Census of flower-visiting bumble bees
Bumble bees were collected in the site for two summers, 2018 and 2019. We conducted 15 surveys at five- or six-days intervals for the two years, 5 times in July in 2018 and 10 times from July to August in 2019. Each survey was conducted on transects (50 m in length and 2 m in width) selected for sampling floral resources and their visiting bumble bees across the flowering season; transects were located to include a representative diversity and abundance of floral resources, and used again until there were no flowers in it. New transects were added to the field survey as floral resources changed in the meadow, and at peak flowering there were about 20 transects. From 0930 to 1630 H on a warm and dry day favorable to bee activity, we recorded the bumble bees that were visiting flowers by walking along the transects at constant speed for a period of about 10 minutes. On each survey day we completed between 10 and 22 transects according to the abundance and diversity in floral resources at the meadow (see Table S3 for details). Sampling involved collecting all bumble bees with sweep nets, and then the bees were cooled in an ice chest to slow their movement for identification with a hand lens. The abundance of each bumble bee species was calculated as the number per transect. Each bumble bee was identified to species and social caste (queen, worker, male; only 1-2 queens per species were seen). The bees were released after being identified except for those that could not be identified easily in the field, which were stored in centrifuge tubes (5ml) and later pinned to take to the laboratory for identification. For those bees with similar color stripes, the DNA was extracted from a middle leg stored in 100% pure ethanol for identification by COI barcoding (Huang et al., 2015). Proboscis length (the total length of prementum and glossa) was measured with a vernier caliper when the specimens were fresh. All specimens were deposited in the Wuhan Botanical Garden, Chinese Academy of Sciences.
We also recorded the identity of plant species for each visit when its flowers were visited by the bumble bees. Several traits were measured for all the involved flowering plants, e.g. nectar or nectarless, tubular or non-tubular, and the corolla depth for a tubular flower. Depending on the reward of the flowering plant, the bumble bees were classified as nectar collectors when they visited a flower providing nectar (although they may have also collected pollen from the flower) and pollen collectors when they visited a nectarless flower. In fact, the pollen collectors only visited five species of Pedicularis,which provided pollen as the only reward for bumble bees in this meadow. Flower depth referred to the distance from the base of the nectary to the base of the corolla limb, which was measured on fresh and fully open flowers (N = 10) to the nearest 0.01 mm with a vernier caliper.