Modification of bumblebee behavior by floral color change and implications for pollen transfer in Weigela middendorffiana

Abstract

Flowers of Weigela middendorffiana change the color from yellow to red. The previous study revealed that red-phase flowers no longer have sexual function and nectar, and bumblebees selectively visit yellow-phase flowers. The present study examined how retaining color-changed flowers can regulate the foraging behavior of bumblebees and pollen transport among flowers within (geitonogamous pollination) and between (outcrossing pollination) plants and how the behavior is influenced by display size (i.e., number of functional flowers) and visitation frequency. The visitation frequencies of bumblebees to plants and successive flower probes within plants were observed in the field using plants whose flower number and composition of the two color-phase flowers had been manipulated. To evaluate pollination efficiency over multiple pollinator visits, a pollen transport model was constructed based on the observed bumblebee behavior. In the simulation, three flowering patterns associated with display size and existence of color-changed flowers were postulated as follows: Type 1, large display (100 functional flowers) and no retention of color-changed flowers; Type 2, small display (50 functional flowers) and retention of color-changed flowers (50 old flowers), and; Type 3, large display (100 functional flowers) and retention of color-changed flowers (100 old flowers). Color-changed flowers did not contribute to increasing bumblebee attraction at a distance but reduced the number of successive flower probes within plants. Comparisons of pollen transfer between Types 1 and 3 revealed that the retention of color-changed flowers did not influence the total amount of pollen exported when pollinator visits were abundant (>100 visits) but decreased geitonogamous pollination. Comparisons between Types 2 and 3 revealed that the discouragement effect of floral color change on successive probes accelerated in plants with a large display size. Overall, the floral color change strategy contributed to reduce geitonogamous pollination, but its effectiveness was highly sensitive to display size and pollinator frequency.

Appendix

Estimation of preference index

The pollinator’s relative preference for yellow-phase (pre-changed) flowers to red-phase (post-changed) flowers was defined as a preference index (PI). If there was no preference for yellow-phase flowers, PI equals 1. A likelihood for the estimation of PI was produced using observed data in the field (Ida and Kudo 2003). When V _Y visits to yellow-phase flowers and V _R visits to red-phase flowers were observed in exiting N _Y yellow-phase flowers and N _R red-phase flowers within plants and pollinators have a PI, then the binomial distribution becomes the following likelihood function:

$$ L = \prod\limits_{k} {\left\{ {\left( {\begin{array}{*{20}c} {V_{Y,k} + V_{R,k} } \\ {V_{Y,k} } \\ \end{array} } \right)\left( {{\frac{{N_{Y,k} \exp \left( q \right)}}{{N_{Y,k} \exp \left( q \right) + N_{R,k} }}}} \right)^{{V_{Y,k} }} \left( {1 - {\frac{{N_{Y,k} \exp \left( q \right)}}{{N_{Y,k} \exp \left( q \right) + N_{R,k} }}}} \right)^{{V_{R,k} }} } \right\}} , $$

(5)

where PI was represented as exp(q). Then, the log-likelihood is

$$ \log L = \sum\limits_{k} {\left\{ {V_{Y,k} \left( {q + \log \left( {N_{Y,k} } \right)} \right) + V_{R,k} \log \left( {N_{R,k} } \right) - \left( {V_{Y,k} + V_{R,k} } \right)\log \left( {N_{Y,k} \exp \left( q \right) + N_{R,k} } \right)} \right\}} + {\text{constant}} . $$

(6)

The maximum likelihood estimation based on the observation data demonstrated that PI value was approximately 9. Thus, the present study postulated PI = 9.0.