Using pictographs as traits to explore morphological diversity in sharks

Abstract

Body shape is a foundational trait on the differences between species. However, morphological measurements can be simplifying and, for many taxa, can be distorted upon preservation or are difficult to collect due to a species' habit or size. Scientific illustrations, or pictographs, provide information on a species' morphology but are rarely used as traits. Here, we demonstrate the use of pictographs using two shark clades: Lamniformes and Carcharhinidae + Sphyrnidae. After collecting 473 pictographs from 67 species across 12 sources, we used landmarking to show that measurements derived from pictographs do not substantially differ from those garnered from specimens. We then used Elliptical Fourier Analysis and principal components analysis to construct a multivariate morphospace. Using global shape measurements, we evaluated whether substantial variability in body shape was introduced by habitat association, endemism, or illustrator. We found that a species' habitat preference strongly influenced the discovery rate of pictographs and the within‐species similarity. While illustrations varied within a species, only a limited set of illustrators exhibited significant systematic variability. We also demonstrated the utility of pictographs in two common applications. For ancestral trait reconstruction, we developed a simple extension to estimate body shapes from principal components and, in doing so, observed that the Lamnid body plan diverged from the rest of Lamniformes ~100 MYA. For phylogenetic generalized linear mixed models (PGLMM), we found that the pictographs had greater explanatory power than traditional morphological measurements. We used the PGLMM to show that higher endemism across Carcharhinidae + Sphyrnidae taxa correlates with body shapes that have caudal fins with small heterocercal angles and more pronounced second dorsal/anal fins. We concluded that pictographs are likely an undervalued and easy‐to‐digitize data source on a species' body shape with numerous established methods for comparing pictographs and assessing variability.

New Interactive Machine Learning Tool for Marine Image Analysis

“…Extracting the length, perimeter, or area of species from images allows estimation of their size, and/or growth rates [11][12][13][14][15] and extracting count data can be used in species abundance [16], or biodiversity estimates [17]. Previously unknown species behaviour traits can be revealed by tracking individuals through extraction of their x,y coordinates [2], and extraction of 'global shape measures' such as the eccentricity, or curvature, of subjects of interest may allow investigation of morphological diversity [18]. Combining analyses can increase the value of information obtained further, for example extracting both species abundance and individual areas allows estimation of biomass [15,19,20], and investigating sessile species size alongside local biotic or abiotic factors may inform on their behaviour [21,22].…”