Global Change Biology (2010), doi: 10.1111/j.1365-2486.2010.02203.

Nonlinear climate change and Andean feedbacks: an

imminent turning point?
M . B . B U S H , J . A . H A N S E L M A N 1 and W . D . G O S L I N G 2
Department of Biological Sciences, Florida Institute of Technology, 150 W. University Blvd., Melbourne, FL 32901, USA

Abstract
A 370 000-year paleoecological record from Lake Titicaca provides a detailed record of past climate change in which
interglacial periods are seen to have some elements of commonality, but also some key differences. We advance a
conceptual feedback model to account for the observed changes that includes previously ignored lake effects. Today
Lake Titicaca serves to warm the local environment by about 4–5 1C and also to increase rainfall. We observe that as
water levels in the lake are drawn down due to warm, dry, interglacial conditions, there is a possible regional cooling
as the lake effect on local microclimates diminishes. Positive feedback mechanisms promote drying until much of the
lake basin is reduced to salt marsh. Consequently, the usual concept of upslope migration of species with warming
would not be applicable in the Altiplano. If, as projected, the next century brings warmer and drier conditions than
those of today, a tipping point appears to exist within ca. 1–2 1C of current temperatures, where the relatively benign
agricultural conditions of the northern Altiplano would be replaced by inhospitable arid climates. Such a change
would have profound implications for the citizens of the Bolivian capital, La Paz.
Keywords: aridity, charcoal, conservation, fossil pollen, grayscale, Lake Titicaca, positive feedback, warming

Received 21 October 2009; revised version received 16 December 2009 and accepted 17 December 2009

those of MIS 9 and 5e (Van der Hammen & Hooghiem-

Introduction
For the last 5000 years, inhabitants of the Bolivian
Altiplano have taken advantage of the relatively mild,
moist climate adjacent to Lake Titicaca (Erickson, 1988).
Lying at 3810 m elevation, Lake Titicaca is the world’s
highest great lake and produces a local microclimate or
lake effect that allows cultivation of crops, such as
maize, at unusually high elevations. However, paleoe-
cological data show that the size of this and other vast
paleolakes of the Altiplano have been volatile, exhibit-
ing rapid changes in lake level with presumed conco-
mitant changes in microclimate. The 370 thousand year
(ka)-long dataset from Lake Titicaca (Peru/Bolivia)
(Hanselman et al., 2005, in press), reveals that environ-
mental changes associated with the warming of marine
isotope stage (MIS) 9 and 5e are more extreme than of
the Holocene. The trajectory of Andean warming (ca.
1 1C in the last 30 years) Andes suggests a minimum
warming of 1–2 1C by mid century (Christensen et al.,
2007) producing temperatures probably equivalent to
1 Denton, 1989; Stocker, 2000; Bond et al., 2001), and are
Present address: J. A. Hanselman, Department of Biology, West-
field State College, Westfield, MA, USA.
2
Present address: W. D. Gosling, Department of Earth and Envir-
onmental Sciences, CEPSAR, The Open University, Milton Keynes,
UK.
Correspondence: M. B. Bush, tel. 1 1 321 674 7166, e-mail:
mbush@fit.edu
stra, 2003) by ca. 2050.
A common expectation has developed among biolo-
gists that, as the world warms, species will migrate
upslope and poleward (Parmesan & Yohe, 2003). Such
changes are already documented on tropical mountains
particularly where ecotones are strong and range shifts
readily detectable, such as the lower limit of cloud
forest (Pounds et al., 1999; Pounds, 2001). Assumptions
that species will continue to migrate upslope have
raised concerns that as the bioclimatic envelopes defin-
ing the potential range of a species lift off the top of
mountains, these upper-elevation species will go extinct
(Peters & Darling, 1985; Thomas et al., 2004; Malcolm
et al., 2006). Although such extinction may be the case in
some settings, an alternative scenario is one in which,
rather than marching steadily upslope in response to
linear climate change, the system flips to a new state.
Nonlinear changes in climate have been suggested to
underlie millennial-scale climate pulses such as Dans-
gaard-Oeschger cycles and Heinrich events (Broecker &
clearly capable of causing rapid and profound switches
of climatic direction. Such changes would require an
entirely different societal response to conserve liveli-
hoods, agricultural production, and biota, compared
with that of incremental linear changes.
Thus, estimates of future climate change need to take
into consideration phenomena operating from global to

r 2010 Blackwell Publishing Ltd 1

2 M . B . B U S H et al.
local scales, as their interaction, through feedback me-
chanisms, could result in abrupt tipping points that lead
to nonlinear responses in temperature or precipitation.
The paleoecological record becomes our best resource
for studying such complex responses, and attention
needs to focus on times when climate changed rapidly,
and temperatures were warmer-than-modern.
Here we describe, climatic events and feedbacks
associated with past interglacials that were probably
about 1–3 1C warmer-than-present in the Andes. Twice,
a tipping point was reached that led to a profound
alteration of regional ecosystems.
The present and past climates of Lake Titicaca
Lake Titicaca (Bolivia/Peru, 15.5–171S, 68.5–701W) lies at
3810 m above sea level in the Altiplano (Fig. 1) and has a
modern climate that is cool and dry. Cloudiness is much
greater during the wet season than the dry season, with
just 6 h of bright sunlight in January compared with
9.6 h in July (http://waterwiki.net/index.php/Lake_Ti-
ticaca-Poopo). A lake effect is evident in local tempera-
tures with averages at Titicaca about 5 1C higher than
areas remote from the lake. Roche et al. (1992) mapped
temperatures around Titicaca and concluded that the
lake warming effect extended for tens of kilometers
around the lake. Similarly, the lake influences precipita-
tion patterns. Average annual rainfall within the Alti-
plano ranges from ca. 200 mm yr 1 in the south to
400 mm yr 1 in the north, with ca. 890 mm yr 1 over
the main basin of Titicaca. About 80% of the precipita-
tion falls during the period of the South American
summer monsoon (SASM) (December–March) when
the winds are dominated by an easterly flow bringing
Amazonian moisture to the Altiplano (Roche et al., 1992).
Over Lake Titicaca itself, precipitation is supplemented
by night-time convective rains in the wet season.
Climatic oscillations between glacial and interglacial
landscapes were evident in the 370 ka paleoecological
record from Lake Titicaca (Hanselman et al., 2005, in
press). During glacials, cold temperatures resulted in
very little pollen production as the shoreline of Lake
Titicaca became a glacial foreland. The ice masses lay
within 100 m vertically of the lake, but never reached
the shore (Seltzer, 1990; Seltzer et al., 2002). During the
glacial periods, combinations of meltwater and changes
in the precipitation : evaporation (P : E) ratio of open
water bodies resulted in both higher and lower lake
levels than those of today. During highstand events, the
increased overflow of Lake Titicaca into the Rio Desa-
guadero contributed to the flooding of the southern
Altiplano (Ybert, 1992; Clapperton, 1993; Placzek et al.,
2006). Considerable debate surrounds the timing of
highstands that resulted in a series (temporally) of
r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x
paleolakes with shorelines as much as 120 m above
the modern level of the Salar de Uyuni (e.g. Hastenrath
& Kutzbach, 1985; Baker et al., 2001a; Placzek et al., 2006;
Gosling et al., 2008). At the dry extreme, during each of
the four interglacials heightened concentrations of car-
bonate were deposited in the sediments of Lake Titicaca
relative to modern values. Carbonate was deposited as
the level of Titicaca fell below its outlet and the P : E
ratio declined causing salts in the lake to become more
concentrated (Baker et al., 2005). In the Holocene, the
peak calcium carbonate (CaCO3) concentration in sedi-
ments was about 54%, but during prior interglacials it
was as high as 100%, perhaps suggesting longer or more
intense periods of evaporative dominance. These long
oscillations appeared to have followed the pattern of
orbital precession, a finding that was consistent be-
tween Andean and Amazonian lakes and regional
speleothem records (Hooghiemstra et al., 1993; Baker
et al., 2001a; Bush et al., 2002; Fritz et al., 2004; Wang
et al., 2004; Cruz et al., 2005).
Although long-term precessional variability underlay
the lowstand of the mid-Holocene (Baker et al., 2001a),
shorter-term events, ranging from annual to millennial,
were superimposed on this basic pattern (Ekdahl et al.,
2008; Hillyer et al., 2009). Understanding the mid-Holo-
cene dry event provides insights into changes asso-
ciated with MIS 9 and 5e. The mid-Holocene dry
event has been a consistent feature of paleoecological
records from western Amazonia and the Andes. As
knowledge of the event deepened it became apparent
that it should be described as a period of increased
drought probability, rather than a single drought
(Hillyer et al., 2009). Between 9 and 4.4 ka, the net effect
was a drier climate that lowered lake level, but this was
not a time of uniform aridity. Rather, the climate had
shifted toward a drier tendency, marked in the paleoe-
cological record by sediment chemistry and fossil dia-
toms associated with shallow, ephemeral, or saline
lakes. However, the same proxies indicate brief but
frequent wetter periods in which lake levels rose
(Hillyer et al., 2009).
The mid-Holocene drying lowered lake level at Titi-
caca by ca. 85 m (Seltzer et al., 1998; Baker et al., 2001b).
Once the lake level had dropped below the outflow to
the southern basin of Huinaimarca (a 25 m lowering),
further loss of water from the system was due to
evaporation. A hydrological model of this event sug-
gested that P : E ratios could account for the observed
lowering of lake level if windspeed was held constant,
temperature increased by 1 1C and precipitation fell by
40%. Tweaking of the parameters in the hydrological
simulations demonstrated that lake level was not
strongly responsive to variations in temperature (Cross
et al., 2000).
 ANDEAN TIPPING POINTS 3

Fig. 1 Sketch map of the Altiplano in Peru/Bolivia showing the modern lakes (gray shading) and salars (hashing). Also shown on the
main diagram is the highstand associated with paleolake Tauca (short dashed line, after Placzek et al., 2006), and, on the inset, the
lowstands of the Holocene (dashed and dotted line) and MIS 5e (long dashed line, after Baker et al., 2005). Location of core site LT01-2B
marked with star on inset.

Whatever had caused the compound dry events
comprising the mid-Holocene drought ended ca.
5.2 ka, when lake levels started to rebound (Baker
et al., 2001b; Paduano et al., 2003). Most lakes in the
Andes and Amazon were filled close to modern levels
by 4.4 ka (Seltzer et al., 2000; Abbott et al., 2003). This
wet event was apparently due to increased precipitation
that corresponded both to the increased intensity of El
Niño/southern oscillation (ENSO) events and reduced
summer insolation to the northern tropical Atlantic
Ocean (Moy et al., 2002; Riedinger et al., 2002; Rein,
2007).
Thus, the long-term pattern provided by precessional
change was overlain by regional and local factors that
produced sharply defined changes in lake level. Similar
interactions of factors operating on a range of spatial
and temporal scales would be expected to influence
past and future climate.
Methods
Analysis of the fossil pollen data from core LT01-2B followed
standard protocols and has been described in detail elsewhere
(Hanselman et al., 2005). The chronology of the core was
established via a combination of isotopic dating using 14C
A.M.S and U/Th, and wiggle matching to the Vostok ice core
record (Petit et al., 1999) and is described in Fritz et al. (2007,
2008).
Grayscale analysis, in which the relative darkness (black-
est 5 0, whitest 5 255) of sediment bands was recorded. This

r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x

4 M . B . B U S H et al.
technique was applied to the MIS 5e section of the core based
on high-resolution images collected on a Geotek core-logger at
the LacCore facility, and analyzed using IMAGEJ (Rasband,
2005).
Results
The most striking features of the Lake Titicaca paleo-
ecological record are the differences between glacial
and interglacial conditions (Fig. 2). During glacials very
little pollen was deposited and there was an almost
complete absence of fire from the system (Hanselman
et al., in press). However, during each of the intergla-
cials a progression from Polylepis woodland to Puna
vegetation is apparent and fire becomes a regular
feature of the landscape. Changing lake levels are
indicated by the presence of aquatic taxa such as
Myriophyllum and Isöetes. In general, Isöetes could not
withstand the cold peaks of glacials nor the warm
peaks of the interglacials, and therefore is at peak
abundance at the transition between stages. Similarly,
Polylepis occurs primarily in the transitional period
(Gosling et al., 2009) and declines as charcoal concentra-
tions, i.e. fire events, increase in abundance and
frequency.
Another marked pattern is the difference in the pollen
signatures between interglacials. In MIS 9 and 5e,
following peaks of aquatic taxa and Polylepis, Amar-
anthaceae (cf. Chenopodium) dominate the interglacial
flora. In contrast to this pattern, MIS 7 and 1 the glacial
termination and inception stages are similar to those of
MIS 9 and 5e, but there is no Amaranthaceae-domi-
nated stage in the middle of the interglacial.
The grayscale analysis of sediments ascribed to MIS
5e revealed increasing contrast as the matrix of the
sediment changed from gray to black with sharply
defined white laminae composed of CaCO3 (Fig. 3).
Fig. 2 The fossil record of selected pollen and spores from Lake Titicaca core LT01-2B. Note the peak of Myriophyllum in MIS 9 was
truncated, actual peak was 6000% of dry land pollen (after Hanselman et al. in press).
r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x
Discussion
During the early phases of both MIS 9 and 5e the lack of
charcoal in sediments, the upslope migration of Andean
taxa, and occupation of the Altiplano by fire-sensitive
trees such as Polylepis, indicate warm, moist conditions
(Gosling et al., 2009). A simple model of warming
would predict that the vertical migration of trees would
continue until they fully occupied the lake basin (Fig. 4).
However, long before the peak of the interglacial, a
sudden transition was evident in which trees were
replaced by Amaranthaceae and the 280 m modern
water depth of Lake Titicaca fell so much that large
beds of Myriophyllum formed in shallow water (or on
exposed mudflats). As the lake became more saline this
genus was replaced by Amaranthaceae (Hanselman
et al., in press).
The lake sediment deposited at the inferred peak of
MIS 5e was an almost pure precipitate of CaCO3. The
mid-Holocene warming, which resulted in a 85 m low-
ering of lake level was sufficient to deposit some carbo-
nate, but the change in conditions was insufficient to
generate a comparable change in vegetation (Paduano
et al., 2003). Cross et al. (2000) estimated that a 40%
reduction in precipitation for about 4000 years was suffi-
cient to achieve the Holocene drawdown. Here we add to
those insights with a conceptual feedback model that
highlights the importance of local effects in this system.
A few other pieces of information are important for
our model. First, following the transition to an Amar-
anthaceae-dominated system, climate was not constant
and wetter events punctuated the dry state. Evidence
for the wet episodes comes from peaks of Asteraceae
pollen percentages and concentrations rising and falling
in counterpoint to Amaranthaceae. We interpret the
Amaranthaceae to indicate dry times and low lake level,
with Asteraceae representing wetter cycles. Indeed, the
sediment throughout this section of the core is finely

Fig. 3 Grayscale analysis of sediments attributed to MIS 5e in the Lake Titicaca LT01-2B core. An image of the sediment is shown
directly beneath the grayscale values for that section.
laminated (Fig. 3). Dark organic material alternating
with carbonate-rich layers, indicate a lake oscillating
between states and rather rapid and strong climate
change.
These data suggest that more than a simple preces-
sional pattern is represented. ENSO is thought to have
been operating in a broadly similar temporal pattern
to today during much of MIS 5e (Tudhope et al., 2001),
and so it is possible that when the lake is at very
low levels the influence of long ENSO-like cycles on
Andean precipitation become evident. Alternatively,
and perhaps more likely, the oscillations could reflect
thermal changes in the Atlantic Ocean. Baker et al.
(2005) suggested that Bond Cycles were evident in the
Holocene portion of the Lake Titicaca record with cold
conditions in the North Atlantic corresponding to wet
conditions on the Altiplano, and warm conditions cor-
responding to dry times. This suggestion is complimen-
tary to the hypothesis that weakened trade winds
resulting from warming of the subtropical North Atlan-
tic would weaken the transport mechanism that pumps
Atlantic moisture into Amazonia, the South American
low level jet (SALLJ), and induce aridity in Amazonia
and the Andes (Marengo et al., 2008; Zeng et al., 2008).
This mechanism probably caused the Amazon drought
of AD 2005, and here we suggest that it may also have
been important at longer timescales in response to
persistent thermal dipoles in the Atlantic Ocean (Bond
et al., 2001; Baker et al., 2005).
Conceptual models of nonlinear climate change in the
Altiplano
Our conceptual model starts with precessional forcing
that promotes warm conditions in western Amazonia
r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x
ANDEAN TIPPING POINTS 5
and the Andes. The large-scale presence of ice changes
the feedbacks, and so glacial and interglacial versions
are presented.
Interglacial
Warming during the deglacial period increased convec-
tion and exported moisture from Amazonia via SASM
to the Altiplano. A combination of increased precipita-
tion and runoff from meltwater facilitated formation of
a paleolake. Hence the largest paleolakes would be
predicted to occur within the deglacial period. Periodi-
cally, late Pleistocene paleolakes occupied much of the
Altiplano, with paleolake Tauca (Fig. 1) forming during
the last deglaciation. Lake Tauca occupied 450 000 km2
(i.e. ca. 6  the area of modern Titicaca) and would have
exerted a considerable lake effect on the climate of the
southern Altiplano. However, this warm wet interval
was cut short by an ‘oceanic forcing’ (Fig. 5a), which
could have been due to warming of either the eastern
equatorial Pacific (EEP) inducing El Niño, a warming in
the subtropical Atlantic weakening the SALLJ, or some
combination of these phenomena. If the SASM wea-
kened (December–March) and skies were less cloudy,
i.e. more direct sunlight than the present average
6 h day 1, evaporation would increase. As drought
gripped the Altiplano a positive feedback mechanism
could have exacerbated this effect.
The paleolake would have begun to contract and as it
did so the warm, moist, lake effect would have been
reduced. A decreased lake area would lead to reduced
humidity and thereby increased evaporation, which
would have reinforced reduction of lake area. Although
humidity provokes convective rains over Titicaca, in
general nights over the adjacent land are cloudless. If
6 M . B . B U S H et al.
(a)
(b)
(c)
Fig. 4 Schematic diagram showing (a) the modern system of
Titicaca, (b) a hypothetical migration response due to warming
assuming that species would migrate upslope, and (c) the out-
come of warming in prior interglacials (MIS 5e and 9) in which
forest did not migrate to fill the basin, but the ecosystem dried
out to become a salt marsh.
nightly low temperatures over the land warmed, the
temperature differential between the lake and the land
would have lessened (currently 4 1C with the lake being
warmer than the land at night). One consequence of
reducing that differential would have been to weaken
the lake breeze (2–4 m s 1 as an average modern flow)
that promotes night-time convective rain over the lake,
further reducing inputs of water and thereby accentu-
ating the drawdown of the lake.
The lowstand phase would be reversed by external
forcing such as a change in the dominant phase of
r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x
ENSO or in the temperature regime of the subtropical
Atlantic.
Glacial
The cool conditions of the glacial probably reduced the
importance of convective activity (Fig. 5b). As the ice
mass grew so the influence of katabatic winds coming
from the ice would have increased their microclimatic
influence. The cold descending air would have encoun-
tered a relatively warm lake and created fog in the base
of the valleys. Thus, the system would have been wet,
cold, and light-limited for much of the year. Fog over
the lake surface and generally cold air would combine
to reduce evaporation. Warmer oscillations would have
added meltwater to accentuate highstands. The feed-
back loop that linked precession to height of lake can
serve to draw down the lake in the absence of ice (i.e. in
an interglacial), but it requires the meltwater compo-
nent to boost lake level from this forcing to create a
highstand. Thus once the ice mass was established
precessional patterns became evident, but as the ice
mass melted, there was a much weaker precessional
signature of wetting.
Lake level effects and climate change
It is an important realization that the modern conditions
of the Altiplano are part of a cycle from high to low lake
level change and that modern conditions are not a
‘norm.’ The cycle does not follow a regular rhythm as
it is subject to multiple forcings, some of which are, at
best, quasiperiodic, i.e. Bond Cycles. Nor are the cycles
of lake level variability of equal amplitude, i.e. they
produce high or lowstands of different magnitudes.
Thus during MIS 5e, the cycle was strong taking the
Altiplano to a drier extreme. In looking for an approx-
imate analog for MIS 5e conditions in Lake Titicaca,
some similarities might exist with Lake Poopó, which is
a shallow saline lake that receives about 390 mm of
precipitation annually. As the lake receives some input
from Titicaca via the Rio Desaguadero, 390 mm becomes
a minimum amount of precipitation that can sustain an
open body of water on the Altiplano. For Lake Titicaca
to flip to a saline lake, such as Lake Poopó, a reduction
of 450% precipitation might be needed.
We hypothesize that as evaporation increased and
lake level dropped, it triggered the positive feedback of
aridification, in which temperatures dropped due to
loss of lake effect, and reduced lake surface area re-
sulted in less convective rain falling basinwide. Just a
25 m lowering of lake level ceases drainage into the
Huinaimarca sub-basin. That sub-basin would dry out
(this is known from paleoecological data; Gosling et al.,
 ANDEAN TIPPING POINTS 7

Fig. 5 Conceptual model of feedbacks that result in nonlinear responses in climate change in the Altiplano of Peru/Bolivia. (a)
Interglacial feedbacks, note that deglacial feedbacks are grayed and (b) glacial feedbacks.

2008), the overflow into the Desaguadero River would
have stopped and the downstream system of Lake
Poopó would have dried up. Consequently, a warmer
interglacial instead of stimulating local temperature
increase, and increased convective precipitation, might
at its peak have induced falling temperatures and
aridity in this basin.
Some quantification of when this tipping point
occurred can be made based on moist air adiabatic
lapse rates and observations of tree line. In the absence
of anthropogenic clearance, modern tree-line in the
Andes would probably lie between 3300 and 3700 m
elevation (Wille et al., 2002; Di Pasquale et al., 2008), i.e.
300–500 m below Lake Titicaca. Moist air adiabatic lapse
rates in this section of the Andes are ca. 5.2 1C per
1000 m of vertical ascent. As the trees never fully
invaded the shoreline, the tip to drier conditions must
have occurred within 1–2 1C of modern tempera-
tures. Thus, anthropogenic climate change could drive
this system past that the tipping point, which judging
from the abrupt rise in Amaranthaceae pollen in MIS 9
and 5e, comes with very little warning.

r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x

8 M . B . B U S H et al.
This pattern of aridity was observed in both MIS 9
and 5e. MIS 7 does not appear to have been quite so
extreme and its cycle toward aridity was curtailed
before the most extreme lake draw down was reached.
This observation is entirely consistent with marine and
ice core records that show MIS 7 to have been a
protracted, multipeak interglacial, but with less extreme
conditions than MIS 9 and 5e. If, as is projected, climates
warm by 3–6 1C this century, the positive feedback
toward aridity might be renewed. Although the last
1000 years have probably been the wettest of the
Holocene in the Andes, warming of the subtropical
north Atlantic may have already begun, and has been
suggested to have induced the AD 2005 Amazonian
drought (Marengo et al., 2008; Zeng et al., 2008). Indeed,
aerosols, e.g. SO2 and fine particulate carbon, may have
mitigated subtropical Atlantic warming, and as the
loading of these pollutants is reduced, warming may
accelerate (Cox et al., 2008). If that trend becomes
persistent it would probably increase fire frequency,
decrease forest cover, and result in reduced moisture
transport into the Andes (da Silva et al., 2008; Cochrane
& Barber, 2009). Similarly, if the sea surface temperature
(SST) changes predicted by the HADCM3LC happen,
then a warmer EEP would probably result in Andean
drought. Of these two processes ENSO variation ap-
pears to be dominant, as effects of subtropical Atlantic
temperature are clearest when ENSO is weak (Zeng
et al., 2008). However, the systems are not mutually
exclusive and perhaps the fluctuating pulses of wet and
dry conditions within the overall dry interglacial con-
ditions reflect a synergy between these systems.
One factor that will differ substantially between past
lowstands on the Altiplano and the modern state is that
it will occur in a landscape containing a capital city (La
Paz, Bolivia) and approximately 2 200 000 human inha-
bitants.
Ecosystem and societal adaptation
The predictable ecosystem response to a lowering of
Lake Titicaca and desiccation of the Altiplano would be
a stalling of the anticipated upslope migration of forest
and the expansion of xerophytic species. Erickson
(1988) has argued that humans overcome climate
change in the Andes through cultural adaptation and
landscape manipulation. While this appears to be true
of the relatively subtle changes in climate of the last
3000 years, the response of humans to the millennial
scale mid-Holocene droughts in the Altiplano was to
abandon the region (Núñez et al., 2002). Speculating
about how humans could maintain crop productivity
and water supply is beyond the scope of this paper,
though the twin obstacles of increasingly concentrated
r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x
salts in surface waters and decreased water availability
(both from reduced precipitation and loss of ice caps)
would be substantial hurdles to overcome.
In the seminatural systems of the Andes, populations
of species that were stressed by the drier conditions
would be predicted to go locally extinct, survive in
moist microrefugia, or undergo selection to genotypes
that could withstand drought. Emphasizing the con-
servation of potential microrefugia and ecosystem con-
nectivity may become an important component of
strategies to safeguard some at-risk species.
Biomass and fuel load are inextricably linked in the
high Andes, and fire is a landscape transformer. Planta-
tions of trees that trap moisture could mitigate some
drought effects, as shading helps to retain soil moisture
and organic material, plus the structure of the plants
serves to intercept moisture from ground-level cloud.
Thus such stands, while increasing biomass, could
reduce fire risk. However, Eucalyptus, which is the most
widely planted tree in the high Andes, wicks prodi-
gious quantities of moisture out of the soil, effectively
drying the landscape, and is highly combustible (Luzar,
2007). Eucalyptus plantations increase the probability
that the wettest locations would become fire-prone
and unsuitable for many of the indigenous elements
that might otherwise occupy these settings. A more
desirable candidate for afforestation is Polylepis, but
control of fire would be essential for their ability to
mature (Cierjacks et al., 2008).
Conclusion
Cycles of lake level have characterized at least the last
370 ka on the Peru/Bolivian Altiplano. Orbital forcing
has been a basic pacemaker of events, but superimposed
on this pattern are changes in moisture availability
driven by oceanic processes, which are themselves par-
tially controlled by orbital patterns. Nonlinear feedbacks
appear to have induced rapid landscape change.
Changes in lake area have significant climatic effects on
the Altiplano and positive feedbacks are envisaged that
lead both to high and low lake stands.
In MIS 9 and 5e, a trend toward warm wet conditions
was initiated by changes in precession, but suddenly
altered by a change in the moisture balance. We hy-
pothesize that oceanic forcing induced droughts that
were then accentuated through positive feedback me-
chanisms, leading to a profound change in local eco-
systems. Eventually, outside forcing broke the feedback
cycle and once again reversed these trends.
On the Altiplano, contracting lake area during inter-
glacial warm periods may have initiated feedback ef-
fects so strong that they reversed the thermal trend of
the progression toward warming. A similar turning

point where the dominant signature shifting from
warming to aridity may occur again as regional tem-
peratures rise 1–2 1C. Despite this turning point poten-
tially reducing local temperatures, it offers no solace to
conservation of natural resources as the accompanying
droughts would cause major dislocations of natural and
human communities.
A key lesson to learn from this study is that although
species have already started to expand their ranges
upslope on tropical mountains this pattern could be
abruptly halted, even reversed, as a result of nonlinear,
local, responses to ongoing global warming. A critical
difference between MIS 9 and 5e in South America
relative to the Holocene is the presence of a human
population. Our prognoses of climate change are of
direct relevance to the livelihoods of more than 2
million people living in the Peruvian and Bolivian
Andes, as they will influence their choice of land use,
occupancy of marginal lands, and which kind of lands
become marginal.
The modern relevance of this study is emphasized by
the recent push to use the Andes for carbon sequestra-
tion. The need to understand the future capability of the
land to support forest or maintain soil moisture vital for
carbon sequestration is critical to the long-term success
of such ventures. In addition to its global benefits,
agroforestry may play a role in delaying or alleviating
the fullest local effects of these droughts. However,
choice of species to be planted and control of fire are
as important as the area that becomes forested.
Acknowledgements
This work was supported by Grant NSF-ATM 0317539. Andy
Lloyd of the Open University is thanked for preparing Fig. 1.
This work is a product of the Andes Biodiversity and Ecosystem
Research Group (ABERG) and is publication #6 of the Florida
Institute for Technology, Institute for Research on Global Climate
Change.
References
Abbott MB, Wolfe BB, Wolfe AP et al. (2003) Holocene paleohydrology and glacial
history of the central Andes using multiproxy lake sediment studies. Palaeogeo-
graphy, Palaeoclimatology, Palaeoecology, 194, 123–138.
Baker PA, Fritz SC, Garland J, Ekdahl E (2005) Holocene hydrologic variation at Lake
Titicaca, Bolivia/Peru, and its relationship to North Atlantic climate variation.
Journal of Quaternary Science, 20, 655–662.
Baker PA, Rigsby CA, Seltzer GO, Fritz SC, Lowenstein TK, Bacher NP, Veliz C (2001a)
Tropical climate changes at millennial and orbital timescales on the Bolivian
Altiplano. Nature, 409, 698–701.
Baker PA, Seltzer GO, Fritz SC et al. (2001b) The history of South American tropical
precipitation for the past 25,000 years. Science, 291, 640–643.
Bond G, Kromer B, Beer J et al. (2001) Persistent solar influence on North Atlantic
climate during the Holocene. Science, 294, 2130–2136.
Broecker WS, Denton GH (1989) The role of ocean–atmosphere reorganizations in
glacial cycles. Geochimica et Cosmochimica Acta, 53, 2465–2501.
Bush MB, Miller MC, de Oliveira PE, Colinvaux PA (2002) Orbital forcing signal in
sediments of two Amazonian lakes. Journal of Paleolimnology, 27, 341–352.
r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x
ANDEAN TIPPING POINTS 9
Christensen JH, Hewitson B, Busuioc A et al. (2007) Regional climate projections. In:
Climate Change 2007: the Physical Science Basis. Contribution of Working Group I to the
Fourth Assessment Report of the Intergovernmental Panel on Climate. Change. (eds
Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, Tignor M, Miller
HL), pp. 847–940. Cambridge University Press, New York.
Cierjacks A, Salgado S, Wesche K, Hensen I (2008) Post-fire population dynamics of
two tree species in high-altitude Polylepis forests of Central Ecuador. Biotropica, 40,
176–182.
Clapperton CW (1993) Quaternary Geology and Geomorphology of South America. Elsevier,
Amsterdam.
Cochrane MA, Barber CP (2009) Climate change, human land use and future fires in
the Amazon. Global Change Biology, 15, 601–612.
Cox PM, Harris PP, Huntingford C et al. (2008) Increasing risk of Amazonian drought
due to decreasing aerosol pollution. Nature, 453, 212–215.
Cross SL, Baker PA, Seltzer GO, Fritz SC, Dunbar RB (2000) A new estimate of the
Holocene lowstand level of Lake Titicaca, central Andes, and implications for
tropical palaeohydrology. The Holocene, 10, 21–32.
Cruz FW Jr, Burns SJ, Karmann I et al. (2005) Insolation-driven changes in atmospheric
circulation over the past 116,000 years in subtropical Brazil. Nature, 434, 63–66.
da Silva RR, Werth D, Avissar R (2008) Regional impacts of future land-cover changes
on the Amazon basin wet-season climate. Journal of Climate, 21, 1153–1170.
Di Pasquale G, Marziano M, Impagliazzo S, Lubritto C, De Natale A, Bader MY (2008)
The Holocene treeline in the northern Andes (Ecuador): first evidence from soil
charcoal. Palaeogeography, Palaeoclimatology, Palaeoecology, 259, 17–34.
Ekdahl EJ, Fritz SC, Baker PA, Rigsby CA, Coley K (2008) Holocene multidecadal- to
millennial-scale hydrologic variability on the South American Altiplano. The
Holocene, 18, 867–876.
Erickson CL (1988) Raised field agriculture in the Lake Titicaca Basin: putting ancient
agriculture back to work. Expedition, 30, 8–16.
Fritz SC, Baker PA, Lowenstein TK et al. (2004) Hydrologic variation during the last
170,000 years in the southern hemisphere tropics of South America. Quaternary
Research, 61, 95–104.
Fritz SC, Baker PA, Seltzer GO, Ballantyne A, Tapia P, Cheng H, Edwards RL (2008)
Corrigendum to ‘‘Quaternary glaciation and hydrologic variation in the South
American tropics as reconstructed from the Lake Titicaca drilling project’’ [Qua-
ternary Research 68 (2007) 410–420]. Quaternary Research, 69, 342.
Fritz SC, Baker PA, Seltzer GO, Ballantyne A, Tapia PM, Cheng H, Edwards RL (2007)
Quaternary glaciation and hydrologic variation in the South American tropics
as reconstructed from the Lake Titicaca drilling project. Quaternary Research, 68,
410–420.
Gosling WD, Bush MB, Hanselman JA, Chepstow-Lusty AJ (2008) Glacial–interglacial
changes in moisture balance and the impact on vegetation in the southern hemisphere
tropical Andes (Bolivia/Peru). Palaeogeography, Palaeoclimatology, Palaeoecology, 259,
35–50.
Gosling WD, Hanselman JA, Knox C, Valencia BG, Bush MB (2009) Long term drivers
of change in Polylepis woodland distribution in the central Andes. Journal of
Vegetation Science, 20, 1041–1052.
Hanselman JA, Bush MB, Gosling WD, Collins A, Knox C (in press) A 370,000-year
record of vegetation and fire history around Lake Titicaca (Bolivia/Peru). Palaeo-
geography, Palaeoclimatology, Palaeoecology.
Hanselman JA, Gosling WD, Paduano GM, Bush MB (2005) Contrasting pollen
histories of MIS 5e and the Holocene from Lake Titicaca (Bolivia/Peru). Journal of
Quaternary Science, 20, 663–670.
Hastenrath S, Kutzbach J (1985) Late Pleistocene climate and water budget of the
South American Altiplano. Quaternary Research, 24, 249–256.
Hillyer R, Valencia BG, Bush MB, Silman MR, Steinitz-Kannan M (2009) A 24,
900-year paleolimnological history from the Peruvian Andes. Quaternary Research,
71, 71–82.
Hooghiemstra H, Melice JL, Berger A, Shackleton NJ (1993) Frequency spectra and
paleoclimatic variability of the high-resolution 30–1450 ka Funza I pollen record
(Eastern Cordillera, Colombia). Quaternary Science Reviews, 12, 141–156.
Luzar J (2007) The political ecology of a ‘‘forest transition’’: Eucalyptus forestry in the
Southern Peruvian Andes. Ethnobotany Research and Applications, 5, 85–93.
Malcolm JR, Liu C, Neilson RP, Hansen L, Hannah L (2006) Global warming and
extinctions of endemic species from biodiversity hotspots. Conservation Biology, 20,
538–548.
Marengo JA, Nobre CA, Tomasella J et al. (2008) The Drought of Amazonia in 2005.
Journal of Climate, 21, 495–516.
Moy CM, Seltzer GO, Rodbell DT, Anderson DM (2002) Variability of El Nino/
Southern Oscillation activity at millennial timescales during the Holocene epoch.
Nature, 420, 162–164.
10 M . B . B U S H et al.
Núñez L, Grosjean M, Cartajena I (2002) Human occupations and climate change in
the Puna de Atacama, Chile. Science, 298, 821–824.
Paduano GM, Bush MB, Baker PA, Fritz SC, Seltzer GO (2003) A vegetation and fire
history of Lake Titicaca since the Last Glacial Maximum. Palaeogeography, Palaeo-
climatology, Palaeoecology, 194, 259–279.
Parmesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts
across natural systems. Nature, 421, 37–42.
Peters RL, Darling JDS (1985) The greenhouse effect and nature reserves. BioScience, 35,
707–711.
Petit JR, Jouzel J, Raynaud D et al. (1999) Climate and atmospheric history of the past
420,000 years from the Vostok ice core, Antarctica. Nature, 399, 429–436.
Placzek C, Quade J, Patchett PJ (2006) Geochronology and stratigraphy of late
Pleistocene lake cycles on the southern Bolivian Altiplano: implications for causes
of tropical climate change. Geological Society of America Bulletin, 118, 515–532.
Pounds JA (2001) Climate and amphibian declines. Nature, 410, 639–640.
Pounds JA, Fogden MPL, Campbell JH (1999) Biological response to climate change on
a tropical mountain. Nature, 398, 611–615.
Rasband WS (2005) ImageJ 1.32j. National Institute of Health, USA.
Rein B (2007) How do the 1982/83 and 1997/98 El Niños rank in a geological record
from Peru? Quaternary International, 161, 56–66.
Riedinger MA, Steinitz-Kannan M, Last WM, Brenner M (2002) A 6100 14C yr record
of El Nino activity from the Galapagos Islands. Journal of Paleolimnology, 27, 1–7.
Roche MA, Bourges JC, Mattos R (1992) Climatology and hydrology of the Lake
Titicaca basin. In: Lake Titicaca: A Synthesis of Limnological Knowledge (eds Dejoux C,
Iltis A), pp. 63–83. Kluwer Academic Press, Boston.
Seltzer G, Rodbell D, Burns S (2000) Isotopic evidence for late Quaternary climatic
change in tropical South America. Geology, 28, 35–38.
r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02203.x
Seltzer GO (1990) Recent glacial history and paleoclimate of the Peruvian–Bolivian
Andes. Quaternary Science Reviews, 9, 137–152.
Seltzer GO, Cross S, Baker P, Dunbar R, Fritz S (1998) High-resolution seismic
reflection profiles from Lake Titicaca, Peru/Bolivia. Evidence for Holocene aridity
in the tropical Andes. Geology, 26, 167–170.
Seltzer GO, Rodbell DT, Baker PA, Fritz SC, Tapia PM, Rowe HD, Dunbar RB (2002)
Early deglaciation in the tropical Andes. Science, 298, 1685–1686.
Stocker T (2000) Past and future reorganizations in the climate system. Quaternary
Science Reviews, 19, 301–319.
Thomas CD, Cameron A, Green RE et al. (2004) Extinction risk from climate change.
Nature, 427, 145–148.
Tudhope AW, Chilcott CP, McCulloch MT et al. (2001) Variability in the El Nino-
Southern Oscillation through a glacial–interglacial cycle. Science, 291, 1511–1517.
Van der Hammen T, Hooghiemstra H (2003) Interglacial–glacial Fuquene-3 pollen
record from Colombia: an Eemian to Holocene climate record. Global and Planetary
Change, 36, 181–199.
Wang X, Auler AS, Edwards RL et al. (2004) Wet periods in northeastern Brazil over the
past 210 kyr linked to distant climate anomalies. Nature, 432, 740–743.
Wille M, Hooghiemstra H, Hofstede R, Fehse J, Sevink J (2002) Upper forest line
reconstruction in a deforested area in northern Ecuador based on pollen and
vegetation analysis. Journal of Tropical Ecology, 18, 409–440.
Ybert JP (1992) Ancient lake environments as deduced from pollen analysis. In: Lake
Titicaca. A Synthesis of Limnological Knowledge (eds DeJoux C, Iltis A), pp. 49–62.
Kluwer Academic Publishers, Boston.
Zeng N, Yoon JH, Marengo JA, Subramaniam A, Nobre CA, Mariotti A, Neelin JD
(2008) Causes and impacts of the 2005 Amazon drought. Environmental Research
Letters, 3, 014002, doi: 10.1088/1748-9326/1083/1081/014002.