Another Page Written in the History of Tetrapod Biodiversity
The story of vertebrate evolution over the last 250 million years is one of the most remarkable, and most complex to unravel, stories of all time. Throughout the ages, extraordinary species and groups have come and gone, and we are left now with only a fingerprint of times long forgotten. Recreating and detecting macroevolutionary patterns within and between vertebrate groups using the fossil record involves an excruciating amount of work, due to the massive amount of data required to be sampled, and the potential number of parameters that could influence biological trends.
Continuing work from his MSc thesis, Roland Sookias (and Roger Benson and Richard Butler) continues to make a name for himself by rigorously analysing the terrestrial tetrapod record within the late Palaeozoic and early Mesozoic. His latest paper [free to access!] extends the analysis of his first (looking at the interaction of intrinsic traits (i.e., body mass) within and between clades of tetrapods around the time that dinosaurs began their ascent) by looking at the impact of extrinsic (i.e., environmental) parameters on body size.
Sookias et al. chose three abiotic parameters (oxygen concentration, temperature, and land area) that have previously been associated with body size evolution. The reason for selecting these three is that they are critical in understanding the historical impact of climatic variation on contemporary biotas, and thus can be used to understand and predict faunal responses to climatic fluctuations. The aim is to see if variation in any environmental factor corresponds with models of body size variation within tetrapods, with the null hypotheses being a consistent lack of correlation between any variables.
The methods emlpoyed are pretty hardcore, requiring use of several R (a statistical programming language) packages integrated with the body mass data used in Roland’s previous paper, a previously published set of Cenozoic mammal body masses, and several other data sets for the environmental variables. Carbon dioxide isotope concentrations were used as proxies for temperature. Maximum and mean sizes within groups (e.g., archosauromorphs, therapsids, and less inclusive clades within) were calculated using femoral length, a widely used and accepted proxy. Sookias et al. however correctly acknowledge that their data assumes an unbiased fossil record, which immediately constrains their results in terms of reliability (e.g., smaller tetrapods are preserved less, so imposes a skew on the mean size).
Most conducted analyses were best explained by the null models, implying that there is no correlation between the tested environmental variables and maximum and mean body mass, contradicting much previous work. This is a pretty neat finding! However, the logical subsequent step to this is not that which is stated, i.e., that because these three abiotic variables do not correspond with body mass, it implies that group-specific biological factors favour body size evolution. There are other abiotic factors out there, such as latitude, sea-level, and non-clade-specific factors such as floral diversity and biomass. I wouldn’t have made that suggestion unless I had rigorously tested it in the same manner as conducted in the rest of the study. Just sayin’..
The studies reveal that with a ‘no autoregressive’ model, the three variables do however correspond with mammalian body size evolution during the Cenozoic. Somewhat weirdly, these results are dismissed (possibly rightly so) as being the potential products of serial correlation (type 1 error), in which case, why were the analyses conducted? These results are also not recovered when the Eocene data is removed (due to the Eocene Thermal Maximum?), thus the authors suggest that the environmental parameters imposed a maximum threshold on mammal body size, they did not ultimately have a causal effect. I agree with Paul Barrett’s comment in the Nature article covering this study, that based on this we shouldn’t ignore possible synchrony between changes in environment and physiology.
A similar correspondence to that mentioned above is recovered between raw archosauromorph size and oxygen concentration between the Permian and Jurassic, however this too is dismissed as false due to inflated type 1 error due to serial correlation (“probably”). This ‘lack of correlation’ is then seen as due cause to speculate on the intrinsic biotic factors within lineages (such as avian-like breathing systems) that may have controlled the evolution of lineage-specific body mass trends. As Sookias et al. state, finer scale aspects such as within-lineage evolution and other ecological aspects were not looked at, and I look forward to seeing how this study develops along these paths. This is where the real meaty details lie, as opposed to looking within a taxonomic context on a pretty broad spatio-temporal scale. Also, phylogenetic correlation is something that would need to be corrected for in future (i.e., body mass similarity due to common ancestry), as well as sampling biases.
All in all, this is a cracking second study to come out of an MSc thesis, and really raises the bar for students on the course this year (something that I’ll be drilling into them this weekend), as well as other Palaeo students in higher education. The data forming the core of these analyses is open to so much more, and I look forward, as I’m sure many others do, to how Roland plays with this data in future. I’d like to see more sampling and phylogenetic correction, so that true signals are not occluded, and correspondence with more ecological parameters and traits.
References
Sookias, R. B. et al. (2012) Biology, not environment, drives major patterns in maximum tetrapod size through time, Proceedings of the Royal Society, Biology Letters, doi:10.1098/rspl.2012.0060
Sookias, R. B. et al. (2012) Rise of dinosaurs reveals major body size transitions are driven by passive processes of trait evolution, Proceedings of the Royal Society, Biology Letters, doi:10.1098/rspb.2011.2441
Hi Jon,
Nice post! I have a couple of quick comments on a couple of more critical points you made.
Re: external vs. intrinsic driving factors. You are correct, there are a few potentially testable additional factors out there, but we have already indirectly tested sea level (which is to a first approximation the inverse of exposed non-marine area). Palaeolatitude would obviously impact on maximum body sizes in different areas of the globe within a single time slice, but it is unclear to me how it could drive changes between two time slices given that continental drift is a relatively slow process. I don’t really get the biomass suggestion either. In any case, we tested the three things that have by far most widely considered as possible controls on maximum body size, and they all fail. We then support the idea that intrinsic clade-specific biological constraints are more important in controlling maximum body size because there is a large, rapidly growing, and convincing body of literature that suggests this – nicely summarised by the Sander et al. (2010) review of dinosaur gigantism in Biological Reviews.
As for the serial correlation issue, there are two points here. First, Roland carried out analyses that lack an autoregressive model (AR0 – this means they don’t take serial correlation and trend into account, which can inflate correlation coefficients when comparing time series statistically) because we wanted to duplicate the results of Smith et al. (2010), who found correlations between Cenozoic mammal size and land area/temperature, but did not deal with serial correlation in their data series. Second, it is standard in these kind of model-fitting approaches to use and compare models that don’t take into account serial correlation (AR0) and models that do (AR1, AR2 etc.) because we don’t a priori know if serial correlation is a problem or not. The model-fitting approach checks whether models that deal with serial correlation are better than those that do not, and in this case AR1 models were always better than AR0 models. So, I think it is perfectly correct to report AR0 and AR1 results, but to note that the AR1 results are better supported than those using AR0.
Cheers for the comments Richard – I did think it was perhaps being a little overly-critical, but I hate writing just about results, so tend to squeeze a bit too hard sometimes!
With the biomass idea, it was more in terms of the co-evolution of plants and contemporary herbivores (did increasing global plant biomass drive body size?), or something. Pretty impossible to test though, I guess.
Yep, Sander et al.’s work is quite thorough! I’m working on looking at correspondence between snout shape (intake rate, bite size etc.) and body mass atm in ornithopods and ruminants, should hopefully have something to report there shortly too.
Cool wrt the modelling approach – you guys obviously know what you were doing! It just seemed a little odd at first to me.