Introduction
The El Niño–Southern Oscillation (ENSO) phenomenon is the largest known single source of global interannual climatic variability (Reference Diaz, N, Diaz, Markgraf and NiñoDiaz and Markgraf, 1992), as illustrated by the worldwide climate anomalies that are associated with the atmospheric and oceanic fluctuations in the ENSO center of action, the tropical Indian–Pacific Ocean (Reference Glantz, Glantz, Katz and NichollsGlantz, 1991; Reference Diaz, Markgraf, Diaz, Markgraf and NiñoDiaz and Kiladis, 1992). the majority of studies have understandably focused on the ENSO center of action in the tropical Pacific and climate teleconnections in the low to mid-latitudes.
Instrumental linkages between the high southern latitudes and the tropics demonstrate the existence of an Antarctic– ENSO climate connection in several different ways: annual temperatures at South Pole are positively correlated to annual values of the Southern Oscillation Index (SOI) of the previous year (Reference Savage, Stearns and WeidnerSavage and others, 1988; Reference Smith and Stearns.Smith and Stearns, 1993); an eastward shift in Amundsen Sea low (ASL) pressure is noted during El Niño events from 1980 to 1990 (Reference Cullather, Bromwich and Van Woert.Cullather and others, 1996; Reference Bromwich, Rogers, Kållberg, Cullather, White and JBromwich and others, 2000); analysis of satellite-derived Antarctic sea-ice extent data reveals El Niño-associated periodicities (Reference GloersenGloersen, 1995); key sea-ice regions in the Pacific sector are positively related to the SOI (Reference CarletonCarleton,1989; Reference Simmonds and JackaSimmonds andJacka,1995); and sea-ice concentrations are reduced in the Ross Sea region during El Niño events (Reference Ledley and HuangLedley and Huang,1997).
Most attempts at developing a reliable paleoclimate proxy for ENSO have focused on the tropics and subtropics (Reference StahleStahle and others, 1998). Notably, historical records originating from the west coast of South America provide a reliable proxy of El Niño events well beyond the instrumental record (Reference Quinn, Diaz, Markgraf and NiñoQuinn, 1992; Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992), significantly increasing the temporal coverage of ENSO variations. In addition, an ice core covering the time period 1922–84 from South Pole and examined by Reference Legrand and Feniet-SaigneLegrand and Feniet-Saigne (1991) yielded evidence that sub-annually resolved methanesulfonate (MS) and non-sea-salt sulfate (SO4 2–), both of which are produced from the photo-oxidation of the marine biogenic species dimethylsulfide (Reference Charlson, Lovelock, Andreae and WarrenCharlson and others, 1987; Reference Saigne and LegrandSaigne and Legrand, 1987), clearly show perturbations that are associated with El Niño events in the historical record (Reference Quinn, Diaz, Markgraf and NiñoQuinn and others, 1987).
Ice-core measurements of MS have also been used to determine past fluctuations in Antarctic sea ice at coastal sites (Reference Welch, Mayewski and WhitlowWelch and others, 1993; Reference PhillipsPhillips, 1998). These studies reveal a positive relationship between increased MS and increased sea-ice extent in the adjacent ocean sector, most likely because of the increased productivity due to the growth of phytoplankton in and under the ice, as well as at the ice edges (Reference Bunt, F and WoodBunt and Wood, 1963; Reference Clarke and FClarke and Ackley, 1984; Reference Nelson, Smith, Gordon and ANelson and others, 1987; Reference Welch, Mayewski and WhitlowWelch and others, 1993). Studies of Antarctic sea-ice extent are of great importance not only because of their potential link to ENSO, but also because Antarctic sea-ice variations have been shown to have a major influence on Southern Hemisphere atmospheric circulation patterns (Reference CarletonCarleton,1989) such as the Antarctic circumpolar wave (Reference White and PetersonWhite and Peterson, 1996).
Methods
A 71 m, high-resolution ice core, covering the period 1487– 1992, was collected at South Pole (snow accumulation rate ~8.0±2.0cma–1 w.e.) at a site ~1.5 km upwind in the designated clean-air sector in 1995 (henceforth SP95). SP95 was shipped frozen to the Climate Change Research Center, University of New Hampshire, U.S.A., for processing and analysis. Ultra-clean procedures (Reference Buck, Mayewski, Spencer, Whitlow, Twickler and BarrettBuck and others, 1992) were used to section the core at ~2 cm intervals to obtain subannual resolution (8.2 samples per year, 1900–92; 6.5 samples per year, 1700–1899;5.7 samples per year, 1487–1699). the core was sub-annually dated using seasonal maxima in several chemical species (SO4 2–, Cl–,Na+,Mg2+) (Fig.1) (Reference MeyersonMeyerson, 1999) in order to focus on the MS record. Cation( Na+,Mg2+) concentrations were determined using ion chromatography with a Dionex CS12 column, 20 mM methanesulfonic acid eluent and a 0.50 mL sample loop. Anion (Cl–, SO4 2–) concentrations were determined using a AS11 column, 6.0mM NaOH eluent and a 0.25 mL sample loop. MS concentrations were obtained with a AS11 column with 0.1 mM NaOH eluent and a 1.50 mL sample loop. the resultant annually dated MS time series is used to develop a sea-ice proxy series and to extend the previous 1922–84 South Pole MS–El Niño relationship (Reference Legrand and Feniet-SaigneLegrand and Feniet-Saigne,1991) over the time period 1487–1992.
The 1487–1992 SP95MS record is characterized by high-frequency interannual variations superimposed on longer (decadal and lower) frequency patterns (Fig. 2). It is also characterized by increasing values from the mid-20th century to the present, and limited migration from summer to winter layers below ~3.5m (AD 1980). the decreasing trend in SP95MS from surface concentrations of 25 ppb to an average background concentration level of 5 ppb at 10.5 m depth (AD1940) (Fig. 2) is similar to the MS decrease in theVostok (Antarctica) ice core (snow accumulation rate ~2.2 g cm–2 a–1) (Fig.3) from 20 ppb to a baseline of 2 ppb at ~5.5 m depth that is attributed to post-depositional modification (Reference Wagnon, Delmas and LegrandWagnon and others, 1999). Thus we assume that there is some post-depositional loss of MS in the upper 10.5 m (AD1940) of the SP95 core, but believe it is limited to the uppermost portion of the SP95 MS record, as illustrated by the increased mean and variance from 1940 to 1992 (10.42, 23.79, respectively) relative to 1487–1940 (4.99, 4.79, respectively). A t test of mean MS concentrations for 1940–92 and 1487–1939 demonstrates that the distribution of MS values from 1487 to 1940 is significantly different from 1940–92, and leads us to conclude that further post-depositional loss of MS does not affect SP95 MS concentrations below 10.5 m depth (~AD1940). Migration of MS has also been noted in the Siple Dome ice core (Reference Kreutz, Mayewski, Whitlow and S. TwicklerKreutz and others, 1998) and other Antarctic coastal sites (Mulvaney and others, 1992), but limited movement of summer-deposited MS into adjacent winter layers at this site has been demonstrated not to affect the investigation of climate signals at the annual scale and higher (Reference Kreutz, Mayewski, Whitlow and S. TwicklerKreutz and others, 1998; Reference Legrand and C. Pasteur.Legrand and Pasteur, 1998).
Other prominent features in the MS record are steps in the baseline concentration at ~1575, ~1620 and ~1940 (Fig. 2). the ~1940 step is synchronous with the prolonged El Niño conditions from July 1939 to June 1942 (negativev alues in the SOI) recorded as three relatively strong El Niño events in the historical record (1939, 1940–41, 1943) (Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992). This step is also observed in other records: Tarawa (Kiribati) coral δ18O proxy series for the SOI (Reference Cole, Fairbanks and TCole and others, 1993); annually resolved Southern Hemisphere temperature series (Reference Hansen and LebedeffHansen and Lebedeff,1988); and South Pole ice-core deuterium (Reference Jouzel, Merlivat, Petit and LoriusJouzel and others, 1983). Collectively, these records indicate a regional shift in the climate
system at ~1940. Other large steps in MS at ~1575 and ~1620 were also accompanied by strong El Niño events (1574,1618–19) (Reference Quinn, Diaz, Markgraf and NiñoQuinn and Neal,1992).
Exploratory empirical orthogonal function (EOF) decomposition analysis was conducted on the SP95 MS record, Amundsen–Ross (AR) sea-ice extent anomalies (185–245˚E), the SOI, and South Pole surface pressure and temperature anomalies for the period 1973–92 in order to statistically identify associated variances in the Antarctic– ENSO–sea-ice system. the SP95 MS series was first resampled to 12 samples per year for correlation with the monthly SOI, sea-ice and South Pole meteorological series. This allows the MS series to be matched with the monthly instrumental series for the EOF decomposition. the resampling of high-resolution proxy data to match monthly instrumental series (e.g. SOI, meteorological series) is demonstrated by Reference Cole, Fairbanks and TCole and others (1993) in their discussion of the Tarawa coral record. Sea-ice-extent monthly anomaly time series developed from monthly latitudinal sea-ice extent records compiled by Jacka (Reference Jacka1983; with updates) for every 10˚ of longitude from U.S. Navy and U.S.National Oceanic and Atmospheric Administration (NOAA) Joint Ice Center satellite-derived maps. Monthly SOI values, obtained from the Australian Bureau of Meteorology, were calculated using the Troup definition of the SOI (standardized anomaly of Tahiti minus Darw in mean monthly sea-level pressure). South Pole Station (SPS) surface pressure time series were obtained from Reference Jacka, Christou and JJacka and others (1984) (1957–77) and from the Climate Monitoring and Diagnostics Laboratory of the NOAA(1977–92). SPS surface temperature monthly time series were obtained from Jacka and others (Reference Jacka, Christou and J1984; with updates).
The rationale behind investigating the AR sea-ice sector (185–245˚E) is based on previous work in this region. First, this sector has been identified as an air-mass source region for the West Antarctic ice sheet, of which South Pole is situated on the southern flank (Reference Cullather, Bromwich and Van Woert.Cullather and others, 1996; Hogan, 1997; Reference Bromwich, Rogers, Kållberg, Cullather, White and JBromwich and others, 2000). Additionally, there are multiple studies documenting the instrumental relationships between sea ice and ENSO in the AR region/southeastern Pacific (Reference CarletonCarleton,1989; Reference Simmonds and JackaSimmonds and Jacka, 1995; Reference Ledley and HuangLedley and Huang, 1997). the AR sector is also the adjacent region to South Pole in terms of distance, and this has been shown to be important for MS ice-core series (Reference Welch, Mayewski and WhitlowWelch and others, 1993; Reference PhillipsPhillips, 1998). the time period of study was constrained by sea-ice extent (satellite measurements started in 1973) and the SP95 core (most recent year 1992). All records were smoothed with a 7-point running mean to remove high-frequency variations ≤6months and thus reduce noise in the monthly data, providing a better indication of possible climate behavior (Reference Trenberth and Hoar.Trenberth and Hoar,1996).
Results and Discussion
Variance decomposition
The EOF results show that three main modes of variability are identified by the EOF decomposition, representing in total 78% of the variance of the ice-core, sea-ice and surface meteorological time series. EOF 1 explains 32% of the total variance in all the series and represents a positive relationship between 47% of MS, 65% of sea-ice extent and 47% of the SOI (Fig. 4). EOF 2 (30% overall variance) represents 73% of the SPS surface pressure anomalies and 72% of the temperature anomalies. EOF 3 explains 16% of the overall total variance and reveals an inverse relationship between MS and the SOI at 37% and 38% explained variance, respectively (Fig.4).
The highest overall variance among the explored records is represented by EOF 1which displays a positive relationship between SP95 MS concentrations, AR region sea-ice extent and the SOI (Fig. 4). Studies of ice-core MS records from two Antarctic coastal sites, Newall Glacier and Law Dome (Reference Welch, Mayewski and WhitlowWelch and others, 1993; Reference PhillipsPhillips, 1998) (Fig. 3), also reveal a positive relationship between sea-ice latitudinal extent and increased MS. EOF 2 associates 75% of the variance in both the SPS surface pressure and temperature series. Previous studies linking the SOI to South Pole meteorological data identify lagged responses in surface temperature (Reference Savage, Stearns and WeidnerSavage and others, 1988; Reference Smith and Stearns.Smith and Stearns, 1993) and pressure (Reference Smith and Stearns.Smith and Stearns, 1993), which are not taken into account with EOF decomposition analysis. EOF 3 reveals an inverse relationship between MS and the SOI (Fig. 4). This association (i.e. increased MS and negative values of the SOI) supports statistically the association between the extreme MS events in the SP84 ice-core record (Reference Legrand and Feniet-SaigneLegrand and Feniet-Saigne, 1991) and El Niño events (Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992). the most notable structures of EOF 3 are the extreme events found in both the SP95 MS and the SOI, which correspond to the El Niño events in 1982–83 and 1987.
The separate and distinct MS signals in the SP95 core identified by EOF 1 and EOF 3 in the variance decomposition demonstrate that the SP95 glaciochemical MS is influenced by air masses that travel over sea-ice-related sources of dimethylsulfide (DMS) (EOF 1). During El Niño events (EOF 3), as suggested by Reference Legrand and Feniet-SaigneLegrand and Feniet-Saigne (1991), the DMS in the high latitudes is either enhanced by productivity or more efficiently transported to South Pole.
We suggest, based on the EOF results, that all sources of MS reaching South Pole are in the high-latitude oceans surrounding Antarctica. An earlier hypothesis, based on the latitudinal distribution of MS and nssSO4 2–, submits that the majority of MS deposited at South Pole during non-El Niño years was from more temperate latitudes (Reference Legrand and Feniet-SaigneLegrand and Feniet-Saigne, 1991; Reference Legrand, Feniet-Saigne, Saltzman and GermainLegrand and others, 1992). Further investigation of MS and nssSO4 2– has revealed that the prime latitudinal source band for Antarctic sulfur is from DMS that is south of 50–60˚S (Reference Legrand and C. Pasteur.Legrand and Pasteur, 1998; Reference MinikinMinikin and others, 1998). A more recent study, however, shows evidence that inferring the latitudinal origin of DMS with the ratios of MS to nssSO4 2–might be misleading because high-latitude ratio values are also seen at higher altitudes in the low latitudes (Reference DibbDibb and others, 1999).
Methanesulfonate–sea-ice reconstruction
The dominant mode of MS deposition to South Pole, as represented by the first EOF loading series, is from sea-ice sources in the AR sector (Fig. 3). Sea-ice-associated MS at South Pole is also positively correlated with the SOI in EOF 1. This relationship suggests that this MS is incorporated with air masses that travel over the AR sea-ice sector (185–245˚E) and is eventually precipitated out in the West Antarctic sector, where it is most likely incorporated into snow at South Pole. There is additional MS deposition to South Pole (EOF 3) during El Niño events from a source that is statistically unrelated to sea ice in the AR sector, notably the open ocean and/or another sea-ice source. Analysis of air flow into Antarctica shows that relatively warm air appears to enter the southern part of the polar plateau through a topographic low between the summit of the Transantarctic Mountains and the polar plateau (Hogan, 1997) (Fig. 3). Furthermore, marine aerosols circulating around lows in the Bellingshausen–Amundsen–Ross Seas area frequently reach the surface of the polar plateau as a consequence of this topographic low. An eastward shift in the ASL during El Niño events (Reference Cullather, Bromwich and Van Woert.Cullather and others, 1996) would increase the atmospheric content of MS in the southeastern Pacific sector (~270˚ E). We assume that upslope transport of this MS-laden air mass in the southerly direction through the topographic low (Hogan,1997) contributes to the MS concentrations at South Pole. Therefore, the SP95 glaciochemical MS time series (1487–1992) not only provides an estimate of sea-ice extent in the AR region, but also provides a marker for El Niño-associated changes in high-latitude circulation.
EOF 1provides an estimate for background values in the SP95 MS, AR sea-ice extent and the SOI over the time period 1973–92 and demonstrates a direct relationship with sea-ice extent in the AR region (Fig. 4). Before this relationship is extended over the 500 year record of SP95 MS, the possible loss of MS in the upper portion of the core contributing to the downward trend from the present to ~1940 is addressed. the two time periods separated by the break, 1487–1940 and 1940–92, were standardized separately by subtracting out their relative means, and then dividing by their respective standard deviations. Justification for treating the MS series as two distributions (1487–1940,1940–92) stems from the t-test results that demonstrate that these two periods have significantly different means.
A strong correlation between EOF 1 (MS–sea-ice loading series) and a ~50 year smoothed MS robust-spline curve demonstrates that the robust spline approximates the EOF calibration signal (Fig. 4). the ~500 year SP95 MS proxy for sea-ice extent in the AR sector is dominated by multi-decadal variability with a distinct shift at ~AD 1800 from higher-frequency variations (1550–1800) to lower-frequency variations (1800–1940) (Fig. 2). the MS proxy for sea ice also reveals periods of increased sea-ice extent relative to the present (1800–50,1900–40) as well as periods of decreased sea-ice extent relative to the present (1550–1610, 1660–1710, 1760–1800). There is also relatively increased sea-ice extent from ~AD1800 to 1940 as compared to 1550–1800. Longer records of Antarctic sea-ice extent have been ascertained from whaling records (de la Mare, 1997), but the number of observations in the Amundsen and Ross Seas results in low temporal resolution and therefore precludes accurate comparison.
Methanesulfonate–El Niño reconstruction
Reference Legrand and Feniet-SaigneLegrand and Feniet-Saigne (1991) demonstrated in their seminal work that increases in South Pole MS were associatedwith El Niño events in the historical chronology (Reference Quinn, Diaz, Markgraf and NiñoQuinn and others,1987). A separate study (Reference IsakssonIsaksson,1994) confirmed this finding by using an MS record from Dronning Maud Land, on the polar plateau, covering the period 1865–1992. Our exploratory EOF analysis also identifies, and significantly expands the findings of, an inverse relationship between SP95 MS and the SOI that statistically supports these previous findings. Visual inspection of the SP95 MS record shows that there is a strong association between elevatedMS and El Niño events in the historical record (Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992) within the dating error of the ice core (Fig. 2).
To extract an El Niño chronology from the SP95 MS record, the series was first detrended by subtracting out the ~50 year smoothed MS robust spline that represents the sea-ice-associatedMS described earlier. the time periods (1940– 92,1487–1940) were then standardized as described earlier, to account for MS loss. SP95 standardized MS residuals ≥1 standard deviation (σ) were selected as El Niño-related MS peaks (Fig. 2) and were then correlated at ±2 years to the historical record of El Niño events (Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992). We used a truncation level of 1σ to objectively ensure that only the elevated levels of MS, comparable to those identified by Reference Legrand and Feniet-SaigneLegrand and Feniet-Saigne (1991) are considered (16% of the MS values). We compared the standardized MS residual peaks ≥1σ to the historical record of El Niño events (Quinn and Neal, 1992) to test the validity of SP95 as a proxy for El Niño. the MS peaks ≥1σ that correlate to El Niño events for 1922–84 and 1860–1992 validate previous studies on polar plateau MS–El Niño relationships (Reference Legrand and Feniet-SaigneLegrand and Feniet-Saigne,1991; Reference IsakssonIsaksson,1994).
The results of this correlation are displayed by century at the bottom of Figure 2. the results show that the MS proxy captures 88% of the M+ and higher El Niño events in the Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992 record for the 1900–92 period. the capture results of the MS proxy decline to 68% for 1800–99 and to <50% for the time period 1600–1800. the capture results for the earliest portion of the MS proxy, 1525–99, increase to 67%. While the SP95 MS–El Niño proxy does not have a one-to-one reconstruction of the historical chronology of El Niño events (Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992), it does provide an estimate of the polar expression of El Niño back to 500 years BP. As such, the SP95 record is valuable in the reconstruction of global teleconnection patterns of climate change associated with the El Niño– Southern Oscillation (ENSO) system.
Conclusions
From the SP95MS–ElNiño proxy we suggest that either the capture efficiency of high-level MS decreased from 1600 to 1800 or the El Niño impact on the polar latitudes declined during this period. We do not believe that the explanation relates to capture efficiency since the capture efficiency in the earliest portion of the record, the 16th century, is comparable to that in the 19th century. We therefore assume that the lower correlation between the SP95 MS–El Niño proxy and the historical record from 1600–1800 demonstrates that there could be a decoupling between the tropical ENSO system and high-latitude circulation at a time when variations in the MS–sea-ice proxy are relatively more variable (1550–1800). Decoupling between the ENSO system and the high latitudes indicates large-scale changes in Southern Hemisphere climate during this period relative to the remainder of the record. It is notable that this decoupling occurs not only during periods when sea-ice extent is less but also when sea-ice variability is high. An example of this decoupling at a finer scale (monthly) is the regime shift in correlation from positive to negative seen at ~1990 between the SOI and West Antarctic precipitation (Reference Cullather, Bromwich and Van Woert.Cullather and others, 1996; Reference Bromwich, Rogers, Kållberg, Cullather, White and JBromwich and others, 2000).
Frequency analysis on the historical El Niño chronologies from South America (Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992) and the Nile region (Reference Quinn, Diaz, Markgraf and NiñoQuinn, 1992) reveals significant changes as a function of temperature, notably increased El Niño activity during the period of the Little Ice Age (LIA; nominally 1400–1900) (Reference Anderson, Diaz, Markgraf and ElNiñoAnderson, 1992) and decreased El Niño activity during the Medieval Warm Period (nominally 950– 1250) (Reference Anderson, Diaz, Markgraf and ElNiñoAnderson, 1992; de Putter and others, 1998). the SP95 MS–sea-ice proxy is indicative of regional temperatures within the LIA period in the southeastern Pacific sea-ice sector for the last 500 years. the shift at ~1800 towards generally cooler conditions (increased sea-ice extent) in this region is concurrent with an increase in the frequency of El Niño events in the SP95 MS–El Niño proxy, as well as an increase in the frequency of events in the independent historical El Niño record (Reference Quinn, T. Neal, Bradley and JonesQuinn and Neal, 1992) (Fig. 2). This multi-decadal to centennial-scale relationship in the SP95 record between cooler (warmer) conditions and increased (decreased) occurrence of El Niño events supports the previous study (Reference Anderson, Diaz, Markgraf and ElNiñoAnderson, 1992). the short-term trend in the instrumental SOI since 1970 towards more El Niño events (Reference Trenberth and Hoar.Trenberth and Hoar,1997) is associated with a rise in average global temperatures. This departure in the longer-term El Niño–temperature relationship may suggest a different mode of operation at present, or that ENSO–temperature associations can vary on annual to decadal and centennial scales.
Acknowledgements
We thank: M. Battle, M. Bender,T. Sowers, D. Giles and J. Kyn. for collecting the SP95 core during their fieldwork supported by the U.S. National Science Foundation; C. M. Zdanowicz and R. Boudro. for technical assistance; the Climate Change Research Center ion chromatography laboratory staff; J. Horsman for freezer assistance; J. Dibb, K. Hibbard, B. Keim, D. Z. Smith and C. Wak. for reviewing the manuscript; S. Palmer for graphical support; and the two anonymous referees for their thorough reviews and detailed comments that helped to improve our manuscript.