Biodiversity - Extinction by numbers

Abstract

How large will be the loss of species through human activities? And over what time period might that loss unfold? Habitat destruction is the leading cause of species extinction. Generally, many of the species found across large areas of a given habitat are represented in smaller areas of it. So habitat loss initially causes few extinctions, then many only as the last remnants of habitat are destroyed. Thus, at current rates of habitat destruction, the peak of extinctions might not occur for decades. But we should not be complacent. On page 853 of this issue, Myers et al.1 document an uneven, highly clumped, distribution of vulnerable species over the world's land surface. Within these 'biodiversity hotspots', habitats are already disproportionately reduced.

Full text

H Iis expected to become largely molecular
(H2) before forming stars. Star formation is
also an inefficient process, so lots of molec-
ular gas should be left over.
By far the most abundant chemical in a
molecular cloud is H2, but it does not radiate
detectable energy at typical cloud tempera-
tures of 20–50 K. Like other astronomers,
Braine et al. have to infer the amount of H2
by observing radiation from accompanying
trace molecules, principally carbon monox-
ide (CO). Crucial to this procedure is the
assumed ratio of H2 to CO in the cloud. This
is known to depend on the chemical abun-
dance in the surroundings, in this case the
TDG. The few existing studies of TDGs sug-
gest that their abundances are sufficiently
like those in our Galaxy to justify Braine et
al.’s use of the ‘standard’ CO-to-H2 conver-
sion factor derived for molecular clouds near
the Sun.
Many tidal tails have been searched for
CO (refs 11,12), but it has been detected in
only three13–15. Two serious impediments
prevent us from associating these with true
TDGs. First, the detected CO is nearby and
moving rather slowly with respect to the
probable parent galaxies. It is not obviously
escaping. Admittedly the situation is unclear
because as usual the velocity is measured in
only one direction, and the spatial separa-
tion relies on only two coordinates. Second,
although the CO appears to accompany the
atomic gas, the observations do not distin-
guish between molecules that were pulled
out of the parent galaxy or that were formed
within the ejected H I. The latter is favoured
by observations suggesting that H Iin most
large galaxies extends farther from the centre
of mass than CO, and so should be easier to
remove. Such a distinction is vital if one
wants the ejected material to look like an
irregular dwarf galaxy after leaving the site of
the interaction — a journey requiring more
than 108 years. In the accepted picture, con-
tinuing star formation demands ongoing
formation of molecular gas.
Braine et al.4present the first reasonably
strong case that some TDGs can synthesize
new molecular clouds. The CO in their
clouds is far removed from the parent galaxy
and moving rapidly. It also appears to be
concentrated at the same location as the H I,
and to share its velocity. Coincidence in loca-
tion and motion is expected if the molecules
formed in situ from the atomic gas in the
runaway cloud, but difficult to understand
otherwise. The velocities and distributions
of the H Iand CO are sufficiently similar to
support the case for in situ formation. My
chief reservation is that the CO emission is
sampled at only a few spots — certainly
enough to suggest that it follows the atomic
gas, but too few to make a solid case. More
complete sampling would help.
It is unlikely that all irregular dwarf galax-
ies were once TDGs — for one thing most
dwarfs have fewer heavy elements than the
TDGs measured so far. To explain even the
most element-rich dwarfs, TDGs must sur-
vive as long as 1010 years — the age of a typical
galaxy. The discovery of stars 106–107years
old at the ends of some 108-year-old tidal
tails doesn’t guarantee long-term stability.
Braine et al. propose that in situ formation of
molecules is a good indicator of stability.
I would agree that a gas cloud that is break-
ing up is not a good place to form molecules.
But tidally stripped material is clumpy, prob-
ably on smaller scales than we have yet
probed. It is possible that molecules could
still form within small, stable clumps, how-
ever dispersed. Computer simulations of
interactions cannot yet resolve this issue;
meanwhile I remain sceptical.
One final question: why hasn’t CO been
seen in other tidal features, such as the merg-
ing galaxies shown in Fig. 1? If one assumes
the most reasonable value for the CO-to-H2
conversion factor, then earlier upper limits
on detection are generally consistent with
the CO signals seen by Braine and colleagues.
Again, additional (and more sensitive) data
are needed. The detections of molecular gas
in TDGs are certain to stimulate such obser-
vations, and to lead to a better understanding
of the formation of dwarf galaxies. ■
Gary Welch is in the Department of Astronomy and
Physics, Saint Mary’s University, Halifax, Nova
Scotia, B3H 3C3, Canada.
e-mail: gwelch@orion.stmarys.ca
1. van den Bergh, S., Abraham, R. G., Ellis, R. S., Tanvir, N. R. &
Glazebrook, K. G. Astron. J. 112, 359–368 (1996).
2. Balland, C., Silk, J. & Schaeffer, R. Astrophys. J. 497, 541–554
(1998).
3. Governato, F., Gardner, J. P., Stadel, J., Quinn, T. & Lake, G.
Astron. J. 117, 1651–1656 (1999).
4. Braine, J., Lisenfeld, U., Due, P.-A. & Leon, S. Natu re 403,
867–869 (2000).
5. Toomre, A. & Toomre, J. Astrophys. J. 178, 623–666 (1972).
6. Zwicky, F. Ergeb. Exakt. Naturwiss. 29, 344–385 (1956).
7. Barnes, J. E. & Hernquist, L. Nature 360, 715–717 (1992).
8. Elmegreen, B. G., Kaufman, M. & Thomasson, M. Astrophys. J.
412, 90–98 (1993).
9. Hunter, D. A. Publ. Astron. Soc. Pacif. 109, 937–950 (1997).
10.Duc, P.-A. & Mirabel, I. F. in Proc. 186th Symp. Int.
Astronomical Union (eds Barnes, J. E. & Sanders, D. B.) 5
(Kluwer, Dordrecht, 1999).
11.Young, J. S., Tacconi, L. J. & Scoville, N. Z. Astrophys. J. 269,
136–146 (1983).
12.Smith, B. J. & Higdon, J. L. Astron. J. 108, 837–843 (1994).
13.Brouillet, N., Henkel, C. & Baudry, A. Astron. Astrophys. 262,
L5–L8 (1992).
14.Walter, F. & Heithausen, A. Astrophys. J. 519, L69–L72 (1999).
15.Smith, B. J., Struck, C., Kenney, J. D. P. & Jogee, S. Astron. J.
117, 1237–1248 (1999).
16. Hibbard, J. G. & van Gorkum, J. H. Astron. J. 111, 655–695 (1996).
news and views
NATURE
|
VOL 403
|
24 FEBRUARY 2000
|
www.nature.com 843
Figure 1 Birthplace of dwarf galaxies?
a, Interacting galaxies in the NGC4676 system,
showing the remarkably straight tail of stars
pulled from one of the two galaxies16. The image
is a greyscale negative of visible light. The black
scale bar shows a distance of 90,000 light years
— about the size of our own Galaxy. b, The same
system seen with a hydrogen filter, so that
atomic hydrogen appears blue, whereas
hydrogen that has been ionized (presumably by
young stars) appears red. Curiously, the
molecular gas expected to accompany these stars
is not found in the tail. Images of similar tails
analysed by Braine et al.4do show signs of
molecular hydrogen, suggesting that these tails
are involved in continuous star formation.
How large will be the loss of species
through human activities? And over
what time period might that loss
unfold? Habitat destruction is the leading
cause of species extinction. Generally, many
of the species found across large areas of a
given habitat are represented in smaller
areas of it. So habitat loss initially causes few
extinctions, then many only as the last rem-
nants of habitat are destroyed. Thus, at cur-
rent rates of habitat destruction, the peak of
extinctions might not occur for decades.
But we should not be complacent. On page
853 of this issue, Myers et al.1document an
uneven, highly clumped, distribution of
vulnerable species over the world’s land sur-
face. Within these ‘biodiversity hotspots’,
habitats are already disproportionately
reduced.
Conservatively, there are about seven
million species of eukaryote2— a definition
encompassing most organisms that would
be generally recognized as plants or animals
but excluding bacteria, for instance. Most
Biodiversity
Extinction by numbers
Stuart L. Pimm and Peter Raven
© 2000 Macmillan Magazines Ltd

of these seven million are animals and about
85% are terrestrial.
Humanity is rapidly destroying habitats
that are most species-rich. About two-thirds
of all species occur in the tropics, largely in
the tropical humid forests3. These forests
originally covered between 14 million and
18 million square kilometres, depending on
the exact definition, and about half of the
original area remains4. Much of the rain-
forest reduction is recent, and clearing now
eliminates about 1 million square kilometres
every 5 to 10 years4–6. Burning and selective
logging severely damages several times the
area that is cleared5,6.
To convert habitat loss to species loss, the
principles of island ecology are applied to the
terrestrial ‘islands’ that remain in a ‘sea’ of
converted land7. The relationship between
number of species and island area is nonlin-
ear, and from this one can predict how many
species should become extinct as the size of
the forest islands shrinks. These doomed spe-
cies do not disappear immediately, however.
How does one go about calculating the
rate of species extinctions from habitat frag-
ments? There have been only a few such esti-
mates, but projections based on a species
survivorship curve with a half-life of roughly
50 years seem reasonable8. Combining the
rate of habitat loss, the species-to-area rela-
tionship and the survivorship curve gives a
crude extinction curve (curve a in Box 1).
From this, we would expect that current
extinction rates should be modest — on the
order of a thousand species per decade, per
million species, a figure that matches other
estimates9.
Because the species–area curve is non-
linear, the clearing to date of half of the
humid forests is predicted to eliminate only
15% of the species that they contain. The
time delays before extinction mean that
many more species should be ‘threatened’
than have already become extinct; that is,
they are thought likely to become extinct in
the wild in the medium-term future. At least
12% of all plants10 and 11% of all birds11
come into this category.
Of course, clearing the remaining half of
the forests would eliminate the other 85% of
species that they contain. The extinction
curve should accelerate rapidly to a peak by
the middle of the twenty-first century if the
rate of forest clearing remains constant. But
it will be upon us sooner if that rate is
increasing — as seems probable4,6.
Once the extinction peak has passed, the
extinction curve declines into the twenty-
second century as species are lost from the
remaining fragments of habitat. The relative
height of the peak depends critically on the
amount of habitat that remains. A value of
5% of remaining habitat (see Table 1 on page
854) would protect about 50% of all the
forests’ species; smaller percentages would
lead to smaller estimates of surviving species.
Modest tinkering with parameters does
not alter the ‘fewer extinctions now, many
more later’ feature of the extinction curve
(curve a in Box 1). But the calculations of
Myers et al.1do. They show that roughly
30–50% of plant, amphibian, reptile, mam-
mal and bird species occur in 25 hotspots
that individually occupy no more than 2% of
the ice-free land surface (see the map on
page 853). That is, terrestrial species with
small geographical ranges are numerous and
they have highly clumped distributions.
Myers et al. exclude the oceans from their
analysis. But there, too, fishes and other
organisms dependent on coral reefs are simi-
larly concentrated12.
Habitat destruction acts like a cookie
cutter stamping out poorly mixed dough9.
Species found only within the stamped-out
area are themselves stamped out. Those
found more widely are not. Moreover,
species with small ranges are typically
scarcer within their ranges than are more
widely distributed species, making them yet
more vulnerable. Consequently, even ran-
dom destruction would create centres of
extinction that match the concentrations of
small-ranged species — the hotspots9.
Worse, however, Myers et al. show that
the cookie cutter is not random — it is
malevolent. In the 17 tropical forest areas
designated as biodiversity hotspots, only
12% of the original primary vegetation
remains, compared with about 50% for
tropical forests as a whole. Even within those
hotspots, the areas richest in endemic plant
species — species that are found there, and
only there — have proportionately the least
remaining vegetation and the smallest areas
currently protected.
Applying the species–area relationship to
the individual hotspots gives the prediction
that 18% of all their species will eventually
become extinct if all of the remaining habi-
tats within hotspots were quickly protected
(curve c in Box 1). Assuming that the higher-
than-average rate of habitat loss in these hot-
spots continues for another decade until only
the areas currently protected remain (curve
b in Box 1), these hotspots would eventually
lose about 40% of all their species. All of
these projections ignore other effects on bio-
diversity, such as the possibly adverse impact
of global warming, and the introduction of
alien species, which is a well-documented
cause of extinction of native species.
Unless there is immediate action to sal-
vage the remaining unprotected hotspot
areas, the species losses will more than dou-
ble. There is, however, a glimmer of light in
this gloomy picture. High densities of small-
ranged species make many species vulnera-
ble to extinction. But equally this pattern
allows both minds and budgets to be con-
centrated on the prevention of nature’s
untimely end. According to Myers et al.,
these areas constitute only a little more than
one million square kilometres. Protecting
them is necessary, but not sufficient. Unless
the large remaining areas of humid tropical
forests are also protected, extinctions of
those species that are still wide-ranging
news and views
844 NATURE
|
VOL 403
|
24 FEBRUARY 2000
|
www.nature.com
Three projections of how
numbers of species extinctions
in tropical forests may unfold
from forest clearance. Curve a is
the extinction curve on current
estimates, not taking into
account biodiversity hotspots.
According to the relationship
Sn/So
4(
Ao
/
An
)0.25 (see refs 6–8),
as habitat is reduced from an
original area of
Ao
to
An
,
An
will
hold
Sn
viable species in year
n
from an original total of
So
. The
So
–
Sn
doomed species will die
off with a half-life of 50 years7.
With a constant rate of forest
clearance, this curve takes
time to peak because of the
nonlinear relationship between
species and area, and the time
lags for species to become
extinct.
Myers
et al.
1 identify 25
biodiversity hotspots around
the world, of which 17 are in
tropical forests. These areas
have already suffered
disproportionate loss of
primary vegetation, meaning
that the many species they
contain are under particular
threat of extinction. If all
remaining habitat in hotspots is
saved (as shown in curve c),
some 18% of their species
will be lost. The same half-life
for currently threatened species
is used as in curve a. However,
if the hotspots are cleared in
the next decade to the point
where only currently protected
areas are saved (curve b) then
the total extinctions will be
higher. S. L. P. & P. R.
0
10,000
20,000
30,000
40,000
50,000
Extinctions per million species per decade
2000 2020 2040 2060 2080 2100
Decade
a, Tropical forest
extinction curve
on current
estimates
c, Extinction curve if
all hotspots are saved
b, Extinction curve if
protected hotspots
are saved
Box 1: Extinctions in tropical forests, 2000–2100
© 2000 Macmillan Magazines Ltd

used for measuring femtosecond pulses and
is a close cousin of the pump–probe tech-
nique used in femtochemistry.
For autocorrelation measurements, a
femtosecond pulse is split into two at a beam
splitter. (A beam-splitter functions like a
window at night. In a lighted room you can
see your reflection in the window while
simultaneously being able to see outside.)
The two beams are sent through different
paths, and usually recombine within a crys-
tal with nonlinear optical properties.
Because the harmonic light produced by the
nonlinear crystal is stronger when the two
pulses are overlapped, observing the signal
strength as a function of the difference in the
path length of the two beams gives a
measurement of the pulse’s duration.
Unfortunately, the short duration and
wavelength of attosecond pulses means that
neither traditional beam splitters nor nonlin-
ear crystals are suitable. Papadogiannis et al.
should exceed those in the hotspots within a
few decades (Box 1). ■
Stuart L. Pimm is at the Center for Environmental
Research and Conservation, MC 5556, Columbia
University, 1200 Amsterdam Avenue, New York,
New York 10027, USA.
e-mail: stuartpimm@aol.com
Peter Raven is at the Missouri Botanical Garden,
PO Box 299, St Louis, Missouri 63166, USA.
e-mail: raven@mobot.org
1. Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca,
G. A. B. & Kent, J. Nature 403, 853–858 (2000).
2. May, R. M. in Nature and Human Society (ed. Raven, P. H.)
(Natl Acad. Sci. Press, Washington DC, 2000).
3. Raven, P. H. (ed.) Research Priorities in Tropical Biology (Natl
Acad. Sci. Press, Washington DC, 1980).
4. Skole, D. & Tucker, C. J. Science 260, 1905–1910 (1993).
5. Nepstead, D. C. et al. Nature398, 505–508 (1999).
6. Cochrane, M. A. et al. Science 284, 1832–1835 (1999).
7. Brooks, T. M., Pimm, S. L. & Collar, N. J. Conserv. Biol. 11,
382–384 (1997).
8. Brooks, T. M., Pimm, S. L. & Oyugi, J. O. Conserv. Biol. 13,
1140–1150 (1999).
9. Pimm, S. L., Russell, G. J., Gittleman, J. L. & Brooks, T. M.
Science 269, 347–350 (1995).
10.Walter, K. S. & Gillett, H. J. IUCN Red List of Threatened Plants
(IUCN, Gland, Switzerland, 1998).
11.Collar, N. J., Crosby, M. J. & Stattersfield,A. J. Birds to Watch 2
(Smithsonian Inst. Press, Washington DC, 1994).
12.McAllister, D., Schueler, F. W., Roberts C. M. & Hawkins, J. P. in
Advances in Mapping the Diversity of Nature (ed. Miller, R.)
155–175 (Chapman & Hall, London, 1994).
have sidestepped these problems by splitting
the femtosecond pulse before the attosecond
pulse is produced, and using a rare gas for the
dual purpose of producing and measuring
the attosecond pulses. All characteristics
necessary for measurement are present, but
because production and measurement are
entwined, their measurement is not com-
pletely transparent, so the method is contro-
versial. But the controversy will not last for
long because the basic physics behind the
measurements is well understood.
Although current research is naturally
focused on the production and measure-
ment of attosecond pulses, it is important to
look at the future direction of attosecond sci-
ence. For one thing, it will benefit from the
experience gained in previous experiments
with ultrashort pulses. This is because we
have been performing indirect attosecond
experiments (often referred to as strong-
field science) for a decade or more and the
necessary tools are well developed. For
example, normal visible laser pulses contain
electric fields that change significantly dur-
ing 100 attoseconds (Fig. 1). The electric
field of the light pulse is proportional to the
force that the field exerts on any electrically
charged particle. With modern laser tech-
nology, the forces can be very large and
precisely controlled. So, hidden within the
interactions of intense visible laser light with
matter are attosecond or near-attosecond
phenomena induced by the laser field.
Indeed, indirect attosecond science can
explain the attosecond pulses produced by
Papadogiannis et al. As the electric field of
the laser becomes strong, one of the electrons
is pulled free from an atom in the argon gas.
Once free, it moves in response to the strong
force of the laser; first it is driven away from
the ion, then back. Its path can be compared
to that of a lifeboat launched from a ship in a
stormy sea. The ship (or ion) from which it
detached is an obstacle that remains in the
area and with which it can collide. As with the
lifeboat and ship analogy, the wave deter-
mines the possible time of collision. In the
violent electron–ion collision that may
occur, very-short-wavelength radiation can
be emitted. So the precise synchronization of
the individual attosecond pulses with the
much-longer-wavelength radiation that
produced them (Fig. 1) is not an accident,
but is forced by the field. Quantum mechan-
ics adds ‘fuzziness’ to this essentially classical
description, but does not change it much.
No experiments have yet been performed
with attosecond pulses, and we cannot even
produce an isolated pulse. There are, howev-
er, many ideas and proposals waiting in the
wings, all involving increasingly precise con-
trol over oscillations of the strong laser field.
Once attosecond pulses can be produced
routinely, indirect and direct attosecond sci-
ence will become increasingly integrated.
Whereas the goal behind the development
news and views
NATURE
|
VOL 403
|
24 FEBRUARY 2000
|
www.nature.com 845
For the past five years, scientists have
stood on the threshold of generating
attosecond laser pulses but have been
unable to cross it. (An attosecond — 10118 s
— is to one second as one second is to the
age of the Universe.) This may have finally
changed with the publication of a paper by
Papadogiannis et al. (Phys. Rev. Lett. 83,
4289–4292; 1999), who claim to have
measured trains of attosecond pulses. The
previous record for the shortest laser pulse
was 4.5210115 s (4.5 femtoseconds). Pulses
in the femtosecond range led to the develop-
ment of femtochemistry — making it possi-
ble to study chemical reactions in real time
— for which the 1999 Nobel Prize in Chem-
istry was awarded to Ahmed Zewail. But the
new science that will ultimately emerge from
attosecond research will have its own unique
drive.
The approach that Papadogiannis et al.
use for generating attosecond pulses has been
under investigation for some time. They use
the short-wavelength harmonics generated
when rare gases (such as argon) ionize as a
result of irradiation from an intense femto-
second pulse. Harmonics occur at multiples
of the frequency of the original femtosecond
pulse. Next, the authors select a set of these
harmonics, which theory indicates should
combine to produce a train of pulses about
100 attoseconds in duration (Fig. 1). Such
pulses have probably already been created in
many laboratories, but no one has been able
to measure them accurately.
Papadogiannis et al. may eventually be
recognized as the parents of experimental
attosecond science because they have actual-
ly measured the duration of these pulses.
Their measurement process is experimental-
ly simple, but theoretically complex. This is
because the production of the attosecond
pulses is intrinsically entwined with the
measurement. They use a technique influ-
enced by autocorrelation, which is widely
Figure 1 Train of attosecond pulses similar to that
produced by Papadogiannis et al. Here, the initial
femtosecond pulse (red) is much shorter than the
one that they used, and the higher-frequency
harmonic radiation (blue) is much more intense
than in their experiment. The offset between the
peak of the initial pulse and of the harmonic
radiation illustrates the delay in the harmonic
emission imposed by the laser field oscillation.
–3
Electric field
–1.0
–0.5
–0.0
0.5
1.0
0
Time (femtoseconds, 10–15 s)
3
Laser physics
Attosecond pulses at last
Paul Corkum
© 2000 Macmillan Magazines Ltd