced in the afternoon (Figure 6). No difference was detected between the visitation rates of two plant species attracting a different pollinator spectrum before and after noon: C. smaragdula visiting Canavalia maritima (F = 0.711, P = 0.41); Zosterops japonicus visiting Cordia subcordata (F = 2.801, P = 0.12).
Figure 5. (a) Comparison of visitation rate (means ± SE) of Apis cerana to whole-day flowering plant species between morning and afternoon. These plant species have a pollinator spectrum overlapping with morning flowering plant species; (b) Comparison of visitation rate (means ± SE) of Apis cerana among morning flowering plant species (white) and afternoon flowering plant species (black). * and different letters represent that significant difference are detected using one-way ANOVA, P < 0.05. For (b), significant differences according to the post hoc Tukey tests are indicated by different letters (Tukey test: P < 0.05).
Figure 6. Comparison of visitation rate (means ± SE) of (a) Ceratina lieftincki and (b) Braunapis puangensis to whole-day flowering plant species between morning and afternoon. These plant species have a pollinator spectrum overlapping with morning flowering plant species. * represents significant difference detected using one-way ANOVA, P < 0.05.
Discussion
The data obtained demonstrate that most plant–pollinator interactions occurred during the morning, which is likely the result of the domination of morning flowering plant species in the Yongxing Island community. The observed interactions in the morning and afternoon might be explained by the sampling design, where observation time was not relatively homogeneously distributed across the main period of pollinator activity (Schwarz et al. Reference Schwarz, Vázquez, CaraDonna, Knight, Benadi, Dormann and Fründ2020). Early flowering plant species had high visitation rates by pollinators, thus early flower closure and late flower opening may have a significant impact on plant–pollinator networks. In our model, the quantitative dissimilarity of plant–pollinator interaction between morning and afternoon networks was lower when only whole-day flowering plants were considered. Moreover, the morning subweb presented higher values for most network indices when compared with the afternoon subweb, thereby the temporal variation of network structure within a single day might be influenced by the presence of half-day flowering plants closing their flowers around noon or opening their flowers during the afternoon.
In the plant–pollinator community of our study, pollinators did not present a clear pattern of daily activity; their foraging activity declined in the afternoon but did not completely diminish. This contrasts with the results of Fründ et al. (Reference Fründ, Dormann and Tscharntke2011) for a community in Southern Germany, where a number of pollinator species specialized on Cichorioideae were absent during the afternoon, while their flowers closed around noon. According to other studies, the within-day variation in insect activity patterns was considered likely due to abiotic or biotic factors (Lienhard et al. Reference Lienhard, Mirwald, Hötzl, Kranner and Kastberger2010, Stone et al. Reference Stone, Gilbert, Willmer, Potts, Semida and Zalat1999). In the community of the present study, the lower afternoon foraging activity might be due to the morning peak of insect activity rather than a preference for morning half-day flowering plants. Although the differences in species-level specialization and species strength of whole-day flowering plant species and various functional pollinator groups were insignificant between morning and afternoon networks, these two indices of plant and pollinator species still showed a certain level of temporal variation within a single day. Thus, no stable role of these plant and pollinator species could be established over a one-day timescale, suggesting that plant species share a pool of generalist pollinators, which provide a certain level of pollination and plant population maintenance on oceanic islands.
In our case study, a sizeable proportion of plant species (35%) were only flowering either in the morning or the afternoon, thus reducing competition for pollination services. It is considered that the half-day flowering pattern of some plant species allows for relying on the same pollination resource, and contributes to maintaining community stability in the long term. This study showed no changes in active pollinator species throughout the day, but the same pollinator species shifted from morning flowering plants to afternoon flowering plants maximizing the utilization of floral resources throughout the day. Morning and afternoon flowering plants both mainly relied on pollination by Apidae, which may generate fierce competition for pollination services. Correspondingly, visitation rates of whole-day flowering plants by A. cerana, C. lieftincki and B. puangensis showed no significant differences between the times of day. Nevertheless, the flower visitation rate of M. argentea by B. puangensis was reduced during the afternoon. One potential reason is that M. argentea floral resources are significantly larger in the morning (Wang et al., unpublished data). In addition, the target of A. cerana visits changed from morning flowering plants to afternoon flowering plants, as evidenced by the significantly higher visitation rate of afternoon flowering S. cannabina by A. cerana, as compared with morning flowering plants. These results were inconsistent with those of Fründ et al. (Reference Fründ, Dormann and Tscharntke2011), namely that whole-day flowering plant species were in competition for pollinators of morning flowering plants, and that shared pollinators showed lower visitation rates in the morning. In the community subject to the present work, competition for pollinators among co-flowering plants sharing pollinators could be reduced due to the temporal segregation between morning and afternoon flowering species. In a similar manner, pollinators exploit Dalechampia heteromorpha that opens and produces pollen during the early morning hours, and subsequently switch to D. scandens that flowers exclusively in the afternoon, thereby minimizing competition between the two sympatric plant species (Armbruster & Herzig Reference Armbruster and Herzig1984). Overlap in the flowering times of mass-flowering dipterocarp species is significantly reduced by their staggering flowering periods, resulting in reduced competition for pollinators (Ashton et al. Reference Ashton, Givnish and Appanah1988). Consequently, on oceanic islands, the temporal dynamics of mutualistic and competitive interactions may be different from the patterns suggested for larger temporal scales. Hence competition for pollinators may be reduced among co-flowering plant species, while competition of pollinators is probably more pronounced between morning and afternoon flowering plants.
Theoretical work on pollination networks has shown that generalist core species make networks stable and tolerant to extinctions (Bascompte & Jordano Reference Bascompte and Jordano2007, Emer et al. Reference Emer, Memmott, Vaughan, Montoya and Tylianakis2016). Our analyses, however, question the assumption of functional redundancy. For example, pollinators need different plant species for providing resources at different times of the day. Nevertheless, the presence of morning or afternoon flowering suggests a flexible link structure, enabling pollinators to establish new links within the same plant community. Changes in plant–pollinator interactions over a one-day timescale should be considered in the interpretation of network data without whole-day sampling, as important interactions may be missed if sampling is restricted to a specific part of the day, and cross-day turnover may be overestimated in studies with varied sampling times (Fründ et al. Reference Fründ, Dormann and Tscharntke2011). The within-day shift pattern featured in plant–pollinator interactions may also occur in other communities, as many plant–pollinator networks are characterized by a similar circadian variation of interactions, and a high number of plant species show intra-day flower opening and pollen availability patterns (Fründ et al. Reference Fründ, Dormann and Tscharntke2011, Hoehn et al. Reference Hoehn, Tscharntke, Tylianakis and Steffan-Dewenter2008). In summary, the dynamics of flower opening and closure time need to be considered when designing field surveys involving pollinators, flower availability, and plant–pollinator interactions.
On oceanic islands, temporal plant–pollinator network dynamics may enhance biodiversity maintenance through temporal niche partitioning (Abdala-Roberts et al. Reference Abdala-Roberts, Parra-Tabla and Navarro2007), which may also manifest on a daily basis. As a result of reduced competition for pollinators, respective plant communities are expected to benefit from this pattern. Owing to the poor dispersal ability of insects, islands have a depauperate pollinator fauna and a relative low pollinator/plant ratio in turn (Bernardello et al. Reference Bernardello, Anderson, Stuessy and Crawford2001, Dupont et al. Reference Dupont, Hansen and Olesen2003, Gillespie & Roderick Reference Gillespie and Roderick2002, Stuessy et al. Reference Stuessy, Crawford, López-Sepúlveda and Ruiz2017, Traveset et al. Reference Traveset, Tur, Trøjelsgaard, Heleno, Castro-Urgal and Olesen2016, Trøjelsgaard & Olesen Reference Trøjelsgaard and Olesen2013). The half-day flowering pattern in oceanic island communities might thus be attributed to the comparatively small number of pollinators and subsequent high competition for pollinator services. Flowering periods limited to morning or afternoon might help to reduce the amount of resources required for open flowers, such as that of water (Ashman & Schoen Reference Ashman and Schoen1994), and may improve the visiting accuracy of pollinators by guiding them to flower specimens in need of pollination, thus enhancing pollination efficiency and effectiveness in these biogeographic areas. Since only one oceanic island was investigated in this study, more islands need to be targeted to explore the effect of within-day flowering time limit and other factors on plant–pollinator network structure.
Acknowledgements
We thank the following experts for identifying insect specimens: Prof. Zhu Chaodong and Yuan Feng, the Key Laboratory of Zoological Systematics and Evolution, Chinese Academy of Sciences, Beijing (Hymenoptera); Prof. Jia Fenglong, Zhongshan University (Diptera); Prof. Wang Min, South China Agricultural University (Lepidoptera). We highly appreciate the logistical support for fieldwork from Mr Long Zhenhua, Mr Huo Da and Mr Li Daning of Xisha Ocean observation and research station, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China. The authors would like to thank TopEdit (www.topeditsci.com) for its linguistic assistance during the preparation of this manuscript.
Financial support
This research was supported by National Natural Science Foundation of China (grant number 31800447), the Chinese Academy of Sciences (grant number XDA13020504) and Natural Science Foundation of Guangdong Province (grant number 2018A030310385).
Supplementary materials
For supplementary material for this article, please visit https://doi.org/10.1017/S0266467421000055
