Research Article

Annually Resolved Ice Core Records of Tropical Climate Variability over the Past ~1800 Years

See allHide authors and affiliations

Science  24 May 2013:
Vol. 340, Issue 6135, pp. 945-950
DOI: 10.1126/science.1234210

Quelccaya Ice Cap

Ice cores drilled in the ice sheets of Greenland and Antarctica are some of the most important sources of information about the paleoclimate of high latitudes. Comparable records from the tropics are rare, however, because there are so few locations at which long-lived, undisturbed ice can be found. Thompson et al. (p. 945, published online 4 April) report results obtained from one of the few such sites, the Quelccaya ice cap in the Peruvian Andes. The annually resolved data, extending back 1800 years, provide a detailed chronicle of changes in the isotopic composition of the oxygen in the ice, which are related to the sea surface temperature of the water's source. Analyses of a collection of major ions such as ammonium and nitrate reveal how atmospheric circulation in the region varied over that period. Finally, the radiocarbon content of ancient plants—recently exposed by the retreat of the ice sheet—reveals that Quelccaya has not been smaller for at least six thousand years.


Ice cores from low latitudes can provide a wealth of unique information about past climate in the tropics, but they are difficult to recover and few exist. Here, we report annually resolved ice core records from the Quelccaya ice cap (5670 meters above sea level) in Peru that extend back ~1800 years and provide a high-resolution record of climate variability there. Oxygen isotopic ratios (δ18O) are linked to sea surface temperatures in the tropical eastern Pacific, whereas concentrations of ammonium and nitrate document the dominant role played by the migration of the Intertropical Convergence Zone in the region of the tropical Andes. Quelccaya continues to retreat and thin. Radiocarbon dates on wetland plants exposed along its retreating margins indicate that it has not been smaller for at least six millennia.

Ice cores from high-altitude tropical glaciers offer long-term perspectives on the variability of precipitation, temperature, aridity, and atmospheric and sea surface conditions at low latitudes. Most meteorological and climatic disturbances affecting Earth’s surface and lower atmosphere originate in or are amplified by ocean-atmosphere interactions in tropical latitudes. As Earth’s “heat engine,” the warmest atmospheric and sea surface temperatures (SSTs) occur there. This energy drives intense convective precipitation and is crucial for the evolution of phenomena such as El Niño–Southern Oscillation (ENSO), the monsoonal systems of Asia and Africa, and, on intraannual time scales, hurricanes and other tropical disturbances that distribute equatorial heat energy poleward. ENSO dominates tropical climate variability. It is linked to the position of the Intertropical Convergence Zone (ITCZ), and its associated teleconnections affect the strength and direction of air masses and storm tracks, variations in convective activity that control flooding and drought, and modulation of tropical storm intensities. There are few high mountain glaciers in these regions and even fewer that preserve detailed histories of this variability. Unfortunately, most are now rapidly shrinking, and unique records are being lost. The potential impacts on water resources have social and economic consequences that underpin the imperative to understand the drivers and responses of past and present tropical climate variability.

The first ice cores drilled from the Quelccaya ice cap (QIC) [13°56′S, 70°50′W, 5670 m above sea level (masl)], in 1983 (13), could not be returned frozen to the laboratory and instead were cut into samples that were melted and bottled in the field. In 2003, two additional cores were drilled to bedrock on the QIC (Fig. 1). The Summit Dome (5670 masl) core (QSD, 168.68 m) and the North Dome (5600 masl) core (QND, 128.57 m) were returned frozen and are stored at –30°C at Ohio State University’s Byrd Polar Research Center. Minimal postdepositional reworking of the snow surface, even during the wet season, results in the distinct annual layers (fig. S1A) used to reconstruct an ~1800-year climate history. Details about the construction of the time scale, extracting annually resolved information, and reconstructing the net annual accumulation are provided in the supplementary text. The oxygen isotopic ratio (δ18O) records for the 2003 QSD and QND cores, separated by 1.92 km, are highly correlated (r = 0.898, P < 0.0001 for decadal averages; table S1). In light of their similarity, documented in the supplementary text, all subsequent discussions are based on the QSD core.

Fig. 1 Location of the QIC, Peru, and other ice fields and features discussed in the text.

Also included are the upper (500 hPa) and lower (850 hPa) level atmospheric circulation in the austral summer [December-January-February (DJF)] (38).

These records (Fig. 2 and fig. S2), based on freshly cut ice samples, are precisely dated and include δ18O and concentrations of insoluble dust and major anions (F, Cl, SO42–, and NO3) and cations (Na+, NH4+, K+, Mg2+, and Ca2+). The reproducibility of decadal averages of δ18O (1983 Core 1 versus the 2003 QSD core) from 1000 to 1982 CE (fig. S3A) is excellent (r = 0.856, P < 0.0001). Details of the reproducibility of δ18O and net accumulation among the four cores, two in 1983 and two in 2003, over the past 1000 years are in the supplementary text (fig. S3, A and B, and table S1).

Fig. 2 Decadal averages of δ18O, net accumulation, insoluble dust, ammonium, and nitrate in the QSD ice core.

Specific climatological periods discussed in the text are shaded and identified. The asterisk on the dust profile indicates the 1600 CE eruption of Huaynaputina (Peru). ppb indicates parts per billion; m w.e. a–1 indicates meters of water equivalent per year.

The Record

Figure 2 presents the decadal averages of δ18O, net accumulation, insoluble dust, ammonium (NH4+), and nitrate (NO3) measured in the QSD core. Decadal averages of all other species are shown in fig. S2. Because annual resolution is preserved to ~160 m (683 CE) (supplementary materials), the annual net accumulation (An) record terminates there. These histories detail changes in climatic and environmental conditions in the tropical Andes over the past 18 centuries.

Four distinct climatic stages are evident. Table S2 presents the average values of all QSD core constituents for the different time periods discussed in the paper. Before 1100 CE, most constituents show little variability, although accumulation and insoluble dust concentrations are slightly higher and decline slowly toward 1100 CE. Simultaneously, δ18O becomes modestly depleted in 18O.

The Medieval Climate Anomaly (MCA) is characterized by below-average An, consistent with hydroclimate reconstructions for this region (4), and δ18O is more variable and modestly enriched [0.26 per mil (‰)] relative to the long-term average (–17.92‰) (table S2). MCA ice contains no visible evidence of surface melting or smoothing of the δ18O annual signal that, as discussed below, characterizes the QIC core since 1991. For just 4 decades near the end of the MCA, the QSD core contains high levels of ammonium (NH4+) and nitrate (NO3) that are not contemporaneous with δ18O evidence of strong warming or any notable increase in An. During the MCA, the climate across South America appears highly variable, possibly reflecting the observed dipole in humidity over the northern and southern Amazon Basin (5). The result is that wetter conditions in the northeast tend to be contemporaneous with drier conditions in southern Amazonia. For example, a high-resolution record of El Niño flooding from a marine core off the coast of Peru at 12°S indicates intense aridity from 800 to 1250 CE (6), whereas the Cariaco Basin (Venezuela) titanium record (7) reflects wetter-than-average conditions in the northeast Amazon Basin from 950 to 1450 CE.

The most prominent feature in the QIC record is the Little Ice Age (LIA). Early in the 16th century, concentrations of ammonium, nitrate, calcium, magnesium, and sulfate began to increase (Fig. 2, fig. S2, and table S2). The period from ~1520 to 1880, identified as the LIA, is characterized by lower δ18O values that remained low until the late 19th century. The net accumulation trend is distinctive. In ~1520 CE, An increased rapidly but declined abruptly in ~1680 CE from record highs in the Early LIA (1520 to 1680 CE) to record lows (~1800 CE) in the Late LIA (1681 to 1880 CE). Over much of the LIA, concentrations of most ionic species were persistently high but declined rapidly near the end as An began to increase.

With the onset of the Current Warm Period (CWP, 1880 CE to the present), both δ18O and An increased, whereas all aerosol concentrations have remained low over much of the CWP. Over the past 30 years, most aerosol concentrations have increased severalfold primarily because of postdepositional surface melting and percolation through the firn pack. The firn-ice transition is at 17.7 m and corresponds to 1992; however, isotopic smoothing in the top of the record indicates that, since the onset of the melting and retreat of the ice cap in ~1991, meltwater has percolated down to the 1980 level [see figure 4 in (8)].

Linkages with the Tropical Pacific

The seasonal temperature range in the tropics is only a few degrees, whereas the seasonal differences in the δ18O of Andean snowfall are much larger, often up to 20‰. Because 70 to 80% of the precipitation falling on Quelccaya arrives during the wet season, the δ18O history reflects primarily conditions during austral summer. Decadal averages of precipitation amount and δ18O are not strongly related (fig. S3). Twenty-one-year running correlations between δ18O and An alternate between positive and negative values over multiple decades (fig. S4), and very few coefficients are significant at the 95% level. Essentially, there is no consistent, long-term statistically significant relationship between the amount of precipitation and its δ18O signature. In general, low-latitude isotopic ratios yield a climate signal reflecting a variety of hydrologic and thermal influences in the broad geographic region that supplies moisture to the high glaciated mountains (9).

Although the moisture source for Quelccaya precipitation is primarily from the tropical Atlantic via the Amazon Basin, Vuille et al. (10) demonstrated that Pacific SSTs exert the dominant control on interannual δ18O variability preserved in tropical Andean ice cores. This occurs via the expansion of the tropical troposphere associated with a warm tropical Pacific and enhanced westerly flow over the tropical Andes or via the shrinkage of the tropical troposphere associated with a cool tropical Pacific, and enhanced easterly flow over the subtropical Andes. These upper-level wind anomalies force the low-level moisture flow over the Andes and thereby link oceanic forcing and climate variability on interannual time scales.

Bradley et al. (11) demonstrated that δ18O in the Sajama (Bolivia) ice core is more closely linked to SSTs, and hence to ENSO variability, across the equatorial Pacific Ocean even though the moisture source for precipitation is the Atlantic. A δ18O composite of three Peruvian ice cores and one Himalayan ice core from 1856 to 1996 CE [see figure 3 in (8)] is strongly linked (r = 0.73; P < 0.0001, for 5-year moving averages) to the NINO4 extended reconstructed SSTs [ERSST (12)]. This strong relationship between ice core isotopic records throughout the tropics and tropical SSTs likely reflects the dominance of tropical evaporation in determining water vapor flux into the atmosphere (13). This also provides a likely explanation of the large-scale isotopic links among low-latitude, high-altitude ice core records (14).

Figure 3A illustrates the spatial distribution of the correlations between the annual QSD δ18O and ERSST records between 80°N and 80°S from 1870 to 2009 CE (15, 16). The 2003 ice core record was extended to 2009 by using subsequently collected pit and shallow core samples. The annual δ18O values of QIC precipitation are positively related to equatorial SSTs in the mid to eastern tropical Pacific Basin (solid line encloses P < 0.001). Correlations between the thermal-year (July to June) averages of δ18O in the QSD core and ERSSTs in the NINO4, NINO3.4, and NINO3 regions were examined. The NINO4 SSTs are most strongly related to the QSD δ18O (r = 0.55 for annual data; r = 0.61 for 3-year averages; P < 0.0001).

Fig. 3 QSD δ18O and SST comparisons.

(A) Spatial distribution of correlations between the annual QSD δ18O and ERSST records [(15, 16) and] between 30°N and 30°S from 1870 to 2009 CE. (B) Three-year running means of the thermal-year averages of QSD δ18O and SSTs in the NINO4 region [(12) and] from 1871 to 2009 CE. (C) ~1800 year SST history reconstructed by using the QSD δ18O/SST regression equation (fig. S5) for data in (B).

Figure 3B shows the 3-year running means of the NINO4 SST index (12) and QSD δ18O for thermal years 1870-1871 to 2009-2010. Given their strong correlation (r = 0.61, P < 0.0001; fig. S5), the resulting SST-δ18O transfer function (SSTNINO4 = 0.19 × δ18O + 29.64) is used to reconstruct a contemporaneous SST history for the NINO4 region (Fig. 3C) for the past 18 centuries. The reconstructed SSTs range between 25.8° and 26.6°C (decadal averages) or 25.0° to 27.1°C (annual data) with temperatures in the region depressed by 0.2°C during the LIA (1520 to 1880 CE) relative to the 20th-century average.

This reconstruction assumes that the QSD δ18O–NINO4 SST relationship has been stationary over the past 1800 years, which is not the nature of climate processes. The spatial distribution of the most highly correlated fields is not stationary through time (fig. S6). From near the end of the LIA (1870 CE) to 1900 CE, the enriched values of QSD δ18O are strongly related to higher SSTs in the eastern Pacific region along the equator and extending southward. Note the strong positive correlation in the equatorial Atlantic at this time as well. Over the next 50 years (1901 to 1950 CE), the δ18O-SST correlations weaken, and the region of strongest positive correlations moves northward. In the past ~60 years (1951 to 2009 CE) during which anthropogenic forcing has increased (17), the region of strongest correlation increases in coherence and migrates north of the equator. The movement of the fields of significant δ18O-SST correlation into the Northern Hemisphere over the CWP likely reflects the post-LIA northward migration of the ITCZ over the tropical Pacific Basin, consistent with other evidence, including lake sediments from central equatorial Pacific islands (18).

A comprehensive review of various proxy indicators from diverse tropical sites (19) suggests a widespread southward migration of the ITCZ over the Holocene (their figure 12-5) affecting the Pacific, Indian, and Atlantic basins. In general, wetter conditions in the northern tropics gave way to more arid conditions in the late Holocene, whereas the southern tropics experienced the reverse trend. This plus evidence from coral records from the tropical Pacific (20) indicate that the movement of the ITCZ was coherent between the Atlantic and Pacific oceans (19). The climatological controls on snow accumulation and the isotopic chemistry of precipitation on the QIC are likely to change with the position of the ITCZ. Thus, the QSD records should provide additional details of the ITCZ migration.

ITCZ Migration and Tropical Linkages

As discussed above, the tropical eastern Pacific SSTs and atmospheric circulation influence the stable isotopic composition of the precipitation falling on Quelccaya, as well as on other central Andean glaciers (11, 21). However, much of the precipitation falling on the eastern side of the Peruvian Andes is produced by deep convection in the tropical Amazon Basin during the austral summer. Much of the water vapor originates in the tropical Atlantic, as it has since the Last Glacial Maximum. This is confirmed by mountain snow lines that were tilted toward the Amazon Basin then as they are now (22).

Elevated concentrations of nitrate (NO3) and ammonium (NH4+) and 18O depleted isotopes in the QSD during the LIA are contemporaneous with decreasing percentages of Ti in the Cariaco Basin record off the coast of Venezuela at 10°N (Fig. 4). Reduced Ti concentrations indicate decreased precipitation and runoff from the northeast coast of tropical South America (7). Conversely, elevated NO3 and NH4+ likely reflect more moist conditions to the south over the Amazon Basin. These aerosol species originate in part as products from ammonia (NH3), a gas produced by biological activity in the soil and vegetation in the Amazon Basin (23, 24). NO3 and NH4+ are correlated over the entire QSD core (r = 0.76, P < 0.0001 for decadal averages, from 230 to 1990 CE) and covary most strongly (r = 0.92, P < 0.0001) over the LIA. QSD NH4+ and therefore NO3 concentrations are significantly correlated with precipitable water in the southwest Amazon Basin from 1949 to 2002 (Fig. 5).

Fig. 4 Decadal averages of nitrate, ammonium, δ18O, net accumulation from QIC’s QSD, δ18O and nitrate from Huascarán in northern Peru (34), and Ti from Cariaco Basin (7).

Note reversed axis for Ti. Illimani ice cap ammonium data are courtesy of M. Schwikowski (Paul Scherrer Institut).

Fig. 5

Correlation fields between National Centers for Environmental Prediction/National Center for Atmospheric Research summertime (December to February) precipitable water (39) and annual ammonium concentrations from 1949 to 2002 in the QSD.

Both the Cariaco Basin and the Caribbean Sea lie along the path of moisture flow to the central Andes and experienced lower SSTs during the LIA (25, 26). The Quelccaya δ18O values also suggest lower SSTs in the tropical Pacific (Fig. 3C) at this time, indicating a LIA cooling in both ocean basins. Hydrological conditions were more regionally variable, and the QSD records suggest more complex large-scale linkages, possibly reflecting the recent northward migration of the ITCZ. However, over the Amazon Basin the ITCZ is not well defined but appears as a wide area of continental convection that is connected with the convergence zones over both the tropical Pacific and Atlantic oceans. Thus, its migration is not a simple shift into subtropical latitudes (27), and the associated precipitation anomalies are more complex. For example, lake (28) and speleothem (29) records in northern Peru indicate wet conditions, whereas speleothem records in northeastern Brazil (30) suggest dry conditions during the LIA.

Water levels measured in Lake Titicaca show reasonable similarities with the Quelccaya composite An before 1975 (fig. S7). Thereafter, mass wasting on the QIC has led to decreasing net accumulation. Marengo (5) demonstrated that even under current warmer conditions precipitation cycles in the northern and southern Brazilian Amazon Basin are out of phase, with both showing precipitation trend reversals in the mid-1940s and 1970s. Both QSD An and Lake Titicaca water level records closely reflect the timing and magnitude of changes in the Northern Amazonia Rainfall Index from 1929 to 1998 [figure 3 in (5)], which the author links to changes in oceanic and atmospheric circulation patterns and SSTs in the tropical central and eastern Pacific. These patterns argue not only that the QSD An represents the larger-scale regional pattern in Amazonian precipitation but that the NO3 and NH4+ signals likely originate in the Amazon Basin northeast of the QIC (Fig. 5) as a result of northeast airflow at the 850-mbar level (Fig. 1). Figure 4 illustrates that over the past ~1800 years the LIA is the most prominent feature in three ice-core records (Huascarán, Quelccaya, and Illimani; Fig. 1) collected along a north-to-south Andean transect. The relationships among these records suggest large-scale regional climate variability that requires multiple ice cores to disentangle the complex atmospheric and hydrological dynamics of the region.

The QIC accumulation history represents a much larger regional signal of precipitation variability. An on Quelccaya (Fig. 2) was well above average in the first half of the LIA (1520 to 1680 CE) and much reduced during the second half (1680 to 1880 CE). The An record is very similar to the 400-year climate history based on pollen (Poaceae/Asteraceae or P/A ratios) from the Sajama ice cap (Bolivia) 350 km to the southwest (31) and is consistent with nearby records of LIA glacier advance (32). The highest An rate in the QSD record is contemporaneous with the maximum LIA glacial extent (1630 to 1680 CE) in Bolivia and Peru, suggesting a cool and humid period followed by drier and gradually warming conditions as glaciers retreated. The humidity of the Bolivian/Peruvian Altiplano is reflected over long time scales by the nearby Lake Titicaca record (33). Thus, the QSD An record reflects larger regional changes in precipitation and indicates that Lake Titicaca’s water level was likely above average from ~1520 to ~1680, after which it declined to much lower levels until rising again in the 20th century.

Throughout the LIA, elevated concentrations of NO3 and NH4+ on Quelccaya are contemporaneous with reduced Ti deposition in the Cariaco Basin (7) (Fig. 4), which indicates arid conditions in the northeast portion of the Amazon Basin. If the relationship between precipitable water and QSD NH4+ and NO3 (Fig. 5) persisted over the LIA, then wetter conditions in the southwest Amazon Basin just upwind of Quelccaya would be contemporaneous with drier conditions in northeast Brazil (30), consistent with a southward migration of the ITCZ (fig. S6), resulting in a region of intense continental convection over tropical Amazon Basin. Strong similarities exist between the oxygen isotope records from Quelccaya and Huascarán to the north; however, the differences between the QSD NH4+ record and the Huascarán NO3 (34) and Illimani NH4+ (24) records tell a different story (Fig. 5), suggesting different sources for these species.

The QIC NO3 and NH4 concentrations reflect hydrological conditions in the Amazon Basin northeast of Quelccaya (Fig. 5) with moisture advected to the site by northeasterly winds at the 850 hPa level (Fig. 1) (below the boundary layer). However, the moisture bringing the δ18O, the NO3 deposited on Huascarán (6050 masl), and the NH4+ deposited on Illimani (6300 masl) arrives via winds at the 500-hPa level originating from the east-southeast. Because both Illimani and Huascarán are located outside the area where QSD and NH4+ are correlated with precipitable water (Fig. 5), this points to the complexity of precipitation and chemistry patterns in the Andes. Additional linkages are discussed in the supplementary text, but, whatever the ultimate driver, NH4+ cannot be interpreted as a large-scale regional temperature recorder in all tropical ice cores.

Modern Retreat of the Margin of the Quelccaya Ice Cap

Nearly annual field observations confirm that since 1978 the QIC has been retreating along its margins (14). Radiocarbon-dated wetland plants exposed by the retreating ice provide temporal constraints on both its advance and retreat. Twenty wetland plants exposed by the retreating margin of the QIC were collected between 2004 and 2007 next to a meltwater lake (site A in Fig. 6) that started forming in 1985 on the west side of the ice cap. Radiocarbon dates for the plants [4676 ± 41 years before the present (yr B.P.)] indicate that the ice cap is smaller than it has been in almost five millennia (14, 35). In 2011 as the ice cap continued to retreat, fresh plant remains were uncovered on the eastern side of the expanding North Lake (site B in Fig. 6), while most of the plants exposed in 2002 had already decayed because of their lack of woody tissue. Radiocarbon dates on these newly exposed plants, still in growth position, average 6298 ± 35 yr B.P. (table S3), ~1600 years older than the plants collected on the west side of the lake, indicating that Quelccaya is now smaller than it was 6000 years ago. Moreover, the plant ages confirm that the advance of the QIC ~6000 years ago was much slower, ~300 m over ~1600 years, than its current rate of retreat, ~300 m in 25 years. δ18O is not correlated with net accumulation on decadal and longer time scales (Fig. 2 and figs. S3 and S4) but is highly correlated with SSTs in the eastern equatorial Pacific (Fig. 3, A and B). For the past century, Quelccaya’s net accumulation has been above average, while δ18O has been enriched (Fig. 2), suggesting that the current retreat is driven in part by its warming environment. The rapidity of Quelccaya’s retreat may reflect snow-ice feedbacks that are considered instrumental (36) for rapidly increasing temperature trends near the 0°C isotherm during the 20th century. The accelerating retreat of Quelccaya and other tropical ice fields (8) is consistent with model predictions for vertical amplification of temperature in the tropics (37) and has serious implications for those living in these areas.

Fig. 6

Locations of recently exposed plants collected along the western side of the QIC.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Tables S1 to S3

References (4042)

References and Notes

  1. Acknowledgments: Funding was provided by NSF Paleoclimate Program award ATM-0318430 and the Ohio State University’s Climate, Water and Carbon Program. We thank all the field and laboratory team members, many from the Byrd Polar Research Center, who have worked so diligently over the years to acquire these ice cores and extract their preserved climate histories. We acknowledge the efforts of our Peruvian colleagues from the Servicio Nacional de Meteorologia e Hidrologia, C. Portocarrero, and our mountaineers, led by B. Vicencio, who cooperated with us to make the project possible. Accelerator mass spectrometry 14C measurements were made at the National Ocean Sciences Accelerator Mass Spectrometry Facility (Woods Hole Oceanographic Institution). This is Byrd Polar Research Center Contribution Number 1430. The data are archived at the National Oceanic and Atmospheric Administration World Data Center-A for Paleoclimatology:
View Abstract

Stay Connected to Science

Navigate This Article