Tropical and mid-latitude forcing of continental Antarctic temperatures

Future changes in atmospheric circulation and associated modes of variability are a major source of uncertainty in climate projections. Nowhere is this issue more acute than across the mid-latitudes to high latitudes of the Southern Hemisphere (SH), which over the last few decades have experienced extreme and regionally variable trends in precipitation, ocean circulation and temperature, with major implications for Antarctic ice melt and surface mass balance. Unfortunately there is a relative dearth of observational data, limiting our understanding of the driving mechanism(s). Here we report a new 130-year annually resolved record of δD – a proxy for temperature – from the geographic South Pole where we find a significant influence from extratropical pressure anomalies which act as “gatekeepers” to the meridional exchange of air masses. Reanalysis of global atmospheric circulation suggests these pressure anomalies play a significant influence on midto high-latitude SH climate, modulated by the tropical Pacific Ocean. This work adds to a growing body of literature confirming the important roles of tropical and mid-latitude atmospheric circulation variability on Antarctic temperatures. Our findings suggest that future increasing tropical warmth will strengthen meridional circulation, exaggerating current trends, with potentially significant impacts on Antarctic surface mass balance.


Introduction
Uncertainty surrounding future changes in atmospheric circulation is exacerbated in the Southern Hemisphere (SH) by the acutely short baseline of observations available only since the onset of continuous monitoring during the mid-20th Century and satellite 20 era (post-1979) (Bracegirdle et al., 2015;Delworth and Zeng, 2014;Hobbs and Raphael, 2010;Marshall, 2003), despite the extreme and often contrasting trends in precipitation, ocean circulation and temperature (Gille, 2014), and Antarctic ice melt (Rye et al., 2014) which it has experienced in recent decades. In the high latitudes of the SH, the limited available data suggest opposing trends over the last few Introduction warming (up to 2.5 • C since 1950) (Ding et al., 2011), observations and reanalysis data have provided contrasting trends over continental East Antarctica (Jones and Lister, 2014). A potentially confounding factor is the pronounced temperature inversion that exists over much of the high altitude plateau for most of the year (up to 10 • C over 30 m) (Hudson and Brandt, 2005), making the interpretation of surface reanalysis 5 data problematic (Fréville et al., 2014). In marked contrast, radiosonde observations obtained through the upper atmosphere appear to provide a more robust signal, with pronounced warming observed across the mid-troposphere (centred on 650 hPa) during all seasons above the interior of East Antarctica, independent of the trends at the surface but limited to the last five decades (Screen and Simmonds, 2012;Turner et al., 2006). Previous studies on Antarctic climate have suggested a significant low-and midlatitude influence (Ding et al., 2011;Jouzel et al., 1983;Screen and Simmonds, 2012;Shaheen et al., 2013). For instance, an early study at the South Pole suggested snow δD closely mirrored the Pacific Decadal Oscillation over the previous century (Ekaykin 15 et al., 2004). More recently, a relationship between surface temperature increases over West Antarctica, global sea surface temperature (SST), and atmospheric circulation has recently been identified (Ding et al., 2011), implying that recent warming may be coupled to the tropical Pacific. Records extending the mid-troposphere temperature record (Screen and Simmonds, 2012;Turner et al., 2006) are therefore needed to 20 better understand the driving mechanism(s) of warming over the continent.
Fortunately, clear-sky precipitation (or diamond dust) -the isotopic signal of which preserves the temperature during formation in the mid-troposphere -dominates the total precipitation over much of the continental interior (Casey et al., 2014;Ekaykin et al., 2004;Hou et al., 2007); unlike coastal areas where different precipitation sources 25 contribute to significant amounts of accumulation, diamond dust is continuously formed in the low atmosphere on the plateau during clear days, where most of the crystals are generated in the relatively moist upper part of the surface inversion (approximately 600-700 hPa) (Ekaykin et al., 2004;Fujita and Abe, 2006;Walden et al., 2003) these precipitation events, thick plates/short column crystals < 150 µm are commonly observed in the South Pole atmosphere Hogan, 1978, 1976;Walden et al., 2003), often producing optical effects, such as halos. Crucially, although the quantities of this type of precipitation on a daily basis can be relatively low (Bromwich, 1988;Massom et al., 2004), estimates suggest that over a year diamond dust accounts for 5 50 to 91 % of accumulation on the plateau, including the South Pole (Casey et al., 2014;Ekaykin et al., 2004;Fujita and Abe, 2006;Hou et al., 2007). Diamond dust on the plateau therefore offers an opportunity to investigate long-term forcing of Antarctic temperatures, supplementing radiosonde data in the region.

10
In January 2012 a 2-metre snow pit was dug and contiguous snow samples were taken at 5 cm intervals at the South Geographic Pole (Fig. 1). The snow pit was extended to 7 m depth by a snow core (sub-sampled at 2.5 cm) obtained using a 7.5 cm diameter Kovacs corer (hand cored). δD analysis on the samples was undertaken on an IsoPrime mass spectrometer at the Antarctic Climate and Ecosystems Cooperative
The 2012 core overlapped with a previously reported δD sequence for the South

25
Geographic Pole across the year CE 1977(Jouzel et al., 1983. The chronology for the older part of the sequence was based on visible stratigraphic observations, deuterium 4022 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | maxima, tritium and β peaks (the latter across the 1960s and 1970s) (Jouzel et al., 1983), providing a robust chronology back to 1887 ( Fig. S1 in the Supplement). To investigate multidecadal trends and reduce the impact of any missing years in the new combined South Pole record, we used a 30-year smoothing through the sequence.

5
The resulting composite record of temperature-sensitive δD therefore extends the radiosonde record back before 1961 (Screen and Simmonds, 2012;Turner et al., 2006) (Fig. 3). Prior to the 1940s we observe high inter-annual variability but with relatively stable long-term (30-year running mean) δD values, implying temperatures did not vary significantly in the mid-troposphere over this period. From the 1940s through to the 1960s, however, we identify a trend to lower temperatures (depleted deuterium content) that was reversed in the late 1960s and sustained up to present day.
To explore the potential role of SH circulation on temperatures over the South Geographic  the low pressure anomalies associated here with warming over the South Geographic Pole and wider Antarctica may be associated with the zonal wave-3 pattern (ZW3) (Raphael, 2004) which plays a significant role in meridional flow across 45-55 • S on interannual to decadal timescales (Steig et al., 2009). ZW3 has been linked to changing sea ice concentration in the Southern Ocean (Raphael, 2007), the delivery of 20 rainfall to southwest Australia (van Ommen and Morgan, 2010) and recent warming over West Antarctica (Steig et al., 2009). Estimates obtained across the period of satellite observations (post-1979) suggest ZW3 explains some 8 % of SH circulation but hemispherically-averaged reconstructions only extend back to the 1960s (Raphael, 2004), hampering our understanding of its significance on multi-decadal timescales.

25
Long-term changes in ZW3 and its influence on the SH remain highly uncertain, with limited observational data in a marine-dominated realm. Recent work has suggested ZW3 provides a mechanism for the meridional exchange of air masses, implying a strengthening since the 1970s (Schneider et al., 2004;van  2010), a trend coupled models struggle to reproduce (Holland and Raphael, 2006;Landrum et al., 2012;Steig et al., 2009). We observe fields at 700 hPa composited for 10th percentile (negative) surface pressure anomalies in the southeastern Indian Ocean (panels a-e of Figs. 6 and S4) and southwestern Pacific Ocean (panels a-e of Figs. 7 and S5) are associated with 5 an increase in geopotential height (GPH) over the Antarctic continent. Both the ERA (Figs. 6 and 7) and 20CR (Figs. S4 and S5) generate similar patterns, providing confidence that large scale processes are being appropriately captured. By examining the meridional mass streamfunction (Fig. 7), we find that in comparison with the situation in the intermediate state, the negative anomaly state shows a weakening (reduced volume) of the Polar cell and a strengthening (increased volume) of the Ferrel cell (particularly on the poleward side of the cell), consistent with increased poleward heat transport. In marked contrast, the opposite pattern is observed with positive pressure anomalies in the mid-latitudes, with a stronger Polar cell (particularly with the SIO region) and weaker Ferrel cell (most marked with the SWP), suggesting 15 a greater dynamical barrier to poleward heat transport.
To explore whether this relationship has changed over time (pre-satellite era) we undertook wavelet coherence analysis between the Pacific and Indian Ocean pressure anomalies as extracted from the 20CR with our newly obtained South Geographic Pole δD record (Fig. 9). Focusing on the second half of the twentieth century (considered 20 to have more reliable reconstructive ability over the mid latitudes of the SH) (Compo et al., 2011), we observe a significant shift in the dominance of the different pressure anomalies. During the period of the 1940s through to the late 1960s, low-pressure anomalies centered on the south Indian Ocean lead warming over the South Pole across the periodicity 6-12 years (arrows pointing up and to the left); in the southwest 25 Pacific, the pressure anomalies did not lead but were in antiphase over the same periodicity but did not lead (arrows pointing to the left). But in the 1970s, this coherence ended in the south Indian Ocean and the leading pressure anomaly switched to the southwest Pacific (arrows pointing up), paralleling the 30-year trend in these  (Fig. 3), suggesting a switch in the dominant influence of different regions over continental Antarctic temperatures. Further work is required to extend the observational record of the mid-latitudes beyond the mid-1940s. Intriguingly, the spatial plots of pressure anomalies vs. Antarctic temperature suggest the tropical Pacific may play a role. During positive pressure anomalies, 5 the eastern Pacific also experiences higher sea level pressure and is associated with cooling over continental East Antarctica (Fig. 4b). Conversely, with increasing El Niño-Southern Oscillation (ENSO) variance (Fig. 3) (McGregor et al., 2013), the southwest Pacific pressure anomaly apparently weakens (Fig. 4a). Previous studies have demonstrated tropical Pacific variability has a significant influence on the mid-to 10 high-latitudes (Turner, 2004;Vance et al., 2012), particularly over the west Antarctic where atmospheric teleconnections are relatively strong (Bromwich et al., 2000;Ding et al., 2011). Whilst the southwest Pacific pressure anomaly exhibits a significant relationship with annual average GPH in the tropical west Pacific (Fig. 7), absolute differences in the tropics are relatively small (Fig. 8) and may be masked by seasonal 15 changes. To investigate this more fully, we undertook seasonal correlations between changes in southwest Pacific pressure, GPH (700 hPa) and sea surface temperatures across the period 1979-2013 (Fig. 10). We observe a significant inverse relationship with the tropical west Pacific between September and May but find no significant correlations during the Austral winter (June to August). Specifically, we find low 20 pressure over the southwest Pacific is associated with the centres of low pressure in the south Indian Ocean and the southwest Atlantic Ocean, high temperatures in the Amundsen Sea and the central and east tropical Pacific, and relatively high pressure over the tropical west Pacific (Fig. 10) from the tropical west Pacific down to the Antarctic via the southwest Pacific during the summer (Fig. 10). Circulation changes associated with El Niño appear to lead to a relative weakening of the pressure anomalies in the southwest Pacific Ocean (Fig. 3), strengthening the Ferrel cell while weakening the Polar cell (Fig. 8). In contrast, the negative pressure anomaly in the tropical east Pacific identified in Fig. 4b is consistent 5 with the La Niña-like conditions and an intensified Walker Cell circulation, increasing atmospheric pressure in the east. Importantly, the recently observed increase in ENSO variability over the last three decades (with an associated increase in temperature over the Antarctic) implies the recent dominance of the southwest Pacific pressure anomaly over meridional air exchange in the mid-latitudes (Fig. 3) is exceptional in the context of the past six centuries (McGregor et al., 2013). If correct, the observed links to Antarctic temperatures suggest long-term changes in the tropics may in fact play a significant role in driving high-latitude temperatures (Figs. 4 and 10), with potential implications for past climate states (Pedro et al., 2011;Visser et al., 2003).

15
By extending historical observations over the South Pole and in combination with reanalysis products we find that meridional circulation changes associated with centres of pressure anomalies are part of a broader change observed over recent decades. Of particular note is the marked decrease in rainfall in southwest Australia since the 1970s. Our results demonstrate this trend is part of a hemispheric pattern of alternating shows the opposite relationship i.e. the area of negative temperature anomalies produced by compositing months of positive surface pressure anomalies (thresholded at the 90th percentile). The grid spacing is 15 • in longitude and latitude. For reference, the area of the Antarctic continent is 14 million km 2 and the area poleward of 65 • S is 25 million km 2 .