1. Introduction
The southern annular mode (SAM), also called the Antarctic Oscillation or zonal index, is the leading mode of variability of the extratropical circulation in the Southern Hemisphere (SH) on time scales from weeks to decades (e.g., Trenberth 1979; Szeredi and Karoly 1987; Kidson 1988; Karoly et al. 1996; Hartmann and Lo 1998; Gong and Wang 1999; Robinson 2000; Thompson and Wallace 2000; Marshall 2003; Deser et al. 2012). The SAM is characterized by a nearly zonally symmetric seesaw pattern of pressure/geopotential height anomalies between the high latitudes and the midlatitudes in the SH extratropics (Fig. 1), which is dominantly caused by a meridional swing in the position of the eddy-driven westerly jet that typically sits at ~50°S. Because baroclinicity from the westerly jet is the main source for growth of extratropical weather systems, the SAM is tied with north–south shifts of storm tracks (e.g., Karoly 1990; Kidson and Sinclair 1995; Lorenz and Hartmann 2001; Rashid and Simmonds 2004). The SAM therefore plays a primary role for variability of weather and climate of the SH, impacting rainfall, surface air and ocean temperatures, and even upper ocean heat and momentum transport (e.g., Silvestri and Vera 2003; Reason and Rouault 2005; Gillett et al. 2006; Sen Gupta and England 2007; Hendon et al. 2007; Meneghini et al. 2007; Hendon et al. 2013).
(top) Regression pattern of MSLP onto (bottom) the standardized Gong and Wang (1999) SAM index derived from the NCEP–NCAR surface pressure reanalysis (Kalnay et al. 1996). A 3-month running mean was applied. In the upper panel the contour interval is 0.5 hPa and the regression is scaled by one standard deviation (1σ) of the SAM index to preserve the unit.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
The SAM is understood to develop through internal atmospheric dynamics, and this is reflected in its decorrelation time of ~10 days (e.g., Robinson 2000, and references therein). Furthermore, the basic characteristics of SAM variability, including its temporal variability and spatial structure, are captured in models without any low-frequency forcing (e.g., Limpasuvan and Hartmann 1999; Rashid and Simmonds 2004). Positive feedback between transient eddies and the eddy-driven jet is understood as a critical mechanism for determining the meridional displacement and decorrelation time scale of the eddy-driven jet and, therefore, for determining the phase of SAM and its persistence (e.g., Karoly 1990; Limpasuvan and Hartmann 1999; Lorenz and Hartmann 2001; Rashid and Simmonds 2004, 2005; Gerber and Vallis 2007; Hartmann 2007; Kidston et al. 2010).
Although low frequency (i.e., from monthly to seasonal) variations of the SAM do occur owing to unforced internal variability, slow variations of tropical sea surface temperatures (SSTs), especially those associated with the El Niño–Southern Oscillation (ENSO) during the SH warm seasons, also promote variations in the SAM (e.g., Karoly 1989; Seager et al. 2003; Silvestri and Vera 2003; Zhou and Yu 2004; Codron 2005; L'Heureux and Thompson 2006; Lu et al. 2008; Chen et al. 2008; Gong et al. 2010; Fogt et al. 2011; Lim et al. 2011). A basic dynamical explanation of this association is that equatorial SST anomalies during ENSO can directly cause changes in the intensity and latitudinal extent of the Hadley circulation (e.g., Oort and Yienger 1996; Seager et al. 2003; Hartmann 2007). These changes then lead to changes in the subtropical westerly jet, thereby shifting the critical latitudes at which extratropical baroclinic waves break as they propagate to low latitudes (e.g., Seager et al. 2003; L'Heureux and Thompson 2006; Lu et al. 2008; Chen et al. 2008). Hence, eddy momentum flux convergence and the eddy-driven jet shift meridionally from their climatological positions (toward the equator during El Niño), which is then expressed in the surface pressure/geopotential fields as a swing of the SAM (e.g., toward the negative phase of the SAM during El Niño).
This chain of interactions appears to happen most favorably during the SH summer when the mean structure of the SH atmospheric circulation is most zonally symmetric, the subtropical jet is indistinguishable from the eddy-driven jet, and the SST anomalies associated with ENSO (here we are referring to traditional, eastern Pacific ENSO) typically mature (e.g., Karoly 1989). The lack of a relationship between ENSO and the SAM in austral winter is consistent with the findings of Lu et al. (2008), who found that shifts of the eddy-driven jet are decoupled from variations of the subtropical jet and the Hadley cell in austral winter. Barns and Hartmann (2010) also found no active feedback between transient eddies and the zonal mean flow when the subtropical jet dominates over the eddy-driven jet, as occurs during the SH winter.
The relationship between the SAM (or meridional shifts of the eddy-driven jet) and ENSO hints that low-frequency variations of the SAM should be predictable at least during austral summer. Nevertheless, predictability and prediction of seasonal variations of the SAM has been paid little attention despite obvious potential benefits from skillful seasonal prediction of the SAM for regional climate in the Southern Hemisphere. Hence, this study is aimed at investigating the predictability of the low frequency component of the SAM in association with ENSO and assessing the prediction capability of seasonal variations of the SAM, using the Australian Bureau of Meteorology (BoM) dynamical seasonal climate prediction system [the Predictive Ocean and Atmosphere Model for Australia (POAMA)].
The observational datasets, forecast model configuration, and retrospective forecasts (reforecasts/hindcasts) are described in section 2. We then review and diagnose the relationship between the SAM and tropical SST variations associated with ENSO using the reanalysis datasets in the period 1980–2010 in section 3. In section 4 the forecast model abilities to predict ENSO and to simulate the teleconnection of tropical SSTs to the SAM will be examined, and in section 5 forecast skill of predicting seasonal variations of the SAM will be explored. Finally, concluding remarks will be provided in section 6.
2. Data and methodology
a. Forecast model and verification data
The POAMA is an atmosphere–ocean coupled system used for making operational seasonal climate predictions in Australia. POAMA consists of the BoM spectral Atmospheric Model version 3 (BAM3) (Colman et al. 2005) and the Australian Community Ocean Model version 2 (ACOM2) (Schiller et al. 2002; Oke et al. 2005) based on the Geophysical Fluid Dynamics Laboratory (GFDL) Modular Ocean Model, version 2 (MOM2) (Pacanowski 1996). The atmosphere and ocean models are coupled using the Ocean Atmosphere Sea Ice Soil (OASIS) coupler (Valke et al. 2000). The atmospheric model is run with triangular truncation at wavenumber 47 (~300 km grid) and 17 vertical levels. The zonal resolution of the ocean model is 2°, and the meridional spacing is 0.5° within 8° of the equator, increasing gradually to 1.5° near the poles. The ocean model has 25 levels in the vertical with 12 levels in the top 185 m.
Forecasts from the current operational version of POAMA (version 2; i.e., POAMA2) are initialized from observed atmospheric and oceanic states. The atmosphere model is initialized with realistic atmosphere and land initial conditions generated from a nudging scheme called Atmosphere and Land Initialisation (ALI) (Hudson et al. 2011), and the ocean model is initialized with ensemble ocean initial conditions generated from the BoM's newly developed POAMA Ensemble Ocean Data Assimilation System (PEODAS) (Yin et al. 2011).
The ALI generates a set of atmosphere–land initial states by nudging zonal and meridional winds, temperature, and humidity from the atmospheric model of POAMA toward an observationally based analysis (Hudson et al. 2011). ALI nudges to the reanalyses from the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) (Uppala et al. 2005) for the earlier period of hindcasts (1980–2001) and toward the analyses from the BoM operational global NWP system for the later period of hindcasts (2002–10) and for the real-time forecasts. The resultant atmospheric fields are similar to the observationally based analyses, causing less initial shock than if the ERA-40 or the NWP analyses are used directly as initial conditions. The land surface conditions (soil moisture and temperature) are also initialized by bringing them into balance with the conditions of the atmospheric surface fields produced by ALI.
In PEODAS available in situ temperature and salinity observations are assimilated (Yin et al. 2011). PEODAS generates an ensemble of ocean initial conditions covering the range of observational errors in surface wind stress and heat fluxes on the intraseasonal time scale, and these conditions are directly used to make ensemble forecasts. The PEODAS ocean analyses are produced every 3 days.
To address forecast model drift that negatively impacts the representation of the ENSO in the model, the ensemble of forecasts of POAMA2 is generated using three slightly different model versions that produce different degrees of model mean state bias. Two of the three versions use a different treatment of shallow convection (that results in less climate drift at longer lead times), and one of the two uses an explicit flux correction scheme to control the model climate drift (Lim et al. 2010, Wang et al. 2011).
For each of the three versions an ensemble of 10 members is initialized on the first day of each month for 1960–2010. The central member is initialized from the ALI and PEODAS initial conditions, and the other nine members are initialized using ocean perturbations as provided by PEODAS. Monthly anomalies from the hindcasts were computed relative to each of the three model versions' monthly climatology. An ensemble mean forecast was obtained by averaging anomalies using all 30 ensemble members from the three versions of POAMA2, which was found to be more skillful in predicting seasonal climate than an ensemble-mean forecast comprising 10 ensemble members of any single version of POAMA2.
Hindcasts of POAMA2 for the period of 1980–2010 are verified against the SST analyses from Hurrell et al. (2008), which is the combined product of the Hadley Centre Sea Ice and Sea Surface Temperature dataset (HadISST) (Rayner et al. 2003) for pre-1982 and Reynolds Optimum Interpolation (OI) SST version 2 (Reynolds et al. 2002) for post-1982. Mean sea level pressure (MSLP) and zonal wind forecasts are verified against the National Centers for Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) reanalysis (Kalnay et al. 1996).
SAM forecasts from POAMA2 are also compared to the forecasts from two earlier versions of POAMA. The primary differences for the earlier versions of POAMA are the methods of atmosphere and ocean initialization and ensemble generation. In POAMA, version 1.5a (POAMA1.5a), which was the first operational version of POAMA during 2002–06, the ocean initial conditions were provided by a univariate (temperature only) ocean assimilation system (Smith et al. 1991), and the atmosphere and land surface initial conditions were drawn from an Atmospheric Model Intercomparison Project (AMIP)-style1 integration of the atmospheric model (Hudson et al. 2011). Its hindcasts consist of an ensemble of three members that were initialized with three different atmospheric conditions generated from three different AMIP runs. In the POAMA1.5b version (which was operational during 2007–11), the same univariate assimilation system as for POAMA1.5a was used for ocean conditions, but the atmosphere and land surface were initialized with realistic conditions generated from ALI. The POAMA1.5b hindcasts consist of an ensemble of 10 members that were generated by using successively 6 h earlier atmosphere/land surface initial conditions from ALI with a single ocean initial state. For comparison of forecast skill among the three model systems (POAMA versions 1.5a, 1.5b, and 2), ensemble mean forecasts using just three members are used for the common period 1980–2005.
b. SAM and ENSO indices
Variations of the SAM are depicted by the difference of the normalized anomalies of zonally averaged MSLP at 40° and 65°S following Gong and Wang (1999) using 3-month mean data. The observed time series of SAM is derived using the monthly NCEP–NCAR reanalysis (Kalnay et al. 1996). The seasonal (3-month mean) SAM index (SAMI) time series and the associated anomalous MSLP pattern (obtained by regressing MSLP anomalies onto the standardized SAMI) are shown in Fig. 1. During the positive phase of the SAM (high SAM), positive anomalies of MSLP occur equatorward of 50°S and negative anomalies occur poleward of 50°S (e.g., Kidson 1988; Karoly 1990; Thompson and Wallace 2000). The opposite occurs during the negative phase of the SAM (low SAM).
To confirm the link between the SAM and the meridional vacillation of the eddy-driven jet, we conducted EOF analysis on the zonal-mean zonal winds south of 20°S at 1000–100-hPa levels. The first EOF mode of zonal-mean zonal wind variability clearly depicts the equivalent barotropic meridional shift of the eddy-driven westerly jet all year round (Fig. 2) and its principal component (projection time series) is highly correlated with the SAMI in all seasons (Table 1).
Regression of zonal-mean zonal winds onto the first EOF of zonal mean zonal winds over 1000–100-hPa levels south of 20°S (U_EOF1, shading) overlaid with the climatological zonal-mean zonal winds (contours) in austral autumn (MAM), winter (JJA), spring (SON), and summer (DJF) seasons. Color shading interval is 0.5 m s−1 and the contour interval is 5 m s−1. Regressions were scaled by 1σ of the principal component time series of U_EOF1 of the four seasons.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
Explained variance (%) by the first EOF of tropospheric zonal-mean zonal winds (U_EOF1, analysis domain is south of 20°S) for 3-month mean data of each season for the period 1980–2010. The correlation between the principal component time series of U_EOF1 (U_PC1) and the SAMI in each season is given in the right-hand column.
Reflecting its nature of internally driven atmospheric process, the SAMI shows high season-to-season variability (Fig. 1, bottom panel). However, superimposed on this year-to-year variability is a substantial upward trend (0.4 standard deviation per decade with statistical significance at the 99% confidence level). A similar trend is also found in the SAM index based on station data provided by the British Antarctic Survey (http://www.nerc-bas.ac.uk/icd/gjma/sam.html; Marshall 2003) but the trend is weaker (0.25 standard deviations per decade over the same 31-yr period).
In exploring the connection of the SAM with the SST anomalies due to ENSO, we consider the SST anomalies associated with two different flavors of ENSO. Traditional or canonical ENSO (also referred to as eastern Pacific or cold-tongue ENSO) refers to ENSO events whereby the maximum SST anomaly occurs in the far eastern equatorial Pacific (hereafter, referred to as cold-tongue ENSO). Cold-tongue ENSO is depicted by the first EOF of tropical Pacific SST variability (SSTPC1; Fig. 3, left panels). The EOF analysis was computed in the domain 20°S–20°N, 120°E–60°W using 3-month running mean SST data for 1980–2010. Major cold-tongue El Niño events occurred in 1982–83, 1987–88, and 1997–98 whereas cold-tongue La Niña events occurred in 1988–89, 1998–2000, 2008, and 2010. El Niño events that have maximum SST anomaly in the central and western Pacific (e.g., Ashok et al. 2007a; Kao and Yu 2009; Kug et al. 2009) are referred to as El Niño Modoki, central Pacific, or warm-pool El Niño (hereafter, referred to as warm-pool El Niño). We monitor warm-pool El Niño by using the second EOF of tropical Pacific SST variability (SSTPC2; Fig. 3, right panels). As pointed out in earlier studies (e.g., Ashok et al. 2007a; Weng et al. 2007), SSTPC2 demonstrates multiyear variations with persistent positive amplitudes in 1990–96 and 2002–06 and persistent negative amplitudes in 1997–2000. Previous studies have reported that the impact on the atmosphere from these two types of ENSO is very different (Hoerling et al. 1997; Weng et al. 2007; Wang and Hendon 2007; Kim et al. 2009), and a recent study by Ding et al. (2012) proposed a significant relationship between SAM and SST variations in the central and western Pacific during austral winter. Therefore, it is worth exploring the association of the SAM with these different flavors of ENSO.
(top) Regression patterns of SST anomalies onto (bottom) the principal component (PC) time series of the (left) first and (right) second EOF of SSTs in the domain 20°S–20°N, 120°E–60°W. A 3-month running mean was applied. In the upper panels the contour interval is 0.3°C. Regressions were scaled by 1σ of the respective PC time series.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
3. Observed relationship between SAM and ENSO
The relationship between the SAM and the SST variations associated with the two different types of ENSO is quantified by the correlation of the SAMI with the two SSTPC time series (Fig. 4). Statistically significant correlations at the 95% confidence level (c.l.) are found between the SAMI and SSTPC1 (cold-tongue El Niño) in the austral spring and summer seasons [October–December (OND) to December–February (DJF)], consistent with the findings of earlier studies that showed a relationship between SAM and traditional ENSO during austral summer (e.g., Seager et al. 2003; Silvestri and Vera 2003; L'Heureux and Thompson 2006; Fogt et al. 2011). Interestingly, we also see that the SAMI is related to SSTPC2 (warm-pool El Niño) during the austral winter seasons [May–July (MJJ) to June–August (JJA)] although the correlation coefficients are smaller than those with SSTPC1 in the warm seasons.
Correlation of the SAMI with the PC time series of the first two leading EOFs of tropical Pacific SSTs (SSTPC1 and SSTPC2). The x-axis label indicates 3-month seasons. Correlation coefficients greater than 0.37 (0.30) are significant at the 95% (90%) confidence level (c.l.) based on a two-tailed Student's t test with 31 independent samples [i.e., degrees of freedom (d.f.) = 31; see Wilks (2006), his Eq. (5.12)]. The thin dotted horizontal line indicates the confidence threshold for 95% significance.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
This seasonal dependence of the relationship of the SAM with the two different flavors of El Niño is confirmed by the regression patterns of SSTs onto the inversely signed SAMI for JJA and November–January (NDJ), which are the seasons when the relationship of SAM with SSTPC2 and with SSTPC1 are strong, respectively (Fig. 5, upper panels). The sign of the SAMI is flipped here so that the SST patterns correspond to warm phases of ENSO (which is typically related to the low phase of SAM). For reference, the regression of SSTs onto SSTPC1 and SSTPC2 is also shown in Fig. 5 (lower panels). It is clearly seen in the figure that the SST anomaly associated with the negative phase of SAM in austral winter (JJA) highly resembles that associated with warm-pool El Niño. Also, the SST anomaly associated with the negative phase of SAM in austral summer (NDJ) highly resembles that associated with cold-tongue El Niño.
Regression of SST anomalies onto (a),(c) the standardized and inverted SAMI and (b),(d) the standardized SSTPC2 and SSTPC1 in (left) JJA and (right) NDJ. The color shading interval is 0.2°C. Stippling denotes statistical significance at the 95% c.l. assessed by a two-tailed Student's t test with d.f. = 29 (d.f. = N − J − 1 for regression, where N is independent sample size and J is the number of predictors). Regressions were scaled for 1σ of SAMI and SSTPCs 1 and 2 to preserve the unit, respectively.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
A similar set of regressions but for zonal winds at the 300-hPa level is shown in Fig. 6. In austral winter there is a strong similarity between the extratropical zonal wind anomalies associated with SSTPC2 and associated with the inverted SAMI (Fig. 6, left panels). Likewise, in austral summer a strong similarity is found between the zonal wind anomalies associated with SSTPC1 and associated with the inverted SAMI (Fig. 6, right panels). The analyses shown in Fig. 6 again confirm that the SAM is related to the two different flavors of ENSO in the SH summer and winter.
Regression of 300-hPa level zonal wind anomalies onto (a),(c) the standardized and inverted SAMI and (b),(d) the standardized SSTPC2 and SSTPC1 in (left) JJA and (right) NDJ. The color shading interval is 0.5 m s−1. Stippling denotes statistical significance at the 95% c.l. assessed by a two-tailed Student's t test with d.f = 29. Regressions were scaled for 1σ of SAMI and SSTPCs1 and 2 to preserve the unit, respectively.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
Although the mechanism for the promotion of the SAM by the cold-tongue El Niño in austral summer is well established (e.g., Seager et al. 2003; L'Heureux and Thompson 2006; Lu et al. 2008; Chen et al. 2008), the mechanism for promotion of the SAM by the warm-pool El Niño in austral winter has yet to be established. Is the mechanism for promotion of SAM by the warm-pool El Niño in austral winter similar to that for promotion of SAM by the cold-tongue El Niño in summer?
To address this question we examine the anomalies of the zonal mean temperature, mass streamfunction, zonal-mean zonal winds, and zonal-mean transient eddy momentum flux convergence2 associated with the warm-pool El Niño in JJA (Fig. 7, left panels) and cold-tongue El Niño in NDJ (Fig. 7, right panels). These anomalies are developed by regression onto SSTPC2 in austral winter (JJA) and SSTPC1 in austral summer (NDJ).
Regression of anomalies of (a),(e) zonal mean temperature, (b),(f) mass streamfunction, (c),(g) zonal mean zonal winds, and (d),(h) zonal mean transient eddy momentum flux convergence onto (left) SSTPC2 in JJA and (right) SSTPC1 in NDJ. The color shadings denote the regression and the contour lines denote the climatology in JJA and NDJ. Stippling denotes statistical significance at the 95% c.l. assessed by a two-tailed Student's t test with d.f. = 29. Regressions were scaled for 1σ of SSTPC1 and SSTPC2 to preserve the units, respectively.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
The mechanism of the summertime shift toward the negative phase of SAM during cold-tongue El Niño (or toward the positive phase of SAM during cold-tongue La Niña), as envisioned by Seager et al. (2003), is apparent in Fig. 7 (right panels). During cold-tongue El Niño, tropospheric warming occurs in the tropics (Fig. 7e) in response to the equatorial SST warming in the eastern Pacific (Fig. 5d), leading to an intensification and equatorward contraction of the Hadley circulation (Fig. 7f). Therefore, the subtropical westerly jet intensifies as a result of the Coriolis torque arising from increased upper-level outflow at low latitudes (Fig. 7g). These stronger low-latitude westerlies allow equatorward-propagating baroclinic waves to penetrate farther into the tropics before encountering their critical latitudes where they break, thereby causing an equatorward shift of the extratropical eddy momentum flux convergence (Fig. 7h) that feeds westerly momentum into the equatorward side of the eddy-driven jet; thus, the eddy-driven jet shifts equatorward (Fig. 7g), resulting in a shift toward the low phase of the SAM (e.g., L'Heureux and Thompson 2006; Lu et al. 2008). In response to the anomalous eddy momentum flux convergence, a meridional circulation is induced, which by continuity drives upward motion along about 45°S, where the troposphere cools adiabatically, and downward motion along about 25°S, where the troposphere warms adiabatically and acts to spread poleward the diabatically forced warming near the equator (Fig. 7e) [Seager et al. 2003, their Eqs. (3)–(7)]. The adiabatic warming and cooling increase the meridional temperature gradient in the 20°–40°S latitude band, which means a strengthening of vertical wind shear (by thermal wind balance), and therefore increases eddy generation there, resulting in an increase of eddy momentum flux convergence that feeds westerlies into the mean flow in the midlatitudes of 30°–45°S (Fig. 7h). These baroclinic processes thus can further promote the low phase of the SAM.
In contrast to the cold-tongue El Niño during austral summer, during the warm-pool El Niño in austral winter the zonal mean temperature shows an overall cooling in the tropics (Fig. 7a), despite the occurrence of positive SST anomalies in the central Pacific as seen in Fig. 5b. Furthermore, the austral winter Hadley circulation intensifies while slightly expanding southward (Fig. 7b) rather than contracting as is associated with El Niño in austral summer. This intensification and expansion increases the westerlies on the poleward side of the subtropical jet at 30°–40°S, acting to shift the subtropical jet poleward (Fig. 7c). A poleward shift of the subtropical westerlies should act to cause extratropical baroclinic eddies to break at higher latitudes; therefore, the eddy momentum flux divergence pattern in the low latitudes should shift poleward of its climatological position, which is the opposite change from that during austral warm season cold-tongue El Niño described above. Thus, we would expect the warm-pool El Niño to promote the high phase of the SAM rather than the low phase. However, inspection of both anomalous westerlies (Fig. 7c) and anomalous eddy momentum flux convergence (Fig. 7d) indicates that increased westerlies on the poleward side of the subtropical jet are not only supported by the anomalous Hadley circulation but also supported by anomalous momentum flux convergence in midlatitudes (30°–45°S), and westerlies and eddy momentum flux convergence both decrease in higher latitudes (45°–65°S) at the center and poleward of the eddy-driven jet, resulting in the negative phase of the SAM. What can explain this contradictory behavior whereby a shift to high SAM during the SH winter is expected to be promoted by the poleward expansion of the upper branch of the Hadley cell but the changes in eddy momentum flux convergence and the eddy-driven jet would suggest shifts to low SAM?
In answering this question, we note that the subtropical jet not only determines critical latitudes but also is a source of baroclinic eddy activity in the SH winter season. Simmonds and Keay (2000), Nakamura and Shimpo (2004), Hoskins and Hodges (2005), and Lim and Simmonds (2007) demonstrated that in austral winter in the Indo-Pacific region, where the subtropical jet is most pronounced and distinctive from the eddy-driven jet, extratropical westerlies associated with the subtropical jet act as a source of eddy activity. The climatology of the transient eddy momentum flux convergence/divergence shown in Figs. 7d,h also confirms that eddy activity and its interaction with the mean flow occurs over a much broader latitude band in JJA than NDJ when the two jets coalesce (e.g., Kim and Lee 2004). Therefore, the poleward shift and strengthening of the subtropical jet induced by warm-pool El Niños in austral winter can promote increased eddy activity on the poleward side of the subtropical jet; hence, directly causing an anomalous convergence of the eddy momentum flux on the poleward side of the subtropical jet but on the equatorward side of the eddy-driven jet; thus, acting to shift the eddy-driven jet equatorward; and, so, promoting a shift to the low phase of the SAM. Furthermore, the meridional circulation induced by the anomalous eddy momentum convergence/divergence pattern in Fig. 7d causes adiabatic warming at ~30°S but adiabatic cooling at ~45°S as evidenced in Fig. 7a, which increases the meridional temperature gradient in the midlatitudes but decreases it in the high latitudes and therefore promotes eddy generation in the midlatitudes but suppresses it in the high latitudes, respectively, also resulting in the low phase of the SAM. Interestingly, this chain of processes is consistent with the results of Lee and Kim (2003), whose idealized model experiments show that an intensified and poleward-shifted subtropical jet can draw the eddy-driven jet equatorward (i.e., a shift toward low SAM).
Regression of tropospheric zonal mean temperature onto the inverted SAMI also suggests a contrasting behavior with tropical temperatures between austral summer and winter (Fig. 8). Low SAM is associated with tropical cooling in JJA but with tropical warming in NDJ although the tropical temperature variance accounted for by the SAM is small. This contrast in tropical temperature anomalies for SAM during summer versus winter appears to have not been obviously documented in the literature.
Regression of anomalies of zonal mean temperature onto the inverted SAMI in (left) JJA and (right) NDJ displayed with color shading. Contours denote the climatology in JJA and NDJ. Stippling denotes statistical significance at the 95% c.l. assessed by a two-tailed Student's t test with d.f. = 29. Regressions were scaled for 1σ of SAMI to preserve the unit.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
Although we have suggested a plausible explanation for how the warm-pool El Niño promotes high SAM in austral winter, an interesting question is raised as to why SAM is not promoted by the cold-tongue ENSO in the SH winter. Cold-tongue La Niñas are associated with stronger zonal mean cooling in the tropics than are warm-pool El Niños, and this cooling also causes the Hadley circulation to expand and the subtropical jet to shift poleward. However, the extratropical zonal wind anomalies in the SH winter during cold-tongue ENSO are primarily zonally asymmetric owing to the presence of a strong Rossby wave that propagates poleward into the SH high latitudes from the central-eastern Pacific, which is known as the Pacific–South American (PSA) pattern (e.g., Karoly 1989). Hence, the zonally asymmetric circulation driven by the tropical heating/cooling associated with cold-tongue ENSO dominates the extratropical circulation in the SH winter (e.g., Cai et al. 2011) and masks any shift in SAM resulting from shifts in subtropical jet that is directly thermally driven from the tropics.
The above analysis, which outlines the seasonality of the relationship of SAM with tropical SST anomalies due to ENSO, implies that there should be seasonal predictability of the SAM arising from predictability of the two flavors of ENSO. We can estimate an upper limit of predictability of seasonal variations of the SAM by computing a simultaneous multiple correlation between the detrended SAMI and detrended SSTPC1 and SSTPC2 (Fig. 9). That is, we estimate how much of the seasonal variance of the SAMI we can predict if we know the behavior of SSTPC1 and SSTPC2 perfectly. Two periods of SAM predictability are inferred in Fig. 9: the extended austral winter (May–August) and the extended austral summer (October–January). Similar results are obtained if we use the station-data-based SAM index from the British Antarctic Survey data and use the data without detrending (not shown). Therefore, a dynamical coupled model forecast system such as POAMA2 should be able to exploit this implied predictability in order to make seasonal predictions of the SAM. To do so, the forecast model will need to have skill predicting the tropical SST anomalies associated with the two different flavors of ENSO and also faithfully represent the SAM and its teleconnection with tropical SSTs. We assess these capabilities in the following section prior to assessing actual forecast skill for seasonal prediction of the SAM in section 5.
Multiple correlation coefficients (R2) of SSTPC1 and SSTPC2 with SAMI. Detrended time series were used. Thick solid curves indicate statistically significant R2 at the 90% confidence level tested by an F test given two predictors and 31 samples (d.f. = 28) (Wilks 2006).
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
4. Representation of the SAM and its teleconnection with ENSO in POAMA2
EOF analysis of the forecast zonal mean zonal winds in the troposphere (850–200 hPa) suggests that the spatial structure of the SAM anomalies is well depicted by POAMA2 even at long lead time. This is demonstrated by the high pattern correlation between the observed and predicted leading EOF patterns (Table 2). However, the model first EOF mode is too dominant, explaining about 65% of the total variance on average, but reaching up to 80% in the austral summer (Table 3, cf. Table 1). Also, the correlation between the model SAMI (as given by the forecast zonal mean pressure difference between 40° and 65°S) and the model leading EOF of zonal-mean zonal winds is higher than the observed counterpart. That is, the correlation using model data is greater than 0.9 almost all year round (Table 3) whereas the correlation using observations averages about 0.8, ranging from 0.7 in JAS to 0.9 in OND (Table 1). Hence, interpretation of the following analysis needs to be tempered by the knowledge that the simulated SAM variability is more dominant in the model SH extratropical circulation than in reality.
Pattern correlation between the first EOF pattern of the forecast zonal-mean zonal winds at 0-, 3-, and 6-month lead times (LT) and the observed counterpart. The pattern correlation was computed with the ensemble mean forecasts for each 3-month season and then averaged over 12 seasons.
Explained variance (%) by the first EOF of forecast tropospheric zonal mean zonal wind variability for the period 1980–2010 (middle column) and the correlation between the resultant principal component times series (U_PC1) and the forecast SAMI (right column) at 0-, 3-, and 6-month lead times. The EOF was computed using the ensemble mean forecast whereas the correlation between the forecast U_PC1 and the SAMI was obtained as the mean of each ensemble member correlation.
We assess the simulated relationship between the SAM and ENSO in the forecast model using 3-month lead time forecasts (e.g., forecasts for MJJ are initialized on 1 February). We chose a 3-month lead time in order to exclude the influence of observed atmospheric initial conditions on the teleconnection, which is substantial at short lead times, and therefore to better depict the model's internal ability to simulate the teleconnection. We compute the regression of SSTs and tropospheric zonal mean temperature with the inverted SAMI in the forecasts for each season but display only MJJ and NDJ in Fig. 10. Variations of the SAM simulated by POAMA2 are, indeed, realistically related to tropical SSTs: the negative phase of SAM is associated with the cold-tongue El Niño and tropospheric warming over the tropics in the late austral spring to summer (Fig. 10, right panels), whereas during the late austral autumn the negative phase of SAM is associated with the warm-pool El Niño and tropospheric cooling over the tropics (Fig. 10, left panels).
Regression of anomalies of (top) SSTs and (bottom) zonal mean tropospheric temperatures onto inverted SAMI using 3-month lead POAMA2 ensemble mean forecasts for MJJ and NDJ (i.e., forecasts initialized on 1 February and 1 August, respectively). The color shading interval is 0.2°C for SSTs and 0.1°C for tropospheric temperatures. Stippling denotes statistical significance at the 95% c.l. assessed by a two-tailed Student's t test with d.f. = 29. Regressions were scaled for 1σ of forecast SAMI to preserve the unit.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
However, the model falls short in capturing the relationship of warm-pool ENSO and SAM in JJA when the relationship is strongest in the observations (not shown). This is likely due to the model's bias in the teleconnection of ENSO to the SH extratropics in JJA, for which the model simulates too zonally symmetric pressure/geopotential height anomalies related to the cold-tongue ENSO than the observation at long lead times (e.g., Lim et al. 2009a). Consequently, the model winter SAM is dominantly driven by the cold-tongue ENSO, which would negatively impact predictive skill of the SAM in JJA. The POAMA model also has meridional biases in the simulation of zonal winds and surface pressure associated with SAM, which was described by Marshall et al. (2012) in detail.
In part, these biases in zonal winds and their teleconnection to ENSO might be attributed to the coarse horizontal and vertical resolution of the POAMA's atmosphere model (~300 km with 17 vertical levels), which might be especially problematic in the SH winter when baroclinic eddy activity is maximum and its feedback with the mean flow (which requires depiction of wave breaking at small scales) is strong (e.g., Simmonds and Keay 2000; Ashok et al. 2007b, 2009). Increased horizontal and vertical resolution of the atmosphere model is one of the priorities in the development of the next version of POAMA. Another possible source of the bias in the teleconnection between cold-tongue ENSO and SAM in austral winter is POAMA's inability to simulate Rossby wave propagation associated with the Indian Ocean dipole mode (IOD) (Saji et al. 1999) in the SH winter (not shown). Because the IOD generally co-occurs with the cold-tongue ENSO (Cai et al. 2011), improvements in the simulation of the Rossby wave propagation toward the SH high latitudes from the Indian Ocean during cold-tongue ENSO should reduce the spurious zonal symmetry of extratropical circulation during cold-tongue ENSO in the model.
Nevertheless, it is still encouraging to see that POAMA2 captures the key relationship of the SAM with tropical SST, which is the foundation for seasonal predictability of the SAM. We thus expect that predictive skill of the SAM will hinge on POAMA's ability to predict cold-tongue and warm-pool El Niños. However, given the model biases, we anticipate stronger predictive skill in austral summer, when the SAM is associated with cold-tongue El Niño, and that the full potential predictability will not be achieved in austral winter, when the SAM is associated with warm-pool El Niño, for which POAMA appears to have a harder time simulating the correct linkage.
The skill of predicting SST anomalies associated with the two types of ENSO is assessed by projecting the ensemble mean seasonally averaged SST forecasts onto the observed SST EOF1 and EOF2 patterns shown in Fig. 3. The projection coefficients for EOF1 and EOF2 from the forecast SSTs are then correlated with the observed PCs (Fig. 11). As reported by Hendon et al. (2009) and Lim et al. (2009b), high skill of POAMA forecasts is seen for both SSTPC1 and SSTPC2 to lead times of 6 months, which is much higher than the skill of persistence forecasts that were obtained by projecting observed monthly SSTs in previous months onto the observed EOF patterns (e.g., persistence forecasts for 2010 DJF SSTPC1 and SSTPC2 at 0 lead time were obtained by projecting the observed November 2010 SSTs onto the SST EOF1 and EOF2 patterns).
Correlation skill of the predicted SSTPC1 and SSTPC2 from POAMA2 (thick curves) and from persistence forecasts (thin curves) displayed as a function of forecast lead time.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
These results suggest that forecasts from POAMA2 meet the necessary conditions for skillful seasonal prediction of the SAM.
5. Seasonal prediction skill of the SAM
Seasonal predictive skill of the SAM is summarized, in Fig. 12, as a function of forecast start time and lead time. Here, we assess skill using correlation and rms error (RMSE) using the ensemble mean forecasts of the SAMI for the hindcast period 1980–2010. For reference, RMSE skill is assessed relative to persistence3 and climatological forecasts. The slanted lines in Fig. 12 indicate constant verification seasons starting from the month labeled on the y axis. For instance, the slanted line starting next to the forecast start time 6 indicates skill for forecasts that verify in JJA and were initialized with 0–5-month lead times. As expected from the SAM being largely an atmospheric internal process, predictive skill of the SAM is moderate and largely confined to the shortest lead time. However, consistent with our analysis of potential predictability of the SAM based on its seasonal relationship with tropical SSTs, two distinctive periods of good skill at longer lead times are apparent: austral late autumn to early winter when skill extends to lead times of 4–5 months (skillful forecasts at the long lead times for these seasons are shown with gray boxes on the upper right side of the diagram) and austral spring to early summer when skill extends to lead times of up to 6 months.
(top) Correlation of the observed SAMI with (a) POAMA2 forecasts and (b) persistence forecasts (see the text for more details of persistence forecasts). Correlation skill that is statistically significant at the 95% c.l. assessed by a one-sided Student's t test given 31 independent samples is shaded. (A one-sided Student's t test was used for the forecast skill assessment as only positive correlation between the observation and forecasts can mean some skill.) (bottom) Skill score of RMSE of POAMA2 forecasts relative to (c) persistence forecasts and (d) climatological forecasts of seasonal SAM. Gray color shading interval indicates % improvement of POAMA2 over the reference forecasts that is statistically significant at the 95% c.l. The statistical significance was assessed by a bootstrapping method (Wilks 2006) that uses 1000 samples and rejects a null hypothesis (RMSEpoama2 ≥ RMSEreference) at the 5% level. Sloping dotted lines denote constant verification time.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
These times of high seasonal prediction skill appear to confirm that POAMA2 is able to exploit the forcing of the SAM by slow variations of boundary conditions due to two distinctive flavors of ENSO. The POAMA2 correlation skill in those high predictability seasons suggests that about 10%–25% variance of the observed SAM is captured by the forecasts, which compares well to the predictability available by the relationship of SAM with the two types of ENSO as shown in Fig. 9.
During those high predictability seasons, the correlation skill of POAMA2 is much higher than observed monthly SAM values persisting from the previous season (Fig. 12b). Likewise, RMSEs are much smaller than those of the persistence forecasts and, therefore, the RMSE skill score relative to persistence forecasts [i.e., (1 − RMSEpoama2/RMSEpersistenceF) × 100 (%)] is high (Fig. 12c). The RMSEs are also smaller than those of the climatological forecasts (Fig. 12d) during the austral late autumn to early winter and the spring to early summer periods when the correlation score is high. However, the improvement of POAMA2 over climatological forecasts is not big, which indicates that predicting the amplitude of SAM is more of a challenge than predicting its phase.
High skill of predicting seasonal variations of the SAM is also detected at zero lead time4 for almost all start months (Fig. 12a). Further investigation reveals that this high skill mainly derives from the first month of the forecasts. This skill at the shortest lead time suggests that the POAMA model is able to represent some aspects of the future evolution of the SAM that depend largely on the atmospheric initial conditions. To understand the role of atmospheric initial conditions, we compare the prediction skill of monthly SAM anomalies from the three different POAMA versions that use different initialization/ensemble generation strategies, as discussed in section 2. For an unbiased assessment across the three versions of the POAMA system that all use a different ensemble size, we form three-member ensemble mean forecasts at zero lead time for the period of 1980–2005 from each of POAMA1.5a, POAMA1.5b, and POAMA2. Recall from section 2 that POAMA1.5a atmospheric initial conditions have no real information about subseasonal variability, whereas POAMA1.5b and POAMA2 both use realistic atmosphere and land initial conditions generated from the ALI nudging scheme. The key difference of POAMA2 from POAMA1.5b is the additional use of improved ocean initial conditions generated from PEODAS and a different ensemble generation strategy.
Figure 13 displays the correlation skill for the SAMI for the first month of the forecast from the three systems. We also include correlation skill of the persistence forecasts that use the observed SAM value of the last day of the preceding month as a predictor for the SAM of the coming month (e.g., we use the observed SAM value on 31 December as a prediction for January SAM). A big skill jump is seen when realistic atmospheric initial conditions are used in POAMA1.5b and POAMA2 compared to when AMIP atmospheric initial conditions are used in POAMA1.5a. Also, the skill of POAMA1.5b and POAMA2 is substantially higher than the persistence forecasts, which implies that the model is extracting useful information about the dynamical evolution of the atmosphere associated with the SAM from the initial conditions. The skill difference between POAMA1.5b and POAMA2 is very small, with perhaps POAMA1.5b skill being slightly higher than POAMA2, presumably as a result of more spread in the first month of the forecast owing to the use of perturbed atmospheric initial conditions in POAMA1.5b compared to perturbed ocean initial conditions in POAMA2. Hence, improved ocean initial conditions do not appear to play a role in the prediction of the SAM at the shortest lead time. These results highlight that high-quality atmospheric initial conditions and properly incorporating uncertainties in the atmospheric conditions via ensemble forecasts are critical for monthly prediction of the SAM at short lead time.
Correlation skill of 0 lead time forecasts of monthly SAM from the persistence forecasts (see the text for the details), POAMA1.5a, POAMA1.5b, and POAMA2. Gray color shading indicates correlation skill that is statistically significant at the 95% c.l. assessed by a one-sided Student's t test given 31 independent samples.
Citation: Journal of Climate 26, 20; 10.1175/JCLI-D-13-00006.1
6. Conclusions
We have demonstrated seasonal predictability of the SAM based on its relationship with SST variations associated with different flavors of El Niño/La Niña. Two distinctive periods of predictability of seasonal variations of the SAM are found: austral winter when the SAM is related to the warm-pool (or Modoki/central Pacific) ENSO and austral early summer when the SAM is related to the cold-tongue (or canonical/eastern Pacific) ENSO. Exploiting this relationship with the two flavors of ENSO, the Australian Bureau of Meteorology dynamical seasonal prediction system POAMA2 demonstrates notable predictive skill for seasonal variations of the SAM to a lead time of 6 months in the spring to early summer seasons. Less remarkable but equally significant predictive skill is demonstrated for a lead time of 4–5 months in autumn to early winter. This capability derives from the skillful prediction of the SST variations associated with ENSO and a good simulation of the teleconnection to the SAM from the tropical SST anomalies. We furthermore showed that prediction of monthly SAM at 0-month lead time derives largely from realistic atmospheric initial conditions.
Our results for predictability and predictive skill of seasonal variations of the SAM in austral warm seasons are consistent with previous studies that have shown a statistically significant relationship between SAM variations and cold-tongue (canonical) El Niño during austral summer. The mechanism for this interaction, as postulated by the earlier studies such as Seager et al. (2003), L'Heureux and Thompson (2006), and Lu et al. (2008), is that the direct forcing of low-latitude tropospheric zonal mean diabatic heating by El Niño–related SST anomalies drives a change in the angular momentum transport in the Hadley circulation and, thus, an increase in low-latitude westerlies in the subtropical jet. This change in the subtropical westerlies leads to a change in the critical latitudes to which extratropical eddies propagate, thereby causing an equatorward shift of the transient eddy momentum flux convergence. This shift in the convergence of the eddy momentum flux feeds anomalous westerly momentum to the extratropical eddy-driven jet, which results in increased vertical wind shear and resultant growth of eddies equatorward of their climatological location, hence causing a swing in the SAM to its negative phase. The opposite occurs during canonical La Niña.
We have shown that the warm-pool El Niño also promotes the negative phase of the SAM during austral late autumn–winter seasons but not by shifting the subtropical westerlies toward the equator, as during El Niño in austral summer. Rather, during the austral cold seasons the warm-pool El Niño results in a poleward shift of the subtropical jet (i.e., increase of westerlies on the poleward flank of the subtropical jet). In the SH winter the westerly anomalies on the poleward flank of the subtropical jet act as a secondary source of baroclinic eddies in addition to the higher-latitude westerlies. This means that more baroclinic eddy activity occurs in the midlatitudes on the poleward side of the subtropical jet during warm-pool El Niño in austral winter, resulting in anomalous eddy momentum flux convergence on the equatorward side of the main eddy-driven jet, therefore causing the eddy-driven jet to shift equatorward and resulting in a swing to the negative phase of SAM. The opposite occurs during the warm-pool La Niña.
We have emphasized that there are two key differences between what happens during canonical El Niño in austral summer and what happens during warm-pool El Niño in austral winter: the austral wintertime warm-pool El Niño is associated with zonal mean cooling in the tropics, whereas the austral summertime canonical El Niño is associated with zonal mean heating in the tropics. The subtropical jet is not distinct during the summer (there is only an eddy-driven jet) so the impact of the zonal mean heating is directly on the eddy-driven jet, whereas in winter the zonal mean tropical heating affects the subtropical jet, which, being distinctly separated in latitude from the eddy-driven jet, then may allow for an opposite change in the meridional displacement of the eddy-driven jet. Importantly, the POAMA forecast model faithfully depicts this seasonality of the mechanism for the promotion of the SAM by tropical SSTs.
We have noted the role of tropical heating for driving changes in the wintertime subtropical jet, which then acts as an anomalous source of baroclinicity that can then bring about changes in the higher-latitude eddy-driven jet. This mechanism for interaction of the eddy-driven and subtropical jets is consistent with that proposed by Lee and Kim (2003). However, we cannot exclude the promotion of the SAM by anomalous tropical SSTs in the central-western Pacific during winter via wave–mean flow interaction as a result of spatially localized forcing of wave activity suggested by Ding et al. (2012). Carefully designed numerical experiments forced with different flavors of El Niños should be conducted to gain a better understanding of the detailed dynamics of the interaction among the tropical boundary forcing, the subtropical jet, and the eddy-driven jet. Also, the possibility of an asymmetric relationship during the warm phase versus cold phase of ENSO with the SAM and its impact on the seasonal prediction of SAM (e.g., Fogt et al. 2011) should be properly investigated in a future study.
Although this study has demonstrated predictive capability of the SAM with the current coupled model forecast system, there is large potential for improvement. For instance, the capability of the POAMA2 system is limited by its coarse vertical resolution in the stratosphere (two levels) and prescribed climatological ozone concentration and sea ice extents that should presumably give some extra predictability to SAM, at least for short lead times and especially in the austral spring season when troposphere–stratosphere coupling is strongest. For instance, Roff et al. (2011) demonstrate that improved stratosphere resolution results in improved prediction of the SAM at lead times 3–4 weeks in austral late spring. S.-W. Son et al. (2013, unpublished manuscript) also show a strong control of the stratospheric polar vortex on subsequent development of the SAM in spring, which thus suggests an additional source of short lead (0–1 month) predictability if the stratosphere is properly resolved and initialized.
Atmospheric initialization/ensemble generation strategy also appears to be important for short lead time forecasts of monthly variations of the SAM. Predictability of SAM variations in the first month of the forecast clearly derives from atmospheric initial conditions. To improve atmospheric initialization and generate more dynamically balanced ocean–atmosphere initial states, the Australian Bureau of Meteorology is developing a coupled data assimilation system using a breeding technique to generate spread (e.g., Okely et al. 2011), which is expected to bring some improvements on prediction of the SAM.
Acknowledgments
The authors thank Professors D. Karoly and I. Simmonds (University of Melbourne) and Dr. K. Ashok (Indian Institute of Tropical Meteorology) for their constructive reviews. We gratefully acknowledge comments by Drs. P. Hope and M. Wheeler in CAWCR on an earlier version of the manuscript. This research was supported by the South Eastern Australian Climate Initiative (SEACI; www.seaci.org), the Victorian Climate Initiative (VicCI), and the Managing Climate Variability Program (MCV; http://www.managingclimate.gov.au/).
REFERENCES
Ashok, K., S. Behera, A. S. Rao, H. Weng, and T. Yamagata, 2007a: El Niño Modoki and its possible teleconnection. J. Geophys. Res., 112, C11007, doi:10.1029/2006JC003798.
Ashok, K., H. Nakamura, and T. Yamagata, 2007b: Impacts of ENSO and Indian Ocean dipole events on the Southern Hemisphere storm-track activity during austral winter. J. Climate, 20, 3147–3163.
Ashok, K., C.-Y. Tam, and W.-J. Lee, 2009: ENSO Modoki impact on the Southern Hemisphere storm track activity during extended austral winter. Geophys. Res. Lett., 36, L12705, doi:10.1029/2009GL038847.
Barns, E. A., and D. L. Hartmann, 2010: Dynamical feedbacks of the southern annular mode in winter and summer. J. Atmos. Sci., 67, 2320–2330.
Cai, W., P. V. Rensch, T. Cowan, and H. H. Hendon, 2011: Teleconnection pathways for ENSO and the IOD and the mechanisms for impacts on Australian rainfall. J. Climate, 24, 3910–3923.
Chen, G., J. Lu, and D. M. W. Frierson, 2008: Phase speed spectra and the latitude of surface westerlies: Interannual variability and global warming trend. J. Climate, 21, 5942–5959.
Codron, F., 2005: Relation between annular modes and the mean state: Southern Hemisphere summer. J. Climate, 18, 320–330.
Colman, R., and Coauthors, 2005: BMRC Atmospheric Model (BAM) version 3.0: Comparison with mean climatology. Bureau of Meteorology Research Centre Research Rep. 108, 32 pp. [Available online at http://www.bom.gov.au/bmrc/pubs/researchreports/researchreports.htm.]
Deser, C., A. Philips, V. Bourdette, and H. Teng, 2012: Uncertainty in climate change projections: The role of internal variability. Climate Dyn., 38, 527–546.
Ding, Q., E. J. Steig, D. S. Battisti, and J. M. Wallace, 2012: Influence of the tropics on the southern annular mode. J. Climate, 25, 6330–6348.
Fogt, R. L., D. H. Bromwich, and K. M. Hines, 2011: Understanding the SAM influence on the South Pacific ENSO teleconnection. Climate Dyn., 36, 1555–1576.
Gates, W. L., 1992: AMIP: The Atmospheric Model Intercomparison Project. Bull. Amer. Meteor. Soc., 73, 1962–1970.
Gerber, E. P., and G. K. Vallis, 2007: Eddy–zonal flow interactions and the persistence of the zonal index. J. Atmos. Sci., 64, 3296–3311.
Gillett, N. P., T. D. Kell, and P. D. Jones, 2006: Regional climate impacts of the southern annular mode. Geophys. Res. Lett., 33, L23704, doi:10.1029/2006GL027721.
Gong, D., and S. Wang, 1999: Definition of Antarctic Oscillation index. Geophys. Res. Lett., 26, 459–462, doi:10.1029/1999GL900003.
Gong, T., S. B. Feldstein, and D. Luo, 2010: The impact of ENSO on wave breaking and southern annular mode events. J. Atmos. Sci., 67, 2854–2870.
Hartmann, D. L., 2007: The atmospheric general circulation and its variability. J. Meteor. Soc. Japan, 85B, 123–143.
Hartmann, D. L., and F. Lo, 1998: Wave-driven zonal flow vacillation in the Southern Hemisphere. J. Atmos. Sci., 55, 1303–1315.
Hendon, H. H., D. W. J. Thompson, and M. C. Wheeler, 2007: Australian rainfall and surface temperature variations associated with the Southern Hemisphere annular mode. J. Climate, 20, 2452–2467.
Hendon, H. H., E.-P. Lim, G. Wang, O. Alves, and D. Hudson, 2009: Prospects for predicting two flavours of El Niño. Geophys. Res. Lett., 36, L19713, doi:10.1029/2009GL040100.
Hendon, H. H., E.-P. Lim, J. M. Arblaster, and D. L. T. Anderson, 2013: Causes and predictability of the record wet east Australian spring 2010. Climate Dyn., doi:10.1007/s00382-013-1700-5, in press.
Hoerling, M. P., M. A. Kumar, and M. Zhong, 1997: El Niño, La Niña, and the nonlinearity of their teleconnections. J. Climate, 10, 1769–1786.
Hoskins, B. J., and K. I. Hodges, 2005: A new perspective on Southern Hemisphere storm tracks. J. Climate, 18, 4108–4129.
Hudson, D., O. Alves, H. H. Hendon, and G. Wang, 2011: The impact of atmospheric initialisation on seasonal prediction of tropical Pacific SST. Climate Dyn., 36, 1155–1171.
Hurrell, J. W., J. J. Hack, D. Shea, J. M. Caron, and J. Rosinski, 2008: A new sea surface temperature and sea ice boundary data set for the Community Atmosphere Model. J. Climate, 21, 5145–5153.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Kao, H.-Y., and J.-Y. Yu, 2009: Contrasting eastern-Pacific and central-Pacific types of ENSO. J. Climate, 22, 615–632.
Karoly, D. J., 1989: Southern Hemisphere circulation features associated with El Niño–Southern Oscillation events. J. Climate, 2, 1239–1252.
Karoly, D. J., 1990: The role of transient eddies in low-frequency zonal variations of the Southern Hemisphere circulation. Tellus, 42A, 41–50.
Karoly, D. J., P. Hope, and D. Jones, 1996: Decadal variations of the Southern Hemisphere circulation. Int. J. Climatol., 16, 723–738.
Kidson, J. W., 1988: Indices of the Southern Hemisphere zonal wind. J. Climate, 1, 183–194.
Kidson, J. W., and M. R. Sinclair, 1995: The influence of persistent anomalies on Southern Hemisphere storm tracks. J. Climate, 8, 1938–1950.
Kidston, J., D. M. W. Frierson, J. A. Renwick, and G. K. Vallis, 2010: Observations, simulations, and dynamics of jet stream variability and annular modes. J. Climate, 23, 6186–6199.
Kim, H.-K., and S. Lee, 2004: The wave–zonal mean flow interaction in the Southern Hemisphere. J. Atmos. Sci., 61, 1055–1067.
Kim, H.-M., P. J. Webster, and J. A. Curry, 2009: Impact of shifting patterns of Pacific Ocean warming on North Atlantic tropical cyclones. Science, 325, 77–80.
Kug, J.-S., F.-F. Jin, and S.-I. An, 2009: Two types of El Niño events: Cold tongue El Niño and warm pool El Niño. J. Climate, 22, 1499–1515.
Lee, S., and H.-K. Kim, 2003: The dynamical relationship between subtropical and eddy-driven jets. J. Atmos. Sci., 60, 1490–1503.
L'Heureux, M. L., and D. W. J. Thompson, 2006: Observed relationships between the El Niño–Southern Oscillation and the extratropical zonal-mean circulation. J. Climate, 19, 276–287.
Lim, E.-P., and I. Simmonds, 2007: Southern Hemisphere winter extratropical cyclone characteristics and vertical organization observed with the ERA-40 data in 1979–2001. J. Climate, 20, 2675–2690.
Lim, E.-P., H. H. Hendon, O. Alves, Y. Yin, M. Zhao, G. Wang, D. Hudson, and G. Liu, 2009a: Impact of SST bias correction on prediction of ENSO and Australian winter rainfall. CAWCR Research Letters, No. 3, CAWCR, Melbourne, Victoria, Australia, 22–29.
Lim, E.-P., H. H. Hendon, D. Hudson, G. Wang, and O. Alves, 2009b: Dynamical forecast of inter–El Niño variations of tropical SST and Australian spring rainfall. Mon. Wea. Rev., 137, 3796–3810.
Lim, E.-P., H. H. Hendon, O. Alves, Y. Yin, G. Wang, D. Hudson, M. Zhao, and L. Shi, 2010: Dynamical seasonal prediction of tropical Indo-Pacific SST and Australian rainfall with improved ocean initial conditions. CAWCR Tech. Rep. 032, 26 pp.
Lim, E.-P., H. H. Hendon, D. L. T. Anderson, and A. Charles, 2011: Dynamical, statistical–dynamical, and multimodel ensemble forecasts of Australian spring season rainfall. Mon. Wea. Rev., 139, 958–975.
Limpasuvan, V., and D. L. Hartmann, 1999: Eddies and the annular modes of climate variability. Geophys. Res. Lett., 26, 3133–3136.
Lorenz, D. J., and D. L. Hartmann, 2001: Eddy–zonal flow feedback in the Southern Hemisphere. J. Atmos. Sci., 58, 3312–3327.
Lu, J., G. Chen, and D. M. W. Frierson, 2008: Response of the zonal mean atmospheric circulation to El Niño versus global warming. J. Climate, 21, 5835–5851.
Marshall, A. G., D. Hudson, M. C. Wheeler, H. H. Hendon, and O. Alves, 2012: Simulation and prediction of the southern annular mode and its influence on Australian intra-seasonal climate in POAMA. Climate Dyn., 38, 2483–2502.
Marshall, G. J., 2003: Trends in the southern annular mode from observations and reanalyses. J. Climate, 16, 4134–4143.
Meneghini, B., I. Simmonds, and I. Smith, 2007: Association between Australian rainfall and the southern annular mode. Int. J. Climatol., 27, 109–121.
Nakamura, H., and A. Shimpo, 2004: Seasonal variations in the Southern Hemisphere storm tracks and jet streams as revealed in a reanalysis dataset. J. Climate, 17, 1828–1844.
Oke, P. R., A. Schiller, D. A. Griffin, and G. B. Brassington, 2005: Ensemble data assimilation for an eddy-resolving ocean model of the Australian region. Quart. J. Roy. Meteor. Soc., 131, 3301–3311.
Okely, P., O. Alves, D. Hudson, and Y. Yin, 2011: Towards coupled data assimilation: Coupled covariance structures. CAWCR Research Letters, No. 7, CAWCR, Melbourne, Victoria, Australia, 4–9.
Oort, A. H., and J. J. Yienger, 1996: Observed interannual variability in the Hadley circulation and its connection to ENSO. J. Climate, 9, 2751–2767.
Pacanowski, R. C., 1996: MOM2: Documentation, user's guide and reference manual. GFDL Ocean Group Tech. Rep. 3.2, 328 pp.
Rashid, H. A., and I. Simmonds, 2004: Eddy–zonal flow interactions associated with the Southern Hemisphere annular mode: Results from NCEP–DOE reanalysis and a quasi-linear model. J. Atmos. Sci., 61, 873–888.
Rashid, H. A., and I. Simmonds, 2005: Southern Hemisphere annular mode variability and the role of optimal nonmodal growth. J. Atmos. Sci., 62, 1947–1961.
Rayner, N. A., D. E. Parker, E. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res.,108, 4407, doi:10.1029/2002JD002670.
Reason, C. J. C., and M. Rouault, 2005: Links between the Antarctic Oscillation and winter rainfall over western South Africa. Geophys. Res. Lett., 32, L07705, doi:10.1029/2005GL022419.
Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 1609–1625.
Robinson, W. A., 2000: A baroclinic mechanism for the eddy feedback on the zonal index. J. Atmos. Sci., 57, 415–422.
Roff, G., D. W. J. Thompson, and H. Hendon, 2011: Does increasing model stratospheric resolution improve extended-range forecast skill? Geophys. Res. Lett., 38, L05809, doi:10.1029/2010GL046515.
Saji, N., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401, 360–363.
Schiller, A., J. S. Godfrey, P. C. McIntosh, G. Meyers, N. R. Smith, O. Alves, G. Wang, and R. Fiedler, 2002: A new version of the Australian Community Ocean Model for seasonal climate prediction. CSIRO Marine Research Rep. 240, 79 pp.
Seager, R., N. Harnik, and Y. Kushnir, 2003: Mechanisms of hemispherically symmetric climate variability. J. Climate, 16, 2960–2978.
Sen Gupta, A., and M. H. England, 2007: Coupled ocean–atmosphere feedback in the southern annular mode. J. Climate, 20, 3677–3692.
Silvestri, G. E., and C. S. Vera, 2003: Antarctic Oscillation signal on precipitation anomalies over southeastern South America. Geophys. Res. Lett., 30, 2155, doi:10.1029/2003GL018277.
Simmonds, I., and K. Keay, 2000: Mean Southern Hemisphere extratropical cyclone behavior in the 40-year NCEP–NCAR reanalysis. J. Climate, 13, 873–885.
Smith, N. R., J. E. Blomley, and G. Meyers, 1991: A univariate statistical interpolation scheme for subsurface thermal analyses in the tropical oceans. Prog. Oceanogr., 28, 219–256.
Szeredi, I., and D. Karoly, 1987: The horizontal structure of monthly fluctuations of the Southern Hemisphere troposphere from station data. Aust. Meteor. Mag.,35, 119–129.
Thompson, D. W. J., and J. M. Wallace, 2000: Annular modes in the extratropical circulation. Part I: Month-to-month variability. J. Climate, 13, 1000–1016.
Trenberth, K. E., 1979: Interannual variability of the 500-mb zonal flow in the Southern Hemisphere. Mon. Wea. Rev., 107, 1515–1524.
Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131, 2961–3012.
Valke, S., L. Terray, and A. Piacentini, 2000: The OASIS coupled user guide version 2.4. CERFACS Tech. Rep. TR/CMGC/00-10, 85 pp.
Wang, G., and H. H. Hendon, 2007: Sensitivity of Australian rainfall to inter–El Niño variations. J. Climate, 20, 4211–4226.
Wang, G., D. Hudson, Y. Yin, O. Alves, H. Hendon, S. Langford, G. Liu, and F. Tseitkin, 2011: POAMA-2 SST skill assessment and beyond. CAWCR Research Letters, No. 6, CAWCR, Melbourne, Victoria, Australia, 40–46.
Weng, H., K. Ashok, S. K. Behera, S. A. Rao, and T. Yamagata, 2007: Impacts of recent El Niño Modoki on dry/wet conditions in the Pacific rim during boreal summer. Climate Dyn., 29, 113–129.
Wilks, D., 2006: Statistical Methods in the Atmospheric Sciences. Academic Press, 592 pp.
Yin, Y., O. Alves, and P. Oke, 2011: An ensemble ocean data assimilation system for seasonal prediction. Mon. Wea. Rev., 139, 786–808.
Zhou, T., and R. Yu, 2004: Sea-surface temperature induced variability of the southern annular mode in an atmospheric general circulation model. Geophys. Res. Lett., 31, L24206, doi:10.1029/2004GL021473.
In AMIP the atmospheric models are run with observed boundary forcings (Gates 1992).
Transient eddy momentum flux convergence was computed from 
Persistence forecasts shown in Fig. 12 are obtained by using the previous month observed SAM as forecasts for the following season SAM (e.g., the observed SAM in January 2010 is used as a zero lead time prediction for February–April 2010 SAM).
Zero lead time is defined as no time gap between the initialization and the verification of forecasts (e.g., forecast initialized on 1 Jan 1980 and verified for January 1980).