Assessing the relative importance of geographical and ecological drivers of evolution is paramount to understand the diversification of species and traits at the macroevolutionary scale. Here, we use an integrative approach, combining phylogenetics, biogeography, ecology, and quantified phenotypes to investigate the drivers of both species and phenotypic diversification of the iconic Neotropical butterfly genus Morpho. We generated a time-calibrated phylogeny for all known species and inferred historical biogeography. We fitted models of time-dependent (accounting for rate heterogeneity across the phylogeny) and paleoenvironment-dependent diversification (accounting for global effect on the phylogeny). We used geometric morphometrics to assess variation of wing size and shape across the tree, and investigated their dynamics of evolution. We found that the diversification of Morpho is best explained when considering variable diversification rates across the tree, possibly associated with lineages occupying different microhabitat conditions. First, a shift from understory to canopy was characterized by an increased speciation rate partially coupled with an increasing rate of wing shape evolution. Second, the occupation of dense bamboo thickets accompanying a major host-plant shift from dicotyledons towards monocotyledons was associated with a simultaneous diversification rate shift and an evolutionary “jump” of wing size. Our study points to a diversification pattern driven by punctuational ecological changes instead of a global driver or biogeographic history.

Time-calibrated phylogeny:

We concatenated DNA data for one mitochondrial (COI) and four nuclear genes (CAD, EF-1α, GAPDH and MDH). The dataset includes all Morpho species and multiple Morpho helenor representatives. We also included 11 outgroups to root and calibrate the tree. To simultaneously estimate the topology and branching times of the phylogeny we used a Bayesian relaxed-clock approach as implemented in BEAST 1.8.2 (Drummond et al., 2012). To choose the best partitioning strategy and the corresponding substitution models, we ran PartitionFinder 1.1.1 (Lanfear et al., 2012) allowing all possible partitions and models implemented in BEAST. We implemented an uncorrelated lognormal relaxed clock model and Yule process for the tree prior. Given the lack of fossils in the focal clade, we relied on secondary calibrations to calibrate the molecular clock. We used TreeAnnotator 1.8.2 (Drummond et al., 2012) to select the maximum clade credibility (MCC) tree with median age values calculated from the posterior distribution of branch lengths, applying a 20% burn-in. The tree provided here is the resulting MCC tree.

Geometric morphometry:

A total of 911 collection specimens of both sexes and representing all Morpho species were photographed. Wing shape was described using landmarks and semi-landmarks placed at vein intersections and wing margins, respectively (see Chazot et al., 2016 for details), which were superimposed with tpsRelw (Rohlf, 1993). Data provided here are Procrustes coordinates after the superimposition step.