1 Introduction

Climate variability over recent decades in the southern extratropics has been linked to the behaviour of the Southern Hemisphere Annular Mode (SAM) by Thompson and Wallace (2000), Thompson and Solomon (2002) and Kwok and Comiso (2002). The SAM, also referred to as the High-Latitude Mode (HLM) of climate variability (Rogers and van Loon 1982, Kidson 1988), and the Antarctic Annular Oscillation (AAO) by Gong and Wang (1999), describes a near zonally symmetric pattern of tropospheric pressure and winds that displays a meridional dipole between ∼45°S and 60°S. A number of statistical treatments of the instrumental data have been used to define the time series of the zonally symmetric seesaw of atmospheric mass between the mid-latitudes and Antarctica. The strength of the SAM has been characterised by an index computed from the difference in geopotential height (Z) between the mid-latitudes and the polar regions (typically Z 40°S–Z65°S, as in Gong and Wang 1999) for the 500, 850 and 1000 hPa surfaces. Alternatively, Thompson and Wallace (2000) defined the SAM by the leading empirical orthogonal function (EOF) of the 850 hPa geopotential height field.

In the more recent study of Thompson and Solomon (2002) the SAM index is defined as the monthly average of 500 hPa geopotential height anomalies over the Southern Hemisphere south of 60°S (inverted in sign for consistency with the previous definition of the SAM index). When the SAM index is positive the mid-latitude westerlies are latitudinally displaced poleward (towards 60°S) and are associated with a double jet, and when the index is negative the westerlies are displaced equatorward (towards 45°S) together with the polar jet, and an associated weakening of the westerlies close to the Antarctic coast. Geopotential height anomalies over the Antarctic Continent (south of 60°S) and those over the subtropics are out of phase with geopotential height anomalies in the mid-latitudes between 40°S and 50°S (Kiladis and Mo 1998). Climatological studies of the SAM have also focused on understanding the behaviour of weather systems (cyclone tracks, cyclone frequency, anticyclone tracks and blocking), particularly with respect to the phase of the low-frequency variability of the SAM itself (Sinclair et al. 1997).

The instrumental data set in the southern extratropics is almost non-existent before 1957, when a number of Antarctic meteorological stations were established, during the International Geophysical Year (IGY). In the absence of suitable climate proxies for the SAM, our knowledge of SAM behaviour at monthly to annual resolution has been restricted to the decades of instrumental observations. Unfortunately the last four decades of observations are insufficient to examine multidecadal variability of the SAM. However, Antarctic ice cores have considerable potential for reconstructing southern extratropic atmospheric circulation, through the measurement of sea-salt chemistry and stable oxygen isotopes (δ18O and δ D) in the ice (Kreutz et al. 2000; Stenni et al. 2001). Antarctic ice cores retrieved from the most northerly locations have the greatest potential for providing records of mid-latitude atmospheric circulation. Prior research by Curran et al. (1998) demonstrated the strong circulation control on the delivery of sea-salts in precipitation at Law Dome (66°46′S, 112°48′E, 1370 m, Fig. 1), a small ice cap near the most northerly extent of the East Antarctic coast. They showed that the annual cycle in sea-salt (sodium, chloride and magnesium) concentration measured in Law Dome snow was strongly linked to wind speed between 50°–60°S. A further study by Souney et al. (2002) established that the annual average sea-salt concentration in precipitation at Law Dome is strongly correlated to June sea-level pressure (SLP) across East Antarctica (Antarctic High) for the 1957 to 1995 period. Souney et al. (2002) concluded that there was no significant relationship between Na and sea-ice extent at Law Dome. This is largely due to the northerly latitude of the East Antarctic coast, and the corresponding regional minimum in sea-ice extent coupled with minimal seasonal changes in sea-ice extent along this Southern Ocean sector (Gloersen et al. 1992; Jacka 1983; Zwally et al. 2002).

Fig. 1
figure 1

Location map showing the DSS ice core site on Law Dome, and the circum-Antarctic, including the mid-latitude meteorological stations at Campbell Island, Macquarie Island, Kerguelen Island and Marion Island

Full size image

In this study, we report a new seasonal sodium (Na) concentration time series from the Law Dome, spanning 1300–1995 AD. Section 2 describes the Na concentration time series and discusses the annual cycle for Na concentration with respect to Na sources and transport patterns to Law Dome. In Section 3 we statistically analyse the Na concentration time series together with the NCEP/NCAR Reanalysis (NNR) MSLP and wind field data to establish the spatial correlation patterns for interannual variability. We also investigate the seasonal relationship between Na concentration and the SAM index, and the suitability of the Na time series as a proxy for the SAM. In Section 4 we present a calibration of early winter (nominally the average for May, June and July) Na concentrations with the mid-latitude station MSLP data, and present the spatial trends in interannual MSLP anomalies associated with the delivery of high and low Na concentration years at Law Dome. In Section 5 we explore the spectral characteristics of the early winter Na concentration (proxy SAM) time series, and discuss the decadal variability of mid-latitude atmospheric circulation over the South Indian Ocean and southwest Pacific Ocean regions since 1300 AD, with respect to weather system behaviour, particularly cyclone frequency and blocking anticyclones.

2 Law Dome sea-salt concentration data

The 1200 m long Dome Summit South (DSS) ice core was retrieved from Law Dome between 1988 and 1993 (Morgan et al. 1997). This ice core provides a climate record extending to at least ∼80 kyr (Morgan et al. 2002). The DSS site experiences very high snow accumulation rates (0.64 kg m–2 per year). This affords the possibility for quasi-monthly sampling over the last 1–2 kyr BP and seasonal to annual sampling to ∼4 kyr BP. The upper ∼400 m of ice core, representing the past 700 years of snow accumulation was analysed for major ion chemistry (Na+, Ca2+, K+, Mg2+, Cl–, NO3 – and SO4 2–) and stable oxygen isotope composition (δ18O) at a resolution of ∼12 samples per annual increment (Palmer et al. 2001). This data was augmented with data from two additional cores from the same site (DSS97 and DSS99) to provide continuous high-fidelity measurements where the original DSS core quality was affected by drilling difficulties (hereafter we refer to the composite record as the DSS record). Clear annual cycles in ion chemistry and δ18O allow the core to be dated without ambiguity back to 1806 AD, and prior to this, with an error of ± 1 year to 1300 AD (Palmer et al. 2001). The sub-annual time scale was derived by linear interpolation between mid-summer δ18O peaks. Comparison with instrumental temperatures shows that this method produces minimal bias, on average, for the DSS core (van Ommen and Morgan 1997), although in any one year, non-uniformities in accumulation rate (McMorrow et al. 2002) will necessarily produce timing “noise”. However, in most years the Law Dome summit experiences snow accumulation throughout the year with short hiatus periods. The absolute phasing of the summer δ18O peaks is established by comparison with instrumental temperatures and, in conjunction with hydrogen peroxide seasonality, which is tied to the summer solstice (van Ommen and Morgan 1996). These methods have resulted in a robust depth-age time scale that is quasi-monthly in resolution for all measured stable isotope and chemical ions.

Curran et al. (1998) reported the seasonal characteristics of the major ions measured in DSS ice. All the ion species display a clear seasonal cycle from year to year. They reported that the sea-salt concentrations (sodium, chloride and magnesium) measured in the DSS ice core were in close agreement with the sea-salt ratios in seawater, and that Na is a suitable index for sea-salt aerosol input.

Wet deposition by falling snow is the dominant depositional process for sea-salt aerosol species in coastal Antarctica (Wolff et al. 1998). Precipitation at Law Dome is controlled by the passage of maritime air from circumpolar cyclones tracking east and southeast over the Antarctic coast. In addition, Law Dome also receives continental air masses tracking from the interior of East Antarctica. Relatively warm and enriched (more positive) δ18O air masses are advected to Law Dome during periods where strong synoptic meridional pressure gradients exist to the north. Under such conditions, the maritime cyclones rapidly approach the coast and bring high Na aerosol loads with less fractionation of oxygen isotopes. In contrast, low Na aerosol loads and depleted δ18O are associated with the colder continental air masses (McMorrow et al. 2002), and cyclones originating closer to the Antarctic coast. The latter are prevalent in the climatological region of the circumpolar trough and are characterized by slower movement and greater isotopic fractionation. These observations are consistent with previous studies (Legrand and Kirchner 1988; Wagenbach 1996) that proposed increased cyclonicity and enhanced meridional transport from ice-free regions of the Southern Ocean are the principal source and transport mechanism for sea-salt deposition in coastal Antarctica during winter. Hence, the range of synoptic conditions results in measurable fluctuations in Na concentration in Law Dome snow.

The mean (n = 295) annual cycle of Na concentration in the DSS ice core for the period 1700 to 1995 AD is shown in Fig. 2. Na concentration steadily increases during the autumn and early winter (April to July), reaching a peak in spring (September), when the maximum winter sea-ice extent occurs (Curran et al. 1998). Also shown in Fig. 2, is the average annual cycle for the highest (n = 75) and lowest (n = 75) Na concentration years. Both the lowest and highest years show very similar seasonality to the overall mean annual cycle, varying in amplitude, but not in shape. This suggests that the variations in strength and/or frequency that modulate the amplitude of the Na delivery are not accompanied by variations in the seasonal timing of Na events or, indeed of precipitation, since timing changes would affect the shape of the cycle. In the highest concentration years Na levels exceed the peak of the average years for the five months from June to October. We investigate the monthly Na concentration together with monthly MSLP data to determine whether interannual variability in seasonal atmospheric circulation is responsible for variations in the strength of Na delivery.

Fig. 2
figure 2

The mean annual cycle of sodium (Na) concentration in DSS ice for 1700 to 1995 AD (green). Also shown is the mean annual cycle for the 75 highest (red) and 75 lowest (blue) Na concentration years

Full size image

3 Comparison of Law Dome sea-salt data with Southern Hemisphere NNR data

3.1 Correlation analyses with annual average MSLP NNR data

Initial comparisons between the annual average Na concentration in the DSS ice core and annual average MSLP were made using the NCEP/NCAR Reanalysis (NNR) MSLP data set (Kalnay et al. 1996; Kistler et al. 2001). Hines et al. (2000) reported that the NNR MSLP data prior to 1968 is not well constrained in the Antarctic region, because the NNR use few East Antarctic surface data in this period. Since 1968 the interannual variability in Antarctic and circumpolar MSLP data are adequately captured by the NNR. We acknowledge these problems inherent in the NNR data and accordingly, use only MSLP data spanning 1970 to 1995.

The correlation between annual (January to December) average Na concentration and annual average NNR MSLP data is shown in Fig. 3. The spatial correlation pattern of interannual variability reveals a strong relationship between the Na concentration and MSLP over the mid-latitude Indian and Pacific Ocean regions and also with the Antarctic Continent. This is consistent with a typical winter circulation described by a planetary Rossby wave number 3 (Mo and White 1985), with regions of negative correlation corresponding to troughs over the southwest Indian, southwest Pacific Oceans, and the tip of South America.

Fig. 3
figure 3

The spatial pattern of cross-correlation between mean annual NNR sea-level pressure and DSS Na concentration for years 1970 to 1995 AD

Full size image

The region of moderate negative correlation (R = –0.3 to –0.5, 95%–99.5% CL) in the South Indian Ocean bounded by 30° to 50°S and 10° to 80°E represents a key region of mid-latitude cyclogenesis for cyclones tracking to Law Dome, as shown by the cyclone track data of Simmonds and Keay (2000). Cyclogenesis also occurs in the circumpolar belt between 55° and 65°S in the Indian Ocean sector, with cyclolysis occurring near the Law Dome coast when there is strong meridional circulation south of Australia (Simmonds and Keay 2000). The region of positive correlation south of Australia between 90° and 150°E represents the climatological location of ridging. This feature has significant influence on the advection of mid-latitude airmasses and the steering of circumpolar cyclones towards the Law Dome coast.

3.2 Correlation analyses with seasonal average MSLP NNR data

Souney et al. (2002) showed that a maximum correlation exists between DSS annual average Na concentration data and monthly mean MSLP for June at Antarctic stations over the 1958 to 1995 period. In this study, we examined the monthly and seasonal (three month average) relationships between DSS Na concentration and MSLP of mid-latitude and sub-Antarctic stations at the interannual time scale (using the NNR data). Individual months that displayed a wave number 3 correlation pattern were April, May, June, July, September and December. In contrast, the monthly correlation pattern for MSLP and Na concentration during August displayed a well-defined wave number 1 structure that is indicative of a zonal circulation. We found that the maximum correlation value for 3-month average Na data and the corresponding average MSLP observational data for mid latitude and sub-Antarctic stations was for early winter (May, June and July, MJJ). The average MJJ Na concentration accounts for up to 50% (R = 0.71) of the variance in the annual average Na concentration in DSS snow. The correlation of the 3-month average data decreases for the late winter and spring seasons using the June, July and August, and August, September and October combinations. This decrease occurs despite high Na concentration years being characterised by significantly higher concentrations from April to November, when compared to those for low Na concentration years (shown in Fig. 2).

The spatial correlation plot for MJJ is shown in Fig. 4. The spatial correlation pattern of interannual variability is similar to that obtained for the annual average data (shown in Fig. 3), except that a stronger relationship exists between the Na concentration and the atmospheric circulation for MJJ, than for the annual average. The stronger correlation pattern reflects the dominance of the Rossby wave number 3 circulation over winter and its lower amplitude expression during the remainder of the annual cycle (Mo and White 1985). The more defined wave number 3 pattern is exemplified between the South Indian Ocean and the southwest Pacific Ocean sectors. The wave number 3 pattern is typically associated with the third empirical orthogonal function (EOF) of the Southern Hemisphere monthly MSLP anomalies (Sinclair et al. 1997). However, the Rossby wave number 3 pattern in the Indo-Pacific circum-polar regions as described by the spatial pattern in Fig. 4, is also a subdued feature of the leading EOF of the Southern Hemisphere monthly MSLP anomalies (Sinclair et al. 1997; Thompson and Wallace 2000). Fig. 4 also describes an out-of-phase relationship between mid-latitude and high-latitudes that is typical of the leading EOF of the Southern Hemisphere monthly MSLP anomalies (Thompson and Wallace 2000).

Fig. 4
figure 4

The spatial pattern of cross-correlation between mean ‘early winter’ MJJ NNR sea-level pressure and the corresponding MJJ average DSS Na concentration for years 1970 to 1995 AD

Full size image

The pronounced area of high positive correlation (R = +0.4 to +0.45, 95% CL) over the Southern Ocean region south of Australia in Fig. 4 represents the climatological zone of ridging, separating the quasi-stationary centres of mid-latitude, cyclonic activity south of Tasmania and New Zealand (R = –0.35 to –0.55, 99.5% CL), and in the south Indian Ocean (R = –0.5 to –0.55, 99.5% CL). This is also shown by the seasonal correlation of MJJ average Na concentration and the corresponding, zonal and meridional surface winds shown in Figs. 5a, b. High Na delivery to Law Dome is positively correlated (R = +0.4 to +0.5, 99.5% CL) to zonal wind strength in the 30° to 45°S latitude belt across the South Indian Ocean, and negatively correlated (R = –0.4 to –0.6, 99.5% CL) to zonal wind strength in the 45°S to 65°S belt. High sea-salt delivery to Law Dome is also positively correlated (R = +0.4 to +0.6, 99.5% CL) with meridional wind strength in the 120° to 140°E longitudes south of Australia. The combination of ridging and the equatorward expansion of the trough produces a stronger meridional transport from the south Indian mid latitudes to Wilkes Land, which results in anomalously high annual precipitation rates to the east of Law Dome from the steerage of cyclones over the Wilkes Land coast (Goodwin et al. 2003). Interannual precipitation rates at Law Dome are much less variable than on the East Antarctic Plateau due to the location of the former, near the Antarctic coast, and there is no clear correlation between Na concentration and annual snow accumulation. The spatial correlation pattern in Figs. 3 and 4 indicates that interannual Na concentrations at Law Dome are most sensitive to the atmospheric circulation during MJJ, and particularly to the Rossby wave structure (wave number ≥3 and phase) in the circumpolar trough, with respect to the location of Law Dome (110°E).

3.3 Characteristics of circumpolar climate during MJJ

There are a number of factors that are responsible for the strong association between the DSS average Na concentration and MSLP during early winter rather than the whole winter period. These relate to atmospheric circulation and to sea-ice variability.

Different annual cycles in tropospheric temperatures over the Southern Ocean and the Antarctic Continent produce a half-yearly cycle in MSLP and horizontal temperature gradients over the Southern Hemisphere middle and high latitudes that is known as the Semi-Annual Oscillation (SAO) (van Loon 1967; Hurrell and van Loon 1994; Simmonds and Jones 1998). The amplitude of the SAO peaks in the mid-latitude oceans between 45°S to 50°S with a secondary peak over Antarctica (Hurrell and van Loon 1994). The characteristic SAO signal in MSLP in the South Indian Ocean and Southern Ocean region, is the decrease in MSLP between March and June and a strong rise from June to September in the region east of Marion Island to Kerguelen Island (Tyson and Preston-Whyte 2000). Half-yearly variations in baroclinicity (where isotherms and isobars are at an angle to each other) are also a feature of the SAO (Meehl 1991). This semiannual oscillation in baroclinicity is important in understanding the seasonal variability in Na concentrations at DSS, since baroclinic conditions are necessary for the advection of mid-latitude airmasses to the Antarctic coast. At latitudes between 45°S and 60°S, a minimum in the temperature gradient and baroclinicity occurs in MJJ and DJF, whilst a maximum occurs in March–April, and September-October. This is reversed in the mid-latitudes north of 45°S and south of 60°S with maxima occurring during June and December (van Loon 1967). North of 50°S, the strongest westerlies and lowest MSLP occur during June, and the circumpolar trough is at its weakest in the annual cycle (Meehl et al. 1998). Thus during June there is the greatest potential for ridging and meridional transport to Antarctica. Meridional transport and baroclinicity are also significantly influenced by the static stability of the atmosphere. Walland and Simmonds (1999) showed that the weak vertical stability of the mid-latitude atmosphere during early winter (May and June), together with the weakened westerlies in the higher latitudes, combined to produce favourable conditions for the development of baroclinic waves. The maximum baroclinicity in the Indian Ocean sector during MJJ ensures a strong southerly meridional circulation between the mid-latitudes and the Antarctic in the sector south of Australia.

3.4 Correlation analyses with seasonal average SAM index data for MJJ

The above analysis shows that the average Na concentration in Law Dome precipitation is strongly related to the wave number and phase in the mid-latitude and sub-Antarctic MSLP field, on annual time scales and particularly during the MJJ season. Both the spatial correlation patterns for annual average and MJJ Na concentration and the respective MSLP and wind fields shown in Figs. 3, 4 and 5 are consistent with a Rossby wave number 3 circulation, and describe an out-of-phase relationship between mid-latitude and high-latitudes that is typical of the leading EOF of the Southern Hemisphere MSLP field. To test whether the MJJ average DSS Na concentration is a suitable proxy for the SAM we compared the DSS Na concentration data with the SAM index data of Thompson and Solomon (2002) that was derived from averaging the monthly 500 hPa geopotential height anomalies (sign-reversed) from radiosonde data observed at seven Antarctic stations between 1969 and 1995. Figure 6 shows that the variability in the DSS Na concentration (that represents the Indo-Pacific circumpolar circulation) is correlated to the SAM index for the MJJ season (R = 0.53, DOF = N-2 = 25 years, CL>99.5%). Higher (lower) Na concentrations are recorded in years with a negative (positive) SAM index (equatorward shift and weakened westerlies south of Australia) circulation. This relationship confirms that the MJJ average DSS Na concentration is a reliable proxy for SAM conditions, and that it is possible to reconstruct the multi-decadal climate variability associated with the SAM over the past millennium.

Fig. 5
figure 5

a The spatial pattern of cross-correlation between mean ‘early winter’ MJJ NNR zonal wind and the corresponding MJJ average DSS Na concentration for years 1970 to 1995 AD. b The spatial pattern of cross-correlation between mean ‘early winter’ MJJ NNR meridional wind and the corresponding MJJ average DSS Na concentration for years 1970 to 1995 AD

Full size image
Fig. 6
figure 6

Mean ‘early winter’ MJJ Na concentration plotted for each year from 1969 to 1995 AD. Also plotted is the corresponding MJJ average SAM index data of Thompson and Solomon (2002) that was derived from averaging the monthly 500 hPa geopotential height anomalies from radiosonde data observed at seven Antarctic stations between 1969 and 1995

Full size image

4 Calibration of Law Dome sodium data with early winter (MJJ) spatial patterns of mid–latitude MSLP and weather systems

The data analysis and discussion in the preceding sections suggest that the Na concentration in Law Dome precipitation is enhanced by strong meridionality of circumpolar and mid-latitude circulation as indicated by a high Rossby wave number and a negative index state of the SAM. In this section we investigate the interdecadal variability of both the DSS Na concentration and mid-latitude station MSLP for MJJ and examine whether the DSS Na concentration for MJJ can be used as a reliable proxy for decadal mid-latitude circulation variability.

4.1 Calibration to mid–latitude station MSLP

Both the respective spatial correlation patterns for MJJ DSS Na concentration and MJJ MSLP, and MJJ zonal surface winds, shown in Figs 4 and 5a, indicate that the highest correlations occur in the mid-latitudes south of eastern Australia and New Zealand (150° to 170°E) and in the South Indian Ocean sector. Within these regions monthly MSLP station data are available as follows: Macquarie Island (54°S, 158°E) for 1948–2001; Campbell Island (52°S, 169°E) for 1941–2001; Marion Island (46°S, 37°E) for 1948–2002; and Kerguelen Island (49°S, 70°E) for 1951–2000. These time-series permit direct correlation analyses and calibration between station MSLP and the Na series, as follows.

Statistically significant correlations are seen between the MJJ Na series and Campbell Is. MSLP (R = –0.32, P < 0.03, N = 54), Macquarie Is. MSLP (R = –0.36, P < 0.05, N = 48) and Marion Is. MSLP (R = –0.41, P < 0.02, N = 48). Correlation for MSLP at Kerguelen Is. is weak and not significant, due in part to a shorter time series, but more importantly because of its unsuitable geographic location in the NNR SLP-Na correlation field (Fig. 4). Kerguelen Is. lies within the variable boundary region between quasi stationary ridges and troughs in some years.

The station correlations are influenced considerably by the variability at the interannual timescale, with significant variability at periods of 2–5 years, as is discussed in Section 5.1. In considering the multiannual to decadal time scale, it is useful to attenuate the short-term correlations, and so the regression analyses were repeated after low-pass filtering with a half-power cutoff of 5 years (i.e. a Gaussian filter with standard deviation of 0.663 years, roughly equivalent to a moving average of 2.2 years). This introduces autocorrelation into the series which was compensated by reducing the number of effectively independent observations (N′) by the factor τ = –1/(lnρ1) (where ρ1 is the lag-1 autocorrelation of the filtered data). This gives a stronger correlation, but with reduced confidence due to the relative shortness of the series compared with the variability time scale. Significant correlation is seen in the southwest Pacific region at Campbell Is. (R = –0.50, P < 0.025, N′ = 21) and Macquarie Is. (R = –0.48, P < 0.05, N′ = 18), whilst in the southwest Indian region, Marion Is. shows only marginal significance (R = –0.44, P < 0.07, N′ = 18), and Kerguelen, as above, is not significantly correlated. Figure 7 shows the filtered Na MJJ series and MSLP time series (as anomalies) for the four stations.

Fig. 7
figure 7

Law Dome MJJ Na time series 1940–1995 AD (red) with SLP anomalies for four Indo-Pacific, mid-latitude meteorological station sites over the period of available overlapping instrumental data. All data are low-pass filtered with a Gaussian running mean (half-power cutoff of 5 years, as described in the text) to emphasise multi-annual to decadal variability

Full size image

Regression slopes for the filtered time series give sensitivities of –0.65 hPa/(μEq/L) and –0.67 hPa/(μEq/L) for Campbell Is. and Macquarie Is., respectively: the two most strongly correlated sites. For Marion Is., the regression implies a sensitivity of –0.57 hPa/(μEq/L).

4.2 Investigation of MJJ MSLP fields and weather systems associated with interannual variability in Law Dome Na concentration

The spatial distribution in interannual MSLP anomalies in the southern mid-latitudes associated with the SAM are illustrated in Fig. 8 which depicts the NNR MSLP anomalies for MJJ computed for the highest 6 Na years (negative SAM phase), the lowest 6 Na years (positive SAM phase), and the difference between the patterns for the highest 6 Na years minus and the lowest 6 Na years (i.e. approximately the upper and lower quartiles in the Na range for the 1970 to 1995 period). Previous research by Sinclair et al. (1997) has established the relationship between the leading EOF of the Southern Hemisphere MSLP field (High Latitude Mode) and weather systems for each phase. Accordingly we discuss the spatial MSLP features in Fig. 8 with respect to the findings of Sinclair et al. (1997) for the South Indian Ocean and southwest Pacific Ocean regions.

Fig. 8
figure 8

The spatial pattern of ‘early winter’ MJJ NNR sea-level pressure anomalies calculated for: a the 6 highest (1990, 1977, 1992, 1970, 1986 and 1975) Na concentration years at DSS between 1970 to 1995 AD; b the six lowest (1982, 1981, 1995, 1985, 1993 and 1971) Na concentration years; and c from the difference between the six highest and six lowest Na concentration years. The difference plot accentuates the circulation changes between SAM phases, in the mid-latitude trough in the southwest Indian and Pacific Ocean sectors, together with the enhanced ridging in the circum-polar trough centred on 110°E during high Na concentration years at DSS

Full size image

The MSLP anomalies associated with high Na concentration at DSS are shown in Fig. 8a and display a strong wave number 3 pattern. Extremely high cyclone activity in the mid-latitudes between 35°S to 55°S is represented by the low MSLP anomalies of –2 hPa in the region south of eastern Australia and New Zealand, and down to –3 hPa in an elongated region of the South Indian Ocean from 0°E to 80°E. Sinclair et al. (1997) found a linear cyclone response to the ‘High Latitude Mode’ with low MSLP anomalies near 40°S coupled to a 15–20% increase in cyclone frequency. In contrast high MSLP anomalies in Fig. 8a of up to 7 hPa occur across the circumpolar region south of Western Australia to the Antarctic coast, and across the Antarctic Continent. In summary, the climate during high Na years at DSS (negative SAM phase) is similar to that defined by Sinclair et al. (1997) for the HLM. It is typified by weakened westerlies at 55°S, fewer cyclones in the circumpolar trough, close to Antarctica, a weakened subtropical anticyclone over the South Indian Ocean, higher cyclone frequency across the mid-latitude Indian and southwest Pacific Oceans, together with enhanced ridging over the circumpolar trough south of Australia. These are the ideal synoptic conditions for the observed high Na aerosol transport to the Antarctic coast near Law Dome.

In contrast, Fig. 8b displays the MSLP anomalies for the low Na concentration years (positive SAM phase). The most obvious differences are the low MSLP anomalies of –1 to –3 hPa over Antarctica, and a westward rotation of the wave number 3 pattern by 30° to 40°. In the mid-latitudes, an area of high MSLP anomalies up to +5 hPa is centred at 50°S, 200°E (160°W) to the east and south-east of New Zealand. A less intense area of high MSLP anomalies of up to +2 hPa exists over the South Indian Ocean between 60°E to 90°E. A weak low MSLP anomaly of –1 hPa exists over the Tasman Sea, southeastern Australia and the circumpolar trough between 130°E to 150°E. This pattern is consistent with the findings of Sinclair et al. (1997) and defines below average cyclone activity (15–20% less frequency) in the mid-latitude areas east of New Zealand and in the elongated area over the South Indian Ocean near 40°S to 50°S. The substantial reduction in cyclone frequency over the mid-latitude South Indian Ocean is associated with strengthened westerlies south of 55°S close to the Antarctic coast, and results in an equivalent reduction in Na aerosol transport to Law Dome. The positive MSLP anomalies in the South Indian Ocean at 45°S are associated with an increase in anticyclone frequency and suggest a poleward shift in the subtropical anticyclone, also noted by Sinclair et al. (1997).

The positive MSLP anomaly east of New Zealand in Fig. 8b is similar to the pattern described in Sinclair et al. (1997) as a signature for anticyclone blocking centred on 55°S and 180°E, with cyclone tracks deflected northwards over the Tasman Sea. Trenberth and Mo (1985) established that blocking to the southeast of New Zealand was often associated with a similar wave number 3 pattern. This similarity between the MSLP anomaly pattern (Fig. 8b) and the blocking signature in Sinclair et al. (1997) indicates that winters characterised by low Na concentration at DSS may be used as proxies for atmospheric conditions conducive to mid-latitude anticyclonic blocking to the southeast of New Zealand.

The difference in MSLP between the high and low Na concentration years at DSS is shown in Fig. 8c. In summary the high Na (negative SAM index) years at DSS are related to negative MSLP anomalies of 3 to 4 hPa in the South Indian and southwest Pacific Ocean, together with positive MSLP anomalies of 1 to 5 hPa in the sector south of Australia, and higher MSLP anomalies over East Antarctica, when compared to low Na (positive SAM index) years.

5 Decadal variability of early-winter southern mid latitude circulation

We have shown a robust association between the mid-latitude circulation and delivery of sea-salt aerosols to the Law Dome region during early winter. We suggest that the DSS Na concentration data for MJJ can be reliably applied as a proxy for early winter, mid-latitude MSLP changes in the Indo-Pacific sector, thus extending the short instrumental record covering just the past 50 to 60 years, to cover the past 700 years. To facilitate this discussion we converted the low-pass filtered (5 year half power cutoff, as in Sect. 4.1) DSS Na concentration data for MJJ spanning the full period from 1300–1995 AD to anomaly data with respect to the mean 1950–1995 period, which overlaps with the instrumental record. The MJJ DSS Na concentration anomaly data were then converted to proxy MJJ MSLP anomaly data for the mid-latitude southwest Pacific Ocean by applying the mean linear regression relationship for Campbell Is. and Macquarie Is., as described in the previous section (i.e. +1μeq/L = –0.66 hPa). The low-pass filtered proxy MJJ MSLP anomaly data are shown in Fig. 9.

Fig. 9
figure 9

Sodium MJJ time-series from AD 1300–1995 in μEq/L are shown as anomalies from the 1950–1995 mean (dashed line). Data are low filter. Right (inverted) axis shows equivalent MSLP anomalies for the southwest Pacific region (Campbell Is. and Macquarie Is.) using the observed calibration slope of 0.66 hPa/(μEq/L) derived from the instrumental period (see text). Negative (positive), states of the Southern Annular Mode correspond to negative (positive) MSLP anomalies at mid latitudes

Full size image

5.1 Spectral characteristics

The proxy MJJ MSLP anomaly timeseries shows a change in character through the ∼700 year period that is readily apparent in spectral analyses shown in Fig. 10. The series was analysed by the multitaper method (MTM) of spectral estimation and also singular spectrum analysis (SSA), using the UCLA Toolkit (Dettinger et al 1995). Over the full-period, 1300–1995 AD, several spectral features appear with high significance, and are identified by both SSA and MTM. Features above the 99% significance level in the MTM spectrum are seen at the following periods: 2.18 ∼2.33 years, 3.2 years, 3.9 years and 10.5 years. Two further periodicities are evident at 4.6 years and 23 years, but only at high significance (>99%) in the latter portion of the record, from 1700–1995AD. Spectral analysis of the annual average Na series (Souney et al., 2002) showed significant (>99%) features at 25.4 years (close to the 23 years periodicity seen here), and at 36.5 years period. The analysis here of the MJJ data also shows a feature at ∼35 years period, but with lower significance (>95%), and, in common with the 23 years signal, only through the latter portion of the record. Cook et al. (2000) find a significant (>95%) 22-year periodicity in Tasmanian temperatures reconstructed from Huon Pine dendrochronology. They also show that this temperature reconstruction is strongly correlated with mid-latitude sea-surface temperatures in the Indian Ocean. Their result, from a record of ∼3.6 ky length shows that this periodicity is a persistent feature of the regional climate.

Fig. 10
figure 10

Multitaper method spectra (three tapers, time-bandwidth product = 2) of the Na MJJ series for different time-periods. The dashed levels on each plot indicate 99%, 95% and 90% confidence levels for an AR(1) process. Periods for features above 99% CL are labelled in italics

Full size image

The shorter-period signals lie in the general band of variability associated with ENSO, which is not unexpected, given the dominant influence of ENSO on the southwest Pacific and Australian regions and the connection demonstrated here between the Na series and the mid-latitude winter MSLP field.

The 10.5 years periodicity is interesting, because it is very strong, particularly in the early part of the record, and is close to the ∼11 years period of the solar Schwabe-cycle. In SSA decomposition of the Na MJJ series, this signal is captured by the leading non-trend principal components with high significance (>97.5%). The ∼10.5 years signal falls in amplitude and significance in the latter part of the record, from ∼1700AD onward, and this change is accompanied by the appearance of the strong 23 years signal: approximately double in period. Reconstruction of the ∼10.5 years signal (by either SSA or MTM) permits comparison with sunspot data, however direct reliable sunspot observations are only available from 1700 AD. Given the observed weakening in the ∼10.5 years Na periodicity around this time, it is not surprising that the reconstruction does not track the Schwabe component in the sunspot data, and the issue of solar influence remains unclear.

In summary, the changing character of the proxy MJJ MSLP anomaly series between the early (1300–1600 AD) and late (1700–1995 AD) periods, may be described by different spectral signatures: strong components at 2.33 years and 10.5 years in the early period, 2.18 years, 3.9 years, 4.6 years and 23 years in the late period, and a weaker, but persistent 3.2 years signal throughout both.

5.2 Interdecadal mid-latitude climate variability

In this section, we discuss the proxy mid-latitude climate history for the southwest Pacific region, as calibrated from the recent half-century of station MSLP data and the NNR data. We acknowledge that the overlapping time interval between the Law Dome Na data and the station MSLP data is relatively short when compared to the full 700 year record. However, we have shown that interannual to multiannual variability in the early winter mid-latitude atmospheric circulation is the principle mechanism controlling the respective variability in Na delivery to Law Dome in coastal East Antarctica. A recent study by Hall and Visbeck (2002) has shown in both observational data and model output that ocean circulation, sea-ice cover and extent are coupled to atmospheric variability described by the SAM. They found that positive (negative) SAM phases were associated with advection of sea-ice northwards (southwards) and greater (less) sea-ice extent. However, Hall and Visbeck (2002) found in the Southern Ocean sector between 80°E and 180°E that the association between the SAM and sea-ice is very weak. As stated, a previous study by Souney et al. (2002) found no statistical relationship between Na concentration in Law Dome snow and Southern Ocean sea-ice coverage and extent, on monthly, seasonal and interannual time scales. There is no evidence to suggest that the MJJ DSS Na and equivalent proxy MSLP time series is significantly influenced by temporal variability in sea-ice characteristics during the past 700 years. Hence, we interpret the MJJ DSS Na time series as indicative of the behaviour of the atmospheric component of the SAM.

After approximately 1600 AD the interdecadal variability in the reconstructed MSLP is typically ±1 hPa around a mean value close to the 1950–1995 reference mean. The earlier part of the record, up to 1600 AD, is characterised by higher interdecadal variability (±1–2 hPa) and lower MSLP anomalies around –1 hPa (note inverted MSLP axis). The negative MSLP anomalies in this earlier part of the record, correspond to a negative SAM index state (enhanced cyclone frequency in the mid-latitudes) compared to the post 1600 AD period.

The period after 1500 AD is marked by a tendency toward slower variations (23 years) and a weakly-positive mean SAM (enhanced westerlies in the 50° to 65°S zone) compared to the early part of the record. Very low interdecadal variability (±0.5 hPa) and a sustained positive SAM index state occurred during 1518–1551 AD and 1793–1833 AD. We suggest that blocking anticyclones in the New Zealand region and a strengthened subtropical anticyclone over the Indian Ocean characterised these periods, and also during 1886–1903 AD and 1920–1929 AD.

Intervals of high cyclone frequency in the mid-latitudes are defined by peaks in negative MSLP anomaly throughout the record. Periods of extreme mid-latitude cyclone frequency in the South Indian Ocean and southwest Pacific Ocean occur during the 1340s, 1370s, 1580s and from 1910–1916 AD. We interpret the circulation during these periods to be characterised by a high Rossby wave number (3 or greater) with ridging south of Australia, based on our analyses of enhanced Na transport events to Law Dome in the past decade. This interpretation is supported by modelling output on the behaviour of the SAM that indicates a shift towards the negative index state can be enhanced by higher Rossby wave number patterns in the longwave trough (Limpasuvan and Hartmann 2000). Typically, these periods experienced the highest interdecadal variability in the record (±2–3 hPa). The high variability is most likely the product of a change in the phase of the standing wave pattern, and the associated location and persistence of meridional ridging south of Australia and New Zealand.

In summary, the period 1300–1600 AD typically experienced a more meridional circumpolar circulation than the past 400 years, whilst the latter has experienced a strengthening and possible poleward shift in the winter position of the subtropical anticyclone over the Indian Ocean, together with strengthened westerlies south of 55°S, and conditions more conducive for blocking anticyclones southeast of New Zealand.

6 Conclusions

The instrumental record of the circum-Antarctic atmospheric circulation is too short to explore decadal variability. In this study we have demonstrated that the average monthly Na concentration for MJJ in the Law Dome ice core can be robustly applied as a proxy for the early winter mid-latitude MSLP field in the Indo-Pacific region. Fortunately, the longitude of the Law Dome site is sensitive to the phase of the stationary Rossby wave number 3 pattern in the longwave trough, and the meridional circulation from northern Australia to the Antarctic. It is these characteristics, combined with a high snow accumulation rate that have resulted in the site’s suitability to record the variability in mid-latitude cyclone frequency and circumpolar westerly circulation in Indo-Pacific region. Although the temporal overlap between the instrumental monthly MSLP data for stations in the mid-latitudes is restricted to the past 50 years, we believe that the relationships established by linear regression for the respective change in Na concentration and MSLP anomaly are robust and apply for the full 700 year record. This conclusion is based on the linear relationship between the average MJJ Na concentration and the SAM index, and the linear response of cyclone frequency to high-latitude pressure variability (Sinclair et al. 1997). Our interpretation of past circulation changes is dependent upon the secular robustness of the circulation patterns over the last 25 years. Since the winter circulation in the Indo-Australian region is locked to the anticyclonic vorticity and the creation of a subtropical Rossby wave source over the Australian continent (Kiladis and Mo 1998), we believe that our interpretations are valid over the past seven centuries.

The 700 year record of MSLP variability is most significant for understanding mid-latitude cyclone frequency over the South Indian Ocean and the southwest Pacific Ocean, together with the meridional circulation between southern Australia and Antarctica. Whilst our knowledge of the forcing of the SAM is limited, this proxy data set indicates on centennial time scales that the SAM was in a more negative index state prior to 1600 AD, compared to the past 400 years. Our results are consistent with the conclusions of Kreutz et al. (2000) on the analysis of the Siple Dome ice core record from West Antarctica, that the strength of the polar vortex and decadal variability of the present winter circumpolar atmospheric circulation is similar to that since 1450 AD. The observed trend towards the positive SAM index and strengthening tropospheric and stratospheric polar vortex in recent decades has been linked to global warming and anthropogenic-influenced photochemical ozone losses (Fyfe et al. 1999; Kushner et al. 2001; Thompson and Solomon 2002). Our proxy record of interdecadal variability in the SAM includes the recent trend since 1970. However, we interpret that this trend is part of multidecadal variability that characterises the past 700 years.