An Andean ice-core record of a Middle Holocene mega-drought in North Africa and Asia

Abstract

An ice core from the Nevado Huascarán col in the Cordillera Blanca of northern Peru contains high-resolution time series of dust concentrations and size distributions since the end of the last glacial stage. A large dust peak, dated $4500 years ago, is contemporaneous with a widespread and prolonged drought that apparently extended from North Africa to eastern China, evidence of which occurs in historical, archeological and paleoclimatic records. This event may have been associated with several centuries of weak Asian/Indian/African monsoons, possibly linked with a protracted cooling in the North Atlantic. During the second half of the 20th century, high austral-summer dust concentrations in the Huascarán record are significantly correlated with atmospheric conditions, such as sea-level pressure and zonal wind velocities that are consistent with El Niño–Southern Oscillation (ENSO) and positive North Atlantic Oscillation (NAO) indices, and with aridity in North Africa, southwest Asia and the Middle East. Therefore, the dominant submicron fraction of the dust may have been transported by more intense northeasterly trade winds from the African dry regions across the tropical Atlantic during a period of frequent and/or intense ENSO activity. The proposed ENSO conditions that may have been linked with drought in the monsoon region may also have contributed to aridity in tropical South America, including the Cordillera Blanca.

Full text

An Andean ice-core record of a Middle Holocene mega-drought
in North Africa and Asia
Mary E. DAVIS,
1
Lonnie G. THOMPSON
1,2
1
Byrd Polar Research Center, The Ohio State University, 1090 Carmack Road, Columbus, OH 43210-1002, USA
E-mail: davis.3@osu.edu
2
Department of Geological Sciences, The Ohio State University, Columbus, 155 South Oval Mall, Columbus,
OH 43210-1398, USA
ABSTRACT. An ice core from the Nevado Huascara
´
n col in the Cordillera Blanca of northern Peru
contains high-resolution time series of dust concentrations and size distributions since the end of the last
glacial stage. A large dust peak, dated 4500 years ago, is contemporaneous with a widespread and
prolonged drought that apparently extended from North Africa to eastern China, evidence of which
occurs in historical, archeological and paleoclimatic records. This event may have been associated with
several centuries of weak Asian/Indian/African monsoons, possibly linked with a protracted cooling in
the North Atlantic. During the second half of the 20th century, high austral-summer dust concentrations
in the Huascara
´
n record are significantly correlated with atmospheric conditions, such as sea-level
pressure and zonal wind velocities that are consistent with El Nin˜o–Southern Oscillation (ENSO) and
positive North Atlantic Oscillation (NAO) indices, and with aridity in North Africa, southwest Asia and
the Middle East. Therefore, the dominant submicron fraction of the dust may have been transported by
more intense northeasterly trade winds from the African dry regions across the tropical Atlantic during a
period of frequent and/or intense ENSO activity. The proposed ENSO conditions that may have been
linked with drought in the monsoon region may also have contributed to aridity in tropical South
America, including the Cordillera Blanca.
INTRODUCTION
Compared to the last glacial stage (LGS) and the subsequent
deglaciation, the Holocene has been a period of relative
climatic calm during which human populations and civiliza-
tions have flourished and expanded over the Earth. However,
it has not been completely free of the kind of climatic
disruptions that have had detrimental effects on even
powerful societies. One such event, which was apparently
a far-reaching period of aridity, occurred 4000–4500 years
ago (4.0–4.5 kyr
BP), and left its mark on a variety of
archeological and geological records, including those from
excavations in Syria (Weiss and others, 1993), lake levels in
North Africa (Gasse, 1977; Servant and Servant-Vildary,
1980; Gillespie and others, 1983) and western China (Gasse
and others, 1991), speoleothems in Israel (Bar-Matthews and
others, 1999) and eastern China (Wang and others, 2005),
marine cores from the Gulf of Oman (Cullen and others,
2000) and the Indus delta region (Staubwasser and others,
2003), peat deposits in Mongolia (Xiao and others, 2004),
and an ice core from eastern equatorial Africa (Thompson
and others, 2002). According to many of these records, the
onset of the drought was abrupt (possibly within a decade),
lasted for up to several centuries, then ended abruptly.
Although the majority of the documentation for this dry
period comes from the Eastern Hemisphere, evidence for a
similar abrupt climatic episode has been found in tropical
South America. For example, deposits from Lake Titicaca on
the border between Peru and Bolivia (Baker and others,
2001; Tapia and others, 2003) and from the Amazon fan
(Maslin and Burns, 2000) show a Middle Holocene period of
desiccation. Finally, an ice core from a high-altitude Andean
tropical glacier contains a prominent dust event within the
Holocene, with particle concentrations that are several
orders of magnitude higher than all other levels since the
end of the LGS (Thompson, 2000; Thompson and others,
2000). The timing and nature of the dust peak in this ice
core, its possible linkages with the Asian/African climate,
and the possible Middle Holocene climate forcings for such
an event are discussed in this paper.
ICE-CORE ANALYSIS
In the austral winter of 1993 (June–August) a team from the
Byrd Polar Research Center’s (BPRC) Ice Core Paleoclima-
tology Research Group drilled two ice cores to bedrock on
the col of Nevado Huascara
´
n(9806
0
S, 77836
0
W;
6050 m a.s.l.) in the Cordillera Blanca of northern Peru
(Fig. 1). The first core (core 1, 160.4 m long) was cut into
samples that were melted and bottled in the field, but the
second (core 2, 166.1 m long) was returned frozen to BPRC
for high-resolution analyses. This core was cut into three
equivalent sets, each with 4476 samples, and analyzed for
the ratios of
18
Oto
16
O(d
18
O), mineral dust concentration
and size distribution, and anion concentrations (chloride,
nitrate and sulfate) analyses. Core 1, on the other hand, was
analyzed only for d
18
O.
Above the firn/ice transition (33.5 m below surface
(mbs)) the three sets of samples were cut from the center of
the core inside a class 100 clean bench in a laboratory
freezer maintained at –108C. The firn was handled using
several pairs of pre-cleaned latex gloves, and the cut
samples were placed in pre-cleaned polystyrene containers
which were transported to a class 100 clean room where
they were allowed to melt before analysis. Below the firn/ice
transition the dust and anion samples were rinsed with
Millipore reagent-grade deionized water to remove con-
taminants from the surface. The core was cut such that the
Annals of Glaciology 43 200634

sample lengths decreased from 17 to 6 cm in the top 55 m,
then from 5 to 3 cm in the interval 55–113 mbs, and finally
to 2 cm at 113 mbs, which was the size at which they
remained to the bottom of the core.
Dust concentrations and size distributions were measured
using two Coulter Counters (Model TAII), one equipped with
a30mm aperture tube to count particles with diameters
between 0.63 and 16 mm in 14 size ranges, and the second
equipped with a 100 mm aperture tube to count particles
with diameters between 2.0 and 40 mm, also in 14 size
ranges. The two datasets were integrated and the total
concentrations between 0.63 and 40 mm were normalized to
1 mL of melted ice. A Finnigan MAT Delta E mass
spectrometer was used for the analyses of d
18
O. The anion
concentrations, which were analyzed with a Dionex Model
2010 ion chromatograph, were presented initially in
Thompson and others (1995), but will not be discussed
further here.
The d
18
O and dust concentration data are plotted by
depth (1 m averages) in Figure 2a and b, respectively. One
obvious feature in the d
18
O profile is the 6.3% increase from
165.3 mbs to 163 mbs, which is consistent with the Last
Glacial Maximum to Early Holocene isotopic difference
found in other ice cores of similar age (Thompson and others,
1998, table 2). The dust record shows that above 165 mbs
this core is relatively clean, with an average concentration of
15 000 particles per mL. The exception is a large peak
around 157.7 mbs (marked with an asterisk in Fig. 2b) with
an average concentration of 82 000 particles per mL.
TIMESCALE RECONSTRUCTION
Since the 1 year snow accumulation between 1992 and
1993 was 3.30 m (1.74 m ice equivalent), and the 20th-
century net accumulation rate, or net balance, averaged
1.4 m ice equivalent (Henderson, 1996), the upper portion
of the core could be sampled at sub-annual resolution. The
dating for this section was determined by counting the peaks
in dust, nitrate and d
18
O that formed every dry season during
the austral winter from June to September (Thompson and
others, 1995). Thus the annual averages of the measured
parameters were calculated based on thermal years (dry
season to dry season). However, below
AD 1817
(119.26 mbs) the thinning of the annual layers from vertical
flow was such that seasonal resolution of the parameters was
difficult to maintain. However, annual dating was still
possible down to 125.2 mbs (corresponding to the 1719/20
thermal year).
The original Holocene timescale for the Huascara´n ice
core (Thompson and others, 1995) was constructed using an
empirical two-parameter model (Bolzan, 1985; Reeh, 1988),
as shown in Figure 3a. The rate of thinning with depth (p in
the model equation in Fig. 3a) was first calculated for the top
of the core where the timescale was annually resolvable,
using the dates of 1915 (78 years before 1993) at 86.47 mbs
and 1817 (176 years before 1993) at 119.26 mbs as pinning
points, and assuming steady-state conditions with a constant
accumulation of 1.74 m ice equivalent (b in the equation).
The lower 47 m of the core (119.26–166.07 mbs) were dated
by constraining the
18
O depletion at 164.1 mbs (marked by
an arrow in Fig. 2a) as the Younger Dryas (YD) event. This
was ascertained by matching the lowest 3 m of the ice core
with the calibrated
14
C dated record of d
18
OofG. bulloides
from the deep-sea marine core SU81-18, which was drilled
in the tropical North Atlantic off the coast of Portugal (Bard
and others, 1987; Fairbanks, 1989). The midpoint of the YD
event in the Huascara´n record was assigned an age of
12.25 kyr
BP, consistent with the YD age in the layer-counted
Fig. 1. Map of Peru depicting the Andes Mountains (areas above
4000 m a.s.l.) and the location of Nevado Huascara
´
n.
Fig. 2. One-meter averages of (a) d
18
O(%) and (b) dust concen-
trations (0.63–40.0 mm diameters) (10
5
mL
–1
) from Huascara´n
core 2. The arrow in (a) and the asterisk in (b) indicate features
discussed in the text.
Davis and Thompson: Andean ice-core record of a Middle Holocene mega-drought 35

GRIP and GISP2 (Greenland) ice-core records (Johnsen and
others, 1992; Taylor and others, 1993). With this lower
chronological horizon established, the Holocene part of the
Huascara´n record was calculated with the two-parameter
model using the thinning rate (p ¼ 1.253) that was deter-
mined for the upper 119 m of the core (Fig. 3a). In this
chronology, the large Holocene dust event at 157.7 mbs
(shown by the asterisk in Fig. 2b) was calculated at 2 kyr
BP.
Since the LGS portion of the Huascara´n record was
matched with an absolutely dated marine core, the
confidence in the chronology of that part of the record
was reasonable. However, the confidence in the Holocene
dating was lower because of the lack of chronological
calibration between
AD 1817 and 12.25 kyr BP. Several years
later, the Holocene timescale was revised using the analyses
of the isotopic composition of O
2
in the air trapped in
glacier ice bubbles (d
18
O
air
), which were measured by
T. Sowers at Pennsylvania State University. Sowers and
others (1989) demonstrated that d
18
O
air
reflects changes in
the global atmosphere. Since the inter-hemispheric mixing
time is only 1 year, the d
18
O
air
is constant throughout the
atmosphere at any time, even though the turnover period is
1200 years (Bender and others, 1994b). The glacial/inter-
glacial transition (8.0–15.0 kyr
BP) is characterized by
large, virtually simultaneous changes in d
18
O
air
in Green-
land and Antarctic cores that are dated by independent
methods (Sowers and Bender, 1995), with an estimated error
of 600 years in the age–depth relationships. When the ages
derived from these d
18
O
air
measurements (Bender and
others, 1994a) were correlated with the d
18
O
air
data from
samples from the Huascara´n ice core (Fig. 3b), the range of
error for the tropical core was construed as being similar to
the polar records over this transition. The measurements of
the Huascara´n d
18
O
air
provided several time horizons for the
Early Holocene that allowed for the development of a new
age–depth relationship for the entire Holocene by passing a
third-order regression curve through the chronological
horizons from the layer-counted points at the top to the
d
18
O
atm
match points towards the bottom (Fig 3a). The
match points above 9 kyr
BP were not used to calculate the
regression because they covered too large a range of
possible corresponding dates in the Greenland record
(Fig. 3b), but they are plotted on the regression curve as a
confirmation of the accuracy of the age–depth relationship.
An error of 5 years was estimated around the upper
pinning point of the model (
AD 1720), and the range
increased to 600 years at 8 kyr
BP, as determined by the
transfer of the d
18
O
air
-based timescale over the glacial/
interglacial transition as discussed above. The original and
the revised age–depth relationships are compared in
Figure 3a. The new chronology for the Huascara´n record
moved the lower (earlier) boundary of the large Holocene
dust peak (hereafter referred to as the MHDE, or the Middle
Holocene dust event) from its originally reconstructed date
at 2 kyr
BP to 4.5 kyr BP.
THE MIDDLE HOLOCENE DUST EVENT
The dust and d
18
O data for the entire 19 kyr Huascara´n ice-
core record, which are recalculated according to the revised
timescale, are illustrated in Figure 4a and b, respectively, as
100 year averages. Other than the very high dust concen-
tration levels of the LGS (prior to 17 kyr
BP), the MHDE is the
most prominent feature of the dust record. During this time,
the d
18
O was decreasing from its high postglacial levels at
10 kyr
BP. The moisture source for the eastern Peruvian
Andes and for the northern Amazon Basin ultimately is the
tropical North Atlantic, from where the water vapor is
carried by northeasterly trade winds over the ocean, then
recycled over the Amazon forest (Grootes and others, 1989;
Thompson and others, 1995). The d
18
O curve over the last
10 kyr resembles the tropical Northern Hemisphere insola-
tion curve through the Holocene (Berger and Loutre, 1991),
and both the d
18
O and the sea surface temperatures (SSTs) of
the source region may be reflections of the insolation
variations.
The MHDE is prominent in Figure 4b, and detailed (i.e.
every sample) views of the dust concentrations and size
distributions between 3.7 and 5.0 kyr
BP are shown in
Figure 4c–h. In addition to the high concentration
(Fig. 4c), the sizes of the mineral dust in the MHDE also
demonstrate that this is a unique feature in the Huascara´n
Holocene record. Submicron particles (Fig. 4d) constitute
up to 70% of the total concentration between 0.63 and
40 mm in diameter, which is 30–40% higher than pre- and
Fig. 3. (a) Timescale development for Huascara
´
n core 2. The earlier
timescale, which was developed for Thompson and others (1995)
and is depicted by the dashed line, was calculated using the two-
parameter model formula, with h (total length of the core in ice
equivalent) ¼ 136.40 m, b (modern accumulation in ice equiv-
alent) ¼ 1.74 m, p (thinning parameter calculated from this model
for the top of the core where layer counting was possible) ¼ 1.253.
In addition, z is depth in the core, and T is its corresponding age in
years
BP. The revised timescale is depicted by the solid line, and is
discussed in the text. The SU81-18 match points were derived from
matching the d
18
O from the lowest 3 m of the Huascara
´
n core (i.e.
the LGS) with the d
18
O from a tropical North Atlantic marine core.
(b) Matching between d
18
O
atm
in the GISP2 ice core and the
Huascara
´
n ice core. GISP2 data can be downloaded from ftp://
ftp.ncdc.noaa.gov/pub/data/paleo/icecore/greenland/summit/gisp2/
gases/gas.txt.
Davis and Thompson: Andean ice-core record of a Middle Holocene mega-drought36

post-MHDE levels (Fig. 4e). However, the size fraction
>5 mm also increases during this event (Fig. 4f), and even the
concentration of giant particles (>16 mm) ranges over an
order of magnitude higher (Fig. 4g), although the per cent of
this size range is reduced from 3% of the total concentration
in the pre- and post-event levels to 1% within the peak
(Fig. 4h). The concentrations of the particles >5.0 mm
actually increase 100 years before the increase in sub-
micron dust, and remain high for several decades after the
fine-dust levels abruptly drop.
CLIMATIC AND ENVIRONMENTAL CONDITIONS
DURING THE MIDDLE HOLOCENE
There is a large collection of Holocene proxy climate
records from the Andes, the Altiplano and the Amazon Basin
that demonstrate that an arid period occurred during the
time of the MHDE in the Huascara´n ice-core record. Stable-
isotope analyses of planktonic foraminifera from the
Amazon fan show that one of highest
18
O values in the
Holocene (inferring reduced Amazon River flow) occurred
4.5 kyr
BP (Maslin and Burns, 2000), almost contempor-
aneously with a depletion of
18
O in the isotopic record from
Lake Junı´n, Peru (Seltzer and others, 2000). Evidence for
Middle Holocene aridity on the Altiplano comes from Lake
Titicaca on the border between Peru and Bolivia, where the
per cent of fresh-water plankton reached its lowest levels
between 5 and 2 kyr
BP, and the per cent saline diatoms
increased at 6 kyr
BP (
14
C age) and reached a maximum at
3.6–4.0 kyr
BP (
14
C age) (Baker and others, 2001). Tapia
and others (2003) offer a more refined timeline for climate
variation in this interval using the Lake Titicaca record of per
cent saline planktonic taxa, which confirms that although
water levels were low between 3 and 6 kyr
BP, the lowest
levels were achieved at 4.5 kyr
BP.
The significant increase in very large dust particles in the
MHDE in the Huascara´n core is suggestive of aridity in the
Cordillera Blanca. A drier climate may have resulted in
decreasing snowfall which could not keep pace with
sublimation during the dry season, leading to ice recession
on the peaks. There is no evidence in the Huascara´n climate
record (in the stable isotopes or in the visible stratigraphy),
nor in any nearby climate record, that an anomalous
warming occurred during this time.
The MHDE is noteworthy not only because it is a unique
feature in the Huascara´n Holocene record, but also because
it appeared during a time when a global-scale climatic
disruption affected a large region from North Africa through
the Mediterranean area, to as far east as China. The
paleoclimatic, archeological and historical records pre-
viously listed indicate that during the third millennium
BCE
(before the Common Era) a sudden and severe event, which
was probably a protracted drought, occurred in many
regions that had been under the expanded influence of the
Asian/Indian/African monsoon during the first half of the
Holocene. This drought was contemporaneous with the
decline of the Akkadian Empire in Mesopotamia (Weiss and
others, 1993), the failure of Nile River floods that were
coincident with the decline of North African civilizations
(Hassan, 1997) and the end of the Harappan civilization in
the Indus valley (Staubwasser and others, 2003). Whether
the drought was actually instrumental in the collapse of
these societies is a controversial point among archeologists
(e.g. Butzer, 1997), but the evidence is very strong that this
climatic anomaly occurred at or near the same point in time
as these historical events.
Figure 5 shows some of the geological records that
document this Middle Holocene aridity, in addition to the
Huascara´n dust profile. A marine core from the Gulf of
Oman (Cullen and others, 2000) contains an abrupt spike in
carbonates (Fig. 5a) which have been chemically traced to
an archeological site (Tell Leilan) in Syria, the home of the
Akkadian culture (Weiss and others, 1993). The shape of this
profile is similar to that from Huascara´n (Fig. 5b), which
shows a sudden onset and then a more gradual decline
followed by a secondary peak. An ice-core record from
Kilimanjaro, Tanzania, (Thompson and others, 2002) con-
tains an unusually thick, black dust band that is dated
around 4.0 kyr
BP (Fig. 5c), and is indicative of a period that
may have lasted several decades to centuries during which
the Northern Ice Field, presently the largest on the moun-
tain, had dramatically reduced in size. Lake records from
equatorial and North Africa (Gasse, 2000) (Fig. 5d–f)
indicate that water levels had decreased greatly, as did lake
levels from western Tibet (Gasse and others, 1991, 1996),
which are not shown.
Fig. 4. (a, b) Huascara
´
n core 2 d
18
O (a) and dust concentration (b)
records for the last 19 kyr, shown as 100 year averages. The MHDE
is prominent and in (c–h) the data are shown as actual samples
(average length 2 cm). (c) Total dust concentrations from 0.63 to
40 mm; (d) the concentration of submicron dust; (e) the per cent of
submicron dust with respect to total dust; (f–h) concentrations of
large (5–16 mm) (f) and giant (>16 to 40 mm) particles (g) through the
MDHE, and the per cent of large particles with respect to total
dust (h).
Davis and Thompson: Andean ice-core record of a Middle Holocene mega-drought 37

Other records not shown in Figure 5 also contain
evidence of a sudden and marked arid-climate episode at
this time. A speleothem from the Soreq cave in Israel (Bar-
Matthews and others, 1999) shows an abrupt drop in d
13
C,
implying that lower precipitation occurred within the 4.2–
4.5 kyr
BP window. Palynological data from north-central
China suggest a cold, dry period from 3.95 to 4.45 kyr
BP,
which is inferred from a sudden decrease in tree pollen
concentration (Xiao and others, 2004). Morrill and others
(2003) compiled paleoclimatic data from monsoon Asia,
particularly China, to produce a pattern of abrupt mid-
Holocene monsoon failure in this region.
The discrepancy in timing between the onset of the
MHDE in Huascara´n and the carbonate peak in the Gulf of
Oman core (4.5 kyr
BP vs 4.2 kyr BP, respectively) may be the
result of the different techniques used to develop the
timescales for the two records. The marine record was
dated with the aid of
14
C dates from carbonates, whereas the
Holocene timescale for the ice core was based on flow
modeling and curve matching, using d
18
O
air
match points
with well-dated polar cores as previously described. This
allowed for the revised age of the MHDE, but with a range of
uncertainty between 5 years and 600 years. In fact, of all the
records shown in Figure 5, Huascara´n is the only one with a
timescale that was not developed using calibrated
14
C dates.
Despite timing differences, the dust event has the same
general shape in the high-resolution records from Huasca-
ra´n, the Gulf of Oman and Kilimanjaro, i.e. a large, abrupt
dust increase, then a more gradual decrease, followed by a
smaller, secondary peak. It is interesting that the character of
a climatic event should be so similar in records that come
from sites located in different environments on opposite
sides of the Earth. If these records are recording the same
event, by what physical processes are they linked?
THE MODERN LINKS BETWEEN THE MHDE IN
HUASCARA
´
N AND DROUGHT IN THE TROPICAL
EASTERN HEMISPHERE
During the austral summer (December–March), which is the
wet season in the Southern Hemisphere tropics, the Inter-
Tropical Convergence Zone is at its southernmost position
and the northeasterly trade winds move across the equatorial
Atlantic from west Africa to northeastern South America.
The dust that is entrained off the west coast of Africa by
frequent windstorms is caught up by the trade winds, which
are connected to the southern limb of the subtropical Azores
high-pressure system, which in turn is the southern dipole of
the North Atlantic Oscillation (NAO). Moulin and others
(1997) statistically linked the transport of dust from the west
African coast to the strength of the NAO. Much of the dust is
blown back towards the northeast by the clockwise motion
of the subtropical high, but a smaller amount is carried with
the moisture across the tropical North Atlantic, and in fact
Saharan dust which has been transported during the austral
summer has been found in the Amazon Basin (Swap and
others, 1992).
Since the source of Huascara´n’s precipitation is likely the
tropical Atlantic, it is probable that some of the dust that
originates in North Africa is carried by the moist air masses
across the Amazon Basin to the Cordillera Blanca and wet-
deposited on glaciers there (Davis, 2002). The recent
atmospheric processes that facilitate the movement of dust
between the desert regions of North Africa and tropical
South America are illustrated in Figure 6, which shows
correlation fields between the wet-season dust concen-
trations in the Huascara´n core and 58  58 grids of US
National Centers for Environmental Prediction/US National
Center for Atmospheric Research (NCEP/NCAR) re-analysis
data (Kalnay and others, 1996) from December to March of
global mean sea-level pressure (SLP; Fig. 6a) and global
850 mbar zonal winds (Fig. 6b) from 1949 to 1992, the last
full year of the ice-core record. High SLP is associated with
dry conditions, so the significant two-tailed Pearson
correlations (R > 0.3, significant at the 0.05 level) between
SLP in the Sahara, the Sahel, the Bode´le´ region of north-
central Africa (Fig. 6a) and ice-core dust concentration
suggest that aridity here may provide a potent potential
source region for the dust that eventually finds its way to the
Cordillera Blanca. Furthermore, the significant negative
correlations (R < –0.3, significant at the 0.05 level) between
Fig. 5. Comparisons of (a) the Gulf of Oman Middle Holocene
carbonate record on a calibrated
14
C timescale (modified from
Cullen and others, 2000); (b) the Huascara
´
n Holocene dust record;
(c) the dust from the Kilimanjaro ice core (modified from Thompson
and others, 2002); and (d–f) tropical African Holocene lake level
records (modified from Gasse, 2000) from Lake Abhe (Gasse, 1977)
(d), Ziway–Shala system (Gillespie and others, 1983) (e) and Bahr-
el-Ghazal (Servant and Servant-Vildary, 1980) (f). The grey bar
marks the Middle Holocene arid period in the records.
Davis and Thompson: Andean ice-core record of a Middle Holocene mega-drought38

the dust and lower-level easterly winds from North Africa to
the eastern coast of South America (Fig. 6b) indicate that
zonal wind velocities from the potential source tend to be
higher during high-dust years. The correlation field is
negative because easterly winds are recorded in the NCEP/
NCAR database as negative values (Fig. 6c). This relationship
ends just inside the northeast coast of Brazil, perhaps
because the relatively ‘linear’ trade-wind circulation over
the ocean is disrupted by moisture recycling and convective
activity over the Amazon rainforest.
Other regions of significant correlation are also observed
in Figure 6a, such as the negative relationship between
Huascara´n dust and SLP in the far North Atlantic (in the
region of the Icelandic low). When this is viewed in
conjunction with the significant +R field over North Africa,
the Mediterranean and the tropical Atlantic, it contributes to
the evidence that austral summers of high dust concentration
are coincident with boreal winters with positive NAO
indices. In addition, the SLP grid in the tropical western
Pacific shows positive significant correlations with the dust
concentrations, indicating negative Southern Oscillation
Indices (SOI), which are symptomatic of El Nin
˜
o conditions.
Thus, the easterly circulation across the tropical Atlantic
that is influenced by the subtropical node of the NAO
appears to be the link between the mineral dust in the
Huascara´n glacier and the climate in North Africa and the
Middle East. However, the interplay between the strength
and phase of the NAO and the circulation of the African/
Asian monsoon is connected with the climate of this
potential source region for Huascara´n dust. Today during
winters of high NAO indices, the Icelandic low and the
Azores high intensify and move northward, causing the
westerly storm tracks to shift from the Mediterranean region
and North Africa to northern Europe (Fig. 6b). For example,
low winter discharge from the Tigris and Euphrates Rivers is
significantly correlated with the positive phase of the NAO
(Cullen and deMenocal, 2000). Not only does this result in a
drier climate in southern Eurasia and southwest Asia, but the
vigorous northeasterly trade-wind circulation from the
Azores high transports more available dust across the
Atlantic to South America. Positive NAO is also associated
with a more intense Middle Eastern Jet Stream (MEJS) (Yang
and others, 2004), which is an antecedent signal for a
weaker Asian monsoon. When the MEJS is strong in the
winter, an ENSO-type warming appears in the equatorial
eastern Pacific (NINO3). Historically, increases in the
NINO3 SSTs have been significantly correlated with Asian/
Indian monsoon failure, although the correlation has
weakened in recent decades (Kumar and others, 1999).
The development of ENSO conditions during years of
positive NAO is also indicated in Figure 6, which shows
atmospheric circulation patterns indicative of negative SOI.
These observations are qualitative, since currently there is
no statistically significant year-to-year link that has been
noted between ENSO and NAO, although over decadal
timescales there does appear to be a relationship between
NAO and the ENSO-linked Pacific Decadal Oscillation
(PDO) (Readinger, 2003).
Possible teleconnections have been proposed between
the North Atlantic and the African/Asian monsoons during
the last glacial (Overpeck and others, 1996; Schulz and
others, 1998), during the glacial/Holocene transition (Sir-
ocko and others, 1996) and during the Holocene (Gupta and
others, 2003; Wang and others, 2005). Stated simply,
monsoon failures were contemporaneous with North
Atlantic cooling (associated with arctic ice discharge),
which in turn contributed to strengthened westerlies across
Eurasia. Comparisons between marine records from the
Arabian Sea and the North Atlantic show that these
conditions existed several times during the Holocene, and
one of the later episodes occurred between 4.0 and
4.5 kyr
BP (Gupta and others, 2003).
The MHDE therefore may have been connected to a
protracted interval of ENSO-linked African/Asian monsoon
weakening, which was coupled with strengthening of the
Icelandic low and Azores high. Cooling of the North Atlantic
accompanied an ice discharge event, and contributed to
more intense westerlies across northern Eurasia. Currently,
ENSO events are also linked to aridity in the Cordillera
Blanca, and if this was the case in the Middle Holocene, ice
recession on the Huascara´n peaks (which resulted in the
influx of large particles onto the col) may also have been
controlled by these conditions. Because the concentrations
of large dust particles in the MHDE increased before the
influx of the submicron particles, the ENSO conditions
responsible for aridity in the northeast Andes may have
Fig. 6. Significant correlation fields (–0.3 > R > +0.3, significant at
the 0.05 level) between Huascara´n wet-season (austral summer)
dust concentration and NCEP/NCAR re-analysis data (December–
March) for (a) SLP and (b) zonal wind velocity from 1949 to 1992.
Dark/light shading depicts fields of significant positive/negative
correlations. The correlation grids are 58 by 58. (c) A comparison
between the wet-season dust concentrations from 1949 to 1992
and the easterly wind velocities in the region of highest correlation
in north-central Africa illustrates why R is negative in the latitudes
dominated by easterlies.
Davis and Thompson: Andean ice-core record of a Middle Holocene mega-drought 39

preceded the monsoon weakening. In fact, the timing of the
interval of lake desiccation in tropical South America
suggests that it may have predated and postdated the Eastern
Hemisphere drought. The major elements of this near-
global-scale mechanism are illustrated in Figure 7, along
with the locations of some of the sites from where records of
this abrupt and prolonged aridity have been obtained.
CONCLUSIONS
The MHDE may have been the product of a protracted series
of ENSO events that possibly were associated with wide-
spread monsoon failures, which in turn may have been
linked with North Atlantic regional cooling and strength-
ened westerlies over Eurasia. The appearance of abrupt
climatic disruptions in such distant locations as the tropical
Andes, North Africa, the Middle East, southwest to east Asia
and the North Atlantic suggests strong linkages between
high- and low-latitude atmospheric processes. The questions
that remain to be addressed concern: (1) the underlying
cause for this near-global-scale episode, and (2) which
processes, i.e. those in the tropics (ENSO, monsoon
circulation) or those in the North Atlantic (ice rafting, ocean
cooling), might have been the triggers. The ultimate forcing
for abrupt climate change on orbital timescales is thought to
be non-linear responses (through oceanic, vegetative and
possibly cryospheric feedbacks) to linear insolation changes
(Ganopolski and others, 1998; deMenocal and others,
2000). In fact, a series of feedbacks between the low
latitudes and the arctic region may have sustained the
Middle Holocene drought in the tropics over several
hundred years.
ACKNOWLEDGEMENTS
We gratefully acknowledge the participants in the 1993
Huascara´n drilling expedition, B. Koci, V. Mikhalenko,
G. Seltzer, P. Kinder and Pin-Nan Lin, and the Peruvian
mountaineers F. Vicencio Maguina, M. Camones Gonzales
and M. Henostroga Zambrano. We also thank Pin-Nan Lin
for the d
18
O analysis, and K. Henderson whose efforts to
reconstruct the Huascara´n timescale were invaluable. The
Huascara´n program was funded by the US National Oceanic
and Atmospheric Administration. The comments of U. Ruth
and an anonymous reviewer greatly helped to improve this
paper and are very much appreciated. This is Byrd Polar
Research Center contribution No. 1322.
REFERENCES
Baker, P.A. and 8 others. 2001. The history of South American
tropical precipitation for the past 25,000 years. Science,
291(5504), 640–643.
Bar-Matthews, M., A. Ayalon, A. Kaufman and G.J. Wasserburg.
1999. The eastern Mediterranean palaeoclimate as a reflection
of regional events: Soreq cave, Israel. Earth Planet Sci. Lett., 166,
85–95.
Bard, E., M. Arnold, P. Maurice, J. Duprat, J. Moyes and
J.C. Duplessy. 1987. Retreat velocity of the North Atlantic polar
front during the last deglaciation determined by
14
C accelerator
mass spectrometry. Nature, 328, 791–794.
Bender, M. and 6 others. 1994a. Climate correlations between
Greenland and Antarctica during the past 100,000 years.
Nature, 372(6507), 663–666.
Bender, M., T.A. Sowers and L. Labeyrie. 1994b. The Dole effect
and its variations during the last 130,000 years as measured in
the Vostok ice core. Global Biogeochem. Cy., 8(3), 363–376.
Berger, A. and M.F. Loutre. 1991. Insolation values for the climate
of the last 10 million years. Quat. Sci. Rev., 10, 297–318.
Bolzan, J.F. 1985. Ice flow at the Dome C ice divide based on a
deep temperature profile. J. Geophys. Res., 90(D5), 8111–8124.
Butzer, K.W. 1997. Sociopolitical discontinuity in the Near East
c. 2200 BCE: scenarios from Palestine and Egypt. In Dalfes,
H.N., G. Kukla and H. Weiss, eds. Third millennium BC climate
change and Old World collapse. Berlin, etc., Springer, 245–296.
(NATO ASI Series 1.)
Cullen, H.M. and P.B. deMenocal. 2000. North Atlantic influence
on Tigris–Euphrates streamflow. Int. J. Climatol., 20(8), 853–863.
Cullen, H.M., P.B. deMenocal, S. Hemming and G. Hemming.
2000. Climate change and the collapse of the Akkadian empire:
evidence from the deep sea. Geology, 28(4), 379–382.
Fig. 7. Schematic illustrating global-scale mechanisms mentioned in the text that may have been influential in the Middle Holocene aridity
and its appearance in the Huascara
´
n ice core. Other sites where the event is documented in climate records are numbered: 1. Lake Titicaca
(Tapia and others, 2003); 2. Bahr-el-Ghazal (Servant and Servant-Vildary, 1980); 3. Lake Abhe (Gasse, 1977); 4. Ziway–Shala system
(Gillespie and others, 1983); 5. Kilimanjaro (Thompson and others, 2002); 6. Soreq cave (Bar-Matthews and others, 1999); 7. Gulf of Oman
(Cullen and others, 2000); 8. Indus delta (Staubwasser and others, 2003); and 9. Dongge cave (Wang and others, 2005). DJF: December–
February.
Davis and Thompson: Andean ice-core record of a Middle Holocene mega-drought40

Davis, M.E. 2002. Climatic interpretations of eolian dust records
from low-latitude, high-altitude ice cores. (PhD thesis, Ohio
State University.)
deMenocal, P. and 6 others. 2000. Abrupt onset and termination of
the African humid period: rapid climate responses to gradual
insolation forcing. Quat. Sci. Rev., 19, 347–361.
Fairbanks, R.G. 1989. A 17,000-year glacio-eustatic sea level
record: influence of glacial melting rates on the Younger Dryas
event and deep-ocean circulation. Nature, 342(6250), 637–642.
Ganopolski, A., C. Kubatzki, M. Claussen, V. Brovkin and
V. Petoukhov. 1998. The influence of vegetation–atmosphere–
ocean interaction on climate during the mid-Holocene. Science,
280, 1916–1919.
Gasse, F. 1977. Evolution of Lake Abhe´ (Ethiopia and T.F.A.I.) from
70,000 B.P. Nature, 265, 42–45.
Gasse, F. 2000. Hydrological changes in the African tropics since
the Last Glacial Maximum. Quat. Sci. Rev., 19(1–5), 189–211.
Gasse, F. and 12 others. 1991. A 13,000-year climate record from
western Tibet. Nature, 353, 742–745.
Gasse, F., J.C. Fontes, E. van Campo and K. Wei. 1996. Holocene
environmental changes in Bangong Co basin (Western Tibet).
Part 4: Discussion and conclusions. Palaeogeogr., Palaeoclima-
tol., Palaeoecol., 120(1), 79–92.
Gillespie, R., F.A. Street-Perrott and R. Switsur. 1983. Post-glacial
arid episodes in Ethiopia have implications for climate predic-
tion. Nature, 306, 680–683.
Grootes, P.M., M. Stuiver, L.G. Thompson and E. Mosley-
Thompson. 1989. Oxygen isotope changes in tropical ice,
Quelccaya, Peru. J. Geophys. Res., 94(D1), 1187–1194.
Gupta, A.K., D.M. Anderson and J.T. Overpeck. 2003. Abrupt
changes in the Asian southwest monsoon during the Holocene
and their links to the North Atlantic Ocean. Nature, 421(6921),
354–357.
Hassan, F.A. 1997. Nile floods and political disorder in early Egypt.
In Dalfes, H.N., G. Kukla and H. Weiss, eds. Third millennium
BC climate change and Old World collapse. New York, Springer,
1–24. (NATO ASI Series 1.)
Henderson, K.A. 1996. The El Nin
˜
o–Southern Oscillation and other
modes of tropical climate variability as recorded in ice cores
from the Nevado Huascara
´
n Col, Peru. (MS thesis, Ohio State
University.)
Johnsen, S.J. and 9 others. 1992. Irregular glacial interstadials
recorded in a new Greenland ice core. Nature, 359(6393),
311–313
Kalnay, E. and 21 others. 1996. The NCEP/NCAR 40-year reanalysis
project. Bull. Am. Meteorol. Soc., 77(3), 437–471.
Kumar, K.K., B. Rajagopalan and M.A. Cane. 1999. On the
weakening relationship between the Indian monsoon and
ENSO.
Science, 284(5423), 2156–2159.
Maslin, M.A. and S.J. Burns. 2000. Reconstruction of the Amazon
Basin effective moisture availability over the past 14,000 years.
Science, 290(5550), 2285–2287.
Morrill, C., J.T. Overpeck and J.E. Cole. 2003. A synthesis of abrupt
changes in the Asian summer monsoon since the last deglaci-
ation. Holocene, 13(4), 465–476.
Moulin, C., C.E. Lambert, F. Dulac and U. Dayan. 1997. Control of
atmospheric export of dust from North Africa by the North
Atlantic Oscillation. Nature, 387(6634), 691–694.
Overpeck, J.T., D.M. Anderson, S. Trumbore and W.L. Prell. 1996.
The southwest Indian monsoon over the last 18,000 years.
Climate Dyn., 12(3), 213–225.
Readinger, C.R. 2003. The North Atlantic and Pacific decadal
oscillations and their imprint on Greenland’s climate record.
(MS thesis, Ohio State University.)
Reeh, N. 1988. A flow-line model for calculating the surface profile
and the velocity, strain-rate, and stress fields in an ice sheet.
J. Glaciol., 34(116), 46–54.
Schulz, H., U. Rad and H. Erlenkeuser. 1998. Correlation between
Arabian Sea and Greenland climate oscillations of the past
110,000 years. Nature, 393(6680), 54–57.
Seltzer, G.O., D. Rodbell and S. Burns. 2000. Isotopic evidence for
late Quaternary climatic change in tropical South America.
Geology, 28(1), 35–38.
Servant, M. and S. Servant-Vildary. 1980. L’environnement quat-
ernaire du bassin du Tchad. In Williams, M.A.J. and H. Faure,
eds. The Sahara and the Nile. Rotterdam, A.A. Balkema,
133–162.
Sirocko, F., D. Garbe-Scho
¨
nberg, A. McIntyre and B. Molfino.
1996. Teleconnections between the subtropical monsoons and
high-latitude climates during the last deglaciation. Science,
272(5261), 526–529.
Sowers, T. and M. Bender. 1995. Climate records covering the last
deglaciation. Science, 269(5221), 210–214.
Sowers, T., M. Bender and D. Raynaud. 1989. Elemental and
isotopic composition of occluded O
2
and N
2
in polar ice.
J. Geophys. Res., 94(D4), 5137–5150.
Staubwasser, M., F. Sirocko, P.M. Grootes and M. Segl. 2003.
Climate change at the 4.2 ka BP termination of the Indus valley
civilization and Holocene south Asian monsoon variability.
Geophys. Res. Lett., 30(8), 1425. (10.1029/2002GL016822.)
Swap, R., M. Garstang, M. Greco, R. Talbot and P. Ka
˚
llberg. 1992.
Saharan dust in the Amazon basin. Tellus, 44B(2), 133–149.
Tapia, P.M., S.C. Fritz, P.A. Baker, G.O. Seltzer and R.B. Dunbar.
2003. A late Quaternary diatom record of tropical climatic
history from Lake Titicaca (Peru and Bolivia). Palaeogeogr.,
Palaeoclimatol., Palaeoecol., 194(1–3), 139–164.
Taylor, K.C. and 7 others. 1993. The ‘flickering switch’ of Late
Pleistocene climate change. Nature, 361(6411), 432–436.
Thompson, L.G. 2000. Ice core evidence for climate change in
the Tropics: implications for our future. Quat. Sci. Rev., 19(1–5),
19–35.
Thompson, L.G. and 7 others. 1995. Late glacial stage and
Holocene tropical ice core records from Huascara´n, Peru.
Science, 269(5220), 46–50.
Thompson, L.G. and 11 others. 1998. A 25,000-year tropical
climate history from Bolivian ice cores. Science, 282(5395),
1858–1864.
Thompson, L.G., E. Mosley-Thompson and K.A. Henderson.
2000. Ice-core palaeoclimate records in tropical South
America since the Last Glacial Maximum. J. Quat. Sci., 15(4),
377–394.
Thompson, L.G. and 10 others. 2002. Kilimanjaro ice core records:
evidence for Holocene climate change in tropical Africa.
Science, 298(5593), 589–593.
Wang, Y. and 9 others. 2005. The Holocene Asian monsoon: links
to solar changes and north Atlantic climate. Science, 308(5723),
854–857.
Weiss, H. and 6 others. 1993. The genesis and collapse of third
millennium north Mesopotamian civilization. Science,
261(5124), 995–1003.
Xiao, J.L., Q. Xu, T. Nakamura, X. Yang, W. Liang and Y. Inouchi.
2004. Holocene vegetation variation in the Daihai Lake region
of north-central China: a direct indication of the Asian monsoon
climatic history. Quat. Sci. Rev., 23(14–15), 1513–1722.
Yang, S., K.-M. Lau, S.-H. Yoo, J.L. Kinter, K. Miyakoda and
C.-H. Ho. 2004. Upstream subtropical signals preceding the
Asian summer monsoon circulation. J. Climate, 17(21),
4213–4229.
Davis and Thompson: Andean ice-core record of a Middle Holocene mega-drought 41