pollinators limit seed production in an early blooming rare plant: evidence of a mismatch between plant phenology and pollinator emergence

The reproductive ecology of rare plants is seldom studied, yet the persistence of plant populations depends on successful mutualisms with pollinators.  As atmospheric temperatures rise, phenology of plants and pollinators may become mismatched.  We investigated the reproduction of Trifolium barnebyi (Barneby’s Clover), a mat-forming perennial endemic to central Wyoming, USA that grows in the crevices of sandstone bedrock.  Our objectives were to evaluate a method for monitoring changes in cover as well as assess the pollination and seed-set of T. barnebyi.  We established five monitoring transects using a divided frame to estimate percent cover.  We conducted seed-set experiments at three locations to measure self-pollination and the degree to which pollinating insects limited seed production.  We used vane traps and bee bowls to capture pollinators, and examined pollen carried on bees.  Percent cover along transects declined over the 4 year period and was associated with spring precipitation. Trifolium barnebyi did not self-pollinate and relied on pollinators to produce seeds.  The number and mass of viable seeds per flower, and the number of bees captured increased as the season progressed, indicating that more and larger seeds were made when more pollinators were present.  Blooming of T. barnebyi ranged between April and June depending on the microhabitat the plant lived in and we observed much higher seed production in later-blooming plants.  Pollen from T. barnebyi was primarily carried by Andrena bees, although we found smaller amounts of pollen on seven other bee genera.  A mismatch in timing between blooming and pollinating insect emergence could limit seed production in T. barnebyi if plants bloom earlier over time and bee emergence does not follow the same phenology.  Rare plants can be pollinated by rare pollinators showing that conserving these pollinators is crucial for rare plants and early blooming species may be higher at risk.

Methods
Monitoring

To assess the cover and asexual reproduction of T. barnebyi, we monitored five transects annually in July between 2018 and 2021. We established two transects at the Weiser Knoll site (Weiser Slope and Weiser Seedlings) and the Hall Creek site (Hall 1 and Hall 2), and one transect at South Red Canyon (Fig. 1). Each transect consisted of a belt between 6.9 and 18 m in length where a plot frame (30 x 30 cm) with sixteen 7.5 x 7.5 cm cells was placed on the ground along the belt to assess cover (Fig. 2).  We used a plot frame to monitor T. barnebyi because the compact and extensive mats of this species make distinguishing individuals difficult without destructive sampling.  We placed the plot frame on one side of the belt at all sites expect South Red Canyon and Weiser Knoll Seedling, where we placed the frame above and below the belt.  We counted the number of cells with T. barnebyi present within each frame at 30 cm intervals.  Cells were scored as present (1) or absent (0) to calculate the percent of cells in which live T. barnebyi were present.

Seed-Set Experiments

We measured the seed production of T. barnebyi to estimate the degree to which pollinators may limit sexual reproduction.  We selected three areas along the rim of Red Canyon where T. barnebyi is most abundant (North Red Canyon, South Red Canyon and Weiser Knoll) to measure seed-set. We measured seed-set of bagged, open, and hand-pollinated flowers to estimate seed production from self- and cross-pollination.  We selected 20 T. barnebyi plants from 15 April through 16 May 2019. Each plant cluster received one of each treatment.  Bagged treatments restricted pollinator access and measured the degree to which flowers can self-pollinate by placing 1 mm2 mesh bags over flowers before blooming. Open treatments left blooms accessible to local pollinators and measured ambient levels of seed-set.  The hand-pollinated treatment added excess pollen in addition to local pollinators to measure seed production when pollen was not limiting.  Pollen came from plants at least 50 m away and we delicately brushed collected anthers on the stigma of the treatment bloom.  Blooms were bagged with mesh bags before (bagged treatment) or after (open and hand-pollinated treatments) flowers bloomed to contain the developing seeds and were held in place with fishing line weighted with color-coded eye bolts.  For the hand-pollinated treatment, we recorded the number of flowers pollinated, and marked and recorded the flowers not ready for pollination in each flowerhead.  We recorded the date that each flower in the hand-pollinated treatment bloomed and was hand-pollinated.  We monitored treatments and collected fruits when flowerheads were ripe, from 4 June through 17 July 2019.  Flowerheads were placed in paper bags, returned to the laboratory and dried at room temperature.

We cleaned, counted, and weighed seed pods and seeds to estimate the degree to which T. barnebyi self-pollinated or depended on pollinators. Each T. barnebyi flower produced one legume seed pod, and there were multiple flowers per flowerhead.  We counted the number of flowers per flowerhead and the number of ovules per flower using a dissecting microscope.  We weighed the pods and seeds together, because the unfertilized ovules were too small to remove from the pods without damaging. We calculated mean seed mass by dividing total mass by the number of seeds.  We noted seeds that appeared viable by size, and counted and weighed seeds for each plant.  We tested viability of the seeds using Tetrazolium staining, which measures the germinative potential of seeds (Lindenbain 1965).  We placed the seeds between moistened paper towels for 24 hours, cut them to expose the endosperm, and immersed them in tetrazolium solution for 24 hours.  The endosperm of viable seeds turned pink or red indicating respiration, while the endosperm of non-viable seeds remained white.

Pollinators

We collected pollinators at the same sites as the seed-set experiments to estimate which insects pollinate this rare plant.  We deployed seven pollinator stations across sites for 24-48 hours 10 times between 15 April and 21 June 2019.  Pollinator stations consisted of one blue vane trap (vane trap hereafter; colored plastic; 37 cm height by 14 cm diameter; left dry; SpringStar) and three bee bowls (148 mL polystyrene vials; 10 cm  height by 6 cm diameter; Thornton Plastic Co.) painted fluorescent yellow, white or fluorescent blue (Royal Exterior Latex Flat House Paint, Ace Hardware Corp., Oak Brook, Illinois) filled with soapy water.  We recorded the location, date and time we deployed and retrieved pollinator stations, weather, and other notes for each sampling event.  Pollinator stations were used to estimate the abundance and diversity of pollinators within the T. barnebyi population, and to assess which pollinators collect T. barnebyi pollen.  We washed (bee bowls only) and pinned bees and identified them to genus using Michener et al. (1994). 

Pollen

We identified pollen carried on bees to estimate who pollinated T. barnebyi.  We collected and identified (Dorn 2001) flowers from plant species that were blooming at the same time as T. barnebyi and we made a library by mounting pollen grains on slides.  We prepared pollen from the scopa or corbicula of individual bees (one hind leg for all bees except those with scopa on abdomen, mainly Megachilidae). We assumed that the pollen on the scopa or corbicula of bees represented the flowers they had visited based on evidence of pollen identified from 6 body regions of Megachilidae for a rare Penstemon (Tepedino et al. 2006).  We performed acetolysis and stained pollen with Safranin O from flowers and bees to make features on pollen grains clearer before slide mounting (Jones 2014).  We scanned the entire slide to count and identify pollen grains from female bees with scopa (non-cleptoparasites) or corbicula under a compound microscope at 200x. 

Analyses

We used generalized linear models (glm), mixed-effects models (lmer) and linear regression (lm) to estimate differences among variables.  We estimated differences in precipitation, air temperature and cover of T. barnebyi using glm analyzed with the Gamma distribution after viewing histograms of the data.  Differences in the number, mass and viability of seeds were assessed using mixed effects models where plant cluster was the random effect, and treatment and site were fixed effects using lme4 package (Bates et al. 2015).  When a categorical fixed effect had α ≤ 0.05, we calculated estimated marginal means using the emmeans package (Lenth 2021) to estimate which treatments or sites differed.  We used linear regression to estimate how the number of viable seeds changed with the number of pollinators captured and day of year.  We transformed the number and mass of seeds using ln(x+1) because they were not normally distributed or had non-constant variance.  All analyses were done in Program R (R Core Team 2017).