Field observation
Observations of plant phenology and flower-visiting insects were conducted five times each year; on 26–30 May (term 1), 21–23 June (term 2), 13–15 July (term 3), 7–9 August (term 4), and 28–30 August (term 5) in 2017; on 5–8 June (term 1), 29 June (term 2), 29–31 July (term 3), 13–14 August (term 4), and 3–5 September (term 5) in 2018. Air temperature was measured at 1-hr intervals at the height of 1 m above ground using a Tidvit V2 data logger (Hobo, Onset Co., USA) from May 31, 2017 to September 3, 2018 at 3060 m elevation.
For the survey of flowering occurrence and flower production, 21 fixed plots (named TW01-21; 2 × 10 m in size) were established at various habitats (from dry ridge habitat to wet depression habitat) to cover the whole vegetation types in the study site (Fig. S1). In each plot, flowering occurrence and flower number of each species were recorded in the early (term 1), middle (term 3), and late seasons (term 5). Locations of individual plots were illustrated in Figure S1. In total, flower productions of 81 species were recorded across plots during two years. Furthermore, a phenological survey of flowering occurrence was conducted for all entomophilous plant species in this area. In each term (term 1–5), we walked around the whole area in which fixed plots were set (within a 2.5 × 1.0 km area), and recorded flowering species. In total, flowering phenologies of 130 species were recorded throughout the survey periods.
For the survey of flower-visiting insects, we repeated 30-min censuses of flower visitors by walk throughout the season (five terms). The census period in each term was 1–3 continuous days during which 11–30 sets of census were conducted. Each census was conducted on calm days (wind speed was < 4 m/s) during the daytime (8:00–17:00). In total, 106 censuses (53 hrs.) and 102 censuses (51 hrs.) were conducted in 2017 and 2018, respectively. At the beginning of each census, air temperature and relative humidity were measured. Flower-visiting insects were classified into following groups; (1) hymenopteran insects (subdivided into bumble bee, honey bee, solitary bee, wasp, and sawfly), (2) dipteran insects (syrphid fly, dagger fly, other fly), (3) lepidopteran insects (butterfly, skipper, moth), (4) coleopteran insects, (5) hemipteran insects, and (6) other insects. Ants and grasshoppers were excluded from the observation because their activity as pollinators seemed to be small. The number of insects visiting flowers and plant species of the flowers were recorded. In total, visits of 15,127 insects were recorded on the flowers of 105 plant species.
About 93% of observed insects were hymenopteran or dipteran insects (see Results). Based on the composition of visiting insects, the pollination type of individual plant species was classified into the following five groups: (1) bee specialist: > 75% of visitors are hymenopteran insects, (2) fly specialist: > 75% of visitors are dipteran insects, (3) bee generalist: 50–75% of visitors are hymenopteran insects, (4) fly generalist: 50–75% of visitors are dipteran insects, and (5) unclear type: very low visits (< 5 visits) or mixture of several insect groups. In the present study, both bee-specialist and bee-generalist species are categorized as bee-visited species, and both fly-specialist and fly-generalist species are done as fly-visited species.