Saturday, January 8, 2011

The Sauropodomorph Database Blog? Yates' analysis

This will be a crazy month for the Theropod Database Blog, with a comparative lack of theropods.  As part of my study of "Dachongosaurus", I examined Yates' (2007) matrix to see which characters it had.  That analysis has been modified and added to by several authors, so I combined all of these changes to make the new version as most recently published.  This involved-

Adding the 8 characters from the Glacialisaurus description (Smith and Pol, 2007).
Adding the 17 characters from the Chromogisaurus description (Ezcurra, 2010), modifying 215 and 221 (but not 245, 251 or 256), and changing many codings for basal taxa according to that paper.
Changing a few character descriptions and codings (mostly for Yunnanosaurus' skull and Tazoudasaurus' redescription) from Yates et al. (2010).  This meant changing a miscoded state for Seitaad too.
Changing Anchisaurus' codings from Yates (2010).
Adding  Aardonyx (Yates et al., 2010), Adeopapposaurus (Sertich and Loewen, 2010), Chromogisaurus (Ezcurra, 2010), Glacialisaurus (Smith and Pol, 2007), Gryponyx (Yates et al., 2010), Ignavusaurus (Knoll, 2010), herrerasaurid MACN PV-18649a (Ezcurra, 2010), Panphagia (Ezcurra, 2010) and Seitaad (Sertich and Loewen, 2010).

Because these have never been combined before, and could lead to a unique topology, I ran the matrix in PAUP.  The strict consensus is-

               |  |--Herrerasaurus
               |  |--MACN PV-18649a
               |  `--Staurikosaurus
                  |  |--Chindesaurus
                  |  `--+--Eoraptor
                  |     `--Avepoda
                     |  `--Saturnalia
                                       |  |--Unaysaurus
                                       |  `--+--Sellosaurus
                                       |     `--+--Gresslyosaurus
                                       |        `--Plateosaurus
                                          |  |--Eucnemesaurus
                                          |  `--Riojasaurus
                                             |  |--+--Coloradisaurus
                                             |  |  `--+--Lufengosaurus
                                             |  |     `--Glacialisaurus
                                             |  `--+--Gryponyx
                                             |     `--+--Massospondylus
                                             |        `--+--Adeopapposaurus
                                             |           `--Seitaad
                                                                  |  `--Lessemsaurus
                                                                                 |  `--Patagosaurus
                                                                                       |  `--Mamenchisaurus

Camelotia is outside Gongxianosaurus+Gravisauria. 

Unlike Ezcurra's matrix, guaibasaurids aren't necessarily monophyletic (and note pruning Agnosphitys or Panphagia does not make them monophyletic), but are all sauropodomorphs unlike Yates' original.  Also like Ezcurra but unlike Yates, Eoraptor is sister to Avepoda.

In Ignavusaurus' description, it was basal to Efraasia, but here it's more derived.  Unlike Yates et al. (2010) but like Yates (2004), Gryponyx is a massospondylid, and unlike Sertich and Loewen (2010) the Lufengosaurus clade are also massospondylids.  Yunnanosaurus is now basal to Jingshanosaurus


  1. If you have the time, you should try adding all of the other new early sauropodomorphs from the past few years: Asylosaurus, Chuxiongosaurus, Eomamenchisaurus, Lamplughsaura, Pradhania, Sarahsaurus, Spinophorosaurus, Tonganosaurus, Xixiposaurus...

  2. A first step would be to integrate the data from Upchurch and Galton's (2007) matrix, as revised by Yates et al. (2010). Xixiposaurus is already coded for that, after all. It emerges sister to Mussaurus, with that pair sister to Sarahsaurus. Most of Plateosauria is a huge polytomy though. Speaking of Sarahsaurus, I had forgotten about it, but it's already coded for Yates' matrix. Changes the tree somewhat to place riojasaurids outside Plateosauria and make massospondylids largely break down into a polytomy with Yunnanosaurus and other massopods. Aslyosaurus is already included as part of Thecodontosaurus, though it would be interesting to see it coded separately from the holotype dentary. I don't have the pdf of Chuxiongosaurus yet. Eomamenchisaurus and Tonganosaurus seem too derived to be worth it, but Spinophorosaurus could be useful. Lamplughsaura and Pradhania would be interesting, as they've only been coded for earlier versions of each matrix so far.

  3. Even if they're 'too derived', they're worth adding in my honest opinion. The most likely thing that could happen is they'll be in a polytomy with higher sauropods, heh. Do the codings for *Cetiosaurus*, etc. include data from the new descriptive papers (Galton and Knoll, etc.)? Just curious, heh.


  4. I just mean since the matrix was constructed to determine basal relationships, adding any more of the numerous derived sauropods probably wouldn't help determine relationships better. Yates' (2007) Melanorosaurus skull paper lists the sources used. For Cetiosaurus, he used Upchurch and Martin (2002, 2003). Checking the matrix reveals two braincase characters are coded.

  5. A very interesting interim experiment. I say interim because I have been working for the last two years on a thorough overhaul of my 2007 matrix. This includes combining all characters from Upchurch et al. 2007, adding all the new characters that have been proposed since and the scads of new taxa. And making sure that all of the taxa are scored for all of the characters they can be. Since my 2010 revisions for the Aardonyx paper showed that the existing matrices carried quite a few miscorings and other errors, I am rescoring from scratch - quite a laborious process. This is being done on Morphobank and I am also attempting to link as many pictures as I can that actually show the character state being scored. This takes time, so its nice to see at least a sketch of what the result MIGHT look like with a lot of the new data to be included.
    Some issues:
    1. Smith and Pol did not score their new characters for all available taxa - thus the distributions could potentially change, weakening the Lufengo + Glacialo clade. A side note is that some of the derived states for Lufengosaurus were taken from L. magnus not L. huenei. Fine if they are synonyms but this has never been demonstrated with apomorphies. Since one of the characters Smith and Pol use seems to vary between L. magnus and L. huenei this is a non-trivial issue.
    2. I really doubt Ignavusaurus is a non-plateosaurian. I would score some of the characters differently from Knoll, because they don't fully take into account the juvenile nature of the material (e.g. the shape of the distal humerus is looks primitive because the epiphysis is not fully developed (ditto femoral head). Further more Massospondylus should be resscored as polymorphic for the transverse flare of the distal tibia. Masso has only a weakly developed transverse flare and some individuals at the end of the range of variation actually fall back into the plesiomorphic condition of a square-shaped distal tibia. I suspect Ignavu is one of these.
    3. I accept Fraser and Padian's argument that the other small dinosauriform bones found with the Agnosphitys ilium very probably belong to the same taxon. If you include this data I doubt very much Agnosphitys will remain a sauropodomorph.
    4. Riojasaurus should probably be split since there isn't much good evidence placing the skull bearing specimen with the type skeleton, indeed there are quite a few postcranial differences.
    However it is great to see a large and inclusive Massospondylidae, I've always felt that these were the most diverse group of early sauropodomorphs (Sarahasaurus may well go here, as well as 'Gyposaurus' sinensis, if it isn't just a juvie Lufengosaurus).



  6. Thanks very much for the commentary, Adam.

    And that's one reason I haven't taken the time necessary to combine the matrices- only an expert can really do it well. I know from my theropod studies that there are many codings that you have to be familiar with the material to get correct. For instance, Deinonychus is illustrated by Ostrom (1969) as having small anterior dorsal hypapophyses and is coded and described that way by many authors, but Norell and Makovicky (1999) indicate a dorsal has a broken base of a large hyapophysis. Or for sauropodomorphs, if I were naively combining matrices I'd see Riojasaurus was uncoded for premaxillary tooth number by Yates but coded for it by Upchurch et al.. Even if I go the extra distance and check Bonaparte and Pumares' (1995) description to "confirm" there are five teeth, I'd be wrong because Yates et al. (2010) note that personal examination indicates the number cannot be determined. This is another reason the recent trend to leave codable states uncoded annoys me- how can I then tell if an author really thinks the state is ambiguous or is just being lazy? As someone who's not a sauropodomorph expert, I simply can't code taxa with much confidence. That leaves me especially credulous of studies like Peters' which take a ton of taxa and characters he's not intimately familiar with, literature-score them into PAUP and spew out a tree. It's just not that easy, and it's great to see Yates going through the effort to get it right.

    Interesting about Ignavusaurus, and I might try coding the rest of the Agnosphitys material in. "Riojasaurus should probably be split"... Is this the return of Strenusaurus?

  7. "Is this the return of Strenusaurus?"
    If you can show that PULR 56 (the skull specimen) is the same thing as the type of Strenusaurus, then yes....maybe.
    I hasten to add that I don't have any evidence that it is.