Examining how environmental factors influence phenotypic distribution might provide valuable information about local adaptation, divergence, and speciation. The red-yellow Müllerian mimicry ring of Heliconius butterflies displays a wide range of color patterns across the Neotropics and is involved in several hybrid zones, making it an excellent system to study color phenotypic distribution. Using a multiscale distribution strategy, we studied whether different phenotypes of the distantly related species H. erato and H. melpomene, belonging to the red-yellow mimetic ring, are associated with different environmental conditions. We show that environmental gradients (particularly heat and precipitation factors) drive Heliconius phenotypic distributions, but that phenotype and environmental correlations vary with spatial scale. While co-mimics are frequently found in similar environments at a broad scale, patterns at the local level are not necessarily consistent (different variables are the best predictors of phenotypic occurrence in different areas) or congruent (co-mimic pairs show distinct associations with the environment). Thus, large-scale analysis may help to identify how environmental heterogeneity influences broad mimic phenotypic distributions, but local studies are needed to understand the context-dependent biotic, abiotic, and historical mechanisms that drive finer-scale phenotypic shifts.

The large-scale distribution of H. erato and H. melpomene phenotypes within the red-yellow Müllerian mimicry ring was predicted using Species Distribution Modelling. Our data comprised 3,403 occurrence points in total, including samples collected from December 2016 to June 2017 and July to October 2018 using entomological nets at 25 km intervals over four transects in the Brazilian Amazon, totaling 77 locations across 1,905 km. We also included samples from the collection of the Laboratory of Ecology and Systematics of Pollinators and Predators (LESPP - Federal University of Maranhão) and data used in Rosser et al. (2012), available at https://heliconius-maps.github.io/. The variables in our environmental dataset were chosen for their possible direct and indirect effects on Heliconius ecology and evolution. These were prepared using QGIS (QGIS Association 2020) with a 1km resolution.

Our local-scale analysis was performed using solely the samples collected within two transects in the Brazilian Amazon crossing two hybrid zones (based on intermediate color phenotypes). To verify if the four previously identified phenotypes constituted different clusters based on color patterns, we ran color variation tests. We used the patternize R package to perform a Principal Component Analysis (PCA) on high-resolution images of H. erato and H. melpomene races that represented the four phenotypes of the red-yellow mimetic ring (Van Belleghem et al. 2018). We then calculated the wing surface area containing black in the different clusters using one-way ANOVA and Tukey testing to analyze differences in black patterning among phenotypes. Lastly, we used the bahz R package to do cline analysis using a Bayesian method (Thurman, 2019).

For more information, check MARTINS_2022_README.txt