Wednesday, July 23, 2008

Deep Dinosaur Diversity

The paper contends that the dinosaurs developed most of their fundamental diversity back at the end of the Triassic and early Jurassic. This is important because there was a major revision of the terrestrial ecosystems during the Cretaceous. This was called the Cretaceous Terrestrial Revolution (KTR) and it brought about our modern ecosystems by and large. The root of it, ha! pun!, was that the angiosperms - ie flowering plants - evolved. Social insects and other critters radiated as well. The traditional thought has been that the dinosaurs were participating in this: there seemed to be a lot more diversity in the dinos during that time. The new paper questions this and actually states that the dinosaur diversity was more to do with the fact that there are far more readily available Cretaceous fossil beds. This created a sampling bias. It gave the impression that the dinosaurs were diversifying. Or so sayeth the paper.

It looks like an impressive bit of work. Sampling bias has been something I have had concerns about due to the fact that my main paleo interest has been mass extinctions. The severity of the extinctions can only be detected by what fossils are found before and after the events. If there are sampling biases simply because there are too few locales it can cause incorrect impressions of the event. More than one author has suggested that as more temporally contiguous fossil beds are found for certain mass extinctions we will need to do some major revisions to the body count. The main ones of concern here are the Permian (for the terrestrial environment) and Late Triassic.

One concern that I have is about my favourite archosaurs, the ceratopsians seemed to be diversifying alot (!!!) at the end of the Cretaceous. So too it seemed that the hadrosaurs were as well. However, this may have been a localized phenomenon in Laurasia. The ceratopsians of unusual size weren't out and about elsewhere.

Anyways, one of the authors, Dave Hone, has a blog, Archosaur Musings. It's worth a read.


Karl Zimmerman said...

I posted this over there, but it doesn't seem to have cleared. Something is *wrong* with the jpg tree, because it doesn't the same as the PDF that was published.

That said, while I was initially wowed, I'm less so now.

First, while the idea of a supertree is good (it's a move towards the "poll of polls" idea that's improved polling so much these last few election cycles), I think it's helpfulness is limited by not weighing more recent trees as being more important. This is proabably why Rahonavis and Jinfengopteryx end up close to birds despite newer studies suggesting they are a Dromaeosaurid and a Troodontid respectively.

Second, and in some ways much worse, is the diversification shift chart. As far as I can see, a major portion of the thesis rests on if Eshanosaurus is really a Therizinosaur. This species is only known from a partial lower jaw, and may be a prosauropod...indeed, given its age, it seems more likely. I'm not sure besides the original tree of the discoverer how many times it has been coded. But given no other maniraptoriforme is known for 41 million years. Without this one dubious species, five different diversification points could be pushed forwards - all of them significantly forward when Cryolophosaurus, which is now thought to not have been a neotheropod, is excluded.

So, with only two species excluded, instead of 11 of 15 major clades diverging in Late Triassic or Early Jurassic, as few 6 of the 15 did. Practically speaking however, it's more likely the Spinosauroids and the Carnosaurs had diverged from the Coelosaurs some time late in the Early Jurassic however, as bona-fide members of the former two appear by around 175MYA.

Anonymous said...

Hi Will, thanks for the promo and link, lots of visitors coming over.

Karl, if you are goign to weight new trees, how would you do it? There is no obviosu philosophically valid way to do this, so you really have to inlcude all of them. If you read the paper you will see how much went into correcting for all kinds of tree biases, it was done very carefully. In a sense here is a lready a bias in favour of more modern trees as they tend to be bigger.

Second, Eshanosaurs is a therizinosaur, and there is a paper coming on it with new information in this regard.

Thirdly any problems with the analysis is as a result of the tree. The tree is based on the best available data (and remember that a tree is a hypothesis, not an 'anser'). You can't criticse the analysis for not artifically moving or removing taxa based on one analysis, that is pretty much the point of a supertree, it is a consensus.

Karl Zimmerman said...


Thanks for replying to me both here and at your blog. I read Will's blog pretty routinely, but had just found yours the other day. I'm just a laymen who is more educated than most, and felt the need to discuss it where I could, not having a blog of my own. Sorry for any misused terminology (or half-completed sentences) due to hasty posting.

Though a very different system, poll analysis sites like have been working on how to weight different sorts of data based upon timeliness. Assuming we develop better phylogeny as time passes, more recent cladistic trees can be given greater weight, with older trees downgraded. Still, older trees should be left in the mix - just like a less-frequently polled state means you need to factor older polling results, species which are less often put into cladistic trees need to have those trees they appear in weighted more even if they are older.

Of course, polls, and thus polls of polls, have one advantage over cladistic trees - you can actually predict results, and thus judge accuracy, so maybe it's not the best possible comparison to make. I do believe that aggregating data to reduce the effect random error has much potential, and the supertree is a great first step.

I'll look forward to reading about Eshanosaurus in the near future, but I'm still skeptical. The supertree as constructed doesn't even have it as a basal outgroup, but a member of one of the most derived clades (again, I know the tree is only as good as the data). Even if it's not a prosauropod, it could always be something convergent on therizinosaurs. Remember what happened with Shuvosaurus, the purported Triassic Ornithomimid?

Mambo-Bob said...

I've posted this on Dave Hone's blog but I'll post it here as well. The jpg tree is the split-fit supertree presented as Supplementary Figure 4 in the original article and was not used in further analyses because it was not well supported. The decision on this is explained in the manuscript. The standard MRP supertree was used for further analyses as it had the highest support. For some reason the topologies are different between these trees - please don't ask me why as I don't know the answer to that.

In answer to your first point: 'When multiple later works had equal
claim, we included them all, but weighted themin tree searches so that their net contribution was equal to one independent tree. Overall, these filters led to a strong skew in the data towards more recent analyses (figure 1), greatly enhancing the chances of recovering a tree that represents current consensus' (Lloyd et al. in Press, p.2)

In answer to your second point, I've commented this over on Dave's blog already but will repost here, exclusion of Eshanosaurus and an alternative position for Cryolophosaurus will indeed change the tempo of diversification shifts but still will have absolutely no effect to the overall conclusions regarding the KTR.