© 2006 Nature Publishing Group
Widespread amphibian extinctions from
epidemic disease driven by global
J. Alan Pounds1, Martı ´n R. Bustamante2, Luis A. Coloma2, Jamie A. Consuegra3, Michael P. L. Fogden1,
Pru N. Foster4†, Enrique La Marca5, Karen L. Masters6, Andre ´s Merino-Viteri2, Robert Puschendorf7,
Santiago R. Ron2,8, G. Arturo Sa ´nchez-Azofeifa9, Christopher J. Still10& Bruce E. Young11
As the Earth warms, many species are likely to disappear, often because of changing disease dynamics. Here we show
that a recent mass extinction associated with pathogen outbreaks is tied to global warming. Seventeen years ago, in the
mountains of Costa Rica, the Monteverde harlequin frog (Atelopus sp.) vanished along with the golden toad (Bufo
periglenes). An estimated 67% of the 110 or so species of Atelopus, which are endemic to the American tropics, have met
the same fate, and a pathogenic chytrid fungus (Batrachochytrium dendrobatidis) is implicated. Analysing the timing of
losses in relation to changes in sea surface and air temperatures, we conclude with ‘very high confidence’ (.99%,
following the Intergovernmental Panel on Climate Change, IPCC) that large-scale warming is a key factor in the
disappearances. We propose that temperatures at many highland localities are shifting towards the growth optimum of
Batrachochytrium, thus encouraging outbreaks. With climate change promoting infectious disease and eroding
biodiversity, the urgency of reducing greenhouse-gas concentrations is now undeniable.
Humans are altering the Earth’s climate1–4and thus the workings of
livingsystems5–8,including pathogens andtheirhosts9–11. Among the
predicted outcomes is the extinction of many species10,12, but
detecting such an effect is difficult against a backdrop of other
changes, especially habitat destruction. One approach is to focus
on organisms for which current rates of extinction exceed
those expected from habitat loss. Amphibians are a case in point.
Thousands of species have declined, and hundreds are on the brink
of extinction or have already vanished13. The Global Amphibian
Assessment (GAA) lists 427 species as “critically endangered”,
including 122 species that are “possibly extinct”13. A majority of
the former, and nearly all of the latter, have declined even in
seemingly undisturbed environments.
The causes have remained unclear, in part because of their
complexity14–16. Although pathogens are implicated14–28, their
relationship to environmental change is poorly understood. Here
we test the “climate-linked epidemic hypothesis”29–34, which predicts
declines in unusually warm years but does not assume a particular
disease or chain of events. Recent studies have considered this
idea15,18,21,23,28, yet data have not permitted a geographically broad
test that examines landscape alteration, global warming and climate
fluctuations on the timescale of El Nin ˜o. Suffering widespread
extinctions often despite habitat protection, harlequin frogs (Atelo-
pus) afford such a test. A new database, produced by 75 researchers,
documents the case in unprecedented detail, owing to the nature of
these members of the toad family (Bufonidae)26. Brightly coloured
identified. For the first time, data indicate when each of numerous
species was seen for the last time.
Our analyses capitalise on insights gained by alternating between
large and small spatial scales35(Supplementary Fig. 1). Since epi-
demics of Batrachochytrium are implicated in Atelopus extinctions in
Central and South America26, we first explain that the predicted
associationwithwarmyears, if juxtaposed with theory regardingthis
chytrid, is a paradox. We then: (1) assess large-scale altitudinal
patterns of extinction risk with this paradox in mind; (2) consider
determinants of local climate in the case of the golden toad and the
Monteverde harlequin frog to select large-scale temperature signals
for analysing the biological patterns; (3) show that the timing of the
widespread extinctions is strongly tied to these signals; and (4)
explore local climate from a chytrid’s viewpoint to frame a solution
to the paradox.
The climate–chytrid paradox
The climate-linked epidemic hypothesis predictsamphibian declines
in unusually warm years, because shifts in temperature or related
variables often influence disease dynamics9–11. As temperatures rise,
climate fluctuations may cross thresholds for certain pathogens,
triggering outbreaks. Many diseases are expected to become more
lethal, or to spread more readily, as the Earth warms9–11.
1Golden Toad Laboratory for Conservation, Monteverde Cloud Forest Preserve and Tropical Science Center, Santa Elena, Puntarenas 5655-73, Costa Rica.2Museo de Zoologı ´a,
Centro de Biodiversidad y Ambiente, Escuela de Biologı ´a, Pontificia Universidad Cato ´lica del Ecuador, Avenida 12 de Octubre 1076 y Roca, Apartado 17-01-2184, Quito, Ecuador.
3Department of Environmental Science, Barnard College, Columbia University, 3009 Broadway, New York, New York 10027, USA.4Center for Climate Studies Research,
University of Tokyo, Kombaba, 4-6-1, Meguro-ku, Tokyo 153-8904, Japan.5Laboratorio de Biogeografı ´a, Escuela de Geografı ´a, Facultad de Ciencias Forestales y Ambientales,
Universidad de Los Andes, Apartado 116, Me ´rida 5101-A, Venezuela.6Council for International Educational Exchange, Monteverde, Puntarenas 5655-26, Costa Rica.7Escuela de
Biologı ´a, Universidad de Costa Rica, San Pedro, Costa Rica.8Texas Memorial Museum and Department of Integrative Biology, University of Texas, Austin, Texas 78712, USA.
9Department of Earth and Atmospheric Sciences, University of Alberta, Edmonton, Alberta T6G 2E3, Canada.10Department of Geography, 3611 Ellison Hall, University of
California at Santa Barbara, Santa Barbara, California 93106, USA.11NatureServe, Monteverde, Puntarenas 5655-75, Costa Rica. †Present address: Department of Earth Sciences,
University of Bristol, Wills Memorial Building, Queen’s Road, Bristol BS8 1RJ, UK.
Vol 439|12 January 2006|doi:10.1038/nature04246
© 2006 Nature Publishing Group
exception10. This chytrid grows on amphibian skin and produces
aquatic zoospores22,24. Widespread and ranging from deserts and
lowland rainforests to cold mountain tops27, it is sometimes a non-
mortality in highlands or during winter22, and, according to theory,
becomesmore pathogenic at lower temperatures19,22. Hence, the idea
that it causes declines in warm years is paradoxical. Moreover, the
fungus is apparently more lethal under moist conditions24,26, yet, at
many affected sites, warm years are comparatively dry.
Ideas of two sorts could resolve this paradox. First, warm or dry
conditions may stress amphibians, possibly increasing susceptibility
to disease29. Second, warm years could favour Batrachochytrium
directly. The prevailing idea—that lower temperatures benefit the
chytrid19,22—might be an oversimplification of the pathogen’s
response to climate.
Altitudinal patterns of extinction risk
This prevailing idea predicts greater extinction risk for higher-
elevation species. Many are already prone to extinction, because
geographic ranges tend to decrease in size with increasing elevation.
The probability of disappearance might thus be expected to increase
from lowlands to mountain tops.
For a preliminary test with Atelopus, we consider 100 species for
which data indicate the last year of observation (LYO). We recognize
two tiers. According to La Marca et al. (ref. 26), the population data
are sufficient to judge whether tier-one species (n ¼ 51) have
declined, but not tier-two species (n ¼ 49). Throughout our ana-
lyses, patterns are similar for tier one and for tiers one and two
combined. Adding tier two increases error but provides insights. We
score species as having disappeared if the LYO is 1998 or earlier.
The altitudinal patterns are more complex than expected (Fig. 1).
Using a sliding window to assess how the probability of disappear-
ance varies with species’ lower elevational limit, we find three
breakpoints. The percentage of species lost increases sharply at
200m, and again at 1,000m. It decreases, however, at 2,400m, and
thus peaks at middle elevations, suggesting that low temperatures as
well as high ones may limit the impact of Batrachochytrium. The
altitudinal effects remain significant when we control for range size,
which also influences extinction probability. Average range size
decreases from lower to higher zones as defined in Fig. 1, but is
similar for the upper two.
These altitudinal patterns contribute to the severity of losses. For
instance, the zone losing the highest percentage of species had the
have disappeared is weighted by the number of species per zone.
Although extinction probabilities are independent of tier, an
the severely affected mid-elevation species.
GAA data for New World amphibians13suggest similar altitudinal
patterns (Supplementary Fig. 2). The percentage of species extinct or
threatened is largest at middle elevations, even though higher-
elevation species generally have smaller ranges. Clearly, the role of
climate needs re-evaluating.
To select temperature signals, we consider the scale at which local
climates are determined. In Costa Rica’s Monteverde cloud forest,
reduced mist frequency in warm years is associated with shifts
in populations of birds, reptiles and amphibians, including the
disappearance of the golden toad and the Monteverde harlequin
frog31. Whereas nearby lowland deforestation might have influenced
conditions36, temperatures in Central and South America agree
with simulated responses to greenhouse-gas accumulation3. Here
we quantify the extent and timing of deforestation upwind of
Monteverde, model regional climate, and consider how local trends
relate to sea surface temperature (SST) and air temperature (AT) on
We focus onwarming and the growing number of dry days, which
reflects increasing precipitation variability and declining mist fre-
quency31. While the latter probably affects many organisms, impacts
of correlated climatic changes are hard to separate, and not all
Atelopus extinctions have occurred in habitats where mist is vital26.
of climate change, are a likely common denominator.
The chiefly historical deforestation probably enhanced sensitivity
to warming but cannot easily explain the trends. Using LANDSAT
images and aerial photos, we assess changes in a 35-km-wide belt
representing the trade-wind path from the Caribbean shore to the
500-m contour. Clearing through the year 2000 claimed about 38%
thechanges occurredatMonteverde,and9%during 1960–1975.The
area of the San Carlos Plain directly upwind was cleared before 1940
In contrast, global temperatures have climbed steeply since the
early 1970s (refs 1–4). In the tropics, all forest regions have
warmed37,38, and mountain glaciers are rapidly melting39. During
308N–308S, were highly correlated (Fig. 2a). The latter increased by
0.188C per decade, which is triple the average rate of warming for
the twentieth century. It is 18 times the inferred average rate for a
mid-elevation cloud forest in the Andes during the 8,000-year
transition from the ice ages to modern times (Pleistocene–
Holocene)40. It is similarly more rapid than the non-directional
changes of the preceding 30,000 years.
The recent warming, our work suggests, has reduced mist fre-
quency at Monteverde by raising heights of orographic cloud
formation. These altitudes depend on relative humidity in the
trade winds ascending the mountain slopes, and thus on moisture
content and temperature41. In our simulations, large-scale warming
reduces relative humidity locally much more than the observed
deforestation (Supplementary Fig. 3). The growing number of dry
days is consistent. It is correlated with SST in each of six regions:
offshore Caribbean (near Costa Rica), offshore Pacific, equatorial
Figure1|Altitudinal patternsintheAtelopusextinctions. Barsindicatethe
number of species known per altitudinal zone (total n ¼ 96), and the grey-
shaded portions represent the estimated percentage of species lost from
each. This percentage differs among zones (x2¼ 31.4, degrees of
freedom ¼ 3, P , 0.0001), but not between the tier-one species-set and the
species-setthatcombinestiersoneandtwo(Fisherexacttests,P . 0.9).The
double lines indicate the values for each of these two species-sets; the
percentage labels are the averages of the two. The percentage for the zone
affected most severely differs from that of each adjacent zone (Fisher exact
tests, P , 0.036).
NATURE|Vol 439|12 January 2006
© 2006 Nature Publishing Group
Pacific (Nin ˜o-3 region), deep tropics (108N–108S), tropics, and the
globe. A residual trend remains, however, unless we consider the
tropics or the globe (Supplementary Fig. 4). Large-scale and local
climatic changes are strikingly concordant, with fluctuations related
to El Nin ˜o superimposed on the trends (Fig. 2b, c). Thus, analysis
of AT or SST for the tropics should capture temperature shifts
influencing the relevant ecological processes.
Signatures of warming
Accordingly, the biological changes at Monteverde are associated
statisticallywithATandSSTfor thetropicsbutnotwithNin ˜o-region
SST alone (Supplementary Fig. 5). Correlations are evident for the
shift of lower-elevation breeding birds up the mountain slopes, and
for the decline of highland lizards. Likewise, the episodic losses of
amphibians occurred in years that were unusually warm across the
the tropics, we randomize this signal for 1979–1998. In each of
10,000 iterations, we assign the annual means at random to the 20
years and recalculate indices of association. The results confirm
that none of the observed relationships is likely to have arisen
by chance (Supplementary Table 1). Moreover, in only one iteration
are the modelled values all as extreme as the observed ones,
indicating a high probability that large-scale warming is affecting
We examine the timing of the widespread Atelopus extinctions in
ignescens) of Ecuador and the Monteverde harlequin frog suggest the
working hypothesis that species tend to be seen for the last time right
after a relatively warm year21,29,31. Both were last found in 1988,
following a temperature peak in 1987. Before 1988, the Jambato toad
was present during 64% of visits to sites throughout its 10,234-km2
range21. After 1988, it was absent at all sites, implying synchronous
declines across localities. The degree of synchrony, however, differs
among species, and survivors of population crashes persist for
variable lengths of time. Furthermore, the spatial and temporal
coverage of sampling varies, introducing error.
At any rate, the climate-linked epidemic hypothesis predicts an
association between disappearances and warm years, but not a one-
to-one correspondence33. An Atelopus population might survive
despite warm weather if the pathogens are absent from particular
sites within their range, or if they are spreading but have not reached
certain areas. Factors that discourage pathogen transmission, such as
low host-density, may likewise forestall declines10. Although tem-
perature shifts can entrain multiple outbreaks, fast-moving waves of
infection might also synchronize declines across localities, and an
Atelopus population experiencing normal weather could succumb to
a wave set in motion elsewhere.
warming. Like the biological changes at Monteverde, they are
associated statistically with ATand SST for the tropics but not with
Nin ˜o-region SST alone (Fig. 3 & Supplementary Table 2). Around
80% of the species that have disappeared were seen for the last time
right after a relatively warm year. For tier one, and for tiers one and
two combined, we use Monte Carlo methods to generate 10,000
random frequency distributions for comparison with the observed
distributions. These analyses confer ‘very high confidence’ (.99%,
following the IPCC1,5) that the tendencies are not due to chance and
that large-scale warming is a key factor. Patterns for the two species-
sets are comparable despite differences in error rate and time period.
Spanning a longer period, the combined data indicate depletion of
the most vulnerable species by the late 1990s.
Results are consistent when we repeat the analyses with various
subsets of species to consider sources of variation, error and uncer-
tainty (Supplementary Table 2). For instance, because some unde-
scribed species are poorly known, we repeat the analyses including
only described ones. Likewise, we consider occurrence in protected
Figure 2 | AT and SST for the tropics and their relationship to climatic
trends at Monteverde. AT for the tropics (blue line) is correlated with:
a,SSTforthetropics(redline)(r ¼ 0.97,P , 0.0001,n ¼ 51);b,numberof
drydaysinruns$5days(grey-shadedarea)(r ¼ 0.70,P , 0.0001,n ¼ 28);
and c, local daily minimum AT (green line) (r ¼ 0.91, P , 0.0001, n ¼ 24).
from a baseline mean (for 1856–1895 and 1951–1979, respectively).
NATURE|Vol 439|12 January 2006
© 2006 Nature Publishing Group
areas, accessibility of regions, and factors that might influence the
timing of extinction. Several analyses provide independent tests of
our hypothesis. The strength of association between warm years
and disappearances is not related to altitude, latitude or range
size. Accordingly, conclusions are similar for ‘northern’ and
‘southern’ species, and for ‘higher-elevation’ and ‘lower-elevation’
In these analyses, AT or SST for the tropics serves as a relative
index, often registering smaller shifts than local indices. In 1987, the
former averaged 0.658C above the baseline (Fig. 2), whereas local
temperatures relevant to the Jambato toad’s extinction in the high-
lands of Ecuador were almost 2.08C above a century-long mean21.
The difference may reflect, in part, increasing atmospheric moisture,
which can amplify the signal at higher altitudes42,43. Global warming
accelerates evaporation and raises the air’s capacity to hold water. As
water vapour rises and condenses, latent heat is transferred to the
A climate for chytrids
Increased water vapour can also translate into enhanced cloud
cover—often with the help of condensation nuclei from particulate
air pollution (aerosols)44—creating additional feedbacks that influ-
ence surface temperatures45. This may be notable in places where air
rises strongly, such as in mountainous regions. Reducing heat loss at
night, cloud cover adds to nocturnal warming. By impeding solar
radiation, however, it moderates daytime trends and may reverse
them. In many areas, the daily temperature range is declining as the
minimum rises faster than the maximum45.
Such trends are evident in the highlands of Central and South
America46–48. At Monteverde, regardless of the season, the daily
minimum is rising while the daily maximum is falling (Fig. 4a).
For 11 Colombian and Venezuelan stations with quality-checked,
long-term data47, we compare 1941–1970, preceding the Atelopus
extinctions, to the first decade with major losses (1981–1990). The
minimum again shows an increase and the maximum a decrease
(Fig. 4b). At some localities both are rising, but the former dis-
proportionately so46–48. These trends imply increasing cloud cover
that contributes to warming at night but diminishes it during the
Cloudiness should favour the chytrids. These fungi reportedly
grow best at 17 to 258C, peaking at 238C (refs 19, 22, 24). They stop
growing at 288C and die at 308C. Shielding them from excessive
warmth and fostering moist conditions, cloud cover may promote
their survival, growth and reproduction. At Monteverde, where
tures inside sunlit moss mats, bromeliads or leaf litter often exceed
308C (ref. 34). Cloudiness, however, largely shuts down radiant
heating, forcing thermal environments to mirror ambient con-
ditions. Microscale trends can thus dwarf local ambient trends,
which, in some places, might even be of opposite sign. If amphibians
seek warmth to combat infection, increasing cloudiness might
hamper their defences34. In any case, local or microscale cooling
should often benefit the chytrids.
So why should these pathogens flourish in the highlands during
warm years? The answer, we suggest, lies in the difference between
night and day. To consider this difference, we plot daily minimum
and maximum temperatures in relation to altitude for 50 localities,
from Costa Rica to Peru, along with the optimal temperatures for
Figure 3 | Signatures of warming in the Atelopus extinctions. For tier one
(a) and tiers one and two combined (b), the number of species observed for
the last time (black line)is related to AT for the tropics inthe preceding year
(blue line). c, d, Percentage of species observed for the last time following a
relatively warm year exceeds that expected by chance (c, tier one, 83%,
P , 0.002, n ¼ 29; d, tiers one and two, 78%, P , 0.0001, n ¼ 68). e, f, The
same patterns (black lines as in a, b) are not significantly related to
Nin ˜o-region SST (red line). The corresponding percentages do not exceed
those expected by chance (e, tier one, 55%, P . 0.43; f, tiers one and two,
62%, P . 0.12). Temperatures are calculated as in Fig. 2.
NATURE|Vol 439|12 January 2006
© 2006 Nature Publishing Group
growth of Batrachochytrium (Fig. 4c). We also plot the altitudinal
distribution of Atelopus (Fig. 4d). Twopatterns are clear. First, just as
the lowlands are often too warm for the chytrids during the day,
the highlands are often too cool for them at night. Second, most
Atelopus extinctions have occurred atelevations wherethe minimum
temperature is shifting towards the growth optimum for these
pathogens. Thus, we propose the chytrid-thermal-optimum hypoth-
esis, in which daytime cooling (local or microscale) and night time
warming accelerate disease development. The impacts at night may
explain the association with warm years and thereby resolve the
We establish that global climate change is already causing the
extinction of species. Taking our results and recent findings that tie
are an immediate threat to biodiversity. Our study sheds light on the
amphibian-decline mystery by showing that large-scale warming is a
key factor. It also points to a chain of events whereby this warming
may accelerate disease development by translating into local or
microscale temperature shifts—increases and decreases—favourable
to Batrachochytrium. The case illustrates how greenhouse warming
and the resultant intensification of the hydrological cycle, together
with aerosol pollution, may affect life on Earth. Influencing patterns
environments of many organisms, changing ecological interactions
and threatening species survival.
Assessing extinction probability in relation to altitude. We examine the
influence of altitudinal distribution while considering range size. Data indicate
the last year of observation (LYO) for 104 of the 113 species of Atelopus (see
Appendix A in the Supplementary Information). Focusing on recent losses, we
exclude four species known only from historical records (1950 or earlier).
Variables include minimum and maximum elevations26, elevational midpoint
and longest axis of the range polygon (GAA data13). For statistical analyses, we
use logistic regression and contingency tables. Sample sizes vary because some
data are missing, particularly for undescribed species. Although extinction
probability is related to each altitudinal variable, we focus on minimum
elevation, because it indicates whether populations occur at low elevations,
which may provide a refuge from chytridiomycosis. Ranking species by mini-
mum elevation, from lowest to highest, we use a sliding window to compare
extinction probabilities between successive subsets of ten, shifting the window
species has disappeared.
Selecting temperature signals. The regional modelling generates climate-
change scenarios, which we test by analysing local trends in relation to
temperature on varying spatial scales. In the simulations, we prescribe realistic
of an increase in ATand SSTapproximating tropical warming over 1973–2000
and, in a separate run, the 1986–1987 El Nin ˜o. We examine aerial photos from
Costa Rica’s National Geographic Institute for 1960, LANDSAT Multispectral
2000. Detecting forest fragments $0.03km2, and classifying areas as forest if
canopy density is $80%, the techniques49produce land-cover maps at a scale of
1:250,000. We run our simulations with the Regional Atmospheric Modelling
System (RAMS)50at a maximum horizontal resolution of 1.6km. To prescribe
initial and boundary conditions, we use the ECMWF Global Reanalysis
(2.58-resolution ERA-15 Pressure Level Analysis) from the European Centre
for Medium-range Weather Forecasts. The GDTOPO30 digital elevation model
from the US Geological Survey defines terrain. To examine local trends in
relation to temperature on varying scales, we examine SST data (28-resolution)
from the National Oceanic and Atmospheric Administration’s (NOAA’s)
National Centers for Environmental Prediction (NCEP), and AT data
(58-resolution) from the Climate Research Unit (CRU), University of East
Figure 4 | Daily minimum and maximum temperatures and the chytrid-
thermal-optimum hypothesis. a, At Monteverde, the average daily
minimum (green lines) and maximum (black lines) for warmer months
(March–October) and cooler months (November–February) show trends
(jrj $ 0.43, P , 0.038, n ¼ 24). b, In Colombia and Venezuela, station
are mean changes (^s.e.m.). c, For Costa Rica to Peru, annual average daily
minimum (closed circles) and maximum (open circles) vary by altitude.
Dashed lines givereported optimal temperaturerange (red dashedline) and
of Atelopus species per 500-m belt is labelled by upper limit. Six ranged
NATURE|Vol 439|12 January 2006
© 2006 Nature Publishing Group
Anglia. The local data, including tallies of dry days for January–May, are from
1,540m on Monteverde’s upper Pacific slope31.
Testing for a link to global warming. We use resampling methods to analyse
biological patterns in relation to large-scale temperature signals. ‘Warm years’
are above average for the period of analysis. For Monteverde, we examine 1979–
1998, which encompasses the field observations31. To produce a single time
series for anoline lizards, we average data for the two declining species, which
are highly correlated. The Atelopus data are from various independent
studies26. To prevent bias, persons contributing or compiling these data were
not told how they would be analysed in relation to climate. We consider AT for
the tropics over the last year of observation (LYO) of each species, for one, two
and three years before, and for averages across these years. The period of
analysis is defined accordingly. We find an association, however, only for one
year before. To examine the strength of this association in relation to
altitudinal and latitudinal distribution and range size, we use logistic
regression. Latitudinal variables include northern and southern limits and
range midpoint. The analyses of biological patterns in relation to Nin ˜o-region
SST (departures relative to 1950–1979) yield similar conclusions regardless of
the signal examined: Nin ˜o-1 and -2, Nin ˜o-3, these combined, or Nin ˜o-1 and -2
combined with Nin ˜o-3 and -4. We present results for the latter, composite,
Resolving the climate–chytrid paradox. We focus on temperature, since
altitudinal patterns in the declines underscore its importance and because the
extinctions are strongly associated with it. Temperature shifts are presumably
more coherent spatially than attendant climatic changes. Our premise is that
any pathogen with an optimal range for growth and reproduction will be
sensitive to low temperatures as well as high ones, as suggested by the altitudinal
patterns of extinction risk. (see the Supplementary Notes pertaining to “A
climate for chytrids”, which explores the meaning of the observed breakpoints.)
Comparing temperatures for the warmer and cooler months at Monteverde
shows that, seasonally, night time temperatures in the highland tropics often lie
even farther below the optimal range for Batrachochytrium than average
conditions suggest. To examine daily minimum and maximum temperatures
for the 11 Colombian and Venezuelan stations, we use the published averages
for particular decades47. The 50 localities from Costa Rica to Peru mostly
represent inland areas. The corresponding analyses do not control for latitude
or sampling period, yet provide a generalized altitudinal profile of minimum
and maximum temperatures. Numbers of Atelopus species that inhabited the
different altitudinal zones are based on n ¼ 96. Some inhabited more than one
Received 2 June; accepted 21 September 2005.
1. Houghton, J. T. et al. (eds) Climate Change 2001, The Scientific Basis. Third
Assessment Report of the Intergovernmental Panel on Climate Change (Cambridge
Univ. Press, Cambridge, 2001).
Santer, B. D. et al. Contributions of anthropogenic and natural forcing to recent
tropopause height changes. Science 301, 479– -483 (2003).
Stott, P. A. Attribution of regional-scale temperature changes to anthropogenic
and natural causes. Geophys. Res. Lett. 30, 1728– -1731 (2003).
Barnett, T. P. et al. Penetration of human-induced warming into the world’s
oceans. Science 309, 284– -287 (2005).
Parmesan, C. & Yohe, G. A globally coherent fingerprint of climate change
impacts across natural systems. Nature 421, 37– -42 (2003).
Root, T. L. et al. Fingerprints of global warming on wild animals and plants.
Nature 421, 57– -60 (2003).
Root, T. L., MacMynowski, D. P., Mastrandrea, M. D. & Schneider, S. H.
Human-modified temperatures induce species changes: Joint attribution. Proc.
Natl Acad. Sci. USA 102, 7465– -7469 (2005).
Lovejoy, T. & Hannah, L. (eds) Climate Change and Biodiversity (Yale Univ.
Press, New Haven, Connecticut, 2005).
Epstein, P. R. Climate change and emerging infectious diseases. Microbes Infect.
3, 747– -754 (2001).
10. Harvell, C. D. et al. Climate warming and disease risks for terrestrial and
marine biota. Science 296, 2158– -2162 (2002).
11. Rodo ´, X., Pascual, M., Fuchs, G. & Faruque, A. S. G. ENSO and cholera: A
nonstationary link related to climate change? Proc. Natl Acad. Sci. USA 99,
12901– -12906 (2002).
12. Thomas, C. D. et al. Extinction risk from climate change. Nature 427, 145– -148
13. Stuart, S. M. et al. Status and trends of amphibian declines and extinctions
worldwide. Science 306, 1783– -1786 (2004).
14. Alford, R. A. & Richards, S. J. Global amphibian declines: a problem in applied
ecology. Annu. Rev. Ecol. Syst. 30, 133– -165 (1999).
15. Kiesecker, J. M., Blaustein, A. R. & Belden, L. K. Complex causes of amphibian
population declines. Nature 410, 681– -684 (2001).
16. Blaustein, A. R. & Kiesecker, J. M. Complexity in conservation: lessons from the
global decline of amphibian populations. Ecol. Lett. 5, 597– -608 (2002).
17. Berger, L. et al. Chytridiomycosis causes amphibian mortality associated with
population declines in the rain forests of Australia and Central America. Proc.
Natl Acad. Sci. USA 95, 9031– -9036 (1998).
18. Carey, C. & Alexander, M. A. Climate change and amphibian declines: Is there
a link? Divers. Distrib. 9, 111– -121 (2003).
19. Daszak, P., Cunningham, A. A. & Hyatt, A. D. Infectious disease and amphibian
population declines. Divers. Distrib. 9, 141– -150 (2003).
20. Lips, K. R., Green, D. E. & Pappendick, R. Chytridiomycosis in wild frogs from
southern Costa Rica. J. Herpetol. 37, 215– -218 (2003).
21. Ron, S. R., Duellman, W. E., Coloma, L. A. & Bustamante, M. Population decline
of the Jambato toad Atelopus ignescens (Anura: Bufonidae) in the Andes of
Ecuador. J. Herpetol. 37, 116– -126 (2003).
22. Berger, L. et al. Effect of season and temperature on mortality in amphibians
due to chytridiomycosis. Aust. Vet. J. 82, 434– -439 (2004).
23. Burrowes, P. A., Joglar, R. L. & Green, D. E. Potential causes for amphibian
declines in Puerto Rico. Herpetologica 60, 141– -154 (2004).
24. Piotrowski, J. S., Annis, S. L. & Longcore, J. E. Physiology of Batrachochytrium
dendrobatidis, a chytrid pathogen of amphibians. Mycologia 96, 9– -15 (2004).
25. Retallick, R. W. R., McCallum, H. & Speare, R. Endemic infection of the
amphibian chytrid fungus in a frog community post-decline. PLoS Biol. 2,
1966– -1971 (2004).
26. La Marca, E. et al. Catastrophic population declines and extinctions in
Neotropical harlequin frogs (Bufonidae: Atelopus). Biotropica 37, 190– -201
27. Ron, S. R. Predicting the distribution of the amphibian pathogen
Batrachochytrium dendrobatidis in the New World. Biotropica 37, 209– -221
28. Merino-Viteri, A., Coloma, L. A. & Almenda ´riz, A. in Studies on the Andean Frogs
of the Genera Telmatobius and Bratrachophrynus (eds Lavilla, E. O. & De la
Riva, I.) 9– -37 (Asociacio ´n Herpetolo ´gica Espan ˜ola, Valencia, 2005).
29. Pounds, J. A. & Crump, M. L. Amphibian declines and climate disturbance:
the case of the golden toad and the harlequin frog. Conserv. Biol. 8, 72– -85
30. Pounds, J. A., Fogden, M. P. L., Savage, J. M. & Gorman, G. C. Tests of null
models for amphibian declines on a tropical mountain. Conserv. Biol. 11,
1307– -1322 (1997).
31. Pounds, J. A., Fogden, M. P. L. & Campbell, J. H. Biological response to climate
change on a tropical mountain. Nature 398, 611– -615 (1999).
32. Pounds, J. A. in Monteverde: Ecology and Conservation of a Tropical Cloud Forest
(eds Nadkarni, N. M. & Wheelwright, N. T.) 149– -177 (Oxford Univ. Press,
33. Pounds, J. A. Climate and amphibian declines. Nature 410, 639– -640 (2001).
34. Pounds, J. A. & Puschendorf, R. Ecology: Clouded futures. Nature 427, 107– -109
35. Root, T. L. & Schneider, S. H. Ecology and climate: Research strategies and
implications. Science 269, 334– -341 (1995).
36. Lawton, R. O., Nair, U. S., Pielke, R. A. Sr & Welch, R. M. Climatic impact of
tropical lowland deforestation on nearby montane cloud forests. Science 294,
584– -587 (2001).
37. Vuille, M., Bradley, R. S., Werner, M. & Keimig, F. 20thcentury climate change
in the tropical Andes: observations and model results. Clim. Change 59, 75– -99
38. Malhi, Y. & Wright, J. Spatial patterns and recent trends in the climate of
tropical forest regions. Proc. R. Soc. Lond. B 359, 311– -329 (2004).
39. Thompson, L. G. et al. Tropical glacier and ice core evidence of climate change
on annual to millennial time scales. Clim. Change 59, 137– -155 (2003).
40. Bush, M. B., Silman, M. R. & Urrego, D. H. 48,000 years of climate and forest
change in a biodiversity hot spot. Science 303, 827– -829 (2004).
41. Still, C. J., Foster, P. N. & Schneider, S. H. Simulating the effects of climate
change on tropical montane cloud forests. Nature 398, 608– -610 (1999).
42. Wentz, F. J. & Schabel, M. Precise climate monitoring using complementary
satellite data sets. Nature 403, 414– -416 (2000).
43. Ross, R. J. & Elliott, W. P. Radiosonde-based Northern Hemisphere
tropospheric water vapor trends. J. Clim. 14, 1602– -1612 (2001).
44. Penner, J. E. Climate change: The cloud conundrum. Nature 432, 962– -963
45. Dai, A., Trenberth, K. E. & Karl, T. R. Effects of clouds, soil moisture,
precipitation, and water vapor on diurnal temperature range. J. Clim. 12,
2451– -2473 (1999).
46. Go ´mez, I. E. & Ferna ´ndez, W. Variacio ´n interanual de la temperatura en Costa
Rica. Top. Meteor. Oceanogr. 3, 27– -44 (1996).
47. Quintana-Go ´mez, R. A. Trends of maximum and minimum temperatures in
northern South America. J. Clim. 12, 2104– -2112 (1999).
48. Ca ´ceres, L., Mejia, R. & Otaneda, G. in Consecuencias Clima ´ticas e Hidrolo ´gicas
del Evento El Nin ˜o a Escala Regional y Local. Incidencia en America del Sur
(eds Cadier, E., Go ´mez, G., Galarraga, R. & Ferna ´ndez-Ja ´uregui, C.) khtpp://
www.unesco.org.uy/phi/libros/enso/caceres.htmll (UNESCO-PHI, 1998).
49. Sanchez-Azofeifa, G. A., Harriss, R. C. & Skole, D. L. Deforestation in Costa
Rica: a quantitative analysis using remote sensing imagery. Biotropica 33,
378– -384 (2001).
50. Pielke, R. A. et al. A comprehensive meteorological modeling system—RAMS.
Meteorol. Atmos. Phys 49, 69– -91 (1992).
NATURE|Vol 439|12 January 2006
© 2006 Nature Publishing Group Download full-text
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank T. P. Mitchell at JISAO and S. J. Worley at NCAR
for help in obtaining climate data, R. A. Alford, F. Bolan ˜os, J. P. Collins,
R. O. Lawton, K. R. Lips, M. D. Mastrandrea, K. G. Murray, P. Ramı ´rez and
B. D. Santer for discussion, and the many contributors to the Atelopus database.
S. H. Schneider, A. R. Blaustein and C. Parmesan commented on earlier drafts of
the manuscript. The Declining Amphibian Populations Task Force and
Conservation International’s Critically Endangered Neotropical Species Fund
provided partial funding to J.A.P. The Canada Foundation for Innovation and the
Tinker Foundation helped produce the remote-sensing databases. The Research
and Analysis Network for Neotropical Amphibians and the US National Science
Foundation sponsored meetings that facilitated portions of the study.
Author Contributions All authors after the first are listed alphabetically. J.A.P.
conceived, designed and orchestrated the study, conducted most of the
analyses (principally with J.A.C. and K.L.M.), formulated the chytrid-thermal-
optimum hypothesis (with R.P.), and wrote the paper (with editing by J.A.C. and
K.L.M.). M.R.B., L.A.C., M.P.L.F., E.L.M., A.M.-V. and S.R.R. provided key data and
background information. E.L.M. compiled the Atelopus database (with B.E.Y.).
P.N.F. conducted the climate simulations. G.A.S.-A. analysed the remote-sensing
data. C.J.S. helped with the climate analyses and their interpretation. B.E.Y.
obtained funding for meetings, provided logistics, and analysed GAA data for
New World amphibians.
Author Information Reprints and permissions information is available at
npg.nature.com/reprintsandpermissions. The authors declare no competing
financial interests. Correspondence and requests for materials should be
addressed to J.A.P. (firstname.lastname@example.org).
NATURE|Vol 439|12 January 2006