Geophysical Research Letters

Tropical American-Atlantic forcing of austral summertime variability in the southern annular mode

Authors


Corresponding author: B. Yu, Climate Research Division, Environment Canada, Toronto, Ontario, Canada. (bin.yu@ec.gc.ca)

Abstract

[1] The relative importance of tropical versus extratropical forcing on the southern annular mode (SAM) during the austral summer is explored using dry atmospheric model experiments with prescribed atmospheric forcing. The tropical forcing regions are determined based on areas of high observed correlations between atmospheric heating and the SAM. Diagnosed atmospheric forcing over the tropical America-Atlantic is found to be the major tropical driver of the SAM trend and interannual variability. Synoptic eddies tend to reinforce and maintain the SAM-associated circulation anomaly. Contribution of the tropical diagnosed forcing, as a whole, to the SAM trend over 1951–2010 is about one third in strength compared to that of the southern extratropical forcing processes.

1 Introduction

[2] The southern annular mode (SAM) is the leading mode of variability in the Southern Hemisphere (SH) extratropics [Thompson and Wallace, 2000]. It is primarily an internal mode of SH variability, arising from eddy-mean flow feedbacks [e.g., Limpasuvan and Hartmann, 1999; Kug and Jin, 2009]. However, it can also be influenced to some degree by external forcings both in the tropics and in the SH extratropics [e.g., Thompson et al., 2011; Greatbatch et al., 2012]. In the tropics, at interannual time scales, the El Niño–Southern Oscillation (ENSO) projects strongly onto the SAM through changes in the eddy-driven mean meridional circulation [e.g., Seager et al., 2003; L'Heureux and Thompson, 2006]. At longer time scales, studies have shown that long-term trends in sea surface temperatures (SSTs) over the tropical Pacific and Atlantic [e.g., Ding et al., 2012] may contribute to the observed increasing trend in the austral summertime SAM, while SSTs over the Indian Ocean may offset the trend [Li et al., 2010]. In the extratropics, anthropogenic drivers such as stratospheric ozone depletion (see Thompson et al. [2011] for a review), atmospheric aerosols (e.g., Cai and Cowan, 2007], and greenhouse gas increases [e.g., Fyfe et al., 1999; Kushner et al., 2001] have all been linked to long-term trends in the SAM index. However, it remains to quantify the relative contribution of tropical and extratropical influences on the SAM trend and variability, and to determine where the main tropical driver is located.

[3] This study addresses the issues raised above by examining observational data and conducting numerical experiments to isolate and quantify the role of different forcings to the SAM trend and interannual variability. With consideration of the misrepresentation of coupled atmosphere/ocean processes over the tropical western Pacific and Indian Ocean in most AGCMs driven by specified SSTs [e.g., Wang et al., 2005; Copsey et al., 2006], our AGCM experiments are driven by diagnosed atmospheric forcing. The model we used also allows for incorporation of stratospheric influences and effects relevant to atmospheric chemical processes, described below. We concentrate on the DJF (December–February) mean, since the SAM has the greatest seasonal trend in these months [Marshall, 2003; Li et al., 2010].

2 Data, Model, and Diagnostic Methods

[4] The DJF mean observation-based SAM index [Marshall, 2003] from 1958 to 2010 was obtained from http://www.nerc-bas.ac.uk/icd/gjma/sam.html. Years are labeled according to the January dates in this study. The precipitation for the same period was obtained from the NOAA (National Oceanic and Atmospheric Administration) precipitation reconstruction data set (http://www.esrl.noaa.gov/psd/ data/gridded/data.prec.html) [Chen et al., 2002]. ENSO variability is characterized by means of the Niño3.4 index and was obtained from the Climate Prediction Center (CPC, http://www.cpc.ncep. noaa.gov/products/precip/CWlink/MJO/enso.shtml). The National Centers for Environmental Prediction–National Center for Atmospheric Research reanalysis (referred to as NCEP) [Kistler et al., 2001] data for the 60 DJFs from 1951 to 2010 were used to examine the atmospheric circulation and to derive the atmospheric forcing for our model, detailed below, and the vertically integrated diabatic heating [e.g., Yu and Zwiers, 2010].

[5] The model employed is a primitive equation dry atmospheric model [Hall, 2000; Hall et al., 2001], with resolution increased to a horizontal T31 triangular truncation and 10 vertical levels. The model's ability to reproduce the observed atmosphere and annular modes has been demonstrated in these earlier studies and in Lin et al. [2002] and Yu and Lin [2012]. The model uses a time-averaged forcing calculated from the NCEP daily data. The forcing fields were calculated for each DJF to represent the interannual variability. The diagnosed forcing involves all processes that are not resolved by the model's dynamics. Hence, the forcing fields include the diabatic forcing as well as forcing terms in the vorticity and divergence equations, in contrast to the surface forcing such as SST and sea ice. Stratospheric influences and effects relevant to atmospheric chemical processes (e.g., ozone depletion, aerosols, and greenhouse gas forcing) are included in the diagnosed forcing terms. In the extratropics, the forcing may be flow dependent and related to internal atmospheric dynamics, although the effect of the orography in the SH is weak. Because of this, when we call it extratropical forcing, it should be kept in mind that it includes many southern extratropical forcing processes.

[6] The synoptic eddy-mean flow interaction is crucial to reinforce the upper tropospheric circulation [e.g., Lau, 1988; Trenberth and Hurrell, 1994; Kug and Jin, 2009]. Thus, we examine the feedback of synoptic eddies on the anomalous circulation. The synoptic eddy vorticity forcing (Fv), which represents the geopotential tendency, can be written as inline image, where f is the Coriolis parameter, V is the horizontal wind velocity, and ζ is the relative vorticity. The prime indicates the 2–8 day band pass–filtered perturbation and the bar is DJF mean. Fv is obtained by solving the Poisson equation globally with the divergence of eddy vorticity fluxes as the forcing term [e.g., Yu and Lin, 2012].

3 Observed Relationship Between SAM and Tropical Precipitation

[7] The SAM-associated precipitation anomalies are dominated by a pair of opposing centers in the tropical central-western Pacific and positive values in the tropical America-Atlantic (Figure 1). The precipitation anomalies over the tropical America-Atlantic and western Pacific correlate positively with the SAM variability, while the precipitation over the tropical central-eastern Pacific correlates negatively with the SAM. The precipitation anomalies over the tropical Pacific also bear resemblance to those of ENSO [e.g., Wallace et al., 1998], indicating the relationship between the SAM and ENSO variability [e.g., L'Heureux and Thompson, 2006].

Figure 1.

Regression of observed precipitation anomalies on the normalized observation-based SAM index for the 53 DJFs from 1958 to 2010. Contour interval is 0.1 mm/d. The anomalies that are significantly correlated with the SAM index at the 5% level are blue shaded.

[8] The correlation relating the SAM variability and the precipitation anomalies averaged over high regression areas (with values greater than 0.1 in Figure 1) is slightly higher in the tropical America-Atlantic (0.43) than in either the tropical western Pacific (0.33) or the tropical central-eastern Pacific (−0.36). The partial correlation between the precipitation in the tropical America-Atlantic and the SAM is still 0.31 when the effect of the Niño3.4 index is eliminated, significant at the 5% level, indicating that ENSO explains only part of precipitation variability in the tropical America-Atlantic. The SAM-associated vertically integrated diabatic heating is similar to the precipitation anomaly (not shown). In addition, SST cooling over the eastern Pacific and warming over most of the remainder of the tropics is found to occur alongside the SAM trend over 1979–2009 [Ding et al., 2012]. These suggest the potentially tropical forcing influence on the SAM variability.

4 Numerical Results of Tropical/Extratropical Forcings on SAM Variability

[9] Based on the observational evidence, we conducted eight model simulations, including the control run and seven sensitivity experiments with forcings that vary interannually in the tropics and the extratropics (Table 1), to explore different forcing variability influences on the SAM. The tropical forcing regions are determined based on areas of high observed correlations between atmospheric heating and the SAM (Figure 1). An ensemble of 10 members of 3 month integrations was performed for each winter after a 50 day spin up [Yu and Lin, 2012].

Table 1. List and Description of Experimentsa
NameDescriptionInterannual Forcing Domain
  1. a

    The climatologically averaged forcing over the 60 DJFs is used beyond the specified areas in the experiments excluding the control simulation.

CtrControl runGlobal
ExtExtratropical runSouthern extratropics (30°S–90°S)
TrpTropical runTropics (30°S–30°N)
TITropical Indian run(30°S–30°N, 40°E–110°E)
TPTropical Pacific run(30°S–30°N, 110°E–90°W)
TATropical America-Atlantic run(30°S–30°N, 90°W–40°E)
TPwTropical western Pacific run(30°S–30°N, 110°E–150°E)
TPcTropical central-eastern Pacific run(30°S–30°N, 150°E–90°W)

4.1 Simulation of SAM Variability

[10] The SAM variability is represented by the first EOF (empirical orthogonal function) of linearly detrended DJF mean 500 hPa geopotential anomalies (Ф500) over the southern extratropics (20°S–90°S), as characterized in many previous studies. The EOF1 accounts for 34.4% (25.6%) of the total interannual variance over 1951–2010 for the NCEP reanalysis (the Ctr ensemble mean) and is well separated from subsequent EOFs as per the criterion of North et al. [1982]. The SAM structure (Figure 2, upper panels) resembles the earlier results [e.g., Li et al., 2010]. Even though there are slight differences, there is generally good correspondence between the NCEP and the model simulation, notably the large amplitude loadings in the polar regions and midlatitudes, with pattern correlation of 0.95 poleward of 20°S.

Figure 2.

(Upper panels) EOF1 patterns of linearly detrended Ф500 over the southern extratropics (20°S–90°S) for the 60 DJFs from 1951 to 2010 for the (left) NCEP reanalysis and (right) ensemble mean of the control simulation. Contour interval is 80 m2 s−2. (Lower left) The NCEP's SAM index for the 60 DJFs (black) and the normalized observation-based SAM index for the 53 DJFs from 1958 to 2010 (blue). (lower right) The NCEP's SAM index (black) and time series of Ф500 anomalies projected on the NCEP's EOF1 pattern for the ensemble mean of the control simulation (red), together with the correspondingly linear trends. Results of the ensemble mean plus (minus) one intermember standard deviation of the control simulations are also indicated by the light (dark) green curves.

[11] The time series of NCEP's Ф500 anomalies projected onto the NCEP's EOF1 pattern is used as the NCEP's SAM index (Figure 2, lower left). The correlation between this index and the observed SAM index (Figure 2, lower left) is 0.93 over the overlapping 53 DJFs from 1958 to 2010, significant over the 1% level. Hence, the NCEP reanalysis captured most of the observed SAM variability. The lower-right panel of Figure 2 displays the time series of Ф500 anomalies projected on the NCEP's EOF1 for the Ctr ensemble mean, together with its linear trend, and compares them to those of the NCEP's SAM index. Here we project the model results onto the NCEP's leading EOF to facilitate a quantitative comparison between NCEP and the simulations. The NCEP's SAM index exhibits a strengthening over 1951–2010 (trend = 0.38 per decade with significance level over 0.1%; Table 2), consistent with previous studies [e.g., Fogt et al., 2009]. The t test is applied to test the slope parameter of linear trend [e.g., von Storch and Zwiers, 1999]. The series also exhibits marked interannual variability. The control run reasonably well captures the SAM variability. The time series analysis reveals that the Ctr ensemble mean accounts for 76% (0.29/0.38) of the NCEP's trend and 86% (0.93 × 0.93) of the NCEP's variability (Table 2). In addition, the spread of the model simulation between 10 Ctr members is quite small, indicating that the SAM variability in the ensemble mean is mainly caused by the imposed interannual forcing rather than the variability generated by the atmospheric internal dynamics.

Table 2. Linear Trends (1/Decade) of the Time Series of Ф500 Anomalies Projected on the NCEP's EOF1 Pattern over 1951–2010 (SAM trend), the Temporal Correlation Coefficient Relating the Time Series of NCEP's Ф500 Anomalies Projected on the NCEP's EOF1 Pattern and the Time Series of Ф500 Anomalies of the Ensemble Means of the Model Simulations Projected on the Same EOF1 (Corr-Ф500), and the Temporal Correlation Coefficient Relating the Time Series of NCEP's Ф250 (Fv250) Anomalies Projected on the NCEP's SAM-Associated Ф250 (Fv250) Regression Pattern and the Time Series of Ф250 (Fv250) Anomalies of the Ensemble Means of the Model Simulations Projected on the Same Regression Pattern (Corr-Ф250(Corr-Fv250))a
 SAM TrendCorr-Ф500Corr-Ф250Corr-Fv250
  1. a

    Bold values indicate trends or correlations significant at the 5% level, assuming one degree of freedom (DOF) per DJF (58 DOFs).

NCEP0.381.001.001.00
Ctr0.290.930.920.71
Ext0.420.890.870.53
Trp0.140.320.370.34
TI0.040.210.150.11
TP0.05−0.15−0.13−0.09
TA0.110.310.320.38
TPw0.030.130.150.13
TPc0.07−0.17−0.11−0.16

4.2 Contributions of Tropical and Extratropical Forcings

[12] The extratropical forcing accounts for a significant amount of the SAM trend and interannual variability, whereas the tropical forcing only explains a small fraction of them. The analysis of Ф500 anomalies projected on the NCEP's EOF1 pattern for the different forcing simulations reveals that the SAM trend in Ext is 10% larger than in NCEP, while Trp explains 37% of the NCEP's trend (Table 2). The contribution of the southern extratropical forcing to the SAM trend is about three times of magnitude than that of the tropical forcing (0.42/0.14, both are significant at the 5% level). The difference between the sum of SAM trends in Trp and Ext runs and the trend in the control run is also clearly evident. This difference may arise from potential influences of the northern extratropics and/or somehow nonlinear interaction between the tropical and extratropical influences on the SAM trend, which remain to be investigated. In addition, Ext accounts for 79% of the NCEP's SAM variability, whereas Trp explains only 10% of the NCEP's (Table 2).

[13] The SAM trend and variability positively respond to the tropical forcings over the Indian Ocean and the America-Atlantic but negatively respond to the Pacific forcing (Table 2). The TA-driven SAM trend is the largest one among these three, and its associated variability is the only one that is significantly correlated with the NCEP's. This indicates that the tropical American-Atlantic forcing is the main tropical driver of the secular trend and interannual variability of the SAM. In addition, over the tropical Pacific, the western Pacific forcing contributes to the SAM trend and variability, while the central-eastern Pacific forcing offsets them (Table 2).

[14] The SAM is characterized by an equivalent barotropic structure in the troposphere [e.g., Thompson and Wallace, 2000]. Thus, similar geopotential anomalies are apparent at 250 and 500 hPa in association with the SAM variability (c.f. shading in Figure 3, left, with Figure 2, upper left). The SAM-associated 250 hPa synoptic eddy vorticity forcing (Fv250) anomalies collocate well with Ф250 anomalies over the extratropical region (Figure 3, left), indicating that the synoptic eddies are systematically reinforcing and helping to maintain the SAM-related circulation anomaly, consistent with previous studies [e.g., Limpasuvan and Hartmann, 1999; Kug and Jin, 2009]. The right panels of Figure 3 further display the time series of Ф250 and Fv250 anomalies projected on the corresponding NCEP's SAM-associated regression patterns for the ensemble means of the forcing simulations. Similar to that in Ф500, Ext accounts for a large amount of the SAM-associated Ф250 variability, whereas Trp explains only a small fraction of the variability (Figure 3, upper right, and Table 2). Meanwhile, the TA-driven Ф250 variability is the only tropical one that is significantly correlated with the NCEP's. In addition, the time series of Fv250 anomalies in the Ctr, Ext, Trp, and TA runs projected on the NCEP's SAM-associated Fv250 regression pattern correlate positively with that in the NCEP's, significant at the 5% level (Figure 3, lower right, and Table 2), indicating that similar synoptic eddy feedback mechanisms exist in these experiments and in the NCEP reanalysis. Synoptic eddies tend to reinforce and maintain the SAM-related circulation anomalies in these simulations.

Figure 3.

(left panel) Regressions of NCEP's Ф250 (shading in m2 s−2) and Fv250 (contours in 10−5 m2 s−3) anomalies on the NCEP's SAM index. (right panels) Time series of (upper) Ф250 and (lower) Fv250 anomalies projected on the corresponding NCEP's regression patterns (left panel) for the ensemble means of the forcing simulations, together with the corresponding trends.

5 Discussion and Conclusions

[15] Large changes in the global atmospheric observing system in the 1970s [e.g., Bengtsson et al., 2004], such as the introduction of satellite data in 1979, may have an effect on some of our analysis. Nevertheless, the NCEP's SAM index compares well with the observed index and does not exhibit abrupt changes in the late 1970s. In addition, similar SAM variability is seen in the European Centre for Medium-Range Weather Forecasts (not shown), a combination of ERA-40 (1958–1979) and ERA-Interim (1980–2010) reanalysis data [Uppala et al., 2005; Berrisford et al., 2009] for the 53 DJFs. Furthermore, an additional experiment as in Ext but linearly removing the forcing anomalies in association with the tropical precipitation variability, characterized by the leading principal component of the observed tropical precipitation, produces similar SAM variability (trend = 0.42 per decade, Corr-Ф500 = 0.81, as in Table 2) as in Ext. This increases confidence that the SAM variability in Ext is mainly driven by the southern extratropical forcing, and implies that the tropical and extratropical influences on the SAM variability can largely be attributed separately.

[16] In summary, observational evidence indicates that the atmospheric heating over the tropical America-Atlantic and western Pacific–eastern Indian Ocean (tropical central-eastern Pacific) is positively (negatively) correlated with the SAM variability. Numerical experiments with diagnosed atmospheric forcings confirm these relationships and further indicate that the tropical -Atlantic forcing dominates over the forcings in either tropical Pacific or tropical Indian Ocean for the SAM trend and interannual variability. Synoptic eddies systematically reinforce and maintain the SAM-associated circulation anomaly. The contribution of the tropical diagnosed forcing to the SAM trend over 1951–2010 is about one third in strength compared with the forcing associated with southern extratropical processes, such as ozone depletion and greenhouse gas forcing.

Acknowledgments

[17] We are grateful for helpful comments from K. Szeto and J. Cole. We thank two anonymous reviewers for their constructive suggestions and comments, which helped to improve the paper.