Interpretation of Recent Southern Hemisphere Climate Change

See allHide authors and affiliations

Science  03 May 2002:
Vol. 296, Issue 5569, pp. 895-899
DOI: 10.1126/science.1069270


Climate variability in the high-latitude Southern Hemisphere (SH) is dominated by the SH annular mode, a large-scale pattern of variability characterized by fluctuations in the strength of the circumpolar vortex. We present evidence that recent trends in the SH tropospheric circulation can be interpreted as a bias toward the high-index polarity of this pattern, with stronger westerly flow encircling the polar cap. It is argued that the largest and most significant tropospheric trends can be traced to recent trends in the lower stratospheric polar vortex, which are due largely to photochemical ozone losses. During the summer-fall season, the trend toward stronger circumpolar flow has contributed substantially to the observed warming over the Antarctic Peninsula and Patagonia and to the cooling over eastern Antarctica and the Antarctic plateau.

The atmosphere of the SH high latitudes has undergone pronounced changes over the past few decades. Total column ozone losses have exceeded 50% during October throughout the 1990s (1–3), and the Antarctic ozone “hole” reached record physical size during the spring of 2000 (4). The lower polar stratosphere has cooled by ∼10 K during October-November since 1985 (5,6), and the seasonal breakdown of the polar vortex has been remarkably delayed: from early November during the 1970s to late December during the 1990s, in both the troposphere (7) and the lower stratosphere (1,7–9). At the surface, the Antarctic Peninsula has warmed by several K over the past several decades, while the interior of the Antarctic continent has exhibited weak cooling (10, 11). Ice shelves have retreated over the peninsula and sea-ice extent has decreased over the Bellingshausen Sea (12–14), while sea-ice concentration has increased and the length of the sea-ice season has increased over much of eastern Antarctica and the Ross Sea (14–16). Here, we offer evidence that illuminates the connections between these seemingly disparate trends.

We examine climate trends in the high-latitude SH using 30 years (1969–1998) of monthly mean radiosonde data from seven stations located over Antarctica (Table 1), 32 years (1969–2000) of monthly surface temperature data observations, 30 years (1969–1998) of ground-based total column ozone measurements from Halley station, and 22 years (1979–2000) of tropospheric geopotential height data from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis (17). The choice of periods used in the analysis was motivated by the facts that (i) the radiosonde and surface temperature records are most complete after ∼1969, (ii) the radiosonde data are only available through 1998, and (iii) the NCEP/NCAR reanalysis agrees best with the radiosonde data after ∼1979 (18).

Table 1

Antarctic radiosonde stations used in this study (17).

View this table:

The mean atmospheric circulation of the mid-high latitude SH is dominated by a westerly circumpolar vortex that extends from the surface to the stratosphere. The vortex is strongest during midwinter in the stratosphere, when polar temperatures are coldest, and is weakest during the summer months, when the circulation at levels above ∼30 hPa reverses sign and becomes weak easterly. The vortex exhibits considerable variability on month-to-month and year-to-year time scales. As shown in Fig. 1, it has also exhibited a pronounced trend over the past few decades. The temperature and strength of the SH polar vortex were estimated by averaging temperature and geopotential height anomalies, respectively, over the seven Antarctic radiosonde stations listed in Table 1 at levels throughout the depth of the troposphere and stratosphere. Because variability in geopotential height is largest in high latitudes, low geopotential heights over the Antarctic continent are consistent with anomalously strong westerly flow along ∼60°S, and vice versa. In the lower polar stratosphere, the trends are dominated by falling geopotential height and cooling that peaks during the SH spring months but persists throughout the summer. Temperature drops exceeding 6 K per 30 years are evident near 100 hPa from October through December, and geopotential height decreases in excess of 300 m per 30 years are evident at levels upward of 30 hPa from November through December (Table 2). The summertime stratospheric trends are weaker than their springtime counterparts, but nonetheless exceed 1 standard deviation (SD) of the respective monthly time series. The data also reveal a secondary peak in the stratospheric geopotential height and temperature decreases during May.

Figure 1

30-year (1969–98) linear trends in temperature (T, left) and geopotential height (Z, right) averaged over the radiosonde stations listed in Table 1. Trends are plotted as a function of calendar month and vertical level. Contours are drawn at 40 m per 30 years (–60, –20, 20 …) and 1 K per 30 years (–1.5, –0.5, 0.5, … ). Dashed contours de- note geopotential height and temperature decreases. Shading denotes trends that exceed 1 SD of the respective monthly time series.

Table 2

30-year (1969–98) linear trends in geopotential height (Z) and temperature (T) averaged over the radiosonde stations listed in Table 1 for levels indicated (meters per 30 years and K per 30 years). Trends exceeding 1 SD of the respective monthly anomaly time series are in bold type.

View this table:

The substantial trends in the lower stratospheric circulation during the spring months have been documented in previous studies (5–9). Radiative and observational analyses have demonstrated that these trends are dominated by the development of the Antarctic ozone hole (1, 5, 6,9, 19), with smaller but important contributions from increases in well-mixed greenhouse gases (20) and stratospheric water vapor (19). The results in Fig. 1 andTable 2 further reveal that similar trends have occurred in the troposphere and that there are important differences in the timing of the stratospheric and tropospheric trends. Whereas the stratospheric trends peak during November, the most pronounced tropospheric trends occur during the summer months of December-January. Like their stratospheric counterparts, the tropospheric trends exhibit “dual maxima,” with peak values occurring during both summer (December-January) and fall (April-May) (21).

The observed trends in geopotential height over the SH polar regions are consistent with a trend toward the high-index polarity of the SH annular mode (SAM), a large-scale pattern of variability that dominates the SH extratropical circulation on week-to-week and month-to-month time scales [(22–27) the SAM is also referred to as the High-Latitude Mode (23) and the Antarctic Oscillation (27)]. Months corresponding to the high-index polarity of the SAM are characterized by cold polar temperatures, low geopotential height over the polar cap, and strong circumpolar flow along ∼60°S. Months corresponding to the low-index polarity are marked by anomalies in the opposite sense. Like its Northern Hemisphere (NH) counterpart (28), the SAM is evident year-round in the troposphere but is coupled to the circulation of the lower stratosphere during seasons when the stratospheric polar vortex is perturbed by waves dispersing upward from the troposphere (27). Theory predicts that this coupling should occur when the lower stratospheric circulation is westerly but less strong than a threshold value (29). In the SH, these conditions are met when the polar vortex is decaying (late spring/early summer) and when it is building (fall). As shown in Fig. 1and Table 2, it is roughly during these seasons that the trends in the tropospheric circulation are largest.

Time series of geopotential height anomalies in the lower polar stratosphere during November (when the stratospheric trends are largest) were generated by averaging geopotential height anomalies at 30 hPa over all seven Antarctic radiosonde stations. Similarly, time series of lower stratospheric geopotential height and the SAM during December-January (when the tropospheric trends are largest) were found by averaging geopotential height anomalies at 30 and 500 hPa, respectively (30). Consistent with previous studies linking ozone depletion with the observed changes in SH spring circulation (1, 5, 6, 9,19), November values of polar stratospheric geopotential height are strongly correlated with concomitant values of total column ozone from Halley station (Fig. 2 andTable 3). Variability in the lower stratospheric circulation during November is also strongly related to variability in the lower stratospheric circulation during December-January, a relationship that presumably reflects the thermal memory of the lower stratosphere and the photochemical memory inherent in the ozone field (1). In turn, variability in the stratospheric circulation during December-January is strongly coupled with the SAM in the troposphere (Fig. 2 and Table 3). Hence, the results of Figs. 1 and 2 imply not only that the impact of springtime ozone losses on the circulation of the lower stratosphere extends into the SH summer months but also that it extends to the circulation of the troposphere. That the trends in springtime ozone exceed those in 30-hPa height reflects the facts that the response of temperature to reduced ozone is highly nonlinear and is also affected by the slow time scale for radiative relaxation. The direct correlation between November stratospheric circulation anomalies and December-January values of the SAM index is weaker than the concomitant link during December-January but nevertheless exceeds the 95% confidence level (31). The significance of this correlation is not sensitive to shared trends in the time series [see also (32)].

Figure 2

Time series of total column ozone at Halley station and geopotential height anomalies averaged over the radiosonde stations listed in Table 1 for levels and seasons indicated. Negative values denote low values of total column ozone and geopotential height over the pole, and vice versa. Total column ozone is given in Dobson units, and geopotential height is given in SD. In NovemberZ 30, 1 SD corresponds to roughly 250 m; in December-January Z30, roughly 100 m; and in December-January Z 500, roughly 40 m. TheZ 500 index can be viewed as an inverted index of the SAM (30).

Table 3

Correlation statistics for the time series shown inFig. 2 (31).

View this table:

Similar coupling is also evident between the SAM and the strength of the lower stratospheric polar vortex during April-May, when the tropospheric trends exhibit a secondary maximum. The stratospheric trends during this season are not strongly linked with values of total column ozone during the previous spring (33) but are consistent with the onset of winter season cooling, together with depleted ozone during April-May (34, 35).

The trend toward the high-index polarity of the SAM is most pronounced during the December-January and April-May seasons but persists throughout the summer months. The fraction of recent SH high-latitude climate trends that are linearly congruent with the SAM were found by regressing monthly mean values of 500-hPa geopotential height, near-surface (925 hPa) winds, and high-latitude surface temperature observations onto standardized December-May monthly mean values of the reanalysis-based SAM index (30) and then multiplying the resulting regression coefficients by the trend in the SAM index (Fig. 3, right panels). The fraction of the trends that are linearly congruent with the SAM are relatively insensitive to the trends in the time series (32). The high-index polarity of the SAM favors geopotential height decreases throughout the polar cap and anomalously strong westerlies over the Southern Ocean. Because the SAM is accompanied by thermally indirect vertical motions at polar latitudes, its high-index polarity also favors cooling over much of Antarctica (27). An important exception to this cooling is found over the Antarctic Peninsula (27), where anomalously strong westerlies should act to decrease the incidence of cold air outbreaks from the south and lead to increased warm advection from the Southern Ocean. A similar warm patch is observed over the southern tip of South America, but correlations for surface temperatures equatorward of Patagonia were not found to be significant (31), underscoring the fact that the SAM principally influences the polar regions of the SH.

Figure 3

December-May trends (left) and the contribution of the SAM to the trends (right). Top, 22-year (1979–2000) linear trends in 500-hPa geopotential height. Bottom: 32-year (1969–2000) linear trends in surface temperature and 22-year (1979–2000) linear trends in 925-hPa winds. Shading is drawn at 10 m per 30 years for 500-hPa height and at increments of 0.5 K per 30 years for surface temperature. The longest vector corresponds to ∼4 m/s.

The signature of the SAM is strongly reflected in both the pattern and amplitude of recent trends in the SH tropospheric circulation and surface temperatures over the SH high latitudes during the summer-fall season (Fig. 3, left panels). Roughly 44 m of the 51-m-per-22-year (1979–2000) December-May decreases in 500-hPa geopotential height averaged poleward of 65°S, ∼1.0 of the ∼1.1 K per 32 years (1969–2000) December-May cooling averaged over the stations in eastern Antarctica and the pole, and ∼0.7 of the ∼1.4 K per 32 years (1969–2000) warming averaged over the temperature data in the peninsula region are linearly congruent with the recent trend in the SAM index (32). Hence, not only are the marked trends in the tropospheric circulation during the summer-fall season consistent with the recent trend in the SAM, but so is a substantial fraction of the disparate trends in surface temperature anomalies over the Antarctic continent. The link between the SAM and surface temperature trends is consistent with the observed regionally varying trends in Antarctic sea ice (e.g., decreases in sea ice near the peninsula accompanied by increases over eastern Antarctica). That the trend in the SAM accounts for only ∼50% of the warming over the peninsula attests to the importance of other climate change mechanisms over this region.

The observed trend in the SAM toward stronger circumpolar flow is in the same sense as the trends that have dominated the NH extratropical circulation over the past few decades (36–38) but display important differences in their seasonality. Whereas the trends in the SH troposphere peak during the summer and fall seasons, the trends in the NH troposphere peak during the midwinter months January-March (38). The occurrence of positive trends in both the SAM and the NH annular mode (NAM) suggests that the trends reflect processes that transcend the high-latitude climate of a particular hemisphere. Presumably, any climate change mechanism that projects onto the meridional temperature gradient in the middle-high latitudes may affect the polarity of the annular modes. Recent trends in tropical sea-surface temperatures have been shown to affect the NAM (39), and it has been hypothesized (but not yet demonstrated) that a similar link may exist for the SAM (7). Several modeling studies run with increasing greenhouse gases have also simulated the recent trend in the NAM (40–42), and at least three simulations run with increasing greenhouse gases and/or stratospheric ozone losses have derived trends in the modeled SAM that are of the same sign as the observations presented in this paper (42–44). The strong correlation between the SH stratospheric and tropospheric circulations at a time when the SH polar stratosphere is strongly affected by photochemically induced cooling, as revealed in this study, suggests that Antarctic ozone depletion has played an important role in driving secular variability not only in the climate of the high-latitude SH stratosphere but also at the Earth's surface there.

The memory in the temperature of the stratospheric polar vortex from spring through summer and the coupling between the stratosphere and troposphere during the late spring/early summer months suggest that stratospheric anomalies during spring may provide predictive skill for SH tropospheric climate on month-to-month time scales. Similar predictive skill has already been demonstrated in the NH, where large-amplitude anomalies in the strength of the NH wintertime stratospheric polar vortex frequently precede similarly signed anomalies in the lower troposphere (45–47). The statistically significant correlation between the strength of the SH stratospheric polar vortex during November and the tropospheric circulation ∼1–2 months later (Table 3) suggests that a similar link may be important for predicting the climate impacts of the SAM, at least as far north as Patagonia. The “downward propagation” of SH circulation anomalies from the stratosphere to the troposphere during the late spring months is, for example, illustrated by the vertical profile of geopotential height anomalies averaged over the polar cap region during the 2001 spring-summer season (Fig. 4), which constitutes an independent sample from that used in the rest of the analyses. Consistent with the results shown in Figs. 1 and 2, the largest negative anomalies in geopotential height originate at levels above ∼50 hPa during middle November and descend into the troposphere in December. Hence, the month of December 2001 was marked by anomalously high values of the SAM index (Fig. 4, bottom). The time lag of several weeks between anomalies in the stratospheric and tropospheric circulations is consistent with results calculated for the NH (45, 46).

Figure 4

(Top) Geopotential height anomalies (meters) averaged 65°–90°S during spring 2001 based on data from the NCEP-NCAR reanalysis. (Bottom) Standardized values of the SAM index (not inverted), as defined in (30). In the bottom panel, vertical tickmarks denote 1 SD of the SAM index.

We have presented evidence that recent trends in the SH tropospheric circulation and surface temperatures over the Antarctic continent are consistent with a systematic bias toward the high-index polarity of the SAM. The work underscores and clarifies the manner in which high-latitude SH climate change over the past several decades has been characterized not only by changes in hemispheric mean temperature (48) but also by important changes in the high-latitude circulation.

  • * To whom correspondence should be addressed. E-mail: davet{at}


View Abstract

Navigate This Article