Background: Climate change is altering the world’s ecosystems through direct effects of climate warming and precipitation changes, but also indirectly through changes in biotic interactions. For instance, climate-driven changes in plant and/or insect communities may alter plant-pollinator interactions, thereby influencing plant reproductive success and ultimately population dynamics of insect-pollinated plants.

Methods: To better understand how the importance of insect pollination for plant fruit set varies with climate, we experimentally excluded pollinators from the partly selfing key-stone species Vaccinium myrtillus along elevational gradients in the forest-tundra ecotone in central Norway. The study comprised three mountain areas, seven elevational gradients spanning from the climatically relatively benign birch forest to the colder alpine areas above the tree line, and 180 plots of 1 x 1 m, with experimental treatments allocated randomly to plots within sites. Within the experimental plots we counted the number of flowers of V. myrtillus and counted and weighted all fruits, as well as seeds for a selection of fruits.

Results: Excluding pollinators resulted in lower fruit production, as well as reduced fruit and seed mass of Vaccinium myrtillus. In the alpine sites pollinator exclusion resulted in 84 % fewer fruits, 50 % lower fruit weight and 50 % lower seed weight compared to control conditions. Contrary to our expectations, the negative effect of pollinator exclusion was less pronounced in forest compared to alpine sites, suggesting that the importance of insect pollination for seed production is lower at low elevation.

Conclusions: Our findings indicate that the keystone species Vaccinium myrtillus is relatively robust to changes in the pollinator community in a warmer climate, thereby making it less vulnerable to climate-driven changes in plant-pollinator interactions.

Study area

The study was carried out in 2017 and 2018 in three mountain areas, Forollhogna, Dovrefjell and Grødalen in Sunndalsfjella, situated along an east-west gradient in central Norway. In each area except Dovrefjell we established two replicate elevational gradients from the mountain forest to the alpine tundra, and along each gradient we established three experimental sites: one in the mountain birch forest, one at the treeline and one in the open alpine tundra. The sites were established in heathland vegetation with a high abundance of Vaccinium myrtillus. In Dovrefjell we only had one gradient due to few V. myrtillus dominated sites. The mean distance between elevational levels was app. 160 m, and the difference in mean summer temperature was on average 1.0 °C between the forest and treeline sites and 0.6 °C between the treeline and alpine sites.

Experimental design

In each site we established four blocks, approximately 30-100 meters apart (depending on bilberry abundance), each with three 1 x 1 m plots with 3‒5 m distance, resulting in a total of 180 plots. Within each block we applied three experimental treatments to examine the importance of pollinators for V. myrtillus: control, pollinator reduction, and pollinator exclusion. Treatments were randomly assigned to the three plots in each block. The pollinator reduction and exclusion treatments were achieved by placing dome-shaped cages made of two approximately 2.5 m long PVC tubes bent diagonally over the plots. The size of the resulting cages (w × l × h) was approximately 1.5 × 1.5 × 1 m. For the reduction treatment the cages were covered with berry netting with a mesh size of 1.5 × 1.5 cm, through which at least some pollinators, including bumblebees, were able to enter and exit the plots, whereas the exclusion cages were covered with insect netting with a mesh size of 2 × 2 mm, which no flying insects were observed to penetrate. The mesh was fastened to the PVC tubes using plastic strips. To prevent non-flying pollinators from accessing the plots, the netting was fixed to the ground with n-shaped plugs. Initial analyses after one year of treatment suggested that the reduction treatment had a very limited effect, most likely because the mesh size was too large and therefore did not represent a barrier to pollinators. This treatment was therefore discontinued.

At peak flowering time, we counted the number of flowers in each plot. The timing varied along the elevational gradients, with flowering time peaking a week or two earlier in the mountain forest compared to the treeline and alpine sites. When the majority of the fruits were ripe, all fruits from each plot were collected, counted, dried at 60 degrees for 48 hours and weighed. Among the fruits collected in 2017, we randomly picked one mature fruit per plot, and re-wetted and dissected them before counting the number of seeds under a stereo microscope. The seeds were subsequently dried at 60 degrees for 48 hours and weighed.

The experimental treatments could potentially influence e.g. micro-climatic conditions within the mesh cages and thereby bias our results. To check for side-effects of the exclusion cages, we measured temperature using B-series WatchDog B101 8K temperature loggers (Spectrum Technologies Inc., Aurora, USA) and illuminance (lux) using a Hagner EC1 digital luxmeter (B. Hagner AB, Solna, Sweden) in all control and exclusion plots in 2018.  All temperature loggers were placed in white plastic boxes to prevent moisture damage, and the boxes were placed at ground level, shaded by the vegetation. Temperature was recorded every fourth hour from mid-June to mid-August, whereas illuminance was measured once in each plot at peak flowering, making sure to measure all plots in one site on the same day.

More details are found in the manuscript.