Although pollination networks between plants and flower visitors are diverse and flexible, seed production of many plant species is restricted by pollen limitation. Obligate outcrossers often suffer from low pollinator activity or severe interspecific competition for pollinator acquisition among co-flowering species. This study focused on seasonal changes in plant–flower visitor linkages in an alpine ecosystem and examined whether and how this seasonality affected the seed-set of Primula modesta, a self-incompatible distylous herb having long-tubed flowers. First, we recorded the linkages between plants and flower visitors along the snowmelt gradient. Then, pollination experiment was conducted to estimate the degree of pollen limitation over the course of flowering season of P. modesta. Flower visitors were classified by their tongue length based on the morphological matching with P. modesta flowers. As the season progressed, plant–visitor linkages became more diverse and generalized, and the visitation frequency to P. modesta flowers increased. In the later part of the season, however, the seed set of P. modesta was significantly reduced due to severe pollen limitation, presumably because of increased competition for long-tongued pollinators among co-flowering species. The present study revealed that pollinator availability for specialist species may be restricted even when plant–visitor linkages are diverse and generalized as a whole. In the case of P. modesta, morphological matching and competition for pollinators might be the main factors explaining this discrepancy.

Study site

We conducted this study within an area of 150 m × 130 m on the north-facing slope of the Happo-One ridge, northern Japanese Alps, central Japan (N36°42′0″, E137°50′58″). The local topography within the study area gradually changes from fellfield on the wind-blown ridge to snowbed at the bottom of valley; the altitude difference between the two habitats is about 40 m and the slope angle is about 15–20° (see Kameyama et al. 2015a for map). The ridge consists of fellfield vegetation dominated by Pinus pumila, Juniperus communis, Thymus quinquecostatus, Spiraea nipponica, Festuca ovina, and lichens. Meanwhile, the slope is dominated by dwarf shrubs and herbs such as Rhododendron tschonoskii, Rhododendron multiflorum, Carex blepharicarpa, and Schizocodon soldanelloides. The valley consists of herbaceous snowbed vegetation dominated by Carex omiana, Eriophorum vaginatum ssp. fauriei, Trichophorum cespitosum, and Sieversia pentapetala.

Target species

Primula modesta is a distylous alpine herb which produces one flowering stem with three to five tubular flowers, and distributed over a wide range of montane to alpine habitats throughout Japan. Because this species possesses a heteromorphic self-incompatibility system that almost completely prevents self- and intramorph mating, seed-set strongly depends on cross-pollination between two morphs, i.e., long styled pin and short styled thrum flowers. Pollination success of P. modesta may be dependent on long-tongued pollinators, which probe for nectar at the bottom of the floral tube and transfer legitimate pollen grains between reciprocal morphs.

In the study area, P. modesta is distributed widely along the snowmelt gradient, and is one of the earliest bloomers within each topographic site (Kameyama, personal observation). Thus, in the present study, we defined the early, middle, and late season according to the peak flowering time of P. modesta growing on the ridge, slope, and valley, i.e., early June, early July, and late July, respectively.

Community scale linkages between plants and flower visitors

We observed flower visitors within quadrats which were arbitrarily set to include as many flowering patches as possible (1 m × 1 m or 2 m × 2 m depending on flower density). The number of quadrats set during the early, middle, and late season were 3, 9, and 9 in 2016, and 2, 7, and 6 in 2017, respectively. Within each quadrat, we counted the flowers of all entomophilous plant species; apparent anemophilous plant species such as Gyperaceae, Gramineae, Juncaceae, and Pinaceae were excluded. We recorded flower visitors when an insect other than a spider or ant contacted the reproductive organs of flowers. Given that complete identification of flower visitors was impossible in the field, we collected specimens (116 in total) and identified them to the order, genus, or species level wherever possible. Observations were conducted during daylight hours (0830–1450 h) in periods of calm weather. The total observation time and number of observation days during the early season was 6.2 h over 3 d in 2016 and 7.7 h over 2 d in 2017. In the middle season, this was 7.8 h over 6 d in 2016 and 11.8 h over 5 d in 2017, and in the late season it was 11.7 h over 3 d in 2016 and 10.0 h over 3 d in 2017. The total number of visits observed during the early, middle, and late season was 423, 164, and 1171 in 2016, and 56, 788, and 943 in 2017, respectively. Plant–flower visitor linkages were quantified by the frequency of visits to each plant species (flower⁻¹hr⁻¹).

Morphological matching between flower visitors and P. modesta

To consider the morphological matching between flower visitors and P. modesta flowers, we collected 43 pin and 44 thrum flowers (one flower from each plant). The lateral side of each flower was carefully dissected and a digital photograph was taken using a stereomicroscope (Olympus SZ60) with a digital camera (Olympus E-620). The length of the floral tube and the distance between the floral surface and the short reproductive organ (stigma of thrum flower and anther of pin flower) were measured using Image J (Schneider et al. 2012, Rueden et al. 2017). The body lengths and tongue lengths of flower visitors (116 specimens) were also measured for morphological matching with P. modesta flowers using Image J. Based on the measurements, flower visitors were classified as long-tongued (LT), medium-tongued (MT), short-tongued (ST), or other (data deficient).

Seed-set of P. modesta

The degree of pollen limitation was evaluated by comparing the actual and potential seed production of P. modesta growing within a 10 m × 10 m quadrat at the ridge, slope, and valley, which correspond to the early, middle, and late season, respectively. To estimate the actual seed production, we collected several fruits (2.9 on average) from arbitrarily selected plants (18.3 on average) growing within a quadrat. The total number of fruits collected was 168 and 148 in 2016 and 2017, respectively. The potential ability to produce seeds was estimated by artificial pollination experiments. Plants growing within a quadrat was arbitrarily selected (17.7 on average) and covered with fine-meshed nylon nets to exclude flower visitors. For each plant, 1–3 flower buds (1.4 on average) were emasculated just before opening, and other flowers were removed. After flower opening, hand pollination was conducted using legitimate pollen grains collected from a single pollen donor growing at least 5 m away from the recipient plant. All fruits derived from this pollination treatment were harvested just before dehiscence. The total number of fruits harvested was 62 and 87 in 2016 and 2017, respectively. The number of seeds were counted using a stereomicroscope (Olympus SZ60).