22 MAY 2009VOL 324SCIENCEwww.sciencemag.org
clusters whose phones have a different operat-
ing system. As more and more users start using
similar phones and operating systems, the size
of these clusters will increase. Percolation the-
ory predicts that once the market share of any
operating system reaches a critical threshold,
the system undergoes a phase transition and
most of the isolated clusters will merge into a
single large cluster containing a substantial
fraction of mobile phone users. At that point,
an MMS virus will be able to instantaneously
reach most mobile phone users, and con-
sequently, we’ll become seriously concerned
about mobile phone viruses. Wang et al. also
show that this transition point can be accurately
predicted by percolation theory (7). This transi-
tion also explains the relative absence of MMS
mobile phone virus outbreaks: Currently, we are
below the critical threshold, as smartphones still
have a small market share, which is further frag-
mented by the large number of different operat-
ing systems competing in the market.
Yet, consolidation of the mobile phone
industry is unavoidable, which means that the
phase-transition threshold will be inevitably
reached in the near future. Exactly when that
happens depends less on network science than
on market forces, i.e., the rate at which in-
dividuals switch to smartphones. However,
some estimates indicate that within 2 to 3
years, there will be more smartphones than
desktop computers. Thus, now is the time to
start preparing the theoretical knowledge and
tools to deal with this expected major threat to
1. S. Hempel, The Strange Case of the Broad Street Pump:
John Snow and the Mystery of Cholera (Univ. of California
Press, Berkley, CA, 2007).
2. P. Wang, M. C. González, C. A. Hidalgo, A.-L. Barabási,
Science 324, 1071 (2009); published online 2 April 2009
3. M. C. González, C. A. Hidalgo, A.-L. Barabási, Nature 453,
4. R. Lambiotte et al., PhysicaA387, 5317 (2008).
5. C. A. Hidalgo, C. Rodriguez-Sickert, PhysicaA387, 3017
6. J.-P. Onnela et al.,Proc. Natl. Acad. Sci.104, 7332 (2007).
7. R. Cohen, K. Erez, D. ben-Avraham, S. Havlin, Phys. Rev.
Lett. 85, 4626 (2000).
east Asia (1–3). More than 500,000
ha may have been converted
already in the uplands of China,
Laos, Thailand, Vietnam, Cam-
bodia, and Myanmar (see the fig-
ure, panel A). By 2050, the area of
land dedicated to rubber and other
diversified farming systems could
more than double or triple, largely
by replacing lands now occupied
by evergreen broadleaf trees and
swidden-related secondary vegeta-
tion (2). What are the environmen-
tal consequences of this conversion
of vast landscapes to rubber?
The conversion of both primary and sec-
ondary forests to rubber threatens biodiversity
and may result in reduced total carbon bio-
mass (3–5). Negative hydrological conse-
quences are also of concern—for example,
in the Xishuangbanna prefecture of Yunnan
province, China—but current data are too
sparse to quantify the extent of the impacts (3,
6). The effect of conversion to rubber on
catchment or regional hydrology depends, in
part, on the water use of rubber versus that of
the original displaced vegetation. Another fac-
tor is the degree to which rainwater infiltration
ubber plantations are ex-
panding rapidly throughout
montane mainland South-
is reduced when terraces are constructed on
sloping lands (see the figure, panel B). Un-
fortunately, a recent investigation into this
issue in Xishuangbanna was terminated by
regional authorities before sufficient data
were collected (6, 7).
The rapid emergence of rubber is the hall-
mark of a larger land-cover transition that has
been sweeping through montane mainland
Southeast Asia in recent decades: the demise
of swidden cultivation (also referred to as
shifting or slash-and-burn cultivation) (8).
Much of the upland areas that have been con-
verted to rubber in the region are historically
associated with swidden cultivation. Clinging
to the perception that swidden cultivation is a
destructive system that leads only to forest
loss and degradation, governments in South-
east Asia have tried to control or terminate it
through bans, declaration of forest reserves,
forced resettlement, monetary incentives, and
crop substitution programs (9, 10). The un-
controlled expansion of rubber in China was
encouraged in part because it was seen as a
favorable alternative to swiddening. Policies
such as the Sloping Land Conversion Program
supported the planting of rubber, because it
counts as reforestation. Yet such policies have
not always improved environmental condi-
tions. In the case of rubber, homogeneous
monocultures with myriad negative environ-
mental consequences have emerged. This situ-
ation is not new or isolated. The permanent
loss of forest cover through agrarian conver-
sion to oil palm in insular Southeast Asia pro-
vides a parallel (11, 12).
The demise of swidden cultivation in
Southeast Asia may have devastating
The Rubber Juggernaut
Alan D. Ziegler,1Jefferson M. Fox,2Jianchu Xu3
1Department of Geography, National University of
Singapore, Singapore 117570.
Honolulu, HI 96848, USA. 3World Agroforestry Centre,
Kunming Institute of Botany, China. E-mail: adz@
C H I N A
M YA N M A R
V I E T N A M
T H A I LA N D
Swidden versus rubber.(A) Montane mainland Southeast Asia (green shaded areas) is defined here as the lands between
300 and 3000 m above sea level. (B) Most swidden fields and fallows on slopes near villages and roads in this area in
Xishuangbanna have been converted to terraced rubber stands.
CREDITS: (PANEL A) JOHN B. VOGLER/UNIVERSITY OF NORTH CAROLINA, CHARLOTTE.
(PANEL B) ALAN ZIEGLER
Published by AAAS
on June 8, 2009
In retrospect, it has become clear that the Download full-text
environmental impacts of traditional swid-
dening were inconsequential until mountain
populations increased, cropping periods
lengthened, fallow periods became shorter,
and the cultivation of opium as a cash crop
proliferated after the Second World War.
Recent intensification of permanent agricul-
ture has had numerous negative environmen-
tal consequences: Erosion has accelerated and
stream sediment loads have increased where
repetitive cultivation is performed on steep
slopes without appropriate conservation
methods; permanent conversion of hill slopes
and road building have increased the risk of
landslides; irrigation of cash crops in the dry
season has desiccated streams; and use of pes-
ticides and fertilizers to sustain commercial
agriculture has reduced water quality (13–15).
The unrestricted expansion of rubber in
montane mainland Southeast Asia could
have devastating environmental effects. A
reliable assessment of the hydrological
threat requires new data, but time is too
short to wait for results from future catch-
ment monitoring programs aimed at quan-
tifying changes in streamflow variables
caused by rapid land-cover conversion to
rubber. Therefore, studies of rubber evapo-
transpiration and water use, such as those
currently being conducted in Thailand,
Cambodia, and Laos, are becoming increas-
ingly important. A substantial increase in
natural reserve areas could help to reduce
the threats to biodiversity and carbon stocks.
Another possible strategy involves paying
upland farmers to preserve forest resources.
A more realistic approach may be to pro-
mote diversified agroforestry systems in
which cash crops such as rubber and oil
palm play important roles, but are not
planted as monocultures.
References and Notes
1. V. Manivong, R. A. Cramb, Agroforest Syst.74, 113 (2008).
2. J. M. Fox et al. (East-West Center, HI, 2009); see www.
3. J. Qiu, Nature 457, 246 (2009).
4. H. M. Li et al., Biodivers. Conserv. 16, 1731 (2007).
5. H. M. Li et al., Forest Ecol. Manage. 225, 16 (2008).
6. M. T. Guardiola-Claramonte et al., Ecohydrology 1, 13
7. D. Cryanoski, Nature 251, 871 (2008).
8. C. Padoch et al., Geogr. Tidsskr. Dan. J. Geog. 107, 29
9. J. Fox et al., BioScience 50, 521 (2000).
10. D. J. Schmidt-Vogt, Trop. Forest Sci. 13, 748 (2001).
11. J. McMorrow, M. A. Talip. Global Environ. Change 11, 217
12. L. P. Koh, D. S. Wilcove, Nature 448, 993 (2007).
13. R. C. Sidle et al., Forest Ecol. Manage. 224, 199 (2006).
14. T. Forsyth, A. Walker, Forest Guardians, Forest Destroyers
(Univ. of Washington Press, Seattle, WA, 2008).
15. I. Douglas, Geog. Res. 44, 123 (2006).
16. This work was supported by NASA grants NNG04GH59G
and NNX08AL90G and by APN grant ARCP2006-0GNMY.
We acknowledge the contributions of M. Guardiola-
Claramonte, T. W. Giambelluca, J. B. Vogler, P. Troch, and
www.sciencemag.orgSCIENCEVOL 324 22 MAY 2009
infrared part of the seismic spectrum. They are
characterized by weak, if any, wave excitation
at high frequencies because they happen more
slowly than do ordinary, fast earthquakes. The
expectation is that these slow earthquakes
may provide a better understanding of regular
earthquakes, but we are still in the early stages
of understanding them.
Slow earthquakes go by a variety of names
depending on their magnitude and duration:
low-frequency earthquakes with durations of
<1 s (1), very-low-frequency earthquakes last-
ing about 20 s (2), and slow-slip events that
continue for days to months (3). Slow earth-
quakes are frequently accompanied by deep
tremor (4), which itself appears to be a swarm
of low-frequency earthquakes (5). To the
extent that we have been able to discern their
mechanism, slow earthquakes occur as shear
slip events of tectonic plates, just like ordinary
earthquakes (2, 3, 6). Although slow earth-
quakes are located on the same faults that host
ordinary, fast earthquakes, they differ in sev-
he past decade has witnessed the dis-
covery of a family of unusual earth-
quakes in what might be termed the
eral important respects. They grow steadily,
rather than explosively, with time (7), and
their stress drops are low (8).
We’d like to know what relation slow earth-
quakes have with ordinary earthquakes. Their
location on the deep extension of major faults
means that they will increase the shear stress
on the dangerous shallower stretches of these
faults (see the figure). It is therefore important
to know whether slow earthquakes cause an
increase in the likelihood of an adjacent large
earthquake. Slow slip has triggered small
nearby earthquakes in several instances (9,
10). However, there have been many well-
documented episodes of slow slip in Cas-
cadia, Japan, Mexico, Alaska, and Costa Rica,
none of which have been accompanied by a
large earthquake. Strong triggering of tremor
by lunar tides indicates that tremor can be sen-
sitive to small variations in stress. It is reason-
able, then, to speculate that tremor behavior
might vary along with variations in stress dur-
ing the seismic cycle between large earth-
quakes in subduction zones. Slow slip might
even play a role in a slow nucleation process
leading to a large earthquake, but that notion
Tremor can be triggered by waves from
distant earthquakes. Most areas that are
known to experience tremor in California are
of this type (11). One location south of
Parkfield also has tremor that occurs sponta-
neously, without being set off by seismic
waves, but what about the areas of triggered
tremor? Triggered tremor appears to have
much in common with spontaneous tremor,
but the relation to slow slip is not clear. Does
triggered tremor, like spontaneous tremor,
occur by shear slip on the deep extension of
faults, or is some other mechanism involved?
Finding better locations for triggered tremor
will be an important first step in addressing
this question. Location of tremors near
Cholame place them on the deep extension of
the San Andreas Fault (12).
We would like to understand the distribu-
tion and characteristics of tremor and slow
earthquakes more generally. In a few areas,
such as Japan and western North America,
monitoring can detect tremor and slow slip,
but in most areas of the world, seismic sta-
tions are too widely separated to detect it.
Does tremor occur only in tectonically active
areas? Even where tremor and slow slip are
known to occur, there are mysteries in its dis-
tribution. Slow slip is colocated with tremor
in Cascadia (13) and Japan (14), but the over-
lap may only be partial in Mexico (15), and
no tremor at all is observed during slow slip in
New Zealand (16). Spontaneous tremor is
seen under part of the San Andreas Fault but
not in other subduction zones. Triggered
Detection and monitoring of slow earthquakes
may provide a better understanding of
Deep Tremors and Slow Quakes
Gregory C. Beroza1and Satoshi Ide2
1Department of Geophysics, Stanford University, Stanford,
CA 94305, USA. 2Department of Earth and Planetary
Science, University of Tokyo, Tokyo 113-0033, Japan.
Published by AAAS
on June 8, 2009