Patterns of flower-visiting insects depend on flowering phenological shifts along an altitudinal gradient in subalpine moorland ecosystems

Alpine and subalpine moorland ecosystems contain unique plant communities, often with many endemic and threatened species, some of which depend on insect pollination. Although alpine and subalpine moorland ecosystems are vulnerable to climatic change, few studies have investigated flower-visiting insects in such ecosystems and examined the factors regulating plant-pollinator interactions along altitudinal gradients. Here, we explored how altitudinal patterns in flower visitors change according to altitudinal shifts in flowering phenology in subalpine moorland ecosystems in northern Japan. We surveyed flower-visiting insects and flowering plants at five sites differing in altitude in early July (soon after snowmelt) and mid-August (peak growing season). In July, we found a higher visiting frequency by more variable insect orders including dipteran, hymenopteran, coleopteran, and lepidopteran species at the higher altitude sites in association with the mass flowering of Geum pentapetalum and Nephrophyllidium crista-galli. In August, such altitudinal patterns were not observed, and dipteran species dominated across the sites due to the flowering of Narthecium asiaticum and Drosera rotundifolia. Our study provides key baselines for the detection of endangered biotic interactions and extinction risks of moorland plants under ongoing climate change.


Introduction
Alpine and subalpine moorland ecosystems contain unique plant communities, often with large numbers of endemic and threatened species (Galeuchet et al. 2005;Hájková et al. 2006;Sasaki et al. 2013). Such ecosystems also provide important ecosystem services, particularly biodiversity maintenance and cultural services including recreation and aesthetic values (Tomitaka et al. 2021). Many alpine and subalpine plants depend on insects for pollination (Ollerton et al. 2011;Lefebvre et al. 2018), with approximately 78% of plant species pollinated by animals in temperate regions (Ollerton et al. 2011), forming complex, interactive plant-pollinator networks (Bascompte et al. 2003). Because the abundance and diversity of pollinators can be higher in habitats with higher flower abundance (Hegland and Boeke 2006;Hines and Hendrix 2009;Scriven et al. 2013), mountainous moorlands with high plant diversity are considered essential sites for insect foraging.
Alpine and subalpine moorland ecosystems are considered vulnerable to external stressors, such as climatic changes (Chapin et al. 2000;Sasaki et al. 2014;Makishima et al. 2021). Despite the ecological and conservation significance of Japan's alpine and subalpine moorland ecosystems, few studies have investigated the flower-visiting insects in these ecosystems and examined the factors regulating plant-pollinator interactions along altitudinal gradients. Such understanding is essential to provide key baselines for detecting endangered plant-pollinator interactions as well as the extinction risks of moorland plant species subject to ongoing climate change.
In mountainous regions, abiotic factors such as temperature and soil conditions vary with altitude (Beniston 2006), and flowering phenology is delayed in response to drops in temperature with increasing altitude (Cornelius et al. 2013). Indeed, the same plant species show different flowering periods among sites with different altitudes, and clear altitudinal patterns of flowering plant species can be observed (Blionis et al. 2001). Consequently, specialist may be more sensitive to the changes in flowering phenology of their host plant species with different altitudes (Zoller et al. 2002;Mizunaga and Kudo 2017;Lefebvre et al. 2018). Previous studies have generally shown that dipteran flower visitors could be the generalist which dominated at higher altitudes, and that they can more easily adapt to the local environment (Chatelain et al. 2018). Moreover, Lefebvre et al. (2018) clearly demonstrated that Diptera, Coleoptera, and Hymenoptera differed significantly in the angiosperm assemblages they visited.
Here, we explored how altitudinal patterns in flower-visiting insects change according to altitudinal shifts in flowering phenology in subalpine moorland ecosystems in northern Japan. Based on the data from five sites obtained in early July (soon after snowmelt) and mid-August (peak growing season), we describe plant-pollinator interaction networks along an altitudinal gradient and during both periods of observation.

Study area and moorland site characteristics
The study area is located in the Hakkoda Mountains (peak coordinates: 40°41 N, 140°52 E; 1,584 m above sea level [a.s.l.]) in Aomori Prefecture, northern Japan (Fig. 1). According to the Japan Meteorological Agency, the annual maximum snow depth, mean temperature, and mean precipitation between 2009 and 2018 ranged from 3 to 6 m, 5 to 6 °C, and 1,600 to 2,200 mm, respectively, at the Sukayu Meteorological Station (40°38.9 N, 140°50.9 E).
Moorlands are a form of wetland that are most often found in cool climates (Ramsar Convention Bureau 2002). Many moorlands exist in the study area, most of which have developed in depressions in a layer of volcanic ash that settled following a huge eruption during the Pleistocene, and others are created in snow patches on gentle slopes (Muraoka and Takakura 1988). In Japan, the development of subalpine moorlands is largely influenced by spatial patterns of snow accumulation, resulting in an interspersed distribution across the landscape. We selected the following five moorland sites across an altitudinal gradient: (1) Tamoyachi (1,254 m a.s.l.), (2) Kamikenashi (1,217 m a.s.l.), (3) Shimokenashi (1,047 m a.s.l.), (4) Takada (1,005 m a.s.l.), and (5) Tashiro (574 m a.s.l.) sites (Table 1). The moorland plant communities at these sites are dominated by Moliniopsis japonica (Hack.) Hayata, Eriophorum vaginatum L., and Narthecium asiaticum Maxim. (Sasaki et al. 2013). In all cases, taxonomic nomenclature follows the 'Plant Japanese name-Scientific name index YList (BG Plants index: http:// ylist. info/ index. html). General descriptions of the study area and sites are provided by Sasaki et al. (2013).
The studied sites are located within the special protected zone of the Towada-Hachimantai National Park, and, therefore, human disturbance such as building development, harvesting, and mining is restricted by law. Across the study sites, we observed little evidence of deer herbivory based on browsed plants and dung. Mowing and livestock grazing were not conducted at or around these sites during our sampling. Trampling into moorlands is strictly prohibited, but boardwalks are built for hiking at some sites. At these sites, our sampling transects were located far from the boardwalks (at least 50 m apart) to minimize potential disturbance effects. Despite these protections, habitat loss and fragmentation of these moorlands are progressing rapidly, with moorland coverage decreasing by an average of 50.01% over approximately 50 years (Makishima et al. 2021). Indeed, even when direct human impacts are minimal, earlier snowmelt in spring associated with recent climate change is facilitating the expansion of shrubby species alongside moorland habitat loss and fragmentation in Japan's mountainous regions ). However, the drivers of such change remain underexplored and need to be further studied based on long-term observational data.

Flowering plants
We investigated flowering plant species in early June and mid-August in 2019 along four 2 m × 20 m fixed transects (separated by at least 20 m) at each site, with a total of 20 transects across all sites. Abundance (i.e., the number of inflorescences) of flowering plant species, rather than windpollinated plant species, was recorded at each transect.
To examine altitudinal shifts in flowering phenology, additional surveys were carried out at the Tamoyachi (representing high altitude sites) and Takada sites (representing middle altitude sites). We needed to limit the survey of flowering phenology to these two sites due to the limit of sampling efforts. For this, seven 2 m × 2 m plots were randomly placed within each site, and flowering abundance was recorded weekly in each plot for all flowering species from the beginning of June to the end of September 2019.

Flower-visiting insects
In early July and mid-August in 2019, the frequency of flower-visiting insects (i.e., the number of visits to a flower with contact with its stamens or stigma) along each transect was counted by individuals using the protocol of Pollard and Yates (1993). The observation survey was performed by the first author (to ensure consistency) for 40 min per transect (20 min each in the 9-12 AM and 1-4 PM) under sunny and warm conditions. It took one day for the survey at each site. We surveyed in order from the lowest (Tashiro) to highest (Tamoyachi) altitude sites. For post-hoc identification based on morphology and taxonomy at the genus or species level, we recorded pollinator visitation by taking pictures using a single-lens reflex camera (Nikon D5300) with a telephoto lens (AF-P DX NIKKOR 70-300 mm f/4.5-6.3G ED VR). As previously noted, the study sites are located within the special protected zone and, therefore, direct sampling of flower-visiting insects was not possible. We did not distinguish between individual pollinators, but counted the number of times each flower-visiting insect visited each flowering species.

Data analysis
We first generated plant and flower-visitor networks using the "bipartite" package and pooled data from the two 20-min Table 1 Flower abundance (i.e., the number of inflorescence) and richness (species richness of flowering species) in early July and mid-August at five moorland sites in the Hakkoda mountain range, Aomori Prefecture, northern Japan Data were pooled along survey transects at each site. Mean air temperature during growing season (from the beginning of July to the end of September) at each site is also shown Site Altitude ( surveys for each transect at the site level. For each sampling period (early July and mid-August), we regressed the species richness and the visiting frequency of flower-visiting insects along each transect against the number of flower plant species (flowering plant richness) and the total number of inflorescences of flower plant species (flower abundance) as well as altitude (m) using generalized linear mixed models (GLMMs). We used the altitude instead of air temperature because altitude at each site was highly correlated (Pearson's r = -0.79, n = 5) with mean air temperature during growing season (the beginning of July to the end of September in 2019). We did this to avoid multicollinearity problems in the following analysis. To account for the differences in visiting frequency sampled along each transect (Heck et al. 1975), the species richness of flower visitors was rarefied to the lowest visiting frequency for each survey. Before running the GLMMs, we checked for multicollinearity in the explanatory variables, with no variables having a variance inflation factor (VIF) score > 10. For both species richness and visiting frequency, we ran the GLMMs with a Poisson error structure and a square-root link function. We included the transect identity nested within site identity as a random effect in the models. Finally, we visualized the time-series data (i.e., weekly data from the beginning of June to the end of September) of flower abundance of three dominant species in July and August at the Tamoyachi and Takada sites, namely Geum pentapetalum, Nephrophyllidium crista-galli, and N. asiaticum. All data analyses were performed using R software (version 4.0.3; R Development Core Team 2020) using the "bipartite", "vegan", "ggplot2", and "lme4" packages.

Results
We recorded 23 dipteran species or species groups (i.e., spp.) from order to genus level, six hymenopteran species or species groups from superfamily to genus level, four coleopteran species or species groups from subfamily to genus level, and five lepidopteran species from order to genus level (Table 2). In July, we recorded higher visiting frequencies of diverse insect orders, including dipteran, hymenopteran, coleopteran, and lepidopteran species, at the higher altitude sites (Kamikenashi and Tamoyachi; Table 2). In August, however, we found that dipteran species were dominated across all the study sites (Table 2).
In July, G. pentapetalum, N. crista-galli, Schizocodon soldanelloides, N. asiaticum, and Pogonia japonica were the dominant flowering plant species, and Aletris foliata, Ixeridium dentatum, and Menziesia multiflora were the rare plant species (Table 3). The abundance of G. pentapetalum was highest at the higher-altitude sites (Tamoyachi and Kamikenashi), while the abundance of N. crista-galli was most abundant at the middle-altitude sites (Shimokenashi and Takada). N. asiaticum was only observed in flower at the Tashiro site.
In August, N. asiaticum and Drosera rotundifolia. were the dominant flowering plant species along with Parnassia palustris, Sanguisorba tenuifolia, Platanthera tipuloides, Spiranthes sinensis, Inula ciliaris, and Hosta sieboldii (Table 3). Flowering individuals of N. asiaticum were commonly found across all sites in August.
The structure of plant and flower-visitor networks at each site in July and August are shown in Fig. 2 (also see Tables 2 and 3). We observed more diverse plant-pollinator interactions in July at the higher altitude sites (the number of links in the plant-pollinator network was 32, 28, 18, 12, and 9 at Tamoyachi, Kamikenashi, Shimokenashi, Takada, and Tashiro sites, respectively) than in August (18,11,6,41, and 18 at Tamoyachi, Kamikenashi, Shimokenashi, Takada, and Tashiro sites, respectively). For example, in July, across all sites except for Tashiro, G. pentapetalum (F_03), N. crista-galli (F_01), and S. soldanelloides (F_02) were linked with diverse arthropod orders including Hymenoptera, Diptera, Coleoptera, and Lepidoptera (Fig. 2a). The Tashiro site lacked these flowering plant species, and flower visitors were dominated by dipteran species. In August, N. asiaticum (F_04) was the dominant flowering plant species across all the sites, and this species was linked primarily with dipteran species (Fig. 2b).
In July, there were significant positive relationships between the visiting frequency of pollinators and flower abundance, and between visiting frequency and altitude (Table 4). In August, there were significant positive relationships between insect visiting frequency and flower abundance, and between visiting frequency and flowering species richness (Table 4). In both July and August, there were no significant relationships between pollinator species richness and the explanatory variables (Table 4).
Time-series data of flower abundance for the three species (G. pentapetalum, N. crista-galli, and N. asiaticum) dominated both at the Takada (1,005 m a.s.l.) and Tashiro (1,254 m a.s.l.) sites are shown in Fig. 3. Notably, the flowering phenology of G. pentapetalum clearly differed between these sites along with N. crista-galli and N. asiaticum, although there was some overlap.

Discussion
Our observations identified altitudinal patterns in insect flower visitors according to altitudinal variations in flowering phenology in the subalpine moorland ecosystems of northern Japan. In early July, we observed higher visiting frequencies by more diverse insect orders including dipteran, hymenopteran, coleopteran, and lepidopteran species  . Insects were identified based on morphology and taxonomy from order to genus level. Each species or species group is coded as "Order_ID": H, Hymenoptera; D, Diptera; C, Coleoptera; L, Lepidoptera; and O, other order (also see Fig. 2). At the end of table, we also summarized total abundance of species in each order and abundance of all species at higher altitudes than at the lower altitude sites. This is attributed to the high flower biomass of G. pentapetalum, N. crista-galli, and S. soldanelloides (particularly at the Kamikenashi site) at higher-altitude sites. In mid-August, such altitudinal patterns were not observed, and dipteran species dominated across all the sites in association with the flowering of N. asiaticum and D. rotundifolia. In both July and August, we observed a positive relationship between the visiting frequencies of pollinators and flower abundance (Table 4), possibly because pollinators search and select locations with high flower abundance to increase the efficiency of resource acquisition (Hegland and Boeke 2006). This also explains the positive relationship between flower-visiting frequency and altitude in July (Table 4), when flower abundance increased with altitude (Table 3). In August, we observed the highest species richness (Tables 1 and 3) and frequency of flower visitors (Tables 2 and 4) at Takada. Compared to the other sites, Takada is surrounded by wider stretches of moorland and is located near to neighboring moorlands (Fig. 1). Therefore, it is likely that such favorable foraging conditions (i.e., greater accessibility to nectar resources) could enhance the frequency of pollinator visits at this site. In contrast, in July, both the abundance and richness of flowering plant species were low at Takada (Tables 1 and 3) with correspondingly low pollinator visitor frequencies.
The plant-pollinator networks (Fig. 2) suggest that G. pentapetalum (F_03), N. crista-galli (F_01), and S. soldanelloides (F_02) flower mainly in early summer, soon after snowmelt, and attract diverse insect orders. In comparison, N. asiaticum (F_04) mainly flowered during the peak growing season and primarily attracted Diptera. Although uncertainties remain, hymenopteran species may have a greater preference based on flower color of plant species than dipteran species (Branquart and Hemptinne 2000;Campbell et al. 2010;Mizunaga and Kudo 2017). G. pentapetalum and N. crista-galli have white flowers, S. soldanelloides has pink flowers, and N. asiaticum has yellow flowers. Our results suggest a greater preference for G. pentapetalum, N. cristagalli, and S. soldanelloides by hymenopteran species over other species in these moorland ecosystems. In particular, S. soldanelloides at the Kamikenashi site attracted diverse orders mainly including Hymenoptera, Diptera, Coleoptera, and Lepidoptera. These species would thus be fundamental for floral resources, thereby maintaining diverse plantpollinator interaction networks. In contrast, many dipteran species would be generalists (Classen et al. 2020;Zhao et al. 2022), with phenological shifts and associated flower color variations in plant communities across time (i.e., between early July and mid-August) and space (i.e., with altitude) having minor effects on their flower-visiting frequency. Many previous studies have suggested that Diptera predominates over other insect orders, such as Hymenoptera, Coleoptera, and Lepidoptera, at high altitudes. Dipteran species may thus play primary roles in plant-pollinator interactions at high altitudes (Zoller et al. 2002;Mizunaga and Kudo 2017;Lefebvre et al. 2018). Indeed, we found that dipteran species dominated at both higher and lower altitude sites (Fig. 2). Our results suggested that arthropod species turnover between insect orders are relatively low along the altitudinal gradient ( Fig. 2 and Table 2). Dominated dipteran species would have a highly competitive advantage for acquiring floral resources. In addition, shifts in flowering phenology across the sites may prohibit normal visits of diverse insect orders for their hosts during their flowering time (Fig. 3). Note that shifts in flowering phenology might also be attributed to the differences in community composition of flowering plants among sites (Table 3). In this vein, our study could not necessarily distinguish whether differences in altitude or those in community composition of flowering plant species among sites regulated differences in flowering phenology. Nonetheless, in August, the flowering period of the dominated flowering species, N. asiaticum, overlapped between the Tamoyachi and Takada sites (Fig. 3). In addition, flowering individuals of N. asiaticum were commonly found across all sites (Table 3). This likely explains the predominance of Diptera across these sites in August due to the substantial increases in their host plants. Additionally, as our altitudinal gradient was less stratified than in the case of studies in higher mountainous areas, such as the European Alps (e.g., Lefebvre et al. 2018; ranging from 1,000 to 2,700 a.s.l.), environmental limitations on the spatial distributions of insect flower visitors could be weaker. The relatively cool summer temperatures across the Hakkoda sites (ranging 17.34-20.41 Table 2), and the width of the lower bars represents the abundance of each flowering plant species (see Table 3) and September) would have also favored dipteran species. Mizunaga and Kudo (2017) also suggested that dipteran species can be the main pollinators in alpine and subalpine areas because they are adapted to low temperatures. Based on all of these observations, we suggest that altitudinal patterns in flower-visiting insects reflect phenological variations in flowering.
Despite the ecological and conservation significance of Japan's alpine and subalpine moorland ecosystems, little is known about flower-visiting insects in these ecosystems and the factors regulating plant-pollinator interactions along altitudinal gradients. Most concerningly, habitat loss and fragmentation of these moorlands are progressing rapidly in association with climatic warming and associated environmental changes (Makishima et al. 2021). In particular, earlier snowmelt associated with recent warming is likely linked to longer moorland plant growing periods and altered flowering phenology, leading to changes in plant-pollinator interactions. Importantly, our study provides key baseline data for the detection of endangered biotic interactions and extinction risks in moorland plant communities subject to ongoing climate change. and writing. AG: investigation and editing. KU: investigation and editing. TS: conceptualization, investigation, editing, and resources. NM, AG, KU and TS approved the final version of manuscript.
Funding This work was financially supported by a Fostering Joint International Research A award (no. 19KK0393) and a Grant-in-Aid for Scientific Research B (No.18H02221 and No. 20H04380) from the Japanese Ministry of Education, Culture, Sports, Science and Technology.

Declarations
Competing interest None.