More on Biogeographic memory

Pierre Deleporte Pierre.Deleporte at UNIV-RENNES1.FR
Fri Oct 12 17:41:17 CDT 2001

At 10:24 12/10/2001 +0200, Peter Hovenkamp wrote:

>>That the shortest spatial distance is in itself the right thing to take
>>into account for any "biogeographic" purpose seems highly questionable.
>But that was not the original question, I believe. This was - which line to
>draw (either connecting, or dividing) between the distributions of two
>sister groups.
>So the first criterion is taxonomic: sister groups. That excludes many of
>the possible "spatially nearest" neighbours. That solves many of your
>questions, I believe

Sorry if I was not clear enough but my point was not "connect or divide"
sister group distributions, this is somewhat a matter of graphic convention.
My question was: why the shortest distance ? (between sister groups in this
case, networks of extension ranges, connect or divide...). What if this
shortest distance between sister groups crosses a mountain ridge, and you
are dealing with freshwater fishes ?
"Inside a continuous freshwater landscape", the shortest distance will lead
you down the river to a confluent, and then upstream to the location of the
sister group. The closest stream in straight line on the map may not at all
be the nearest adjacent river downstream. Thus shortest distance in
freswhater will not be the shortest line as the crow flies. This way of
evaluating neighbourhood may be justified assuming some constraints on the
biogeographic evolutionary process for the taxon at stake, for instance
considering that fishes do not fly exactly like crows (I apologize, I
could'nt resist...).
Captures of streams by adjacent river basins are geomorphologically
possible, but can hardly be assumed as the general rule to start the
analysis with (would be analogous to assuming generalized horizontal
transfer of characters rather than homology by descent in direct lineage as
the rule for cladistic analysis: no hierarchical pattern could be retrieved
under such a rule).

I just want to emphasize the possibility of practising an "ecological
screening" of the taxa we use, according to different biogeographic
problems at stake.
If I want to test whether the Congo River is a major geographical barrier
for Central African fauna, I will NOT use big river fishes, will I ?
Discarding them requires that fishes of a given group were already strictly
infeodated to life in freshwater in the past, thus rivers were highways and
not barriers to them, but I'm ready to quietly assume this.

I can try to take it another way: suppose you don't find a clear congruent
pattern emerging from your spatial distribution analysis of all
sister-groups in biotas in some region. Maybe there is no traces left of
biogeographical history, or maybe you have mixed up in the analysis taxa
with readily different ecological requirements, and thus their different
historical patterns are confused, but would be revealed in separate
analyses of taxa belonging to different ecological categories, using
different logics adapted to these different ecological categories. Total
relevant evidence for a given approach includes what are relevant data AND
which way these data should be analysed, possibly a specific way.

>>We could say "I don't care", but the fishes could care
>>and we could suspect this.
>By looking at sister-group relationships.

As the crow flies ? (see above...).

>>Extracting and comparing patterns under different logics (or for
>>different taxa with different ecological skills) may also be useful (...)
>Hoping for congruence to emerge...

Yes, but congruence between distribution of taxa with comparable ecological
skills, so that a "barrier inducing vicariance" for instance could
potentially make sense for all of those taxa.

I'll admit that we could analyse spatial pattern for sister groups of
fishes "like that on the map" (shortest distances in straight line), and
then check whether we get a pattern fitting the hierarchy of connecting
rivers or not, if this is what we test.
But I want to emphasize the point that even with such an approach, the fact
that fishes don't fly will have to get into the reflection, sooner or
later... and why not sooner? Comparing the pattern resulting from the
analysis of fishes with RIVER spatial network seems to make sense. The
relation between long-term survival / dispersal abilities of the taxa
(fishes...) and the geological features at stake (rivers, savannas,
mountains, oceans, abyssal hotsprings...) makes sense in my view, and this
involves more argument than merely "sister-groups plus distances".

More than this, under this logic, contrasting "sister groups" becomes
irrelevant when the two groups have readily different ecological skills,
and thus are not assumed to respond to the same "geological" features
(e.g.: stenoece / ubiquitous, crawling / flying...).

I think we can carefully integrate "all relevant evidence" as underlined by
Peter, while at the same time:

- avoiding to fall back in narrative approaches devoided of clear methods
and combining all sorts of arguments an "ad hoc" way with no explicit rules
(we are clearly all trying to get rid of this),

- but also avoiding to apply blindly a unique rule for analysing spatial
distribution of any kind of living features, taking the map as a "tabula
rasa" and shortest distances as obligatorily relevant.

I agree with Grehan, Hovenkamp and others that some approaches seem to take
too little care of the map (limited set of predefined "areas"...), but
others tend to consider the map as a white sheet with just distances
reported on it... maybe we can go beyond this and improve adjustment of the
tools to the problematics.

I must also admit that my fish example transfers the problem from "spatial
nearest neighbour as the crow flies" to "nearest neighbour along
streams"... Maybe we will not escape this "proximity" criterion at some
level, but then let's decipher the implicit model and argue explicitly:
e.g. spatial proximity makes sense because sister groups are generally
assumed to have remained at the same place, roughly both sides of a
barrier, after a vicariance event. The method works (i.e. tells us
something of the past) IF this is the general process.

The presentation of a method in biogeography should ideally begin with
statements of this kind (exactly as for methods for phylogeny
reconstruction: which process is required for the particular kind of
pattern to make sense?). Vicariance biogeography considers processes of
vicariance, dispersal, extinctions (good effort)... we can still question
area definition and delineation, the possible hierarchy of explanation by
processes of V / D / E, ecological screening, exploratory "descriptive"
approaches versus testing of specific hypotheses etc...

Looks trivial? Then such basic information should better be given
immediately, short and clear, when presenting a method, or a new version of
a method. Goals, and assumptions regarding processes, no less. Relevant
data and relevant methods ensue.

>And another two pence ... how many eurocents is that?

This time you got me, Peter: I absolutely can't figure out the beginning of
an answer to that one... sorry, I'm French you know (to my great shame,
simply I was born like that :-)


More information about the Taxacom mailing list