Unsupervised machine learning reveals mimicry complexes in bumblebees occur along a perceptual continuum

Area of Science:

• Evolutionary Biology
• Ecology
• Animal Behavior

Background:

• Müllerian mimicry theory predicts convergence of harmful species onto shared color patterns.
• Mimetic patterns typically form discrete complexes, with limited understanding of transition zones.
• Geographical variation in mimicry and factors driving these zones are understudied.

Purpose of the Study:

• To investigate patterns and processes of Müllerian mimicry in bumblebees across the contiguous United States.
• To quantify color pattern mimicry and analyze transition zones using novel methods.
• To explore factors influencing the dynamics of mimicry transition zones.

Main Methods:

• Integration of large-scale geographical distribution data for bumblebee color patterns.
• Application of an unsupervised machine-learning approach with computer vision for mimicry quantification.
• Analysis of color pattern transition zones in three comimicking, polymorphic bumblebee species.

Main Results:

• Bumblebee color patterns are geographically clustered but imperfectly mimetic.
• Mimicry patterns exhibit gradual spatial transitions, lacking discrete boundaries.
• Transition zones of comimicking species differ in location within a region of poor mimicry.

Conclusions:

• Müllerian mimicry in bumblebees is characterized by spatial clines rather than distinct mimicry rings.
• Transition zone dynamics are influenced by species-specific selective pressures.
• Further research is needed to fully understand the factors shaping mimicry zone evolution.