Do we need the third dimension? Quantifying the effect of the z‐axis in 3D geometric morphometrics based on sailfin silversides (Telmatherinidae)

Abstract

This study investigated the impact of the third dimension in geometric morphometrics (GM) using sailfin silversides (Telmatherinidae) from the Malili Lakes of Sulawesi (Indonesia). The three morphospecies of the monophyletic "roundfin" radiation are laterally compressed and vary in shape traits. The results of 2D and 3D GM were compared and quantified to discuss the advantages and disadvantages of both methods for closely related species and their sexes. This approach focused on the head because it is far more complex and three-dimensionally structured than the trunk or the caudal region. The results revealed no significant benefit concerning repeatability and measurement error in 3D GM compared to 2D GM. The z-axis contributed substantially to the variance of the 3D data set but was irrelevant for discrimination of species and sexes in the approach. Limited gain in information was contrasted by substantially higher effort for 3D compared to the 2D analyses. The study concluded that 2D GM is the more efficient shape analysis approach for discriminating roundfins. Broader studies are needed to test which of the two methods is more efficient in distinguishing laterally compressed fishes in general. For future studies, due to the high investment required, this study recommends carefully evaluating the necessity of 3D GM. If in doubt, this study suggests testing for congruence between 2D and 3D GM with a subsample and consequently applying 2D GM in the case of high congruence.

Hide and seek shark teeth in Random Forests: machine learning applied to Scyliorhinus canicula populations

“…Among landmark and semilandmark features, it is clear that the most discriminant information is contained in two dimensions (x and y), whereas the third dimension brings less information. This implies that running the models with 2D landmark data would have probably achieved similar classification performances as those obtained here, as this has also been evidenced in some traditional geometric morphometric studies ( Cardini, 2014 ; Buser, Sidlauskas & Summers, 2018 ; Wasiljew et al, 2020 ). Nevertheless, 3D landmarking avoids parallax biases ( Cardini, 2014 ) and none of the landmarks and semilandmarks caused a decrease of the metrics, indicating that even though the z dimension contains little information, it still cannot be considered as noise.…”

Alligator appendicular architecture across an ontogenetic niche shift

“…We aimed to account for this issue by examining the humerus and femur in multiple views. While some workers have found 2D and 3D geometric morphometrics to lead to the same results, especially in relatively flat structures such as fish (McWhinnie & Parsons, 2019;Wasiljew, Pfaender, Wipfler, Utama, & Herder, 2020) or mammal mandibles (Cardini, 2014), this is not necessarily the case in structures that have higher 3D complexity (Buser, Sidlauskas, & Summers, 2018;Cardini, 2014;Cardini & Chiapelli, 2020;Hedrick, Antalek-Schrag, Conith, Natanson, & Brennan, 2019). Therefore, the choice of using 2D or 3D geometric morphometrics can be based on the data that are being analyzed.…”

Dots on a screen: The past, present, and future of morphometrics in the study of nonavian dinosaurs

“…Linear morphometric analyses allow incorporation of 3D data (e.g., skull length, skull width) as done by Dodson (1976) for Protoceratops , but do not present the data in a holistic way so it can be difficult to understand which measurements are the primary drivers behind shape change and how they relate to other measurements. Two‐dimensional (2D) geometric morphometrics alleviates the issue of multiple measurements by projecting data into a coordinate system, but 2D analyses require the projection of 3D objects onto 2D representations (Buser et al, 2018; Hedrick, Antalek‐Schrag, et al, 2019; Wasiljew et al, 2020). This has worked quite well when studying ceratopsian squamosals (Maiorino, Farke, Piras, et al, 2013) and mandibles (Maiorino, Farke, Kotsakis, Teresi, & Piras, 2015), which are relatively flattened, but can be problematic when evaluating crania as a whole.…”