Abstract
The drivers of latitudinal differences in the phylogenetic and ecological composition of communities are increasingly studied and understood, but still little is known about the factors underlying morphological differences. High-resolution, three-dimensional morphological data collected using computerized micro-tomography (micro-CT) allows comprehensive comparisons of morphological diversity across latitude. Using marine bivalves as a model system, this study combines 3D shape analysis (based on a new semi-automated procedure for placing landmarks and semilandmarks on shell surfaces) with non-shape traits: centroid size, proportion of shell to soft-tissue volume, and magnitude of shell ornamentation. Analyses conducted on the morphology of 95% of all marine bivalve species from two faunas along the Atlantic coast of North America, the tropical Florida Keys and the boreal Gulf of Maine, show that morphological shifts between these two faunas, and in phylogenetic and ecological subgroups shared between them, occur as changes in total variance with a bounded minimum rather than directional shifts. The dispersion of species in shell-shape morphospace is greater in the Gulf of Maine, which also shows a lower variance in ornamentation and size than the Florida Keys, but the faunas do not differ significantly in the ratio of shell to internal volume. Thus, regional differences conform to hypothesized effects of resource seasonality and predation intensity, but not to carbonate saturation or calcification costs. The overall morphological differences between the regional faunas is largely driven by the loss of ecological functional groups and family-level clades at high latitudes, rather than directional shifts in morphology within the shared groups with latitude. Latitudinal differences in morphology thus represent a complex integration of phylogenetic and ecological factors that are best captured in multivariate analyses across several hierarchical levels.
Published in PLOS ONE.
Towards New Physics at Future Colliders: Machine Learning Optimized Detector and Accelerator Design
Uncovering Patterns in Structure for Voltage Sensing Membrane Proteins with Machine Learning
Finding the likely causes when potential explanatory factors look alike
