Shaping up? A direct comparison between 2D and low-cost 3D shape analysis using African cichlid mandibles

Abstract

Shape is a complex trait which can be investigated through a variety of methods that have been developed over the past century. Currently, ecologists and evolutionary biologists employ the use of geometric morphometrics on 2D images as their standard approach. Recently, there has been increased interest in the use of 3D methods. However, while low-cost 3D methods of data collection are becoming available their potential benefits are often more implied rather than quantified. Using the mandibles from two species of African cichlids (Maylandia zebra and Tropheops BRed Cheek^), this study aimed to evaluate the use of a low-cost 3D method of shape capture versus a range of 2D data sets (termed 'standard', 'even', and 'extended'). Our findings indicated that while both 2D and 3D methods could discriminate differences in species and sexes there was only a slight improvement using 3D when landmark datasets were held even. Further, the standard approaches to data collection that would be taken by most researchers clearly outperformed our 3D approach. Therefore, as 3D methods become more accessible researchers should consider a cost/benefit ratio in terms of the time required to obtain 3D data versus shape information gained.

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

“…25% of the overall variance, the additional discriminating ability appears very limited. These results match with the findings of Álvarez and Perez (2013) and McWhinnie and Parsons (2019), who also did not find any major differences in discriminating between species based on mandibles in 2D and 3D.…”

“…The results imply that none of the two methods is better suited to distinguish between male and female roundfins. A similar result was retrieved by McWhinnie and Parsons (2019), who studied laterally compressed cichlid mandibles. Interestingly, this study’s results do not agree with those of Pfaender et al .…”

“…The first three PCs explained almost 85% of the overall variance in the 2D analysis but only about 75% in the 3D analysis. This is in agreement with the results of Cardini (2014) and McWhinnie and Parsons (2019), who also found lower variance in the first two PCs in their 3D data sets.…”

“…(2018) placed only external landmarks. It is important to note that many internal or asymmetric structures can be reliably landmarked only in 3D and not in 2D (Attard et al, 2018; Hedrick et al, 2019; McWhinnie & Parsons, 2019). Therefore, this study’s results on accuracy cannot be projected onto studies investigating small‐scaled inner or asymmetric structures.…”

“…High similarity in morphospaces between the 2D and 3D analyses was also present in the sculpin heads investigated by Buser et al . (2018), the marmot mandibles analysed by Cardini (2014) and the cichlid mandibles examined by McWhinnie and Parsons (2019). In contrast, the cross‐validated correct classification tests allowed a clear separation between the three morphospecies in both, 2D and 3D ( P = <0.01).…”