Changes in pollinators' flower visits and activities potentially drive a diurnal turnover of plant‐pollinator interactions

Plant‐pollinator interactions are essential to sustaining biodiversity and ecosystem functioning in terrestrial ecosystems and are variable on several spatiotemporal scales. On a fine temporal scale, the responses of pollinators to temperature and floral resource dynamics are likely to be related to the temporal turnover of plant‐pollinator interactions. However, the temporal dynamics of plant‐pollinator interactions on a fine scale are largely unknown. The present study aims to reveal the temporal turnover of plant‐pollinator interactions over a single day and investigates these interactions in semi‐natural grasslands in a Japanese agricultural landscape from the early morning to the evening. The interaction turnover was evaluated as temporal β‐diversity and divided into two components: interaction rewiring and species turnover. Insect species richness and the number of interactions showed hump‐shaped responses to the time of the day, and these responses differed for the different insect groups. Furthermore, the peak time of insect visitation frequency differed among 11 plant species. Interaction turnover (total β‐diversity) in the same day was higher than that in different days. Although interaction rewiring in different days was higher than that in the same day, species turnover had an opposite pattern. Over a single day, the relative importance of interaction rewiring was higher in the morning, while species turnover was higher in the evening. Around noon, interaction rewiring and species turnover were equally important. Therefore, the daily rhythm of pollinator activities and changes in the main floral resources drive the temporal turnover of plant‐pollinator interactions over a single day.

Plant-pollinator interactions are essential for maintaining biodiversity and functioning in terrestrial ecosystems because most plant species, including crops, rely on animal-mediated pollination (Klein et al., 2007;Ollerton et al., 2011).Although this interaction is also treated as a temporal static entity, it has been revealed in recent years that plant-pollinator networks change on various temporal scales (Baldock et al., 2011;CaraDonna et al., 2017CaraDonna et al., , 2021;;CaraDonna & Waser, 2020;Chakraborty et al., 2021;Fründ et al., 2011).
These dynamics within a season and between years are well studied, and phenological matching between pollinators and flowering plants is one of the drivers of these dynamics (CaraDonna et al., 2017;CaraDonna & Waser, 2020).Pollination by insects and foraging of insects occur at shorter temporal scales (e.g., within a single day) than within seasons and between years, and these processes on the fine temporal scale could influence their fitness and long-term persistence.
The temporal dynamics of plant-pollinator interaction networks on a fine scale are likely to help to understand the long-term persistence of plant and pollinator communities.
Within a single day, both plants and pollinators respond to biotic and abiotic contexts and might change those interactions accordingly (CaraDonna et al., 2021;Ogilvie & Forrest, 2017).The daily rhythm of flower opening and closure is a response to abiotic context and is called Linné's floral clock.This daily rhythm is related to light and temperature (Doorn & Meeteren, 2003;Tanaka et al., 1988).In addition to these abiotic contexts, pollination as a biotic context affects flower closure time (Fründ et al., 2011).Fründ et al. (2011) (Pyke, 2016).Indeed, many previous studies indicate that the visitation frequency of pollinators varies within a single day depending on variations in floral resource availability and required resources (e.g., Hart & Eckhart, 2010;Magaly et al., 2017;Schaffer et al., 1979;Stone et al., 1996).Therefore, pollinators would change their flower visitation with an increase and/or a decrease in floral resources of focal plant species (i.e., Adaptive foraging; Valdovinos et al., 2010, 2013, Song & Feldman, 2014).At the same time as these biotic contexts, pollinators respond to some abiotic contexts, such as temperature, solar radiation, and wind speed (Brittain et al., 2013;Rader et al., 2013;Vicens & Bosch, 2000).Moreover, these responses are likely to be variable among pollinating insect species because of differences in optimal weather, especially temperature, for insect activity (Kühsel & Blüthgen, 2015).Depending on this response to temperature, the occurrence and the visitation frequency of specific pollinator species might also be variable across times of the day.Throughout these responses of pollinators to both biotic and abiotic contexts, plant-pollinator interactions turn with the time of the day.Previous studies that have focused on temporal dynamics within a single day investigated only changes in pollinator visitation frequency, pollinator communities, or some properties of plant-pollinator networks and did not explicitly assess the temporal dynamics of interactions at a community scale (e.g., Baldock et al., 2011;Fründ et al., 2011;Hart & Eckhart, 2010;Kikuchi, 1962Kikuchi, , 1963;;Magaly et al., 2017;Stone et al., 1996).Therefore, revealing the temporal turnover of plant-pollinator interactions over a single day is valid to understand these temporal dynamics on a fine scale.
Here, I focused on plant-pollinator interactions in semi-natural grasslands in a Japanese agricultural landscape and investigated those from the early morning to the evening.The temporal turnover of plant-pollinator interactions was evaluated as the temporal β-diversity of networks (Poisot et al., 2012, see also figure 1 in CaraDonna et al., 2017).Moreover, interaction turnover was divided into two components: interaction rewiring and species turnover (Poisot et al., 2012, see also figure 1 in CaraDonna et al., 2017).Interaction rewiring means that a specific species (i.e., insect) changes interacting partners (i.e., plant) to others (i.e., other plants) with the time of the day, which is likely to be related to the response to floral resources.
Species turnover means that species are gained or lost depending on the time of the day and is potentially related to the response to abiotic contexts, especially temperature.To reveal the temporal dynamics of plant-pollinator interactions over a single day, the present study addresses the following three points: (1) temporal changes in insect species richness, the number of interactions, and the link number per insect species, (2) temporal changes of the visitation frequency of insects to each plant, and (3) the relative importance of interaction rewiring and species turnover for different times of the day.

Field observation
I conducted field observation every 2 h from 08:00 to 16:00 for a total of five times per observation day.The observation of plant-insect interactions at each study site was conducted over 6 days (August 30 and 31,and September 1,21,23,and 27).In this period, the range in times of sunrise and sunset was approximately from 05:30 to 18:30.Plantinsect interactions were observed while walking for 10 min at each study site (a total of 300 min at each site).I observed what insect species were visiting and foraging on what plant species and recorded the species name of both insects and plants and their frequency.Flower visits of insects were determined by whether insects were foraging with proboscis extended or bees had their heads in corollas.I recorded morphospecies depending on body size and family for some insect species (e.g., flies) that could not be identified in the field.

Statistical analysis
I analysed the relationships between the time of the day and species richness, the number of interactions, and link numbers per insect species.Moreover, the relative importance of interaction rewiring and species turnover between survey times was compared.To check whether the sampling effort was sufficient, I estimated interaction richness (the number of interactions who visited whom) by calculating abundancebased coverage estimators (ACE, Chao & Lee, 1992, Chao et al., 1993) for each time of the day when I observed plant-insect interactions.
Then, I compared observed interaction richness and estimated interaction richness.Because the proportion sampling (observed/estimated) was 0.73 at the lowest (Table S1), the sampling effort of the present study is likely sufficient.Moreover, the significance of spatial autocorrelation among study sites for insect visitation frequency was also checked by calculating Moran's I statistics (Cliff & Ord, 1973), but no correlation found (Moran's I statistics = À0.22,p-value = 0.40).
The relationships between the times of the day and species richness of insects, the number of interactions between insects and plants, and the link number per insect species were analysed.To analyse these relationships, I developed generalised additive mixed models (GAMMs) using Bayesian modelling.Species richness of insects, the number of interactions, and link number per insect species were used for response variables.The surveyed time of the day was used as an explanatory variable.The Poisson distribution was used for the error term of the species richness model, the negative binomial distribution was used for the error term of the interaction number model, and the gamma distribution was used for the error term of the link number model.As for random terms, I used site ID and survey date.
The relationships between the time of the day and visitation frequency were analysed.The visitation frequency of each insect group and the visitation frequency to each plant species were analysed separately using GAMMs using Bayesian modelling.The plant species that were analysed were selected based on whether the species had more than 50 interactions.The zero-inflated negative binomial and the zeroinflated Poisson distributions were used for error terms in the insect model and plant model, respectively, to cope with zero-inflated data.
These analyses were performed for each insect group and each plant species.The visitation frequency of each insect group and the visitation frequency to each plant species were used for response variables, and the survey time was used as an explanatory variable.SiteID and survey date were used for random terms.
Following Poisot et al. (2012), the absolute turnover of plantpollinator interactions across each surveyed time of the day for each date (i.e., the temporal beta-diversity of plant-pollinator networks within a single day) was quantified using Whittaker's (1960) index: where β int.turn is interaction turnover (i.e., β-diversity of networks) between survey times of the day (e.g., 08:00-10:00 and 12:00-14:00), a is the number of pairwise interactions shared between networks, and b and c are the numbers of pairwise interactions unique to each of the networks.The value of this β-diversity index ranged from 0 to 1; a higher value indicates a larger difference between the two interaction networks.This β-diversity index can be divided into two components, interaction rewiring and species turnover: where β int.rew and β sp.turn are interaction rewiring and species turnover, respectively.Then, both components were calculated according to Poisot et al. (2012).
F I G U R E 1 Map of study sites.The image was obtained from Google Earth.
First, I compared β-diversity values between temporal interaction turnover within a day and that in whole surveyed days.Second, the relative importance of interaction rewiring and species turnover between survey times of the day (e.g., 08:00 vs. 10:00 and 12:00 vs. 14:00) was compared.These comparisons were performed by the linear mixed models (LMMs) using Bayesian modelling.The values of each β-diversity component were arcsin square root transformed for normality, and siteID and date were used for random terms.
The GAMMs and LMMs using Bayesian modelling were performed by the brm function of the 'brms' package (Bürkner et al., 2022).All MCMC sampling was conducted with default values of brm function (2000 iterations, 1000 warmups, 1 thinning, and 4 chains).The convergence of each estimated parameter was confirmed based on the R-hat values.Because all estimated parameters' R-hat values were less than 1.01, all estimated parameters converged.
The predictability of posterior distribution was visually confirmed.The calculation of ACE was conducted using ACE function of the 'fossil' package (Vavrek, 2020).The Moran's I test was performed by the 'spdep' package (Bivand et al., 2022).

RESULTS
I observed 2385 interactions among 90 insect species and 29 plant species (Figure 2, Table S2).The most common insects were bees (both in species richness and abundance), and there were also many hoverflies, other flies, and butterflies (Figure S1a, b).Beetles were more abundant but had lower species richness (Figure S1a, b).For plants visited by these insects, those belonging to Asteraceae and Fabaceae families were the most common flowers (Figure S2).
Species richness, the number of interactions, and link number per species Species richness, the number of interactions, and the link number per insect species were related to the time of the day.The species richness of insects presented a hump-shape response to times of the day (Figure 3a, GAMM; linear component's 95% CI = À0.48 $ 1.66, smoothing component's 95% CI = 0.67 $ 4.31).As with species richness, the number of interactions also showed a hump-shape response to times of the day (Figure 3b, GAMM; linear component's 95% CI = À0.73 $ 2.79, smoothing component's 95% CI = 0.93 $ 5.40).
The patterns showed fewer insects in the morning and the evening but more insects in the evening.On the other hand, the responses of visitation frequency of insects to times of the day differed for different insect groups (Figure S3).Bees, hoverflies, other flies, and butterflies presented hump-shaped responses (Table S3, Figure S3a, d, e, f), while moths presented a U-shaped response (Table S3, Figure S3g).
The other insect groups' visitation frequencies (i.e., wasps, ants, and beetles) were not related to the times of the day (Table S3, Figure S3b, c, h).The frequency of visits by insects to each plant differed for different plant species.Almost all plant species presented a hump-shaped response to times of the day, while A. triphylla var.
japonica presented a U-shaped response (Table S4, Figure S4).Moreover, the peaks of the estimated frequency of visits by insects to plants differed among plant species (Figure 4).

Temporal turnover of plant-pollinator interactions
The interactions between insects and plants changed both within the same day (Figure S5) and among days.The temporal β-diversity in the same day was higher than that on different days (Figure S6).Although the contribution of species turnover in the same day was higher than that on different days (Figure S6, GAMM; 95% CI of estimated coefficient = 0.15 $ 0.22), the contribution of interaction rewiring presented an opposite pattern (Figure S6, GAMM; Interaction rewiring, 95% CI of estimated coefficient = À0.44 $ À0.36, Species turnover, 95% CI of estimated coefficient = 0.43 $ 0.52).

DISCUSSION
The present study demonstrated that plant-pollinator interactions were variable over a single day.Insect species richness and visitation frequency also changed with the time of the day, and the peak time of insect visitation differed for different plant species.Furthermore, the temporal turnover of plant-pollinator interactions was divided into two components; interaction rewiring and species turnover, and the relative importance of these two components differed at different times of the day.Interaction rewiring was more important in the morning, while species turnover was more important in the afternoon.
Around noon, the two components were equally important.
The differences in the relative importance of interaction rewiring versus species turnover (Figure 3) appear to be related to temporal changes in insect activity (Figure 4).In the morning (08:00-10:00), species richness, the interaction number, and the link number per species increased (Figure 3), and interaction rewiring contributed more to the interaction turnover (Figure 4).As time progressed, the temperature increased in the morning (Figure S7), and as temperature increased, insects became more active (Figures S3 and S7) (Kühsel & Blüthgen, 2015).Species richness and the number of interactions would increase because of an increase in insect activity.The results showed that the degree of an increase in the interaction number tended to be larger than that in species richness (Figure 3) and the link number per insect species tended to increase with insect activity (Figures 3 and S3).Consequently, an increase in the number of interactions and the link number per species (i.e., interaction rewiring) would contribute to interaction turnover more than to that species richness (i.e., species turnover).On the other hand, in the afternoon (14:00-16:00), both species richness and interaction number decreased, but the link number per species did not change (Figure 3).
The species turnover contributed more to the interaction turnover in the afternoon (Figure 4).Because temperature decreased in the afternoon (Figure S7b), insects became less active (Figures 3b and S7a) (e.g., Kühsel & Blüthgen, 2015), and species loss would occur (Figure 3a).Species loss caused by reduced activity of insects with a decrease in temperature might be a potential driver of the interaction turnover in the afternoon.Therefore, plant-pollinator interactions potentially turn over in a single day through different processes.Interaction rewiring might drive interaction turnover in the morning when insects become more active, and species turnover is likely to drive it in the afternoon when insects become less active.
The peak time of insect visitation frequency differed for both different insect groups (Figure S3) and different plant species (Figure 4).
These results are consistent with several previous studies (e.g., Hart & Eckhart, 2010;Kikuchi, 1962Kikuchi, , 1963;;Magaly et al., 2017;Stone et al., 1996).This difference in the peak time of visitation frequency appears to be related to the difference in the time when flowers provide floral resources (pollen and/or nectar) among plant species (e.g., Hart & Eckhart, 2010;Pleasants, 1983;Pleasants & Chaplin, 1983;Silva et al., 2004;Willmer & Corbet, 1981;Zimmerman & Pyke, 1986).For example, A. triphylla var japonica presented a U-shaped response to time of the day (Figure 4). A. triphylla var japonica relies on nocturnal pollination by moths and presents a higher amount of sugar in nectar at night (Funamoto & Ohashi, 2017).By these nectar secretions and insect consumption, the amount of floral resources would vary with the time of the day.Flower-visiting insects tend to change foraging flowering plants with time depending on the temporal dynamics of floral resources (e.g., Valdovinos et al., 2016) to obtain more floral resources and achieve better efficiency (Pyke, 2016).Moreover, flower-visiting insects are likely to change their visiting flowers or collecting floral resources with the time of the day (e.g., Hart & Eckhart, 2010).However, the present study did not reveal the temporal dynamics of floral resources.To relate floral resource dynamics directly to differences in the peak time of insect visitation frequency for plant species, an investigation of nectar volume and pollen availability over a single day is needed in further studies.
Plant-pollinator interactions changed in the same day more than on different days (Figure S6).For the two components of interaction turnover, species turnover in the same day contributed to interaction turnover more than that in different days, while interaction rewiring presented an opposite pattern (Figure S6).These results would sug- indicated that artificially pollinated flowers of Crepis capillaris rapidly close, while bagged flowers remained until the late afternoon.As for the pollinating insect perspective, this daily rhythm of flower opening and closure could be translated as resource dynamics.Floral resource dynamics are a key biotic context for pollinator responses.Pollinators should visit flowers with abundant floral resources (i.e., pollen and nectar) to obtain more floral resources and achieve better efficiency The link numbers per insect species presented a different pattern.The link numbers per insect species increased in the morning (08:00-10:00) but did not change after 10:00 (Figure 3c, GAMM; linear component's 95% CI = À1.58 $ À0.07, smoothing component's 95% CI = 0.13 $ 2.48).F I G U R E 2 The plant-pollinator interactions were observed during the whole investigation.The grids coloured grey indicate the presence of interactions.

F
I G U R E 3 The relationship between the time of the day and (a) insect species richness, (b) the number of interactions between plants and insects, and (c) the link number per insect species.The regression lines and 95% credit intervals were estimated by GAMMs using Bayesian modelling.F I G U R E 4 The relationship between the time of the day and the estimated frequency of visits by insects to each plant species.Only plant species with more than 50 interactions are shown.The estimation of visitation frequency was conducted by GAMMs using Bayesian modelling.See Figure S4 for results separated by plant species.
gest that the composition of flower-visiting insects is the same within the season, but insects forage on different plant species within the season.On the other hand, it is suggested that the composition of flower-visiting insects is likely to differ depending on the time or temperature in the same day.Because the link number per insect species appears to differ only in the morning (08:00-10:00, Figure3c), interactions might rewire with time, but the total contribution of interaction rewiring to interaction turnover of whole interaction networks in the same day is likely to be low.The present study suggests that the drivers of plant-pollinator interaction turnover are likely to be different on different temporal scales.Due to the relatively short sampling period (6 days), further studies are necessary to clarify whether this trend is prevalent.AUTHOR CONTRIBUTIONSYuta Nagano: Conceptualization; methodology; software; data curation; investigation; writingoriginal draft; writingreview and editing; visualization; validation; formal analysis; project administration; resources.FI G U R E 5The values of β-diversity of plant-pollinator interactions between survey times of the day.Interaction turnover means the total β-diversity value.* means that the 95% credit interval of the estimated value did not overlap zero, while N.S. means that the 95% CI of the estimated value overlapped zero.These estimations were conducted by LMMs using Bayesian modelling.