|
How long does it take to recover from
a mass extinction event?
To some extent this depends on how "recovery" is defined. It
can be tempting to think of recovery as diversification until taxon count
reaches pre-extinction levels. There are many problems with this approach,
however, not least of which is that by this measure, recovery from the
end-Permian extinction would not have been complete among marine invertebrates
until some time in the Cretaceous. More commonly used measures are the
stabilizing of isotope values representing atmospheric chemistry or nutrient
cycling; or the recovery of specific communities, such as reefs.
This illustration, from Sepkoski's 1998 "Rates of speciation in
the fossil record," is of the end-Ordovician extinction and shows
another definition of recovery. In this paper, Sepkoski defined the recovery
as the point at which the rate of new originations per taxon (which adjusts
for standing diversity) peak and begin to drop off. In the case of this
mass extinction event, that point is the same as the one at which the
taxon count reaches pre-extinction levels. (This is not true of all extinction
events.)
The origination rate peaks about 20 million years after the peak in extinction
rates. Sepkoski considered this to be a long recovery period, and ascribed
the length of the recovery period to the enormity of the mass extinction
event. But long compared to what?
|
 |
 |
How long does it take to recover from
extinction events?
Dr. Sepkoski generously allowed us to work
with the extinction, origination, and total diversity metrics from his
compendia of marine families and marine genera throughout the Phanerozoic.
Jim and I used two different methods to compare
time series of extinction rate metrics to time series of origination rate
metrics. We found that the time series of origination rates were similar
to those of extinction rates, but the peaks and valleys trailed those
of extinction rates by about 10 million years. This was true for both
the familial and generic data sets, and it was true even when we hand-deleted
the mass extinctions from the time series.
Among other things, this means that, on average,
the magnitude of the extinction event doesn't have much effect on the
length of recovery time. It takes about 10 million years for a full recovery,
using the definition of recovery as that of origination rates peaking.
(This illustration is from Kirchner
and Weil, 2000a.)
|
|
Why is recovery time so uniform?
This result is surprising, both because the lag is so uniform, and because
it is so long. So we looked a little further into the extinction and origination
time series. First, we did the same tests as described above, but instead
of comparing the extinction rates to the origination rates, we compared
the extinction rate time series to itself, testing for autocorrelation.
With this dataset, we really couldn't resolve anything going on at lag
times less than five million years. However, at lag times between 5 and
30 Myr, extinction rate metrics (the heavy black lines) showed no autocorrelation
greater than random (the very fine dotted lines represent the random range).
Origination rates turned out to be a different story. They are autocorrelated
to a statistically significant degree up to lags of about ten million
years.
This suggests that origination and extinction are two very different
processes (well, 'duh!' most biologists would say). It also suggests that
the 10-million year lag of origination rates behind extinction rates has
to do with internal properties of the biological processes of diversification.
These figures are from Kirchner
and Weil 2000b. The inset shows the time series analyzed and the long-term
trend that we removed by detrending. The autocorrelations of the non-detrended
data are show in grey -- note that because these trends are high at one
end and low at the other they produce artifactual correlation over the
short ranges we're interested in. The heavy black lines show autocorrelation
in the detrended data. Heavy and dotted lines in both cases show results
obtained through different methods. The fine grey lines show median, 5th
and 95th percentiles of randomized data (the null hypothesis).
|
 |
 |
Why do
origination rates respond so sluggishly?
My theory (may be wrong!) is that extinction is a double-whammy. Each
species is a potential source of evolutionary radiation, which is why
many paleobiologists correct for standing diversity when calculating origination
rates. But each species plays an ecological role as well as an evolutionary
one, and constitutes all or part of a niche for other species in its ecosystem.
Some evolutionary theorists envisage extinction as leaving "open"
niches that are subsequently "filled" by radiation. Instead,
eliminating species destroys niches, and new opportunities for diversification
must spring from diversification itself. This would mean a slow ramping-up
of origination rates as new niches are formed, and a peak and drop-off
in rates when ecosystem structures start to lock in.
Duke on-campus parking is a great analogy. If each parking space were
a niche, the conventional way of looking at it would be that after cars
pull out, different cars pull into the same spaces. However, sometimes
after we pull out, the parking lot is destroyed overnight! (Niches are
gone when enough species are gone.) It's taking quite a while for the
new parking structure to go up, and space-for-space it won't be the same
as the old lots.
Jim
Kirchner (2002) has done some more work that seems to corroborate this.
The pattern is definitely there; inferring the process is riskier and
I don't want to blame him for that!
|
|
Is evolution a self-organized critical
process?
The physicist Per Bak predicted that evolution would turn out to be a
self-organized critical (SOC) process, and that the evidence of this would
be the discovery of a fractal structure in extinction and origination
time series from the fossil record.
Since fractals are autocorrelated, you can probably guess from the above
that the evidence for that is not very good in the case of extinctions.
However, before we had done that work, other authors believed they had
found a strong signature of fractals in the fossil record. Jim Kirchner
and I showed their results stemmed from an error they had made in interpolating
between unevenly spaced data points (this illustration is from Kirchner
and Weil, 1999). But heck, it's still an interesting idea for originations.
I'm not sure that if there were a fractal dimension in these metrics
that it would necessarily indicate SOC -- but we're still playing with
ways to test it ourselves! (Weil and Kirchner, in prep.)
|
 |
|
|
|
|
|
|
