Friday, August 6, 2010

Brief Thoughts on Chromogisaurus and Guaibasauridae

Since SVP 2009 we've been hearing about Ezcurra's idea a broad range of basal saurischians (Guaibasaurus, Saturnalia, Panphagia, Agnosphitys) can be referred to a clade he calls Guaibasauridae, placed in Sauropodomorpha.  Previously Guaibasaurus has emerged as a basal theropod, while Saturnalia and Panphagia were basal sauropodomorphs.  Now thanks to the description of the new guaibasaurid Chromogisaurus by Ezcurra (2010), we have the matrix and character evidence for this arrangement.  His cladogram is-

Strict consensus tree of Ezcurra's (2010) analysis, after Figure 15A in that paper.

My first thought on seeing this is that it's based on Yates' analysis, and indeed it's the version with Glacialisaurus added from Smith and Pol (2007), with 15 added characters and three added taxa (Chromogisaurus, Panphagia and the undescribed herrerasaurid MACN-PV18649a).  Ezcurra also corrected several characters, which was especially useful for Guaibasaurus as it incorporated the referred specimen UFRGS PV0725T.  Yates' analysis is excellent, but the problem here is that it's a sauropodomorph matrix, not a dinosaur matrix.  In other words, it was designed to determine the relationships among sauropodomorphs (which make up at least 37 of the 50 taxa), not the relationships of Herrerasaurus, Eoraptor, Staurikosaurus, Guaibasaurus, theropods, etc..  This is why Neotheropoda is a single OTU, whereas in any analysis of saurischian relationships you'd want it to be split into several taxa.  Similarly, Pisanosaurus would be useful for outgroup comparisons, as would more basal silesaurids like LewisuchusTawa also seems vital for any current current saurischian analysis, though of course it wasn't known when Yates constructed the dataset.  But since this is the matrix we have, let's see what it supports.

"The basic message is that each new analysis does not make prior analyses wrong. One point to Nesbitt et al.'s credit is that they ran constraint analyses to determine just how unlikely alternate arrangements are. We need to keep these in mind whenever we look at a cladogram. Cladograms always show us the MOST parsimonious arrangement, but perhaps it's more useful to be aware of which nodes are SIGNIFICANTLY MORE parsimonious."

Unfortunately, Ezcurra does not run any constraint analyses, so it's hard to know how well supported his new hypotheses are.  He does use Bootstrap and Bremer support, but these are highly influenced by a taxon's completeness so aren't very useful in analyses like this with many partially coded OTUs.  I tried to run Ezcurra's matrix, but it seems that only Crurotarsi is coded for all 378 characters, while all other taxa are coded for only 377.  Based on the codings for the last character, it doesn't seem to be 378, so I'm assuming that's the one that's missing.  So I deleted Crurotarsi's coding for character 378, leaving 377 characters coded for all taxa.  The resulting tree is identical to that in Ezcurra's figure, but is 1181 steps instead of 1186 steps long.

EDIT: I have now received the actual NEXUS file with proper codings for all 378 characters which finds trees 1186 steps long as in Ezcurra's paper, and the results differ slightly.  Those who read this in the first hour of its posting saw incorrect numbers below, though the conclusions remain the same.

So let's examine how other topologies fare, going from most plausible to least.

Making Chindesaurus a herrerasaurian as suggested by several authors is only one step longer, so basically equivocal. Similarly, forcing it to be a sauropodomorph as originally reported is only one step longer.

Ezcurra found Chromogisaurus to be closer to Saturnalia than to Panphagia or Guaibasaurus, and named this clade Saturnaliinae.  It breaks down in two steps, so is very poorly supported.

Placing Panphagia basal to Saturnalia + other sauropodomorphs as in its original description is only three steps longer. This indicates Guaibasauridae is very poorly supported.

Placing Agnosphitys outside of Dinosauria (as in its original description) is only three steps longer.

Making Chindesaurus closer to Herrerasaurus than Staurikosaurus (as in Nesbitt et al., 2009) is also only three steps longer.

Placing Guaibasaurus closer to Neotheropoda than Saturnalia, sauropodomorphs, herrerasaurids and Eoraptor (as in other published analyses) is only four steps longer.  Panphagia and Chromogisaurus remain in a clade with Saturnalia at the base of Sauropodomorpha, while Agnosphitys could go in any eusaurischian position outside Saturnaliinae and Pantydraco + other sauropodomorphs.  Thus another aspect of Guaibasauridae is poorly supported.

Placing Eoraptor outside Eusaurischia as in many analyses is also only four steps longer. Chindesaurus has an uncertain position in Saurischia but outside Sauropodomorpha, while herrerasaurids are more basal.  Notably, this also results in a lack of support for guaibasaurid monophyly. 

Placing Agnosphitys in Theropoda (as in Yates' analyses) is four steps longer.

Making Chindesaurus closer to Staurikosaurus than Herrerasaurus (as in Sereno, 1999) is only four steps longer, so weakly rejected.

Forcing Eoraptor to be a sauropodomorph is also only four steps longer, so is only weakly rejected.

Interestingly, placing Silesaurus in Saurischia is only four steps longer, which is actually more parsimonious than making it an ornithischian.

Placing Silesaurus in Ornithischia is surprisingly only five steps longer, but as noted above including more basal ornithischians and silesaurids might affect this.

Placing herrerasaurids outside Dinosauria, as was popular before 1993, is also only five steps longer.  Less than I thought it would be.  Chindesaurus follows them, but Eoraptor and the other saurischians remain in their usual positions.

Placing Herrerasaurus in Sauropodomorpha is six steps longer, and thus only moderately rejected.  I find this interesting as it is more parsimonious than placing it in Theropoda, but has not been advocated since 1981.  Staurikosaurus is still a herrerasaurid, while Chindesaurus is a slightly more derived sauropodomorph.

Placing Herrerasaurus in Theropoda is seven steps longer, so moderately rejected.  When this happens, there is no resolution in the clade containing Herrerasauridae, Chindesaurus, Eoraptor and Theropoda.

Forcing a polyphyletic Herrerasauridae with Herrerasaurus closer to (at least) theropods than Staurikosaurus results in the latter being an ornithischian, and is only seven steps longer. 

Placing Panphagia or Saturnalia in Theropoda is nine steps longer. Placing Chromogisaurus there is ten steps longer.  Thus they are probably sauropodomorphs.

Forcing a monophyletic Phytodinosauria of Sauropodomorpha+Ornithischia is thirteen steps longer, so strongly rejected, but not by as much as I'd think.  Herrerasaurids, Chindesaurus and Eoraptor form successively closer sister groups to Avepoda.

Then again, enforcing a completely heterodox Theropoda+Ornithischia clade is only fourteen steps longer.

Making Staurikosaurus a sauropodomorph but leaving Herrerasaurus outside the clade (as in Kischlat's papers where he calls the Staurikosaurus+Sauropodomorpha clade Pachypodosauria), is sixteen steps longer, so is strongly rejected.

Placing Marasuchus in Theropoda, as suggested by Olshevsky (they both lack phalanges on metatarsal V- character 352 in this analysis), is 28 steps longer, so is strongly rejected.  Amusingly, Marasuchus drags Herrerasauridae into Theropoda with it and becomes their sister group.  Since Herrerasaurus has a phalanx on metatarsal V, Olshevsky's character doesn't even work in this topology.  In case you're cusrious, it takes 14 steps to make Marasuchus a dinosaur and 8 stepts to make it and Silesaurus switch positions.

What conclusions can we draw from this study?

First, note that some very good recent analyses have recovered relationships that are rejected by as much as seven extra steps in this one.  The Tawa paper's placement of Herrerasauridae inside Theropoda for instance.  This indicates that just about any suggested arrangement of basal dinosauriforms is plausible, except that Saurischia is real, Marasuchus isn't a dinosaur, and Staurikosaurus is not a 'pachypodosaur' to the exclusion of Herrerasaurus.  It also suggests some current ideas (Silesaurus is an ornithischian, herrerasaurids and/or Eoraptor are theropods) are actually less parsimonious than some heterodox alternatives that have not been suggested before in a modern context (Silesaurus is a saurischian, herrerasaurids and/or Eoraptor are sauropodomorphs).  These could be some interesting hypotheses to explore.

But back to the idea at the start of the post- how well supported is Guaibasauridae?  Very poorly.  Agnosphitys can be placed pretty much anywhere, Guaibasaurus goes back to Theropoda with only four steps, and Panphagia is a more basal sauropodomorph with only three steps.  Even confining ourselves to trees less than five steps longer than the most parsimonious one, an approximation of what this dataset indicates would be-

      |  `--Staurikosaurus
            `--other sauropodomorphs

Ezcurra, 2010. A new early dinosaur (Saurischia: Sauropodomorpha) from the Late Triassic of Argentina: A reassessment of dinosaur origin and phylogeny. Journal of Systematic Palaeontology. 8(3), 371-425.

1 comment:

  1. Interesting analysis Mickey, I'll be curious to see how the Guaibasaurid synapomorphies (4 in all relevant taxa, then 3 more which were optimized to be present in their common ancestor) are distributed on a broader scope.