Re: Continental vs. Island extinctions

[MICHAEL HABIB <habib@jhmi.edu>]

I do think there are very strong arguments as to why extinction dynamics on 
continents are expected to differ from those on islands, including the 
importance of predation.  John Bois makes a good, articulate argument on 
several fronts (especially the idea that predators track prey ranges, for which 
there is ample evidence), but I do have several responses:

1. I would argue not so much that species on continents should be less prone to 
extinction as much as those species with large geographic ranges should be less 
prone to extinction (ie. the range the species actually inhabits).  Large 
ranges can only occur on large land masses (or the reverse if the animal is 
aquatic), but not all terrestrial ranges are large.  It has been shown pretty 
soundly that geographic range matters a great deal, and it generally proves to 
be a primary predictor of extinction risk in extant species.  See the work by 
Andy Purvis, John Gittleman, Kate E. Jones, Ian Owens, M. Bennett, Georgina 
Mace, and Marcel Cardillo for some examples.  Others exist as well.  However, I 
will be the first to note that these studies generally did not include measures 
of non-human predation (which is very hard to quantify).  So they show more the 
importance of geographic range than the unimportance of predation.

2. Part of "importance" with regards to threatening processes is frequency.  As 
John Bois points out, invasions on continents are rare, and isolation 
situations are also uncommon.  Therefore, the frequency with which 
predation-mediated extinction events occur on continents should be low.  In 
cases such as the NA/SA interchange, I agree that there _is_ strong evidence 
for predation-mediated extinctions.  But it is also true that situations such 
as the interchange happen very infrequently, even on a geoglogical timescale.

3. The primary reason that islands should not be extrapolated to continents 
with regards to extinction processes is simply that on islands, all geographic 
ranges are small.  On continents, there is a wide variety of range sizes.  
Those species with small ranges are going to be effectively island species, and 
thus be just as prone to extinction as species on real islands (including 
predation effects).  However, some species will have huge geographic ranges, 
and will be very resistant to any threat process with an area of effect smaller 
than their range (including predators with smaller geographic ranges...again 
NOT home ranges).  On oceanic islands, the threat processes will nearly always 
encompass the entire island, and thus the entire range of the species present.  
Species normally capable of wide dispersal will still be stuck with small 
ranges on islands.

4. An individual predator finding a prey item is not stochastic, but the 
probability that a predator's geographic range overlaps with a potential prey 
is somewhat stochastic.  On small islands, invasive predators are nearly 
guaranteed to overlap with the resident prey species in all areas they inhabit. 
 The larger the landmass, the less likely this becomes, at least for those 
continental prey species with large ranges.  There may be an continuous effect. 
 That is, predation may become more important as an extinction threat the 
smaller the landmass becomes, with range becoming less important (land mass is 
probably unimportant statistically on islands, by the way, because there is no 
variance in the land mass size: everything is small).  In the other direction, 
the larger the landmass, the more like a continent it becomes, with species 
ranges being important and predation being only a rare cause of extinction.

In this sense, continents would be one end of a continuum, and thus would be 
"big islands".  Treating them exactly the same as true islands is probably not 
easily supported, however.  My argument for predation/range continuum is 
difficult to test, but not impossible.  If my argument is correct, then 
Madagascar and other large islands should have extinction dynamics somewhere in 
between.

--Mike Habib