Panbiogeography: A recent example

To further illustrate my recent comments on the practice of many phylogeographers of dogmatically rejecting the possibility of long distance dispersal, I would like to discuss a (relatively) recent string of papers dealing with the example of the daisy genus Abrotanella.

Abrotanella is, to our current knowledge, phylogenetically relatively divergent and either the sister group of or on a grade with the entire large tribe Senecioneae (groundsel or fireweed and relatives), and as such probably best accommodated in its own tribe. It is a genus of 21 alpine, often mat-forming species and is found in southeast Australia, Tasmania, New Zealand, New Guinea, South America, and some subantarctic islands.

So basically Australasia and South America. One can see where this is going: A certain type of person will look at a distribution like this and say, Gondwana! This must be the result of continental drift! Molecular phylogenetic data, however, indicated some time ago that the crown group of this genus is too young for its distribution to be explained in this way (Wagstaff et al., 2006), and thus some long distance dispersal must have been involved.

As fantasy author Terry Pratchett once observed, where we define the beginning and end of stories is somewhat arbitrary. For present purposes, I will say...

...it begins the panbiogeographer Michael Heads publishing a paper in which he used Abrotanella as one of several examples to argue that the way dated phylogenies are employed today biases the results towards the conclusion of long distance dispersal (Heads, 2011). Fossils are used to calibrate nodes in the phylogenetic tree, and fossils can only provide minimum clade ages - the group is at least as old as that, but it could still be considerably older without having left any fossils, at least without any that have been found so far.

Heads now argued that it had become practice to "transmogrify" these minimum clade ages into maximum clade ages in Bayesian analysis, and this would lead people to grossly underestimate the potential age of taxa, to conclude that they can only have dispersed long after continental break-ups, and thus to reject vicariance as a biogeographical process. According to him, a less biased approach would be to "not base clock dates on fossils at all" and instead date clades according to "tectonics and geographic distribution". In other words, clade ages should be inferred from dates of continental break-ups under the assumption of vicariance.

One problem with this should be clear immediately: if we want to figure out whether a present pattern of distribution is best explained with vicariance or with dispersal, then this approach would amount to circular reasoning: vicariance is already built into the premises of the study.

From the perspective of formal logic, this should suffice as a rebuttal. But a few colleagues from Sweden and New Zealand did something much better: They took Head's suggestion seriously and tested where it would lead (Swenson et al., 2012). After first carefully explaining that no, the Bayesian priors that are used in these kinds of analyses do not actually "transmogrify" minimum into maximum ages (with a nice figure that could well be labelled "probability distributions for beginners"), they set out to do three dated phylogenetic analyses of Abrotanella, one calibrated with a fossil as usual, and two calibrated with two different tectonic events as suggested by Heads.

The point here was to demonstrate what forcing Abrotanella into an age that makes vicariance plausible would do to the dating of the rest of the phylogeny. The fossil-dated phylogeny shows the usual: Abrotanella is too young for its distribution to be explained by vicariance, and the daisies as a whole have reasonable clade ages. In one of the other two cases, thanks to the power of circular reasoning, Abrotanella is just as old as the vicariance explanation would need it to be; in the other tectonic dating approach, even this minimum hurdle cannot be cleared. But everything else is where it really gets funny.

If the Abrotanella phylogeny is dated with either of the two tectonic events as preferred by Heads, the daisies as a whole are inferred to be either 511 million years or 1.5 billion years old, prompting Swenson et al. to choose one of the most hilarious titles I have ever seen on a scientific paper. Now obviously groups may always be older than their earliest fossils, but 511 million years ago there were no vascular plants, much less daisies, and 1.5 billion years ago the show was run by cyanobacteria!

Since then, Heads (2012) has taken the opportunity to reply. To the best of my ability to make sense of this paper, it seems to be an attempt to show that the sequence of cladogenesis events in Abrotanella would be consistent with the sequence of tectonic events. This would certainly, on its own, and if we knew nothing whatsoever about higher level phylogenetic relationships and clade ages, be a valid approach that might lead to the tentative conclusion that the current distribution of the genus is, well, consistent with vicariance, and nothing more.

But we do in actual fact know a great deal about higher level phylogenetic relationships and clade ages. In effect then, Head's reply appears to simply ignore the problems Swenson et al. have demonstrated. The closest he gets to making an attempt at a rebuttal is some vague hand-waving that implies that rates of sequence evolution are too variable across lineages to be sure of anything.

But that just won't do. If Abrotanella is supposed to be old enough for vicariance, but the daisies as a whole young enough to have originated after, say, the seed plants as a whole, substitution rates would have to be ridiculously low in Abrotanella and/or ridiculously high in the rest of the family. And note that the sequence regions used in these studies are very standard ones whose behaviour is quite well understood.

So for Head's conclusions to work, the rates would have to vary to quite an implausible degree. And what is more, this is pretty much a textbook case of introducing unnecessary ad hoc hypotheses to save an underlying belief that would otherwise not survive an encounter with the evidence. In the present case, wildly unrealistic shifts in substitution rates are invoked to avoid the rather unspectacular conclusion that long distance dispersal is one of several possible biogeographic processes.

Update: minor correction where I wrote 'diversification' instead of 'substitution'.

References

Heads M, 2011. Old taxa on young islands: a critique of the use of island age to date island-endemic clades and calibrate phylogenies. Systematic Biology 60: 204-218.
Heads M, 2012. South Pacific biogeography, tectonic calibration, and pre-drift tectonics: cladogenesis in Abrotanella (Asteraceae). Biological Journal of the Linnean Society 107: 938-952.
Swenson U, Nylinder S, Wagstaff SJ, 2012. Are Asteraceae 1.5 billion years old? A reply to Heads. Systematic Biology 61: 522-532.
Wagstaff SJ, Breitwieser I, Swenson U, 2006. Origin and relationships of the austral genus Abrotanella (Asteraceae) inferred from DNA sequences. Taxon. 55: 95-106.