Manouria emys

Hermanson, Guilherme & Evers, Serjoscha W., 2025, PG- 18: turtles reach adult shell shapes at about 65 % maximum carapace length, Swiss Journal of Palaeontology (47) 144 (1), pp. 1-16 : 7-10

Manouria emys , Trachemys scripta ) could also suggest that shape-growth does not closely follow a Gompertz function. Because of the varying plateauing levels across species, the lower 95% confidence intervals often provide unrealistic low size thresholds (mean = 50%, min= 3%, max = 82%; Supplementary Table 2; Supplementary Fig. 18) for adult shapes, which would imply that adult shapes are reached before sizes typical for sexual maturity. Tus, for our results and discussion, we largely focus on the thresholds implied by the 85% adult shape.

Although the clusters from the cluster analysis are based on the respective shape proxy and shell size, the

A B

shape shape

Multivariate Multivariate

100 200 300 400 100 150 200 250 300

SCL SCL

C D

Orlitia borneensis shape shape

Multivariate Multivariate

200 300 400 500 600 100 300 500 700

SCL SCL

Fig. 3 Ontogenetic shell shape curves for selected species. Bivariate plots of multivariate shape against straight carapace length (SCL, in mm). A Chelus fimbriata , B Trachemys scripta , C Manouria emys , D Orlitia borneensis . The size-range for adult multivariate shape, based on our 85%-adult shape threshold, is indicated by grey boxes.The red vertical line indicates the size threshold at which multivariate shape values reach 85%

of the shape of record-sized specimens,with corresponding female and male percentages of maximum SCL indicated at the top of the line. Datapoints are colour-coded according to their cluster assignments (‘small’,‘intermediate’ or ‘large’), and symbols represent sex.(X/Y%) values at largest datapoint indicates percentage of maximum female/male SCL recorded for the species. Solid lines are Gompertz curves fit across all data points. Dashed horizontal lines represent the lower 95% interval at which the asymptotic multivariate shape value is reached according to the Gompertz functions

‘large’ cluster category does not uniformly fall together with the 85% shape thresholds or with the lower 5% confidence intervals on shape asymptotes. Te 85% shape thresholds nearly ubiquitously fall within the ‘large’ cluster, indicating that the 85% cut-off is quite conservative within the context of multivariate shape and absolute shell size. On the contrary, the lower 5% confidence intervals on shape asymptotes usually fall within the ‘intermediate’ cluster, highlighting the uncertainty of the asymptotic approach to generating size thresholds in the current dataset.

Sexual shape dimorphism

Allometric shape regressions with sex as a covariate show that sexual shape dimorphism is absent in many (N = 15) of the 21 tested turtles (Fig. 4A; Supplementary Table 3). In only six species, we received a significant effect of sex (Fig. 4A). In only two of these, the effect of sex is stronger than the effect of ontogeny, indicating that sex effect changes are less pronounced that ontogenetic shell shape changes (Fig. 4A). Overall, size has a median effect-size of 2.8, and its effect is thus ~ 4.5-fold the effect of sex, which has a median value of 0.6 (Fig. 4B). Shape comparisons of two same-sized female and male specimens of the turtle in which we detected the largest relative sex effects, the geoemydid Leucocephalon yuwonoi , show that shape differences are minor (Fig. 4C). Tese include a somewhat more elongated shell and more concave plastron in the male. However, these differences are minor when compared to ontogenetic shell shape differences in the species (Fig. 4D), which include strong shell elongation, and also when compared to interspecific differences with the closely related Notochelys platynota

(Fig. 4E).

Disparity analysis

Our disparity analysis results lend support to our initial hypothesis of an increased extent of shell shape differentiation throughout ontogeny (Fig. 5). Both of our disparity metrics used (sum of ranges and Procrustes variances) return the same patterns. ‘Small’ turtles exhibit significantly smaller morphological disparity in the shell than both ‘intermediate’ and ‘large’-sized turtles ( Tables 1, 2). However, we find no statistical difference between the disparity ranges of ‘intermediate’ and ‘large’ turtle individuals ( Table 2; Fig. 5B).