Crocodylomorph cranial shape evolution and its relationship with body size and ecology

Crocodylomorpha, which includes living crocodylians and their extinct relatives, has a rich fossil record, extending back for more than 200 million years. Unlike modern semi-aquatic crocodylians, extinct crocodylomorphs exhibited more varied lifestyles, ranging from marine to fully terrestrial forms. This ecological diversity was mirrored by a remarkable morphological disparity, particularly in terms of cranial morphology, which seems to be closely associated with ecological roles in the group. Here, I use geometric morphometrics to comprehensively investigate cranial shape variation and disparity in Crocodylomorpha. I quantitatively assess the relationship between cranial shape and ecology (i.e. terrestrial, aquatic, and semi-aquatic lifestyles), as well as possible allometric shape changes. I also characterise patterns of cranial shape evolution and identify regime shifts. I found a strong link between shape and size, and a significant influence of ecology on the observed shape variation. Terrestrial taxa, particularly notosuchians, have significantly higher disparity, and shifts to more longirostrine regimes are associated with large-bodied aquatic or semi-aquatic species. This demonstrates an intricate relationship between cranial shape, body size and lifestyle in crocodylomorph evolutionary history. Additionally, disparity-through-time analyses were highly sensitive to different phylogenetic hypotheses, suggesting the description of overall patterns among distinct trees. For crocodylomorphs, most results agree in an early peak during the Early Jurassic and another in the middle of the Cretaceous, followed by nearly continuous decline until today. Since only crown-group members survived through the Cenozoic, this decrease in disparity was likely the result of habitat loss, which narrowed down the range of crocodylomorph lifestyles.


0 1
Cranial shape in different crocodylomorph subgroups 3 0 2 Procrustes ANOVA results show that interobserver error accounts for only 1.6% of total 3 0 3 shape variation (Appendix A Table S1), allowing further analyses using the expanded dataset. The aspects of morphology represented by PC1 and PC2 (Fig. 1) are equivalent to those found  using npMANOVA (Appendix A Table S2) reinforces the apparently disparate cranial 3 1 6 morphology of these two groups, as it shows that their morphospaces are significantly 3 1 7 different to one another (p = 0.0015), and also from most of the groups tested (see Appendix 3 1 8 A for further description of morphospace occupation in other crocodylomorph subgroups).

1 9
Cranial shape disparity estimated for different crocodylomorph subgroups revealed 3 2 0 that Notosuchia has the highest cranial shape disparity among all groups assessed (Fig. 2c).

2 1
Crocodylia exhibits a smaller disparity, although slightly higher than the other four groups 3 2 2 (Tethysuchia, Thalattosuchia, non-crocodylian neosuchians, and non-mesoeucrocodylian  from that in all other groups analysed, whereas some other groups have statistically equivalent 3 2 6 disparities (e.g. thalattosuchians and non-mesoeucrocodylian crocodylomorphs, as well as 3 2 7 tethysuchians and non-crocodylian eusuchians). Similar results were recovered when fewer 3 2 8 subsets of taxa were analysed (i.e. Notosuchia, Neosuchia, Thalattosuchia, non-3 2 9 mesoeucrocodylian crocodylomorphs), with notosuchian disparity still higher and 3 3 0 significantly different from the other groups (Appendix A Fig. S5 and Table S4). Even though some crocodylomorph subgroups exhibit morphospaces that are 3 3 2 significantly distinct from other groups, the relatively weak to moderate phylogenetic signal representation of phylogenetic information is incorporated into tangent space (i.e.  In general, significant impacts on disparity-through-time analyses were observed when 3 4 2 distinct tree topologies were used (see Appendix D for plots of all disparity-through-time Figs. S7, S8, and S9; see Appendix D for all plots). Comparisons between the 10 trees within time, and these differences are usually more marked when a greater number of time intervals 3 5 0 1 5 is used (using either the time binning or the time-slicing methods). For example, analyses 3 5 1 using distinct trees within a same phylogenetic scenario (i.e. Thalattosuchia sister to 3 5 2 Crocodyliformes and gavialids within Gavialoidea) disagree on the timing and magnitude of a 3 5 3 disparity peak during the early evolution of the group (Fig. 3). Whereas some trees show this 3 5 4 peak beginning prior to the Triassic-Jurassic (T-J) boundary, other trees yield a later start, 3 5 5 only after the boundary. Other differences include whether there is an increase or a decrease 3 5 6 in disparity from the middle of the Neogene (Eocene) to the Recent, as well as if a peak 3 5 7 observed during the Early Cretaceous corresponds to the highest disparity seen in the group's 3 5 8 entire evolutionary history (Fig. 3). Similarly, the use of alternative phylogenetic scenarios  In general, disparity-through-time analyses using more time intervals (either time bins 3 6 3 or time slices) reconstruct more nuanced changes in disparity, even though they also often 3 6 4 have larger confidence intervals, due to less taxa being included in each time interval (i.e. variation in the timing and magnitude of disparity peaks). For example, the magnitude 3 6 8 estimated for the peak seen at the end of the Early Cretaceous was usually greater when using  sampling methods, most analyses seem to agree on some overall patterns of crocodylomorph 3 7 2 cranial shape disparity through time (Fig. 3, Appendix A Figs. S7, S8, and S9). An early peak in disparity is frequently observed, most often during the Early Jurassic (although sometimes Jurassic, disparity undergoes a continuous increase until the middle of the Cretaceous 3 7 6 (Aptian-Albian), when maximum disparity is reached in most analyses. Subsequently, a near 3 7 7 constant decline is observed during the Late Cretaceous and the Palaeocene, with analyses 3 7 8 only disagreeing whether it continues until the Recent or ceases during the Eocene. In these 3 7 9 latter cases (more frequently seen in analyses using the time-slicing method), a sharp increase 3 8 0 in disparity is seen in the Eocene, but is frequently followed by an equally sharped decrease  These overall patterns resemble those found by Wilberg (2017) in that a clear peak is  variance-based disparities. The first discrepancy arises from the fact that Wilberg (2017) 3 8 7 restricted his study to Crocodyliformes (with the exception of thalattosuchians) and did not crocodylomorph disparity prior to the Jurassic. When using stratigraphic intervals as time bins Aalenian-Bajocian), whereas in most of my analyses a single Jurassic peak was estimated, Albian). Finally, another difference was found in the pattern of disparity from the Eocene to Body size (=centroid size) has a significant (p < 0.005) effect on crocodylomorph cranial 4 0 4 shape, representing nearly 35% of the total observed variation (this is increased to more than 4 0 5 45% when only PC1 is considered; Table 1). This relationship can be visualised in a shape represents more than 70% of all observed shape variation) into crocodylomorph phylogeny 4 0 9 indicates that many of the largest taxa (such as some thalattosuchians and tethysuchians) also 4 1 0 exhibit PC1 values associated to longer rostra, whereas most of the predominantly small-  These results indicate that body size is a strong predictor of cranial shape in the group. Consequently, the morphospace occupation of distinct crocodylomorph subgroups using 4 1 7 "allometry-free" shape data (i.e. from size-adjusted residuals, Fig. 4c) reveal different patterns  morphospace to that of more short-snouted taxa.  Table 1. Procrustes ANOVA results investigating the amount of variation in shape data explained by 4 2 5 body size (=centroid size). SS, sum of squares after 10,000 permutations; MS, mean squares; % of 4 2 6 variation, obtained by dividing the sum of squares of the independent variable (centroid size) by the 4 2 7 total sum of squares; F, F-statistic; p, p-value. *Significant at alpha = 0.05. 4 2 8 Cranial shape and ecology 4 3 0 Procrustes ANOVA results show a significant, although small (5%, p < 0.05), influence of 4 3 1 ecology (=lifestyles) on crocodylomorph cranial shape ( represented by some of the most extreme longirostrine forms (such as the gavialid 4 3 7 Ikanogavialis gameroi), are mainly confined to the region of longirostrine. Semi-aquatic 4 3 8 species are more widespread along the PC1 axis (Fig. 6a), even though their distribution along 4 3 9 the PC2 axis seems to be similar to that observed for aquatic forms. In terms of disparity for terrestrial taxa, mostly associated with regimes of short rostra (θ < −0.1), even though known crocodylomorph taxa, such as "sphenosuchians" and protosuchids.

6 4
The use of different phylogenetic scenarios did not cause significant impacts on these subgroups across different analyses. It is worth mentioning, however, that some SURFACE 4 6 7 analyses, more frequently those with Thalattosuchia placed outside Crocodyliformes, importance of using multiple time-scaled trees for SURFACE analyses. In general, SURFACE analyses identified more regime shifts (usually more than 15 4 7 4 shifts) than bayou, which found less than 15 shifts with highly supported posterior values), which could also be caused by unsuccessful (suboptimal) model fits. Most of the shape variation in crocodylomorph skulls is represented by changes in the snout, 4 9 0 particularly in its length and width (Fig. 1). This is consistent with what was found in taxonomists (e.g. Lydekker, 1888) to erroneously classify crocodylomorphs into different  which are more directly connected to piscivory, highlighting the influence of ecology on the 5 1 7 group's cranial shape evolution.

1 8
Furthermore, snout width also has important biomechanical implications, such as 5 1 9 impacting on hydrodynamic pressure drag (e.g. longirostrine animals compensate the higher  Similarly, other regions of the crocodylomorph skull that vary significantly also have observed in these regions when compared to the snout. Cranial shape and size linked to ecology 5 2 8 We can comprehend crocodylomorph cranial shape evolution within the concept of a 5 2 9 Simpsonian Adaptive Landscape (Simpson, 1944;1953), which is convenient for characterizing macroevolutionary changes, since it includes ideas such as adaptive zones 5 3 1 invasion and quantum evolution (Stanley, 1973;Hansen, 1997;2012). This is consistent with 5 3 2 the methodological approach used here for characterising cranial shape evolution (i.e. bayou 5 3 3 and SURFACE methods), which assumes evolution under an OU process (even though the fit selective pressures associated to these adaptive zones, shifts between macroevolutionary 5 3 9 regimes can possibly drive large-scale patterns of phenotypic evolution.

4 0
In crocodylomorphs, the clear relationship between ecology and cranial shape and size extrinsic factors (e.g. resources availability, such as a predominance of fish as possible preys).

5
The association between cranial shape and diet can also provide insights on the higher 5 5 6 disparity seen in terrestrial taxa (Fig. 5b). Although aquatic and semi-aquatic species also 5 5 7 explore distinct feeding strategies other than piscivory (such as durophagy; Melstrom & Irmis, 2019), a higher variability is exhibited by terrestrial crocodylomorphs, with strategies such as herbivory, omnivory, insectivory and hypercarnivory (Ősi, 2014; disparity seen in terrestrial crocodylomorphs comes from notosuchians, most of which were 5 6 2 terrestrials and displayed exceptionally high cranial disparity (Fig. 2c), mirroring their rich explored, but some hint can be provided by their occurrence temporal and geographically relaxed modes of body size evolution in the group, which could also be the case for other 5 7 1 phenotypic aspects.

7 2
Apart from notosuchians, other crocodylomorph subgroups contribute to the higher modifications related to brachycephaly (e.g. snout length reduction, rounded neurocranial 5 7 8 shape, dorsal rotation of the mandibles, mandibular asymmetry, and tooth loss and/or Overall, disparity-through-time results were highly sensitive to changes in the time sub- Wilberg (2017) could at least partially be explained by the use a single tree in the latter study 6 1 0 (as well as by the different sample sizes and time sub-sampling methods used). Accordingly, 6 1 1 2 6 rather than using a single analysis, perhaps a better way to report the results might be by 6 1 2 describing shared patterns among multiple outputs, as done here.

1 3
Regarding the overall disparity through time results, the peaks and declines observed 6 1 4 are presumably associated to the appearance and extinction of distinct crocodylomorph 6 1 5 subgroups, such as thalattosuchians in the Jurassic and notosuchians in the Cretaceous, as 6 1 6 already pointed out by previous studies (Stubbs et al., 2013;Wilberg, 2017). Some of these 6 1 7 peaks can be more securely be linked to abiotic factors, such as paleotemperature. For (which range from species highly adapted to a fully-aquatic life to terrestrial and nearly 6 2 9 cursorial forms), for which different responses to environmental changes are expected.  This could also help understanding the nearly continuous decline in crocodylomorph 6 3 5 disparity since the Late Cretaceous, which is mainly represented by members of the crown- Crocodylians and Their Relatives: Functional Insights from Ontogeny and Evolution. In  Biology. Preprint available at: https://www.biorxiv.org/content/10.1101/405621v2. approach to the Gavialis problem using geometric morphometric data from crocodylian 8 1 4 braincases and Eustachian systems. PLoS ONE, 9, e105793. Gower, J. C. (1975). Generalized procrustes analysis. Psychometrika, 40, 33-51. Ecology and Evolution, 9, 1755-1763. Hansen, T. F. (1997). Stabilizing selection and the comparative analysis of adaptation.

4 4
Hotelling, H. (1933). Analysis of a complex of statistical variables into principal components.