Content uploaded by Margit Schwikowski
Author content
All content in this area was uploaded by Margit Schwikowski
Content may be subject to copyright.
ENSO signals of the twentieth century in an ice core from Nevado
Illimani, Bolivia
S. Knu¨sel,
1
S. Bru¨tsch, K. A. Henderson, A. S. Palmer, and M. Schwikowski
Paul Scherrer Institute, Villigen, Switzerland
Received 3 September 2004; revised 5 November 2004; accepted 11 November 2004; published 5 January 2005.
[1]An ice core from Nevado Illimani in the eastern Bolivian Altiplano was used to
establish proxy data for past climate conditions. For this purpose a principal component
analysis (PCA) was performed on the records of ionic species of the upper 59.2 m (time
period 1887 ± 6 to 1999 A.D.). The first principal component (PC1) shows high loadings of
mainly dust-related ions. This signal is inversely correlated on a monthly basis with
precipitation on the Bolivian Altiplano. Maxima in the PC1 time series correspond to the
dry season on the Altiplano during the austral winter, whereas minima occur during the wet
season from November to March. Interannual variability of PC1 reveals correlations with
indices of the El Nin˜o–Southern Oscillation (ENSO), showing elevated (reduced) dust
values at the Nevado Illimani site during warm (cold) phases. This relationship may be
explained by below (above) average December–February precipitation in the Bolivian
Altiplano during El Nin˜o (La Nin˜a) episodes. Persistent El Nin˜o events such as those in the
years 1915, 1942, and 1993 A.D. seem to imprint very strong dust signals in the Illimani
ice core. Out of the six PCs extracted, only the PC1 time series is correlated with the
Tropical Southern Atlantic index. Thus we suggest that the major climate variability on the
Bolivian Altiplano as recorded in the Nevado Illimani ice core is more closely related to
conditions in the Pacific, even though the moisture source at this site is ultimately the
Atlantic Ocean.
Citation: Knu¨ sel, S., S. Bru¨tsch, K. A. Henderson, A. S. Palmer, and M. Schwikowski (2005), ENSO signals of the twentieth century
in an ice core from Nevado Illimani, Bolivia, J. Geophys. Res.,110 , D01102, doi:10.1029/2004JD005420.
1. Introduction
[2] There is a strong impact of the climate phenomenon
El Nin˜ o – Southern Oscillation (ENSO) on the South Amer-
ican continent. However, ENSO frequency and strength has
varied in the past, as shown by a number of studies. For
example, during the LGM, an El Nin˜ o –like pattern was
sustained in the tropical Pacific, as deduced from an eastern
equatorial Pacific sediment core [Koutavas et al., 2002].
Paleodata and models both have indicated that Pacific
conditions before 5000 years B.P. were more La Nin˜ a-like
[Bradley et al., 2003a; Koutavas et al., 2002]. Warm phases
of ENSO became more frequent during the Holocene until
1200 years B.P. and have since declined toward the present
[Moy et al., 2002]. However, over the last 30 years warm
phases of ENSO have again become more frequent, intense
and persistent compared to the preceding 100 years [IPCC,
2001]. ENSO variability was also found to be stronger
during the twentieth century as compared to previous glacial
and interglacial times [Tudhope et al., 2001]. Thus the
question has been posed whether the observed increased
ENSO activity is linked to global warming [Allan, 2000,
and references therein].
[3] However, various parts of South America experience
opposite effects of ENSO. Warm phases coincide with
flooding in Ecuador, northern and central Peru, whereas
droughts occur in southern Peru and Bolivia [Thompson et
al., 2000]. It is therefore challenging to disentangle different
teleconnections of ENSO in South America in the past and
to establish temporal changes. Thus different archives are
required in regions of South America that record the
opposing influences of ENSO. Additionally, using a variety
of archives helps in the recognition of local versus regional
sensitivity, and also in the detection of unique (and poten-
tially anomalous) effects in one single archive [Mann, 2002;
Trenberth and Otto-Bliesner, 2003].
[4] Glacial archives provide the advantage of highly
resolved and continuous records. To date, a few ice cores
have been retrieved from the tropics and subtropics of South
America. The interpretation of the d
18
O signal in these ice
cores has been controversial. It is not yet clear whether the
‘‘temperature effect’’ or the ‘‘amount effect’’ controls the
stable isotope signal at high-elevation tropical glaciers.
Most likely, d
18
O in these glaciers is not an indicator of
just one climatic variable, such as surface temperature or
precipitation amount [Vuille et al., 2003a]. Nevertheless,
recent studies showed that the d
18
O signal is correlated with
ENSO variability. Below average sea surface temperatures
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 110, D01102, doi:10.1029/2004JD005420, 2005
1
Also at Department of Chemistry and Biochemistry, University of
Bern, Bern, Switzerland.
Copyright 2005 by the American Geophysical Union.
0148-0227/05/2004JD005420$09.00
D01102 1of14
(SST) across the equatorial Pacific Ocean lead to higher
accumulation rates and more negative d
18
O and vice versa
[Bradley et al., 2003b; Hardy et al., 2003; Hoffmann et al.,
2003; Vuille et al., 2003a, 2003b]. Although the primary
moisture source for precipitation at these glacier sites is the
tropical Atlantic Ocean via the Amazon Basin, the moisture
flux seems to be remotely controlled by the conditions of
the tropical Pacific. The d
18
O records from low-latitude
glaciers in South America may thus provide a proxy for
Pacific climate variability [Vuille et al., 2003b].
[5] Additional evidence for a relationship between the
Pacific and the tropical Andes originates from increased net
ablation observed at Chacaltaya glacier and Zongo glacier
during El Nin˜ o events. During these episodes the tropical
Pacific anomaly is suggested to alter the mass balances of
both glaciers mainly by changes in precipitation, which is
significantly reduced at that time, and by a related decrease
of albedo [Francou et al., 2003; Wagnon et al., 2001]. A
similar picture was formed by the observation of snow line
fluctuations at Sajama, which were found to vary with
ENSO [Arnaud et al., 2001]. A major fraction of the
observed temperature increase in the tropical Andes, caus-
ing a widespread glacier retreat, could be traced to a
concurrent rise in SST in the equatorial Pacific [Vuille et
al., 2003c].
[6] However, ice cores contain more information than the
stable isotope records, namely, concentration records of
trace species. Concentration variations are mainly governed
by changes in emission source strength or atmospheric
circulation. The aim of this study is to investigate if the
records of ionic species can be used as proxy of atmospheric
circulation anomalies such as ENSO and can thus provide
complementary information to the stable isotope records.
For this purpose, ionic species were analyzed in the topmost
59.2 m of one of the two new Illimani ice cores from
Bolivia, recovered in 1999 [Correia et al., 2003; De Angelis
et al., 2003; Knu¨ sel et al., 2003].
[7] Whereas linking a time series of a particular major ion
to climate parameters may be difficult, attempts might prove
more successful for groups of species that are emitted by a
single source, or else transported concurrently from differ-
ent sources. Species with a common origin or transport path
are usually correlated in an ice core, which may necessitate
reducing the data set in such a way that combines those
species that share variance. Data reduction of this kind can
be performed using statistical techniques such as principal
component analysis (PCA) [Bourgeois et al., 2001;
Jackson, 1991; Jones et al., 1998; Kang et al., 2002b;
Souney et al., 2002]. In this study, PCA is applied to the
major ion record of the Illimani ice core in order to establish
proxies for atmospheric circulation anomalies related to
ENSO. The obtained PCs were investigated by correlation
analysis with various instrumental records.
2. Site Description
2.1. Illimani Ice Core
[8] Nevado Illimani (1639
0
S, 6747
0
W) is located on the
eastern side of the Bolivian Altiplano (see Figure 1). In
1999 A.D., two ice cores were drilled at an altitude of
6300 m asl. during a joint expedition involving scientists
from the Institut de Recherche pour le Developpement
(IRD) in France and the Paul Scherrer Institut (PSI) in
Switzerland. The ice cores reached bedrock at depths of
136.7 and 138.7 m, respectively, and were transported in a
frozen state to facilities at both institutes. Previous inves-
tigations on this glacial archive have included dating of the
ice cores, which cover approximately the last 18,000 years
[Knu¨ sel et al., 2003; Ramirez et al., 2003]. Volcanic signals
in the core were detected and analyzed, with special focus
on the layers recording the eruptions of Pinatubo (1991),
Agung (1963), and Tambora (1815) [De Angelis et al.,
2003]. A detailed interpretation of the stable isotope record
showed that in the twentieth century the decadal variability
in this signal is nearly identical for the four Andean ice
Figure 1. (a) Map of South America showing the location
of Nevado Illimani in Bolivia. Reanalysis data with a 5
5grid size covered the area from 76.25W, 26.25Sto
36.25W, 8.75N. The cross marks the location of the saline
lakes, while the white frame depicts the area displayed in
Figure 1b. (b) Map of Bolivia from 16–18S and 66–
69W. The location of Nevado Illimani is indicated together
with the surrounding meteorological stations (black
squares). The nine 0.50.5grids of precipitation data
obtained from the C. J. Willmott and K. Matsuura data set
are labeled in capital letters indicating their orientation
relative to Nevado Illimani. The altitude range is marked by
gray shadings starting with light gray at 2000–3000 m and
becoming darker with increasing altitude (1000 m steps).
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
2of14
D01102
cores available (Huascara´n, Quelccaya, Illimani, and
Sajama) [Hoffmann, 2003; Hoffmann et al., 2003]. The
long-term isotope record from Illimani strongly resembles
that from Huascara´n, indicating wetter/cooler conditions
during glacial times and drier/warmer conditions in the
Early Holocene [Ramirez et al., 2003]. For a number of
trace elements, especially for Cu, As, Zn, Cd, Co, Ni, and
Cr, an anthropogenic impact beginning at the onset of the
twentieth century was identified [Correia et al., 2003]. The
present study describes analyses performed on the 138.7 m
core.
[9] Englacial temperatures were 7C at 10 m depth in
the firn section of the glacier [Ginot et al., 2002], 8.7Cat
50 m, and 8.4C near bedrock [Zweifel, 2000; Ramirez et
al., 2003]. Thus alteration of the ion records by meltwater
percolation is presumed to be negligible. Postdepositional
effects were investigated by way of a surface snow sam-
pling study at the drilling site on Nevado Illimani. This
study (H. Bonnaveira, personal communication, 2003)
showed that despite the high daily sublimation rate of
0.7 to 1.2 mm water equivalents during austral winter
[e.g., Wagnon et al., 2003], the effect on the concentration
of ionic species seems to be weak in comparison to the
seasonal variations. These results confirmed an earlier
conclusion of only limited sublimation influence on the
concentrations of chemical species, deduced from a shallow
Illimani ice core [Ginot et al., 2002]. Thus the data set of
ionic species was used herein without any correction for
postdepositional effects.
2.2. Meteorological Conditions
[10] Climate conditions on the Altiplano are closely
related to the upper-air circulation. Easterly (westerly) zonal
flow favors wet (dry) conditions [Garreaud, 1999;
Garreaud and Aceituno, 2001; Garreaud et al., 2003;
Vuille, 1999]. On the Bolivian Altiplano, rainfall is largely
restricted to November to March [e.g., Johnson, 1976] and
is inherently episodic. Moisture at this latitude originates
from the Amazon Basin and ultimately from the tropical
Atlantic [Vuille et al., 2003b]. For the eastern Altiplano the
austral summer (December, January, February (DJF)) rain-
fall fraction is 50– 60%, with an annual mean precipitation
greater than 700 mm [Garreaud et al., 2003]. DJF-winds in
the eastern part at the 500 hPa level are from E-SE [Vuille,
1999]. Figure 2 illustrates wind directions at the grid
17.5S, 67.5W (center point) during the wet (November
to March) and the dry season (April to October) at the 500
and 600 hPa level.
[11] Dry conditions usually prevail on the Bolivian Alti-
plano during the austral winter (June, July, August (JJA)),
associated with enhanced westerly flow over the entire
region [Vuille, 1999]. The moisture transport from the east
(Amazon Basin) is then restricted to the eastern slope of the
Andes [Garreaud, 1999].
[12] ENSO influences the climate on the Bolivian Alti-
plano in the following manner. During El Nin˜o events, the
zonal wind component at the 200 hPa level over the
Altiplano shows westerly anomalies. These anomalies in-
hibit any significant moisture influx from the east, which
results in precipitation amounts below normal [Garreaud
and Aceituno, 2001; Garreaud et al., 2003; Vuille et al.,
2000]. In contrast, during La Nin˜a episodes, enhanced
upper-air easterlies lead to increased moisture transport
from the interior of the continent to the western Altiplano
[Garreaud, 1999; Vuille et al., 2000, 2001]. This relation-
ship of below average (above average) precipitation during
warm (cold) phases of ENSO is especially pronounced on
the western part of the Altiplano. Along the eastern margin
of the Altiplano, convection still occurs often during El
Nin˜ o events [Vuille et al., 2000, 2001], so that the relation-
ship between ENSO and precipitation here is much weaker.
3. Methods
3.1. Sample Preparation and Analyses
[13] The top 59.2 m of the Illimani ice core were cut in a
cold room at 20C using a modified band saw equipped
with a stainless steel blade and Teflon-covered tabletop and
saw guide. Effective decontamination was simultaneously
achieved, as the outer 1.5 cm of ice core was always removed
in the process. Analysis of the major ions HCOO
,Cl
,
NO
3
,SO
4
2
,C
2
O
4
2
,Na
+
,NH
4
+
, and K
+
was performed
using standard ion chromatography (IC) [Eichler, 2000],
whereas Mg
2+
and Ca
2+
were analyzed by inductively
coupled plasma optical emission spectrometry (ICP-OES).
The samples for ICP-OES were melted and 60 mLofa
solution containing 100 mg mL
1
Cs was added to stabilize
the signal. Sample resolution was roughly constant at 5 cm
down to a depth of 39.7 m, 3 cm between 39.7 and 55 m, and
2.5 cm thereafter. A procedural blank was tested regularly by
cutting and analyzing artificial ice core sections made of
ultrapure water (18.2 MWquality).
[14] After melting, the pH of each sample was deter-
mined using a Metrohm pH meter 605 (sample volume
1 mL, with addition of 10 mL 1 M KCl) equipped with an
8103 Orion Electrode, and was later converted to H
+
concentration. The error in the H
+
concentration was
estimated to ±3 mmol L
1
due to uncertainties in measuring
electrode potentials and the low ionic strength of the melted
sample.
[15] Primarily for the counting of annual layers as well as
the detection or confirmation of volcanic horizons, an
electrical conductivity method (ECM) was applied over
the entire 138.7 m ice core. For this purpose, a pair of
electrodes was drawn across a freshly cut flat surface
prepared from each individual ice core section. Additional
information on ECM and the subsequent data processing are
given in Knu¨ sel et al. [2003].
3.2. Dating and Reconstruction of Accumulation
[16] The ice core was dated using a multi-parameter
approach including counting of annual layers and the
identification of reference horizons supplied by known
volcanic eruptions and the 1964 A.D. tritium maximum
[e.g., Thompson et al., 1998]. Furthermore, this approach
involved dating using the radioactive decay of
210
Pb and
application of firn densification and ice-flow models
[Knu¨ sel et al., 2003]. The uncertainty of the dating was
estimated to ±1 year in the period from 1950– 1999 A.D. At
a depth of 59.2 m, assigned to the year 1887 A.D, the dating
error was ±6 years. This error originated from uncertainties
in the annual layer counting and in the time of arrival of
volcanic signals on Nevado Illimani. The dating applied
here is in agreement within the errors with the age-depth
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
3of14
D01102
relationship established for the other ice core from Illimani
[Correia et al., 2003; Hoffmann et al., 2003].
[17] To obtain a seasonally resolved age-depth relation-
ship, dating by annual layer counting of the five parameters
ECM, Na
+
,Ca
2+
, HCOO
and the dust record (PC1, see
below) was reconciled with the identified reference hori-
zons. These horizons included the tritium activity peak in
1964 A.D., and the volcanic signals attributed to Pinatubo,
1991 A.D., El Chicho´ n, 1982 A.D., Agung, 1963 A.D.,
Nilahue, 1955 A.D., and Tambora, 1815 A.D. [De Angelis
et al., 2003; Knu¨ sel et al., 2003]. According to the well-
known evolution of eruption material from Pinatubo [Trepte
et al., 1993], the arrival of this volcanic signal on Nevado
Illimaniwasascribedtotheaustralspring1991A.D.
Annual layers were assigned using the five parameters
(see above) down to 59.2 m, at which point the IC analysis
ended, leaving only the ECM data available between 59.2
and 73.2 m, the latter being the depth of the Tambora
horizon. In practice, a given annual layer was identified by
meeting the criterion of a minimum three of the five
parameters (as above) indicating clear maxima. Strongest
seasonality was observed in dust-related parameters such as
Ca
2+
and Na
+
. Therefore the maxima of Ca
2+
or Na
+
concentrations were attributed to the middle of the dry
period in June of each year (see section 2.2 and Figure 3).
The depth interval between peaks assigned to June was
divided into 12 bins to obtain a sub-seasonally resolved
record. These bins were determined according to the sea-
sonality of the accumulation expected at Nevado Illimani,
which is illustrated in Figure 3. This seasonality was
derived from the averaged seasonality of La Paz El Alto
(1900–1999 A.D.) by taking into account that the dry-
season accumulation represented 26% of the total yearly
accumulation in 2001/2002 A.D. [De Angelis et al., 2003].
In the following, the term binning is used for the time
linearization process explained above.
[18] The seasonally resolved dating, with peaks of major
ions assigned to June, allowed for the reconstruction of
yearly accumulation at the Nevado Illimani drilling site.
Annual accumulation in m water equivalent (m weq) was
calculated on the basis of June to May of the following year.
The reconstructed accumulation was corrected for layer
thinning in the ice below the firn-ice transition at 37 m.
This correction was performed by dividing the mean accu-
mulation of 0.58 m weq (obtained by the tritium reference
horizon) over the last 35 years (or 34.7 m), by the layer
thickness in any given year [Knu¨ sel et al., 2003]. The series
of these ratios for every year was smoothed by a 20-year
Figure 2. Normalized wind directions at the grid 17.5S, 67.5W (center point) during the wet and the
dry season and at 500 hPa and 600 hPa. These wind directions were derived from reanalysis data over the
time period from 1968 to 2001 A.D.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
4of14
D01102
running mean, thereby accounting only for the mean thin-
ning and not for the year-to-year variations. Each annual
layer was then multiplied by this smoothed ratio to obtain
the reconstructed annual accumulation.
3.3. Principal Component Analysis and Data
Processing
[19] In a temporal sense, core sampling was not uniform
due to a combination of thinning and the stepwise-adjusted
depth resolution (see section 3.1). The number of samples
per year decreased from 20 at the surface to 10 at
59.2 m. As described above, the time series was interpolated
to 12 points per year. Data gaps where fractured ice did not
allow for chemical analysis were removed prior to the PCA.
Gaps occurred at depths of 39.95–40.28 m, 54.52– 54.64 m,
55.0–55.21 m, 55.34– 55.85 m, 57.86 – 58.01 m, and
58.35–58.54 m.
[20] PCA was performed on this interpolated and then
standardized data set, covering the period from 1887 – 1999
A.D., for 11 water-soluble ionic species and 1294 data
points. According to the rules of the 95% limit and the
SCREE-plot [Jackson, 1991], seven PCs should have been
retained. However, only six PCs were interpreted, because
the seventh explained less than 3% of the total variability. It
is important to note that the observations in the time series
are not independent and therefore only allow application of
the PCA for exploratory use [Jackson, 1991]. The robust-
ness of the PCA was tested by selecting every second data
point, performing a PCA on this reduced data set, and then
comparing the results to those obtained by the entire data
set. Good agreement was observed, leading to the conclu-
sion that the PCA performed on this data set was robust.
[21] Additional data processing involved the calculation
of monthly and quarterly standardized anomalies as well as
annual means for the various PCs. Quarterly values had the
advantage of a better time assignment since they are less
influenced by year-to-year variations in the seasonal pre-
cipitation distribution compared to monthly data. In addi-
tion, the quarterly resolved record, with four times the
number of observations as compared to annual values,
results as well in a lower criterion for satisfying significance
tests. Missing values in the time series were replaced prior
to the averaging by the mean monthly values over the time
period available (i.e., from 1888 to 1998 A.D.). Correlation
analyses between PCs and meteorological data (see below)
were performed on a monthly, quarterly and yearly basis.
Anomaly data as well as the yearly values were used for
correlation with climate data. The probability level of these
correlations was 0.05, unless specified otherwise. An aver-
aged seasonality of each principal component was calculated
over the entire record (112 years, see Figure 4).
3.4. Meteorological Data and Climate Indices
[22] Monthly precipitation data from meteorological sta-
tions in Bolivia were retrieved from the Global Historical
Climatology Network (GHCN) [Vose et al., 1992]. The
precipitation data of La Paz El Alto, Patacamaya, and
Chacaltaya were received from SENAMHI (Servicio Nacio-
nal de Meteorologia e Hidrologia). The stations selected
were located in the region around Nevado Illimani between
16–17.5S and 66–68.5W with altitudes ranging be-
tween 2000 and 5300 m asl. (see Figure 1). Before summing
monthly precipitation data to obtain yearly values, missing
data were replaced by the monthly averages over the time
period available. Only 3% of the data set was comprised of
missing values.
[23] Additional monthly precipitation data were obtained
from the terrestrial air temperature and precipitation data set
of Willmott and Matsuura [2001]. This data set was com-
piled from the GHCN and Legates and Willmott [1990]
station records of monthly precipitation data, which were
interpolated onto 0.50.5latitude/longitude grids for the
Figure 3. Seasonal cycle of the precipitation at La Paz El
Alto from 1900 to 1999 A.D (gray bars) and best estimate
of the seasonality at Nevado Illimani (black line) used for
the precipitation-adapted binning. Note that the rainfall is
largely restricted to November to March and that the
transition into the dry season is more abrupt than the gradual
change to the wet period.
Figure 4. Seasonality of the PCs 1–6 averaged from 1888
to 1998 A.D. Note the sharp increase of PC1 toward June
and the gradual decrease toward the end of the year.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
5of14
D01102
time period from 1950– 1999 A.D. Precipitation data from
the nine grids bounded by 16–17.5S and 67– 68.5W
were selected for this study to provide additional data from
regions with sparse coverage by meteorological stations.
These grids are labeled here according to their orientation
relative to the location of Nevado Illimani (see Figure 1).
The grid 16.5–17S, 67.5– 68W containing Nevado Illim-
ani is labeled ‘‘Illimani grid’’.
[24] NCEP Reanalysis data of northern South America
were provided by the NOAA-CIRES Climate Diagnostics
Center, Boulder, Colorado, United States, from their Web
site at http://www.cdc.noaa.gov. The area covered by this
data included 55grids from 76.25W, 26.25Sto
36.25W, 8. 7 5 N and comprised the time period from
1968–2001 A.D. The parameters selected were air temper-
ature, geopotential height, specific humidity and zonal wind
(u-wind) at the 500 hPa level. Standardized anomalies were
calculated for all indices. To create the wind-rose diagrams,
zonal and meridional wind data were retrieved at both
500 hPa and 600 hPa level for the grid 17.5S, 67.5W
(center point) from 1968– 2001 A.D. (see Figure 2).
[25] Climate indices related to ENSO employed here
included the Southern Oscillation Index (SOI) and the sea
surface temperature anomalies (SSTA) in the regions 0–
10S, 80–90W (NINO1/2), 5S–5N, 150–90W
(NINO3), and 5S–5N, 170–120W (NINO3/4) [Kaplan
et al.,1998].Thesedatawereobtainedwithmonthly
resolution. Data gaps in these indices were linearly inter-
polated. The duration of the El Nin˜o events referred herein
was determined as follows. First, the SOI was smoothed by
a five-sample running mean (weighted 1-3-4-3-1). Values of
the smoothed SOI less than 0.5 were then interpreted as
belonging to a particular El Nin˜o event. El Nin˜o episodes
separated by intermediate periods of just one or two
consecutive months with higher values were still considered
to be one single event. Pacific Decadal Oscillation (PDO)
data, obtained from the leading principal component of the
North Pacific Ocean, were from Mantua et al. [1997]. Other
indices investigated were the Tropical Southern Atlantic
(TSA) index, which represents the SSTA in the region 0–
20S and 10E–30W and the Tropical Northern Atlantic
Index (TNA), an analogous SSTA in the region of 5.5–
23.5N and 15–57.5W[Enfield et al., 1999]. Monthly
data of these Atlantic indices were available in the time
period from 1951 to 2000 A.D. at http://www.cdc.noaa.gov/
ClimateIndices/Analysis.
[26] Processing of climate data involved the calculation
of quarterly and annual averages. For precipitation data, the
respective sums were used instead of averages.
4. Results and Interpretation of the Principal
Components
[27] The loadings of the six PCs from the entire data set
of major ions are given in Table 1.
4.1. Illimani Dust Record: PC1
[28] The characteristics of PC1 are the high loadings of
HCOO
,Cl
,SO
4
2
,Na
+
,NH
4
+
,K
+
,Ca
2+
, and Mg
2+
(see
Table 1). This PC explains 50% of the total variance.
[29] Snow sampling studies at the drilling site on Nevado
Illimani showed that there is a strong seasonality in the snow
composition with low concentrations of ionic species during
the wet season (November – March) and high concentrations
during the dry season (May– August). Ions such as NH
4
+
,
SO
4
2
,Ca
2+
,Na
+
, and Cl
were enriched during the dry
season compared to the wet season (H. Bonnaveira, personal
communication, 2003). These ions all show high loadings in
PC1. The Ca
2+
flux was calculated for the dry and the wet
season, employing the reconstructed accumulation values
(as above) split by season as defined by the Ca
2+
median
(0.94 meq L
1
). The dry season (Ca
2+
> median) accounts
for 91% of the total annual flux of Ca
2+
at Nevado Illimani.
Thus Ca
2+
as a tracer for dust shows a highly variable flux
throughout the year, which is responsible for the observed
seasonality in the Illimani Ca
2+
record. Following the snow
sampling study and the consideration of fluxes, peaks in the
PC1 time series were aligned with the dry season. Species
generally associated with dust are SO
4
2
,Cl
,Na
+
,K
+
,Ca
2+
,
and Mg
2+
; however, HCOO
and NH
4
+
also show associated
high loadings in the case of Illimani core. This covariation
may be explained given that PC1 is a mixture of local dust,
saline lake constituents (contributing Na
+
and Cl
), and
biogenic emissions (accounting for HCOO
, and K
+
). These
aerosols might all be enriched during the dry season. The
average seasonality (from 1888 – 1998 A.D.) of the PC1
time series is illustrated in Figure 4.
[30] The PC1 time series, used in the following as a proxy
for dust, is compared to the reconstructed accumulation on
Nevado Illimani (see Figure 5). There is no clear relation-
ship visible. At some places with low accumulation, high
dust values are observed, e.g., in the years 1915–1916 A.D.
or 1938–1939 A.D., but low dust values are also seen to
occur (1961– 1966 A.D.). On a yearly basis (June – May
averages) the accumulation record only correlates with the
TNA (0.29, 1950–1998 A.D.).
[31] The Illimani dust record was further compared with
the precipitation data from nearby meteorological stations.
Figure 3 illustrates the seasonal cycle of the precipitation
measured at the meteorological station in La Paz El Alto
(16.52S, 68.18W, 4050 m asl.), averaged over the period
from 1900–1999 A.D. Precipitation mainly occurs from
November–March and shows a minimum during the austral
winter (dry period). As already stated, the start of the dry
period is generally abrupt compared with the gradual
transition into the wet season, which is visible in Figure 3.
Table 1. Loadings of the PCA of Ionic Species in the Illimani Ice
Core Including the Variance Explained by Each Component
a
Ion PC1 PC2 PC3 PC4 PC5 PC6
HCOO
0.72 0.27 0.38 0.07 0.36 0.29
Cl
0.79 0.41 0.15 0.17 0.12 0.04
NO
3
0.15 0.95 0.14 0.07 0.04 0.06
SO
4
2
0.77 0.00 0.44 0.22 0.27 0.00
C
2
O
4
2
0.53 0.16 0.67 0.34 0.31 0.19
Na
+
0.86 0.01 0.07 0.15 0.34 0.21
NH
4
+
0.81 0.11 0.09 0.22 0.17 0.44
K
+
0.89 0.02 0.11 0.11 0.12 0.19
Ca
2+
0.83 0.12 0.02 0.46 0.19 0.08
Mg
2+
0.79 0.14 0.04 0.54 0.09 0.03
H
+
0.09 0.96 0.14 0.08 0.02 0.06
Variance explained
50% 20% 7.9% 7.3% 4.7% 3.7%
a
Italics indicate the species discussed in the text.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
6of14
D01102
The dust record shows a similar behavior (Figure 4, PC1
and Figure 6a). Dust values increase rapidly in the transition
from the wet to the dry period, and after several months
begin to gradually decline again toward the end of the year.
The 99-year record of precipitation in La Paz El Alto is
correlated with PC1 with a correlation coefficient of 0.35,
which is significant on a 0.0001 probability level (1900–
1998, monthly values, see Figure 6). Thus, on a monthly
basis, the dust record is inversely correlated with precipita-
tion on the Altiplano, recording both the abrupt start of the
dry period and the gradual decline into the wet season.
[32] Interannual variability (correlations using yearly val-
ues) related to precipitation illustrates that out of all selected
meteorological stations (see Figure 1), only the precipitation
Figure 5. Reconstructed annual accumulation (gray dashed line, left scale) in the Illimani ice core and
PC1 (black, right scale). Both accumulation and PC1 were averaged from June to May to obtain yearly
values. PC1 represents the dust record (see text) with high dust values often coincident with low
accumulation.
Figure 6. (a) Precipitation data from grid S (dark gray line with dots) and the meteorological station in
La Paz El Alto (gray dashed line, both right scale). The monthly resolved dust record (PC1) of the
Illimani ice core from 1940 to 1960 A.D is shown in black (left scale). (b) Yearly averaged precipitation
data from grid S (dark gray line with dots) and the meteorological station at Patacamaya (gray dashed
line, both right scale). The dark gray curve (bottom) depicts the monthly resolved PC1 time series with
yearly averages in black. Note that the yearly averages of PC1 here are from January to December.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
7of14
D01102
data of the southern meteorological stations Patacamaya,
1238, and 1243 inversely correlate with the Illimani dust
record. This relationship is illustrated in Figure 6b, which
compares the annually (January–December) averaged
dust (PC1) record to the total yearly precipitation from
Patacamaya. Precipitation data from C. J. Willmott and
K. Matsuura (see section 3.4) at the grids S and SW are
correlated only on a 0.1 probability level. Interestingly,
precipitation data of the Illimani grid do not correlate with
PC1. These findings suggest that the precipitation regime in
the area south of Nevado Illimani influences the dust record,
i.e., less precipitation in this area is related to higher values
in the PC1 time series.
[33] Additional investigations (correlations using monthly
and quarterly standardized anomalies and yearly values)
performed to deduce interannual variations, revealed rela-
tionships between PC1 and ENSO indices such as SOI or
NINO3, and also with the PDO index. Figure 7 shows the
annually averaged PC1 together with the annual means of
the SOI and NINO3 time series. Correlation analyses
between PC1 and the ENSO indices were performed over
the well-dated period from 1950– 1998 A.D, yielding cor-
relation coefficients of 0.35 and 0.36 for SOI and NINO3,
respectively. Also monthly and quarterly anomalies of PC1
showed significant correlation with ENSO indices (r: 0.19
for SOI and 0.16 for NINO3, 1950 – 1997 A.D., quarterly
values). Therefore monthly and quarterly anomalies as well
as annual averages of the PC1 dust record are considered to
be proxies for ENSO on the Altiplano. The relationship of
increased (reduced) dust values indicated by PC1 and warm
(cold) phases of ENSO can be explained by changes in
precipitation patterns during these events. During warm
phases of ENSO, precipitation in primarily the western,
and also in the southern parts of the Bolivian Altiplano is
reduced [Vuille et al., 2000, 2001]. This scenario may lead
to increased dust loading in the air and can imprint a higher
dust value (PC1) in the ice either by way of the precipitation
that remains, and/or by dry deposition. Correlation is
observed between ENSO and grid S precipitation (on a
0.1 level) as well as between grid S and PC1, which might
explain why an ENSO signal is observed here in an ice core
from the eastern Andes. Thus the ENSO signal is initially
expressed in the precipitation amount to the south of
Nevado Illimani, and is suggested to affect Nevado Illimani
in the form of dust transport from this region. This mech-
anism can explain the increased (decreased) dust values
during warm (cold) phases of ENSO.
[34] From studies by Vuille et al. [2000], it is known that
the western part of the Altiplano experienced unexpected
dry La Nin˜ a conditions in the austral summer of 1964/1965
A.D. and wet El Nin˜ o conditions in 1972/1973 and 1976/
1977 A.D. In contrast, on the eastern part of the Altiplano a
single wet El Nin˜ o was observed in 1986/1987 A.D. The
dust record on Nevado Illimani (see Figure 7) shows high
levels during the years 1965, 1973, and 1977, and 1987
A.D. The fact that wet (dry) El Nin˜ o (La Nin˜a) events did
not imprint a signal in PC1 may be explained by ENSO
peaking in varying seasons, e.g., early in 1976 A.D.
Therefore the influence of ENSO on DJF precipitation
could have varied from one event to the next [Vuille et
al., 2000], whereas the dust record seems to be less sensitive
to the timing of ENSO maxima/minima. Another reason
why the Illimani dust record fails to reverse the relationship
with precipitation during certain unusual dry La Nin˜a and
wet El Nin˜ o events may also be that the dust originates
mainly from the south, as indicated above.
[35] The dust record illustrated in Figure 7 shows very
high values in the years 1915, 1942, and 1993 A.D. A closer
examination of the ENSO indices during periods with very
high PC1 values reveals persistent El Nin˜ o events, e.g., in
the years around 1912, 1941, and 1992 A.D. with durations
of 32, 31, and 52 months, respectively (duration estimated
according to section 3.4). This correspondence can be
distinguished in Figure 8, which compares the seasonality
of PC1 with that of the SOI. Thus the years involving long
warm phases of ENSO correspond to those with high PC1/
Figure 7. Comparison of the yearly (January– December) averages of PC1 (black line) with the ENSO
indices SOI (gray, inverse yaxis) and NINO3 (dark gray). The dating error is ±6 years at 1887 A.D. (see
text).
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
8of14
D01102
dust values, and this relationship may suggest that the
Illimani dust record is especially sensitive to persistent El
Nin˜ o events.
[36] The interannual variability as recorded by PC1 is
also related to conditions in the tropical southern Atlantic.
Monthly anomalies and yearly values of TSA and PC1 are
correlated. However, the quarterly anomalies correlate only
on a 0.15 level with TSA. Thus we reason that there is a
relationship between TSA and PC1, however, PC1 is more
strongly related to conditions in the Pacific.
[37] To obtain a more regional picture of the connections
of PC1, correlation analysis was performed for this compo-
nent and the reanalysis data at 500 hPa over northern South
America. PC1 correlates with specific humidity in the
region of the Amazon Basin and southern Brazil (see
Figure 9). The relationship of above-normal dust and
specific humidity values may be explained by an enhanced
rainout over the Amazon Basin, leading to reduced advec-
tion of moist air to the Bolivian Altiplano. This drier climate
allows dust transport from more local dust sources. The PC1
time series also correlates with the temperature at 500 hPa in
the region north of about 1S and west of about 61W (see
Figure 9). Above-average temperature is associated with
above-average dust concentration. This may be an indirect
result of the fact that generally a dry (wet) climate coincides
with warm (cold) conditions and that temperature over the
South American Altiplano follows the conditions in the
Pacific [e.g., Garreaud and Aceituno, 2001; Vuille et al.,
2003b].
4.2. Nitric Acid Puzzle: PC2
[38] PC2 shows high loadings of H
+
and NO
3
and
explains 20% of the total variance. This component essen-
tially represents the contribution from nitric acid. Several
possibilities for the origin of the unusual HNO
3
concen-
trations are discussed below. During the core analysis,
extremely low pH values were observed in several samples
of the 59.2 m long section investigated, the lowest value
being 3.3. By reanalyzing one particular depth range with
anomalously low pH (by cutting new samples from the
remaining core) the corresponding high NO
3
and H
+
concentrations in the record were confirmed. Thus contam-
ination during the sample preparation and analysis can be
excluded.
[39] One hypothesis for the existence of these unusual
high HNO
3
concentrations is that this ice core may have
been contaminated with nitric acid during its transport from
Nevado Illimani to the cold room in Switzerland. Such an
argument could be offered based on the fact that the 136.7 m
core, drilled only 10 m apart from the 138.7 m core, did not
reveal such acidity peaks. Additionally, these unusual peaks
were only observed in the gas-permeable firn portion
(density < 0.82 g cm
3
) of the longer core but not in the
ice. A gas such as NO or NO
2
could be suggested to cause
this contamination.
[40] However, there are also arguments against a contam-
ination hypothesis. There is a gradual, not a stepwise,
transition of the NO
3
or H
+
concentration across successive
core sections that were individually packed and transported.
One example is depicted in Figure 10, showing the conti-
nuity of a large acidity peak between 1991 and 1992 A.D.
This peak covers about two entire ice core sections. Sec-
ondly, the structure of the NO
3
peaks within these zones of
anomalously high concentration still resembles the annual
maxima that were counted down to the 1964 A.D. tritium
reference horizon (34.7 m depth). An age of 1961 ±4 years
Figure 8. (a) Plot of monthly dust values (PC1) during the period from 1888 to 1998 A.D. High dust
values are observed in the years around 1915, 1942, and 1993 A.D. (b) Plot of the SOI in the same period
as Figure 8a. In the years with high dust values, El Nin˜o events with long duration occurred, indicated by
negative SOI indices. Note that the SOI was linearly interpolated from December 1906 to August 1908 as
well as from August 1931 to August 1932. For better visibility, the data are smoothed along the year axis.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
9of14
D01102
A.D. was obtained by NO
3
layer-counting, which is in good
agreement with this reference horizon. Thirdly, if gaseous
diffusion of NO or NO
2
were important, then this diffusion
would be yet more efficient within more porous firn layers,
but especially not in ice lenses. Nevertheless, several of
these unusual high H
+
and NO
3
concentrations do occur
within ice lenses, as illustrated in Figure 10.
[41] Known sources for HNO
3
include the oxidation of
various nitrogen trace gases, originating from lightning or
combustion processes, and NH
3
oxidation. On a monthly
basis, the PC2 time series correlates only with the 14-year
precipitation record at station 1978.
[42] Interannual variability (correlations using monthly
and quarterly standardized anomalies and yearly averages)
of PC2 is related to conditions in the Pacific. PC2 time
series correlates with the ENSO index NINO3 (r: 0.29,
1950–1997 A.D., yearly values), however, not with other
ENSO indices. On a monthly basis (monthly anomaly data)
all ENSO indices correlate with PC2. The number of
correlations of PC2 and ENSO indices is reduced to
SOI, NINO3, and NINO3/4, when applying quarterly
anomaly data. The overall relationship indicates that the
PC2 signal is stronger (weaker) during warm (cold) phases
of ENSO, which may be explained in a similar way as for
PC1, i.e., by way of higher aerosol concentrations in the
drier atmosphere.
[43] Correlation analysis of PC2 with reanalysis data over
northern South America reveals correlation with specific
humidity data north of Illimani, along the border of Bolivia
and Brazil, and in central Brazil. Furthermore, correlation of
PC2 and zonal wind is significant in the region north of
Nevado Illimani and in central Brazil.
4.3. Biogenic Component: PC3
[44] The PC3 component has high loadings of HCOO
and C
2
O
4
2
, explaining 7.9% of the total variance. The
annual cycle of the PC3 time series peaks in March and
July (see Figure 4). The increase toward the March maxima
is sharp, whereas there is a gradual transition of the July
peak toward the end of the year.
[45] On a monthly basis the PC3 time series correlates
inversely with precipitation signals from several climate
stations such as La Paz El Alto (r: 0.08, 1900 – 1998 A.D.)
and Chacaltaya (r: 0.11, 1952–1998 A.D). The biogenic
input on Nevado Illimani is suggested to be indirectly
related to the amount of precipitation. During the dry
season, the biogenic contribution is maximal, probably
due to air being ‘‘saturated’’ with aerosols by the low
precipitation amount.
[46] Interannual variations of this component are related
to the PDO index. However, no correlation with the ENSO
indices is found. This PC may respond to general conditions
in the Pacific.
[47] Correlation analysis of PC3 with the reanalysis
data over northern South America yielded inverse corre-
lation between PC3 and specific humidity in the Amazon
Basin and central Brazil, as illustrated in Figure 11.
Interestingly this figure shows approximately the opposite
pattern of Figure 9, which displayed correlation between
PC1 and specific humidity. Above-average values of PC3
correspond to below-average specific humidity at those
areas. The origin of HCOO
and C
2
O
4
2
is suggested to
lay in the Amazon Basin and in central Brazil. Low
humidity values in these areas probably lead to a reduced
rainout and an increased concentration of those species in
the air which is then advected westward. The humidity in
the source regions has a direct influence on the source
strength of PC3 species. This is not the case for PC1
species, since the dust emissions of the Amazon Basin
seem to be negligible.
4.4. A Mixed Component: PC4
[48] High loadings of Mg
2+
and Ca
2+
are observed in
PC4, which explains 7.3% of the total variance. The peak
maximum of the annual cycle occurs in March (see
Figure 4), whereas the minimum is observed in August
and September. During the austral autumn E-S winds
Figure 9. (top) Correlations of PC1 and specific humidity
(500 hPa) and (bottom) correlations of PC1 and air tempe-
rature (500 hPa) over northern South America using quarterly
anomaly data for the time period from 1968 to 1998 A.D.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
10 of 14
D01102
prevail on the Altiplano, while precipitation declines toward
austral winter [e.g., Johnson, 1976]. Therefore PC4 is
believed to originate from the region E-S of Nevado
Illimani. Apart from PC1, the seasonality of PC4 is notice-
ably strong and shows a minimum during the austral winter,
which may be explained by the prevailing westerly winds
during that season. Suggested sources for these ions include
the Altiplano with several large saline lakes, which could
contribute to the Mg
2+
,Ca
2+
concentrations through erosion
of soil or salt.
[49] On a monthly basis, PC4 is correlated with precip-
itation data of all meteorological stations as well as with the
precipitation of the grids around Illimani. For example, PC4
correlates with the precipitation record at Patacamaya with a
correlation coefficient of 0.29 (1948– 1993 A.D.). High
values for Mg
2+
and Ca
2+
occur whenever abundant pre-
cipitation is observed.
[50] Interannual variability of PC4 is connected to the
PDO. During high values of PDO, below-normal Mg
2+
and
Ca
2+
concentrations are observed.
[51] The reanalysis data over northern South America
form the following picture. Inverse correlation of PC4 and
specific humidity is highest in the region east of Nevado
Illimani (Figure 11). Above-average concentrations of Mg
2+
and Ca
2+
are related to below-average specific humidity in
the Amazon Basin. A possible explanation of this relation-
ship is a reduced rainout activity leading to an increased
PC4 signal.
4.5. PC5
[52] PC5 shows high loadings of HCOO
,Cl
, and Na
+
.
The annual cycle of this signal shows two distinct maxima
in April and August; thus the signal arrives at Nevado
Illimani during austral autumn, and late winter to early
spring. Sources suggested for PC5 are biogenic emissions
and the saline lakes. No precipitation data are correlated
with PC5. Interannual variability seems to be connected
to PDO. Inverse correlation between PC5 and PDO is
observed.
[53] The PC5 time series correlates with specific humidity
values in the region south/southeast of Nevado Illimani,
whereas inverse correlation is observed in the Amazon
Basin and central Brazil (see Figure 11). The inverse
correlation of PC5 and specific humidity in Brazil may be
explained in a similar way as for PC3 and PC4, i.e., by way
of reduced rainout leading to enhanced values at Nevado
Illimani.
4.6. A Wet Season Signal: PC6
[54] This signal shows high loadings of Na
+
,K
+
and
C
2
O
4
2
and arrives on Nevado Illimani during December to
April in the wet period with the prevailing E-S winds (see
Figure 4). This PC explains 3.7% of the total variance.
Sources of Na
+
and K
+
are soil dust and sea salt, whereas
C
2
O
4
2
is a tracer of biogenic emissions. For K
+
biogenic
emissions as well as biomass burning are also proposed as
possible sources [Artaxo et al., 1998; Echalar et al., 1998;
Khalil and Rasmussen, 2003]. Thus the effective aerosol
source(s) of PC6 cannot be attributed.
[55] On a monthly basis, PC6 correlates with precipitation
at all meteorological stations. Therefore high K
+
,Na
+
and
C
2
O
4
2
concentrations are observed during rainy periods at
these stations. Increased biogenic activity during summer
could be the reason for the elevated K
+
and C
2
O
4
2
concentrations. Interannual variability of PC6 was not
related to any of the climate indices applied here.
5. Concluding Remarks
[56] Principal component analysis performed on the ionic
species data set has proven to be a suitable tool for the
search of paleoclimate proxies in the Nevado Illimani ice
core. The results of this analysis may be helpful in the future
interpretation of longer records from that site and in forming
Figure 10. NO
3
(thick black line) and H
+
(thick dark gray dashed line) concentrations from 1989 to
1995 A.D. Ice lenses are indicated by vertical bars shaded in light gray, whereas the junctions between
individually packed ice core sections are represented by black vertical lines. Note that there are no
stratigraphic data given between 1993 and 1994 A.D.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
11 of 14
D01102
a general picture of the climate variability registered by the
Nevado Illimani ice core. Investigations in a similar way
were also performed among others by Kang et al. [2002a] and
Souney et al. [2002] giving insight into climate variability
recorded by ice cores. PCs extracted from ionic records may
provide complementary information to the stable isotope
records and more general information about climate variabil-
ity compared with the accumulation record. Accumulation
reconstruction is sensitive to a correct assignment of annual
layers and itself is subject to local influence by wind erosion,
etc. In contrast, aerosol species arrive at Nevado Illimani with
air masses of different origin and can potentially elucidate
connections over longer distances.
[57] The main results from this investigation emerge from
PC1 as a dust proxy record, registering the seasonal
variations in the precipitation on the Altiplano. Interannual
variability of PC1 is related to various climate indices. Dust
values are correlated with precipitation data south of
Nevado Illimani. PC1 anomaly data and yearly averages
are correlated with ENSO. During El Nin˜o (La Nin˜a)
events, the values of dust in the Illimani ice core are
increased (reduced), probably due to drier (wetter) condi-
tions primarily in the western part of the Altiplano [Vuille et
al., 2001]. In particular, it was deduced that persistent El
Nin˜ o events seem to imprint a strong dust signal in the
Illimani ice core. This response is most strongly expressed
during the long and persistent El Nin˜ o event around 1992
A.D. Thus PC1, explaining about half of the variability
observed in the ionic record of the Illimani ice core, may be
used as a proxy for the influence of ENSO on the Altiplano.
Of all the principal components, only the PC1 time series
correlates with the Atlantic index TSA. Thus we suggest
that the climate variability on the Bolivian Altiplano as
recorded in the Illimani ice core is more closely related to
conditions in the Pacific, as opposed to the Atlantic, even
though the moisture source for Nevado Illimani is generally
the Atlantic Ocean. This finding is consistent with conclu-
sions of Bradley et al. [2003b], Garreaud et al. [2003], and
Vuille et al. [2003b, 2003c].
[58] It is worth considering that the dust signal PC1 as an
ENSO proxy is observed through teleconnections. The
below (above) average precipitation on the Bolivian Alti-
plano is only teleconnected to the warm (cold) SST of the
Pacific, and the dust record itself is also solely indirectly
linked to this precipitation. Thus the dust signal of the entire
Illimani ice core, which covers about 18,000 years [Ramirez
et al., 2003], needs careful investigation in view of possible
changes to the teleconnections over time, in order to
distinguish between the variability of ENSO itself and the
influence of ENSO on the Altiplano.
[59]Acknowledgments. We thank Bernard Francou, Robert Gallaire,
Patrick Ginot, Bernard Pouyaud, Ulrich Schotterer, Felix Stampfli, and
Beni Zweifel for their contribution to the drilling operation. We are grateful
to Rolf Keil and Silvia Ko¨chli for technical assistance and Michael
Mosimann and Michael Mayer of the Statistical Service (University of
Bern) for their support. Also highly appreciated is the effort of Mathias
Vuille and Martin Grosjean for supplying meteorological data as well as
various climate indices. Finally, we acknowledge the valuable comments of
two anonymous reviewers.
References
Allan, R. J. (2000), ENSO and climatic variability in the past 150 years, in
El Nin˜o and the Southern Oscillation, Multiscale Variability and Global
Figure 11. Correlations of (top) PC3, (middle) PC4, and
(bottom) PC5 with specific humidity (500 hPa) over
northern South America using quarterly anomaly data for
the time period from 1968 to 1998 A.D.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
12 of 14
D01102
and Regional Impacts, edited by H. F. Diaz and V. Markgraf, pp. 3 – 55,
Cambridge Univ. Press, New York.
Arnaud, Y., F. Muller, M. Vuille, and P. Ribstein (2001), El Nino – Southern
Oscillation (ENSO) influence on a Sajama volcano glacier (Bolivia) from
1963 to 1998 as seen from Landsat data and aerial photography,
J. Geophys. Res.,106(D16), 17,773– 17,784.
Artaxo, P., E. T. Fernandes, J. V. Martins, M. A. Yamasoe, P. V. Hobbs,
W. Maenhaut, K. M. Longo, and A. Castanho (1998), Large-scale aerosol
source apportionment in Amazonia, J. Geophys. Res.,103(D24),
31,837 – 31,847.
Bourgeois, J. C., K. Gajewski, and R. M. Koerner (2001), Spatial patterns
of pollen deposition in arctic snow, J. Geophys. Res.,106(D6), 5255–
5265.
Bradley, R. S., K. R. Briffa, J. Cole, M. K. Hughes, and T. J. Osborn
(2003a), The climate of the last millennium, in Paleoclimate, Global
Change and the Future, edited by K. Alverson, R. S. Bradley, and
T. F. Pederson, pp. 105 – 149, Springer, New York.
Bradley, R. S., M. Vuille, D. Hardy, and L. G. Thompson (2003b), Low
latitude ice cores record Pacific sea surface temperatures, Geophys. Res.
Lett.,30(4), 1174, doi:10.1029/2002GL016546.
Correia, A., R. Freydier, R. J. Delmas, J. C. Simoes, J. D. Taupin, B. Dupre,
and P. Artaxo (2003), Trace elements in South America aerosol during
20th century inferred from a Nevado Illimani ice core, Eastern Bolivian
Andes (6350 m asl), Atmos. Chem. Phys.,3, 1337 – 1352.
De Angelis, M., J. Simoes, H. Bonnaveira, J. D. Taupin, and R. J.
Delmas (2003), Volcanic eruptions recorded in the Illimani ice core
(Bolivia): 1918 – 1998 and Tambora periods, Atmos. Chem. Phys.,3,
1725 – 1741.
Echalar, F., P. Artaxo, J. V. Martins, M. Yamasoe, F. Gerab, W. Maenhaut,
and B. Holben (1998), Long-term monitoring of atmospheric aerosols in
the Amazon Basin: Source identification and apportionment, J. Geophys.
Res.,103(D24), 31,849– 31,864.
Eichler, A. (2000), Deposition von Spurenstoffen in Firn und Eis alpiner
Gletscher, Untersuchung von Nord-Su¨d-Gradienten, Ph.D. thesis, Dep. of
Chem. and Biochem., Univ. of Bern, Bern, Switzerland.
Enfield, D. B., A. M. Mestas-Nunez, D. A. Mayer, and L. Cid-Serrano
(1999), How ubiquitous is the dipole relationship in tropical Atlantic
sea surface temperatures?, J. Geophys. Res.,104(C4), 7841 – 7848.
Francou, B., M. Vuille, P. Wagnon, J. Mendoza, and J. E. Sicart (2003),
Tropical climate change recorded by a glacier in the central Andes
during the last decades of the twentieth century: Chacaltaya, Bolivia,
16S, J. Geophys. Res.,108(D5), 4154, doi:10.1029/2002JD002959.
Garreaud, R. D. (1999), Multiscale analysis of the summertime preci-
pitation over the central Andes, Mon. Weather Rev.,127(5), 901 –
921.
Garreaud, R. D., and P. Aceituno (2001), Interannual rainfall variability
over the South American Altiplano, J. Clim.,14(12), 2779– 2789.
Garreaud, R. D., M. Vuille, and A. C. Clement (2003), The climate of the
Altiplano: Observed current conditions and mechanisms of past changes,
Paleogeogr. Paleoclimatol. Paleoecol.,194(1 – 3), 5 – 22.
Ginot, P., M. Schwikowski, U. Schotterer, W. Stichler, H. W. Ga¨ggeler,
B. Francou, R. Gallaire, and B. Pouyaud (2002), Potential for climate
variability reconstruction from Andean glaciochemical records, Ann. Gla-
ciol.,35, 443– 450.
Hardy, D. R., M. Vuille, and R. S. Bradley (2003), Variability of snow
accumulation and isotopic composition on Nevado Sajama, Bolivia,
J. Geophys. Res.,108(D22), 4693, doi:10.1029/2003JD003623.
Hoffmann, G. (2003), Taking the pulse of the tropical water cycle, Science,
301(5634), 776– 777.
Hoffmann, G., et al. (2003), Coherent isotope history of Andean ice cores
over the last century, Geophys. Res. Lett.,30(4), 1179, doi:10.1029/
2002GL014870.
Intergovernmental Panel on Climate Change (IPCC) (2001), Climate
Change 2001: Summary for Policymakers, Cambridge Univ. Press,
New York.
Jackson, J. E. (1991), A User’s Guide to Principal Components,John
Wiley, Hoboken, N. J.
Johnson, A. M. (1976), The climate of Peru, Bolivia and Ecuador, in
Climates of Central and South America, edited by W. Schwerdtfeger,
pp. 147– 218, Elsevier, New York.
Jones, P. D., K. R. Briffa, T. P. Barnett, and S. F. B. Tett (1998), High-
resolution palaeoclimatic records for the last millennium: Interpretation,
integration and comparison with General Circulation Model control-run
temperatures, Holocene,8(4), 455– 471.
Kang, S., P. A. Mayewski, D. Qin, Y. Yan, S. Hou, D. Zhang, J. Ren, and
K. Kruetz (2002a), Glaciochemical records from a Mt. Everest ice core:
Relationship to atmospheric circulation over Asia, Atmos. Environ.,
36(21), 3351– 3361.
Kang, S. C., P. A. Mayewski, D. H. Qin, Y. P. Yan, D. Q. Zhang, S. G. Hou,
and J. W. Ren (2002b), Twentieth century increase of atmospheric am-
monia recorded in Mount Everest ice core, J. Geophys. Res.,107(D20),
4595, doi:10.1029/2001JD001413.
Kaplan, A., M. A. Cane, Y. Kushnir, A. C. Clement, M. B. Blumenthal, and
B. Rajagopalan (1998), Analyses of global sea surface temperature
1856 – 1991, J. Geophys. Res.,103(C9), 18,567 – 18,589.
Khalil, M. A. K., and R. A. Rasmussen (2003), Tracers of wood smoke,
Atmos. Environ.,37(9– 10), 1211 – 1222.
Knu¨sel, S., P. Ginot, U. Schotterer, M. Schwikowski, H. W. Ga¨ggeler,
B. Francou, J. R. Petit, J. C. Simo˜es, and J. D. Taupin (2003), Dating
of two nearby ice cores from the Illimani, Bolivia, J. Geophys. Res.,
108(D6), 4181, doi:10.1029/2001JD002028.
Koutavas, A., J. Lynch-Stieglitz, T. M. Marchitto, and J. P. Sachs (2002), El
Nino-like pattern in ice age tropical Pacific sea surface temperature,
Science,297(5579), 226– 230.
Legates, D. R., and C. J. Willmott (1990), Mean seasonal and spatial
variability in gauge-corrected, global precipitation, Int. J. Climatol.,
10(2), 111 – 127.
Mann, M. E. (2002), The value of multiple proxies, Science,297(5586),
1481 – 1482.
Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis
(1997), A Pacific interdecadal climate oscillation with impacts on salmon
production, Bull. Am. Meteorol. Soc.,78(6), 1069 – 1079.
Moy, C. M., G. O. Seltzer, D. T. Rodbell, and D. M. Anderson (2002),
Variability of El Nin˜o/Southern Oscillation activity at millennial time-
scales during the Holocene epoch, Nature,420(6912), 162– 165.
Ramirez, E., et al. (2003), A new Andean deep ice core from
Nevado Illimani (6350 m), Bolivia, Earth Planet. Sci. Lett.,212,
337 – 350.
Souney, J. M., P. A. Mayewski, I. D. Goodwin, L. D. Meeker, V. Morgan,
M. A. J. Curran, T. D. van Ommen, and A. S. Palmer (2002), A 700-year
record of atmospheric circulation developed from the Law Dome ice
core, East Antarctica, J. Geophys. Res.,107(D22), 4608, doi:10.1029/
2002JD002104.
Thompson, L. G., et al. (1998), A 25,000-year tropical climate history from
Bolivian ice cores, Science,282(5395), 1858– 1864.
Thompson, L. G., K. A. Henderson, E. Mosley Thompson, and P. N. Lin
(2000), The tropical ice core record of ENSO, in El Nino and the South-
ern Oscillation, Multiscale Variability and Global and Regional Impacts,
edited by H. F. Diaz and V. Markgraf, pp. 325 – 356, Cambridge Univ.
Press, New York.
Trenberth, K. E., and B. L. Otto-Bliesner (2003), Toward integrated recon-
struction of past climates, Science,300(5619), 589– 591.
Trepte, C. R., R. E. Veiga, and M. P. McCormick (1993), The poleward
dispersal of Mount-Pinatubo volcanic aerosol, J. Geophys. Res.,98(D10),
18,563 – 18,573.
Tudhope, A. W., C. P. Chilcott, M. T. McCulloch, E. R. Cook, J. Chappell,
R. M. Ellam, D. W. Lea, J. M. Lough, and G. B. Shimmield (2001),
Variability in the El Nin˜ o – Southern Oscillation through a glacial-inter-
glacial cycle, Science,291(5508), 1511 – 1517.
Vose, R. S., R. L. Schmoyer, P. M. Steurer, T. C. Peterson, R. Heim,
T. R. Karl, and J. Eischeid (1992), The Global Historical Climatology
Network: Long-term monthly temperature, precipitation, sea level
pressure, and station pressure data, Rep. ORNL/CDIAC-53, NDP-
041, Carbon Dioxide Inf. Anal. Cent., Oak Ridge Natl. Lab., Oak
Ridge, Tenn.
Vuille, M. (1999), Atmospheric circulation over the Bolivian Altiplano
during dry and wet periods and extreme phases of the Southern Oscilla-
tion, Int. J. Climatol.,19(14), 1579– 1600.
Vuille, M., R. S. Bradley, and F. Keimig (2000), Interannual climate varia-
bility in the Central Andes and its relation to tropical Pacific and Atlantic
forcing, J. Geophys. Res.,105(D10), 12,447– 12,460.
Vuille, M., D. R. Hardy, C. Braun, F. Keimig, and R. S. Bradley (2001),
Climate variability on intraseasonal to interannual timescales on the Bo-
livian Altiplano with special emphasis on the Nevado Sajama region,
Ecol. Bolivia,35, 17– 40.
Vuille, M., R. S. Bradley, M. Werner, R. Healy, and F. Keimig (2003a),
Modeling d
18
O in precipitation over the tropical Americas: 1. Interannual
variability and climatic controls, J. Geophys. Res.,108(D6), 4174,
doi:10.1029/2001JD002038.
Vuille, M., R. S. Bradley, R. Healy, M. Werner, D. R. Hardy, L. G.
Thompson, and F. Keimig (2003b), Modeling d
18
O in precipitation
over the tropical Americas: 2. Simulation of the stable isotope signal
in Andean ice cores, J. Geophys. Res.,108(D6), 4175, doi:10.1029/
2001JD002039.
Vuille, M., R. S. Bradley, M. Werner, and F. Keimig (2003c), 20th century
climate change in the tropical Andes: Observations and model results,
Clim. Change,59(1 – 2), 75– 99.
Wagnon, P., P. Ribstein, B. Francou, and J. E. Sicart (2001), Anomalous
heat and mass budget of Glaciar Zongo, Bolivia, during the 1997/98 El
Nin˜o year, J. Glaciol.,47, 21 – 28.
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
13 of 14
D01102
Wagnon, P., J. E. Sicart, E. Berthier, and J. P. Chazarin (2003), Wintertime
high-altitude surface energy balance of a Bolivian glacier, Illimani,
6340 m above sea level, J. Geophys. Res.,108(D6), 4177, doi:10.1029/
2002JD002088.
Willmott, C. J., and K. Matsuura (2001), Terrestrial air temperature
and precipitation: Monthly and annual time series (1950 – 1999),
Version 1.02, Cent. for Clim. Res., Dep. of Geogr., Univ. of Del.,
Newark.
Zweifel, B. (2000), Kernbohrungen in kalten Gletschern (Ice core drilling in
cold glaciers), diploma thesis, Eidg. Techn. Hochsch. Zu¨rich, Zu¨rich,
Switzerland.
S. Bru¨tsch, K. A. Henderson, S. Knu¨sel, A. S. Palmer, and
M. Schwikowski, Paul Scherrer Institute, CH-5232 Villigen PSI,
Switzerland. (margit.schwikowski@psi.ch)
D01102 KNU
¨SEL ET AL.: ENSO SIGNALS IN THE ILLIMANI ICE CORE
14 of 14
D01102
