1. Introduction
Due mainly to sparse observations, the variability of the Southern Hemisphere circulation has not been studied as extensively as its Northern Hemisphere counterpart. Previous efforts on the subject have focused on modes of variability in mid- to high latitudes where variances in geopotential height are large (e.g., Trenberth 1979, 1981; Rogers and van Loon 1982; Mo and van Loon 1984; Mo and White 1985; Kidson 1988a,b; Karoly 1989). Most of these studies employed operational analysis data from the Australian Bureau of Meteorology and from the European Centre for Medium-Range Weather Forecasts. In recent years, reanalysis data from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) have been extensively used. Using this dataset, Kiladis and Mo (1999) revisit modes identified by previous studies and provide a comprehensive review.
On intraseasonal timescales, which often refer to timescales of 10–50 days, the preferred modes are a series of stationary or eastward-propagating wave trains in the mid- to high latitudes. Many recent studies have focused on the properties and the seasonal variations of these modes in terms of their associated atmospheric vertical structure, zonal wavenumber, and tropical teleconnections (Farrara et al. 1989; Ghil and Mo 1991; Kidson 1991; Madden and Julian 1994; Kiladis and Weickmann 1997; Mo and Higgins 1998; Kiladis and Mo 1999). It is found that one of the modes is strongly influenced by tropical convection, which generates wave trains extending across the South Pacific and South America, contributing to the variability on these timescales (e.g., Berbery and Nogues-Paegle 1993; Kiladis and Weickmann 1997; Mo and Higgins 1997, 1998). The pattern is referred to as the Pacific–South American (PSA) pattern (Mo and Ghil 1987; Karoly 1989), by analogy with the Pacific–North American (PNA) pattern in the North Pacific sector (Wallace and Gutzler 1981; Deser and Blackmon 1995; Zhang et al. 1996). Both the PNA and the PSA appear to be consistent with a Rossby wave response to equatorial anomalous heating (Hoskins and Karoly 1981).
On interannual timescales, several preferred modes have been identified using empirical orthogonal function (EOF) analysis. The first mode is the high-latitude mode (HLM; e.g., Mo and White 1985; Mo and Ghil 1987; Kidson and Sinclair 1995; Karoly 1990). This is often referred to as a zonally symmetric or annular mode (e.g., Kidson 1988a; Karoly 1995) due to the feature of north–south out-of-phase variations between the Antarctic and midlatitudes at all longitudes. Given that a similar mode, known as the Arctic Oscillation, operates in the Northern Hemisphere, the HLM is also referred to as the Antarctic Oscillation (Thompson and Wallace 1998, 2000). Other modes include the PSA, a mode with a wavenumber-3 variability pattern in the latitude band of 40°–70°S, and modes with higher zonal wavenumber patterns (e.g. Mo and White 1985; Kiladis and Mo 1999).
There has been a considerable effort in understanding whether the recent observed climate variations are induced by anthropogenic forcing or are part of the natural variability of our climate system. Important issues include how modes of variability will change under anthropogenic forcing and whether the response of the climate system to anthropogenic forcing will project onto modes of internal variability. Knowledge of the statistical properties of these modes over a sufficiently long period is essential for addressing these issues. However, of necessity, most studies on the variability of the Southern Hemisphere circulation are based on data for less than 20 years. Although the recent NCEP–NCAR reanalysis provides data for some 40 years, statistics based on a longer period of time are desirable. Thus, in the foreseeable future, the required statistical properties may have to come from simulations using ocean–atmosphere coupled general circulation models (GCMs). A fundamental question is whether the observed modes of variability are reasonably simulated by climate GCMs. This is the central issue of the present study. A number of studies, for example that by Limpasuvan and Hartmann (2000), have found realistic annular mode structures in the internal variability of atmospheric GCMs. However there has apparently not been a comprehensive analysis of the Southern Hemisphere modes from a coupled model. We examine outputs of simulations using the Commonwealth Scientific and Industrial Research Organisation (CSIRO) climate model.
As mentioned above, variability in the Northern Hemisphere has been studied more extensively. Over the North Pacific sector, a number of issues have been explored concerning the roles of the tropical influence, in situ mid- to high-latitude forcing, and mid- to high-latitude internal dynamical processes in the generation of the PNA mode (see Lau 1997 for a review). It is found that indices of the variability patterns of the 500-hPa geopotential height (Z500) and the North Pacific sea surface temperature (SST) are significantly correlated with each other in the cold season, and both are also correlated with El Niño–Southern Oscillation (ENSO) indices (Horel and Wallace 1981; Trenberth and Paolino 1981). The associated patterns contain elements of the PNA mode. This has led to the “atmospheric bridge” hypothesis: the ENSO anomalies generate perturbations in the atmospheric circulation; the perturbations manifest as a PNA pattern, which in turn causes the midlatitude SST anomalies. This hypothesis is supported by modeling studies (e.g., Lau and Nath 1994; Graham et al. 1994). However, Deser and Blackmon (1995) find, in an EOF analysis of wintertime SST variability over the subtropical and North Pacific domain (20°S–60°N) for the period of 1950–92, that the second mode (after the ENSO mode), the so-called North Pacific mode, is linearly independent of ENSO (over the whole domain), and that the associated atmospheric anomaly also shows a PNA pattern in Z500 anomalies. These features have been confirmed by Zhang et al. (1996). The results from these latter two studies advocate the possibility that the North Pacific mode is one of its own, independent of ENSO, and that PNA is a preferred pattern of atmospheric variability even without tropical forcing. Given that Karoly (1989) found the PSA pattern in a composite analysis of winter anomalies over several ENSO events, similar issues can be raised and explored regarding the relationship between ENSO and modes of the Southern Hemispheric circulation. This is another focus of the paper.
The remainder of this paper is organized as follows. Section 2 describes the CSIRO climate model and the model experiments. The modes of variability in these experiments are described and compared in sections 3 and 4, with a focus on the importance of internal dynamics, air–sea interactions, and ENSO forcing in the generation of these modes. The possible interactions between the model ENSO and the mid- to high-latitude modes are further assessed in section 5. In section 6, we discuss, among other issues, the coherence of variability in the ocean and the atmosphere. A summary is given in section 7.
2. Model and model experiments
The Mark 2 CSIRO coupled model (Gordon and O'Farrell 1997) has the horizontal resolution of the R21 spectral representation, or approximately 3.2° latitude × 5.6° longitude, in both the atmosphere and ocean submodels. The atmospheric GCM has nine vertical sigma levels. It includes a semi-Lagrangian treatment of water vapor transport, dynamic sea ice, and a bare soil and canopy land surface, as well as standard parameterizations of radiation, cloud, precipitation, and the atmospheric boundary layer. In a fully coupled mode, the atmospheric GCM is coupled to an ocean GCM, which has 21 vertical levels, and includes the scheme of Gent and McWilliams (1990). This scheme parameterizes the adiabatic transport effect of subgrid-scale eddies, and replaces the nonphysical horizontal diffusivity as a means of stabilizing the model numerics. The inclusion of the scheme results in a much improved stratification leading to a major reduction in convection at high southern latitudes. A detailed description of the improvements has been given by Hirst et al. (2000).
The three simulations use different versions of the CSIRO climate model. In the first experiment, hereafter referred to as the coupled experiment, the fully coupled ocean–atmosphere GCM is run for 600 years (A. Hirst 1999, personal communication). The experiment is conducted using a version with improved ocean–ice interactions. In the second experiment, hereafter referred to as the climatology experiment, the same atmosphere GCM is forced by SSTs from a fixed climatological annual cycle; in the third experiment, hereafter referred to as the mixed layer experiment, the same atmospheric GCM is coupled to a simple 50-m-deep mixed layer ocean, with SSTs calculated from a mixed layer heat equation, and the role of oceanic processes, such as advection, are not included. The latter two experiments are run for 500 years each, and are those used by Watterson (2001). The three experiments represent three rather different dynamical environments. In the coupled experiment, ENSO and local air–sea coupling are both present; in the climatology experiment, neither of these processes is permitted; and in the mixed layer experiment only local air–sea coupling is allowed.
The Z500 outputs are used to identify major modes. We choose this variable because it is commonly used, thus facilitating comparison with results of previous studies. The analysis domain is from the equator to the South Pole. Unless stated otherwise, we use a covariance-matrix EOF analysis, and the gridpoint anomalies are neither area weighted nor normalized, as in other studies (Kiladis and Mo 1999) allowing intercomparisons. The lack of area weighting means that a greater weighting is given to the high latitudes because there are more grid points per unit area at higher latitudes. In section 6, we will examine modes obtained from normalized anomalies. Because model outputs are saved only in the form of the monthly mean, which does not allow an adequate examination of intraseasonal timescales, we focus mainly on variability on interannual timescales using annual mean anomalies. A climatological mean is formed from the annual mean climatology over the full length of a model simulation. Anomaly fields are then constructed by removing the climatology field from the field for each individual year.
In order to examine whether the identified modes vary significantly from one season to another, EOF analysis is also applied to seasonal mean anomalies, constructed similarly to the annual mean anomalies. We use the December–January–February (DJF), March–April–May (MAM), etc., definition, for the seasonal mean anomalies.
We have also employed Z500 fields from the NCEP–NCAR reanalysis for the period of 1958–98 to help assess the model modes. The reanalysis data are taken as representing observations. Because of the short period of the reanalysis data, monthly anomalies are used to provide more data points; in this case the monthly anomaly fields are formed by subtracting the monthly climatology from the monthly fields, and are compared with equivalent monthly results for the coupled model.
3. EOF modes from the coupled experiment
The left panels of Fig. 1 show patterns of the first three EOFs of the annual mean Z500 anomalies of the coupled experiment. They account for 36.8%, 9.8%, and 7.6%, respectively, of the total variance. The middle panels of Fig. 1 show modes from the monthly anomalies, and they account for 33.8%, 10.1%, and 7.1%, respectively, of the total variance. These modes have been previously referred to as the HLM, wavenumber 3, and the PSA mode, respectively. Statistical testing using the criterion of North et al. (1982) shows that these three modes are statistically distinct: the eigenvalues with error bars for the three modes are well separated from each other and from other modes. The eigenvalue with error bars of the fourth mode in either case overlaps that of EOF5. We therefore confine our attention to the first three EOF modes. The so-called low-latitude mode (LLM) that has been identified by previous studies (e.g., Karoly 1995) has not appeared in this analysis, but will be considered in section 6.
The patterns from the annual mean and monthly anomalies are similar. This suggests that the patterns of variability in the GCM on timescales shorter than 1 year are similar to those on timescales longer than 1 year. This is consistent with the results of Kidson (1988b) who performed a similar analysis admitting periods greater than 50 days, and found that interannual variability can be represented by such modes. A normalization procedure of the analysis ensures that the sum of the squares of the spatial weights (eigenvector) for each mode is equals to 1, allowing the standard deviation of the temporal coefficient to be calculated in terms of a physical unit (m for Z500). Table 1 shows the values for the annual mean and monthly mean results. As expected, the values for the monthly case are bigger than those for the annual mean results.
To compare these modes with those from the NCEP–NCAR reanalysis, EOF analysis is applied to the Z500 anomalies from the reanalysis. The first three EOF modes are displayed in the right panels of Fig. 1. (Note that EOF3 is shown in the second row for comparison with the modeled EOF2.) They account for 36.5%, 11.7%, and 8.3% of the total variance, respectively. These patterns are very similar to those obtained by Kiladis and Mo (1999) using the same data but filtered to retain variability on timescales longer than 50 days. It is seen that EOF3 (EOF2) from the NCEP–NCAR reanalysis appears in the model as EOF2 (EOF3). Thus, the modeled EOF patterns are similar to those from the NCEP–NCAR reanalysis, although the order of the EOFs is different. The standard deviations, in units of meters, of the time series for each mode are shown in Table 1. We see that the standard deviations of the monthly means for the coupled experiment for the HLM and the PSA modes are smaller than those from the “observed” NCEP–NCAR reanalysis, but the difference in the wavenumber-3 mode is small.
To examine the coherence between these modes and the modes of other fields, similar EOF analysis is applied to the annual mean anomalies of mean sea level pressure (MSLP). For the coupled experiment, the first three EOF modes of MSLP are very similar to those of Z500 (figures not shown) and in the same order. They account for 48.8%, 12.2%, and 6.5%, respectively, of the total variance. Correlation coefficients between the corresponding EOF time series range between 0.89 and 0.94. Similarly high correlations between the MSLP and Z500 modes exist in the other two experiments. These high correlation coefficients reflect the equivalent barotropic nature of these modes, all having MSLP signatures in phase with the Z500 anomalies.
EOF modes have been calculated from the modeled seasonal mean anomalies for each season, and the dimensional 1 standard deviation results for DJF and JJA are shown in Fig. 2. For these seasons, and also MAM and SON (not shown), the patterns are broadly similar in structure to those for the annual mean results. This limited seasonality in variability, in contrast to the Northern Hemisphere, has been noted by previous studies. Kiladis and Mo (1999) suggest that this follows from the generally lower amplitude of the mean seasonal cycle in the Southern Hemisphere atmosphere compared to the Northern Hemisphere, due primarily to the weaker continental effects in the meridional temperature gradient. Our results support this feature, as also reflected by the fact that the modes obtained using monthly mean anomalies are similar to those from annual mean anomalies discussed above.
EOF1 is the HLM, or the Antarctic Oscillation, identified and discussed by many previous studies (e.g., Trenberth 1979, 1981; Rogers and van Loon 1982; Mo and White 1985; Mo 1986; Kidson 1986, 1988a; Karoly 1990; Yu and Hartmann 1993; Watterson 2000; Thompson and Wallace 1998, 2000). The characteristics include a strong zonally symmetric component, showing an out-of-phase relationship between heights in a latitude belt just south and north of approximately 50°S. Besides the zonally symmetric nature of the mode, an imprint of a zonal wavenumber-2 and wavenumber-3 pattern is also evident. The left panels of Fig. 2 shows that this mode has comparable anomalies over the Antarctica in all seasons, but the anomalies at mid- to high latitudes are strongest in the Indian Ocean sector in the southern spring (not shown) and summer (DJF) seasons.
Yu and Hartmann (1993) suggested that the maintenance of this mode is the result of driving by the embedded synoptic eddies. They showed that the differences in eddy propagation between a phase when the oscillation is at a positive maximum and a phase when the oscillation is at a negative maximum result in anomalous forcing that helps to maintain zonal wind anomalies against frictional damping, and that this anomalous eddy forcing operates on synoptic timescales, with a changing storm track during the oscillation cycles. A similar mechanism has been demonstrated for the CSIRO model by Watterson (2002).
The modeled EOF2 and the NCEP–NCAR EOF3 (middle row of Fig. 1) both depict a wavenumber-3 pattern and have their maximum anomalies between 50° and 70°S. This wavenumber-3 pattern is another mode operating on a wide range of timescales, ranging from day to day (e.g., Wallace and Hsu 1983; Mo 1986; Mo and Ghil 1987; Kidson 1988b), to intraseasonal (e.g., Mo and White 1985; Karoly et al. 1989; Mo and Higgins 1997, 1998), and to interannual (Mo and van Loon 1984; Karoly et al. 1989). This mode and its interaction on 3–7-yr timescales with the ocean in an early version of the CSIRO coupled model have been described by Cai et al. (1999).
The seasonal EOF2 results (the middle panels of Fig. 2) show that the wavenumber-3 structure is most clear in the southern winter (JJA) and spring (not shown) seasons, consistent with the results of Mo and White (1985). In the other two seasons, the spatial pattern features an imprint of wavenumber-1 structure. Thus, the wavenumber-3 pattern seen in the annual mean anomalies reflects largely the variability pattern of the winter and spring seasons.
The modeled EOF3 and the NCEP–NCAR EOF2 (bottom row of Fig. 1) are the PSA pattern. The pattern displays a wave train starting with a strengthening or weakening of the South Pacific trough to the east of New Zealand, with an opposing anomaly to the west of the Drake Passage and additional downstream centers in the southern mid- to high latitudes of the South Atlantic sector. It emerges as one of the leading circulation patterns in the Southern Hemisphere ranging from daily (Mo and Ghil 1987), intraseasonal (Mo and Higgins 1997, 1998), interannual (Kidson 1988a, 1988b), to interdecadal timescales (Garreaud and Battisti 1999). While it may be forced from the Tropics, it has also been identified as a near-neutral mode of the zonal mean atmospheric flow (Frederiksen and Webster 1988; Frederiksen and Frederiksen 1996); suggesting that the variability of this pattern can be generated by atmospheric internal dynamical processes, a point we shall address further in sections 4 and 5.
The seasonal PSA results (right panels of Fig. 2) show that significant anomalies to the west of the Drake Passage persist throughout the year, but the intensity, in terms of the accompanying midlatitude anomalies, is strongest in the southern winter (JJA) and spring (not shown) seasons. Similar results are found by Kiladis and Mo (1999). In these two seasons, imprints of wavenumber-2 and wavenumber-3 patterns are also evident.
4. EOF modes under different dynamic environments
In the remaining sections we shall focus on modes from annual mean anomalies. The three GCM experiments allow examination of the dynamics of these modes. Comparisons between the coupled experiment and the mixed layer experiment would indicate the effect of oceanic dynamics and the remote effect of model ENSO variability. Comparisons between the climatology experiment and the mixed layer experiment would indicate the effect of the local thermodynamical coupling.
EOF analysis on annual mean Z500 anomalies from the climatology and mixed layer experiments produces similar modes and these modes are in the same order. The spatial anomalies at 1 standard deviation associated with each mode in the three experiments have been calculated as previously, and are depicted in the first three rows (left panels) of Figs. 3, 4, and 5 for the HLM, the wavenumber-3, and the PSA mode, respectively. It is seen that there are only small differences from one experiment to another.
The standard deviations (SD), in terms of meters, of the temporal coefficient for each mode in each experiment are calculated and shown in Table 2. Assuming N normally distributed independent yearly values, the sample variance (×N/σ2, where σ2 is the unknown actual variance) has a χ2(N − 1) distribution, which for large N is approximately normal with variance 2N. The 95% confidence interval for a sample SD is then approximately [1 ± (2/N)1/2]SD. Two values for a given mode are thus not statistically different if the difference is smaller than about 2N−1/2 times the larger value of SD. This is the case for the three modes in Table 2, given N = 500. (Other results in Table 2 are considered later.)
The surface temperature anomalies associated with each mode are shown in the right panels of Figs. 3, 4, and 5. The temperature anomalies are constructed by multiplying the regression coefficients, obtained by regressing surface temperature annual mean anomalies with the time series of an EOF, with the 1 standard deviation of the time series. The generation of temperature anomalies associated with the HLM has been described by previous studies (e.g., Watterson 2000). Over the latitudes where height anomalies and MSLP anomalies are positive, cloud cover reduces and incoming surface heat flux increases, leading to positive surface temperature anomalies. Conversely, over the latitudes with negative height anomalies, total cloud cover increases and less energy penetrates to the surface, leading to the development of negative surface temperature anomalies. This process operates in all three experiments (except for SSTs in the climatology experiment). In the coupled experiment variations in the oceanic Ekman drift also play a role in the generation of temperature anomalies, noted by Watterson (2000), and as will be discussed in section 6.
The relationship between the wavenumber-3 anomalies and the temperature anomaly pattern associated with the wavenumber-3 mode (Fig. 4) is such that positive (negative) centers of surface temperature anomalies lie to the west of the major centers of positive (negative) Z500 anomalies. The generation process for the wavenumber-3 mode in an earlier version of the CSIRO climate model has been described by Cai et al. (1999). Consistent with the barotropic structure and geostrophic balance, northerly low-level wind anomalies prevail to the west of the positive Z500 centers, under which latent and sensible heat transfer from the ocean to the atmosphere decreases, leading to anomalous heating of the ocean and hence positive temperature anomalies. Anomalies of the opposite sign form to the west of negative Z500. Such a process also operates in the generation of temperature anomalies associated with the PSA mode (Fig. 5) and is consistent with the results based on observations (Cai and Baines 2001), which show that the PSA mode plays a significant role in the generation of SST variability on interannual timescales.
For a given mode, the associated temperature anomalies do show some differences from one experiment to another. For example, when the ocean temperatures are not allowed to vary, as in the climatology experiment, the surface temperature anomalies naturally locate in regions over continents and sea ice. In the mixed layer and coupled experiments the temperature anomalies extend to the ocean. In the coupled experiment, temperature anomalies associated with the HLM in the Antarctic Circumpolar Current (ACC) latitudes display stronger zonal uniformity than in the mixed layer Experiment, apparently because of variations in the oceanic Ekman drift, which is absent in the mixed layer experiment. However, these differences do not cause significant differences in the structure of the Z500 modes.
The above results suggest that the ocean dynamics, air–sea coupling, and tropical forcing are not essential for the generation of these modes; in other words, internal atmospheric dynamical processes are capable of generating these modes. However one may argue that, although SSTs in the climatology experiment are not allowed to vary, convection in the Tropics still takes place in this experiment and Rossby wave trains generated by variations in convection could be responsible for these modes in the experiment. To explore this possibility, we correlate gridpoint rainfall anomalies in the climatology experiment with the time series of the three modes of Z500 in the same experiment. Maps of the correlation coefficient are shown in Fig. 6. Maps of the correlation for outgoing longwave radiation are similar, allowing for the opposite polarity. Maps from the two other experiments show similar features. If these modes in the experiment are predominantly driven by tropical convection one would expect the largest correlations to appear in the Tropics. However, for all three modes, the maximum correlations locate at the subtropical and mid- to high latitudes, rather than in the Tropics, reinforcing that these modes are predominantly driven by local internal atmospheric dynamics in the model.
Interestingly, there is a significant correlation between rainfall in the southeastern part of Australia and the HLM. The relation suggests that when the phase of the HLM is as shown in the top left panel (EOF1) of Fig. 1 (hereafter referred to as the positive phase), rainfall there increases. By contrast, there is a tendency for rainfall in southwest western Australia to decrease. This feature appears to be consistent with the result of Pittock (1973), who found that rainfall in the vicinity of Victoria is correlated with an index of the high pressure belt: for a high index value, which corresponds to the positive phase of the HLM (top left panels of Fig. 1), there are anomalous easterlies bringing warm moist air from over the ocean to over the land, conducive to an increased rainfall in the southeastern part of Australia. The reverse is true for southwestern western Australia, leading to the tendency of decreased rainfall.
5. The influence of ENSO
We have seen that the EOF3 of Southern Hemisphere Z500 has a PSA structure, even in the climatology experiment with no ENSO. A similar analysis of the Northern Hemisphere that we have performed produced a PNA mode as EOF2, again in all three experiments. The standard deviations of the associated time series are given in Table 2. The climatology experiment modes support the hypothesis that both PSA and PNA are mainly manifestations of internal atmospheric dynamics, at least in the CSIRO GCM. Nevertheless, there are some differences in each mode between the coupled result and the others (e.g., Fig. 5), potentially associated with ENSO. While the model's El Niño is known to be weak [the amplitude of the modeled Niño-3 temperature index is about one-third of the observed, Hirst et al. (2000)] its influence on the mid- to high latitudes is still worth examining.
a. The modeled ENSO mode
We obtain the ENSO mode in the coupled experiment by applying EOF analysis to the surface temperature variability of ocean grid points within 30°S–30°N. A realistic ENSO structure appears as the first EOF mode, which accounts for 19% of the total variance of the annual means. A power spectra of the time series of this mode shows that most of the energy is concentrated on the timescales longer than 5 yr with conspicuous peaks at 5–8 and 12–17 yr, consistent with the results of Walland et al. (2000). Thus, the model ENSO in this version of the coupled model does not have the observed periodicity of 3–4 yr.
The associated spatial pattern of surface temperature, for a 1 standard deviation anomaly, is shown in Fig. 7a, and extended by regression analysis to cover the entire globe. Most of the associated height anomaly (Fig. 7b) is situated in the mid- to high latitudes. The pattern is indeed similar to the PSA mode in the southern high latitudes, although the amplitudes are smaller than in Fig. 5. The response to ENSO in the north is also similar in structure to the model PNA (not shown), consistent with observational studies (Deser and Blackmon 1995; Zhang et al. 1996; Lau 1997). Figures 7a,b together support the hypothesis of an atmospheric bridge in the generation of the mid- to high-latitude SST anomalies by ENSO in both hemispheres. Figure 7c illustrates that during La Niña events, rainfall in the western Pacific regions increases, but decreases in the central equatorial Pacific region as in observations (Philander 1990), and vice versa during El Niño years.
b. Modes upon removal of ENSO anomalies
One approach to examining mid- to high-latitude internal variability within the coupled model is to apply our hemispheric EOF analyses after removal of the height anomalies that are linearly attributable (using regression as before) to the ENSO mode. The resulting first three “internal” modes for the Southern Hemisphere are similar to those previously obtained when the ENSO anomalies were included in the analysis. The spatial distributions of the 1 standard deviation anomalies associated with the internal HLM, internal wavenumber-3, and internal PSA mode are shown in the bottom row of Figs. 3, 4, and 5, respectively. These are similar to the original coupled results. Likewise, in the Northern Hemisphere, on removal of ENSO-related anomalies, the PNA mode again appears as the second mode.
The standard deviations of the temporal coefficient of each mode are shown in Table 2. For the HLM and the wavenumber-3 mode, the removal of ENSO produces a small decrease in variance, but one that would not be statistically significant by the previous test. However, the difference between the variance of the total PSA mode (i.e., with the ENSO-induced component, 3.35 m, Table 2) and the variance of the internal PSA mode (i.e., without the ENSO-induced component, 3.08 m, Table 2) is significant at the 90% level. For the PNA mode the difference is significant at the 95% level.
c. ENSO and the PSA mode
The enhancement of the PSA amplitude by ENSO is consistent with the similar structure of the response of Z500 to ENSO (Fig. 7a) and the internal PSA mode (Fig. 5, bottom, left panel). Likewise, the associated temperature anomaly patterns both have their largest anomalies situated to the north of the Ross Sea and to the west of the Drake Passage. A noteable difference is the positive anomaly at 60°E in the internal pattern, seen also in the mixed layer result (Fig. 5), which is coincident with a Z500 anomaly.
To the extent that the PSA patterns with and without ENSO are the same, we would anticipate that ENSO would augment the PSA time series by a component proportional to the (normalized) ENSO time series. The proportionality would be given by the projection of the Z500 ENSO Southern Hemisphere response (Fig. 7a) onto EOF3, which is 1.4 m. This value is indeed close to the value of 1.3 m, which would explain the 20% increase in variance of the total PSA mode series over the internal mode series (as in Table 2). Power spectra of the time series of the total PSA mode and the internal PSA mode (Fig. 8) show that there is a considerable decrease in the variance on the timescales longer than 5 yr when the ENSO-induced anomalies are removed (dark-dotted curve). This is consistent with the long periodicity of the model ENSO, discussed previously. It is worth noting that based on the existing model PSA response to ENSO, the position of the PSA as EOF3 in the coupled experiment, rather than EOF2 as in the NCEP data (Table 1), is consistent with the small amplitude of the model ENSO.
The correlation between the ENSO mode time series and that of the (total) PSA mode is 0.38 in the coupled experiment. This is comparable to that obtained from observational indices by Cai and Baines (2001), despite the weakness of the model ENSO. However, the shortness of the observational series may limit the validity of this comparison. In the model, the correlation varies considerably when the data are restricted to a 51-yr sliding window, as shown in Fig. 9. In some multidecadal periods the correlation reaches 0.5, but in others it is 0 or negative. Such variations are consistent with the linear regression model, given the limited number of effectively independent values in a 51-yr window (estimated to be 37; see Watterson 2000).
Overall, the model results support the notion that ENSO adds variance to the internal PSA mode without changing the PSA pattern significantly. Similarly, in the Northern Hemisphere, ENSO adds variance to the internal PNA mode, as shown in Table 2.
d. ENSO's relationship with the wavenumber-3 mode and the HLM
The response of the model southern mid- to high latitudes to ENSO forcing projects only weakly onto both the HLM and the wavenumber-3 mode, the values (for a 1 standard deviation ENSO) being 0.7 and −0.7 m, respectively. These values are closely consistent with the small differences in the variances of the mode time series with or without ENSO, determined from Table 2. Likewise, the spatial anomalies for both Z500 and surface temperature in the top (with ENSO) and bottom (without ENSO) rows of both Figs. 3 and 4 are similar.
The correlation coefficient between the time series of the ENSO mode and the wavenumber-3 mode is −0.28, indicating that there is some relationship between the two modes, nevertheless, which is consistent with previous observational results (e.g., Mo and White 1985).
The correlation coefficient between the 600-yr time series of the ENSO mode and the HLM is even weaker, at 0.08. This small coefficient appears to contrast with previous studies of reanalysis data (e.g., Karoly 1989; Kiladis and Mo 1999), which suggest that ENSO adds significant variance to that of the HLM. It is likely that the weakness of the model ENSO leads to an underestimate of the ENSO influence on the HLM. Again, though, the shortness of the observational record may be an issue, with the model values for a 51-yr sliding window varying from −0.3 to over 0.5.
6. Discussion
We have shown that the major modes of Z500 of the Southern Hemisphere circulation can operate with or without the ocean dynamics, but there is clearly some interaction with the ocean, as evidenced by the related SST anomalies. Do the SSTs and the oceanic circulation display EOF modes that are coherent with these atmospheric modes? What are the associated ocean–atmosphere interaction processes? We have noted that a mode, the LLM (Karoly 1995), has not appeared in our analysis so far. Is our model capable of simulating this mode? These issues will be addressed in this section.
a. Coherence of ocean–atmosphere variability
To address the first issue, we apply EOF analysis over the Southern Hemisphere to raw annual mean SSTs and two oceanic circulation fields from the coupled experiment.
EOF analysis of the SST anomalies reveals that the distribution of the percentage of variance that the SST modes account for is less concentrated than in the Z500 case. EOF1 of SST accounts for 11.8% and has a spatial pattern similar to that associated with the PSA mode shown in the top-right panel of Fig. 5. The time series of this mode is significantly correlated with the time series of both the total PSA (r = 0.58) and the internal PSA (r = 0.31). Thus the SST EOF1 appears to combine anomalies associated with the ENSO mode and anomalies associated with the internal PSA mode. The second EOF mode of SST accounts for 6.8% of the total variance and has a pattern (Fig. 10a) similar to the regression pattern associated with the HLM (top-right panel of Fig. 3). The correlation coefficient between the time series of the SST EOF2 and the HLM is 0.44. The third EOF mode of SST (5.1% of the total variance) has a wavenumber-3 pattern and a correlation coefficient of 0.40 with the wavenumber-3 mode of the Z500 anomalies.
Ocean circulation anomalies coherent with the Z500 modes can be anticipated given the associated low-level winds. In particular, Watterson (2000) demonstrated that the zonal surface wind anomalies associated with the HLM lead to zonal wind stress anomalies. Indeed, in the annual mean from the coupled experiment the correlation between the stress and the HLM (not shown) is around 0.6 at most grid points in the 53°–62°S band, and −0.3 at 33°–43°S. The midlatitude band of small zonal wind stress is thus a region of convergence of anomalous oceanic Ekman transports. Upwelling to the north and south feeds this convergence, resulting in deep overturning. The importance of this process is indicated by an EOF analysis of the anomalies in the oceanic meridional overturning streamfunction from the model. The first EOF mode (Fig. 10b) accounts for 51.1% of the total variance. The correlation between the time series of this mode with the HLM is 0.89.
In addition to the zonal stress, the zonal mean of the meridional surface stress is also perturbed by the HLM (Watterson 2001). We can anticipate that currents in both horizontal directions will be perturbed, by this mode, and also the other Z500 modes. We restrict consideration to the vertically integrated horizontal streamfunction (SF) that is simulated explicitly by the OGCM. The first three SF EOF modes account for 43.8%, 11.3%, and 7.1% percent of the total variance, respectively. Correlation coefficients between the time series of SF EOF1 and the HLM, between SF EOF2 and the PSA mode, and between SF EOF3 and the wavenumber-3 mode are 0.61, 0.56, and 0.49, respectively. The SF EOF1 shows that when the HLM has westerly anomalies over the high latitudes the ACC increases.
In summary, each of the major atmospheric modes is significantly related to an EOF of oceanic variability in both SST and SF, although the only very close relationship that we have identified is between the HLM and the meridional overturning EOF1.
b. Ocean–atmosphere interaction
The relationships just described can attributed to the “one-way” interaction of the atmosphere on the ocean. The driving of SST anomalies by surface heat fluxes associated with the Z500 anomalies has been mentioned in relation with all three modes, and documented elsewhere. The close relationship of the HLM with the overturning EOF1 also leads to an important influence on the SST EOF2. Ekman transport south of the surface convergence zone leads to a more zonally uniform cooling, which extends farther north, than in the mixed layer case (Fig. 3). To the north of the convergence there is some induced warming, which also tends to reinforce the action of the surface heat fluxes. However, this positive feedback effect is partly countered by the upwelling associated with the overturning. The vertical structure of oceanic temperatures in the southern mid- to high latitudes is such that south of about 55°S (approximately the latitudes of the convergence zone), the surface water is cooler than the subsurface water, whereas north of 50°S, the surface water is warmer than the subsurface water. This feature has been described by Cai et al. (2001). As the upwellings take place, the initial negative SST anomalies will be weakened. Thus this is a negative feedback process, without which we might expect a stronger SST EOF, more closely related to HLM.
True air–sea interaction would require a significant influence of such mid- to high-latitude SST anomalies back on the atmosphere. Given the long timescale of SST anomalies, particularly those linked to deep ocean anomalies as in the HLM case, one might expect that if the SST anomalies were to feed back positively on the atmosphere the variance on longer timescales would be greater in the coupled experiment than in the two other experiments. A calculation of the standard deviations of the HLM time series filtered to retain long period variability (using a 5-yr running mean) reveals that this is indeed the case, as in the Watterson (2001) study. However, the difference in variance on these long timescales (periods around 10 yr or more) is only 5% of the total variance, suggesting a small net effect. The differences for the other modes are even smaller.
On the other hand, the larger oceanic heat capacity of the coupled model means that the SSTs will respond to surface energy flux anomalies more slowly than in the mixed layer experiment, and this is evident at certain locations, for example, around 40°S, 50°W in each EOF case (Figs. 3, 4, and 5). One might anticipate (see Barsugli and Battisti 1998) that on shorter (but still interannual) timescales the lower-atmospheric temperature variance will be smaller, the deeper the ocean (with the climatology case having, in effect, an infinitely deep ocean). However, this result does not necessarily transfer to the Z500 modes (Watterson 2001). For example, the standard deviations of the interannual HLM on the shorter timescales (those removed by the 5-yr running mean) are 6.13, 6.17, and 6.27 m, respectively, for the mixed layer, coupled, and climatology experiments, the differences being not statistically significant.
The role of the ocean is an important issue, of course, as it relates to the predictability of atmospheric variability, from ocean anomalies. We have seen some influence of ENSO in the model on the longer-term PSA anomalies, but otherwise limited effect of the ocean on the Z500 modes. The potential predictability associated with local midlatitude interaction is still a subject of research (e.g., Rodwell et al. 1999). Watterson (2001) demonstrates that the predictability associated with the HLM and the related SST anomalies is relatively small. The present study also tends to support the similar conclusions of Bretherton and Battisti (2000) of only limited predictability.
c. The low-latitude mode
The lack of the LLM in our previous (covariance matrix) EOF analysis may be due to several reasons. First, we have reduced the weight of the low latitudes by not area weighting the anomalies. Second, the 500-hPa level is not particularly good for studying this mode, because the internal mode structure in the Tropics may not be well sampled. Third, height anomalies are relatively small in the low latitudes, in part because the atmospheric response to equatorial heating generally does not produce much vorticity in the heating region (Simmons 1982; Sardeshmukh and Hoskins 1988).
We have found, however, that if a correlation-matrix EOF analysis is employed, the LLM appears. We do this by normalizing the annual mean anomalies at each grid point by the local standard deviation before the application of the EOF analysis.
The modes obtained from anomalies of the coupled experiment, normalized after the removal of anomalies linearly attributable to ENSO, are similar to those from the climatology experiment and from the mixed layer experiment. In all the experiments, the HLM, the wavenumber-3, and the PSA mode now appear as EOF1, EOF3, and EOF4, respectively. In comparison with the modes obtained without normalization there is an additional mode that appears as EOF2. The pattern for the coupled experiment is shown in Fig. 11. This additional mode matches the LLM described by Karoly (1995). It resembles the HLM in that both have zonal strips of anomalies that alternate sign meridionally. Thus, the model is also capable of simulating the LLM.
Similar to the HLM, when MSLP and heights are higher than normal, cloud cover decreases, incoming surface heat increases, and warm temperature anomalies develop. Conversely, when MSLP and heights are lower than normal, cloud cover increases and less energy penetrates to the surface, leading to the development of cold surface temperature anomalies. Correlation maps between the time series of the mode and gridpoint surface temperature anomalies show that temperatures over land vary more coherently with the mode than temperatures over the oceans, due probably to the faster responses of temperatures over land.
7. Conclusions
The present study examines the capability of the CSIRO climate model in simulating the observed modes of variability of the Southern Hemisphere circulation. We find that the observed modes of variability are simulated reasonably well in a coupled experiment. These include the HLM (i.e., the Antarctic Oscillation), the PSA mode, and the wavenumber-3 mode. These modes from the coupled experiment are compared with those identified from two other experiments: in one, the atmosphere is forced only by climatological SSTs, and in the other, the atmosphere is coupled to an ocean mixed layer heat equation. All the modes identified in the coupled experiment are reproduced in the two other experiments. Thus, the central result is that the atmospheric internal dynamics are capable of generating these modes.
The coupled experiment produces ENSO cycles allowing examination of possible interactions between ENSO and the mid- to high-latitude variability. The response of the mid- to high-latitude atmosphere circulation to the ENSO forcing mainly projects onto the PSA mode. In spite of its small amplitude, the modeled ENSO variability induces PSA signals that are statistically significant. Many features of the PSA are similar to those associated with the PNA in the Northern Hemisphere, which has been more extensively studied and is also simulated by our model.
The inclusion of full ocean dynamics in the coupled experiment allows examination of the response of the ocean to, and its interaction and coherence with, the atmospheric modes. A focus is placed on the interaction between the oceanic meridional overturning circulation and the HLM, and we show that the HLM and EOF1 of the oceanic meridional overturning vary coherently. However, the oceanic feedback effect is weak, and the oceanic anomalies appear to be mainly driven by anomalies associated with the atmospheric modes.
There are many ways in which the present study can be expanded. Our initial aim has been to address whether the CSIRO model is capable of simulating the observed modes of the Southern Hemisphere circulation. The answer is in the affirmative. The successful simulation of these modes provides confidence that outputs from the climate model may be used to examine the response of these modes to anthropogenic forcing. Such a study is currently under way and will be reported in a future paper.
Acknowledgments
This work was supported by the Australian Greenhouse Office. The efforts of members of the Climate Modeling Program and the Atmospheric Processes Program at CSIRO Atmospheric Research in developing the model used in this study are gratefully acknowledged. We are grateful to A. Hirst for performing the coupled experiment and for reviewing our manuscript before submission. Comments from two reviewers and Dr. F. Zwiers also led to significant improvements.
REFERENCES
Barsugli, J. J., and D. S. Battisti, 1998: The basic effects of atmospheric–ocean thermal coupling on midlatitude variability. J. Atmos. Sci., 55 , 477–493.
Berbery, E. H., and J. Nogues-Paegle, 1993: Intraseasonal fluctuations between the Tropics and extratropics in the Southern Hemisphere. J. Atmos. Sci., 50 , 1950–1965.
Bretherton, C. S., and D. S. Battisti, 2000: An interpretation of the result from atmospheric general circulation models forced by time history of the observed sea surface temperature distribution. Geophys. Res. Lett., 27 , 767–770.
Cai, W., and P. G. Baines, 2001: Forcing of the Antarctic Circumpolar Wave by ENSO teleconnections. J. Geophys. Res., 106 , 9019–9038.
Cai, W., P. G. Baines, and H. B. Gordon, 1999: Southern mid- to high-latitude variability, a zonal wavenumber 3 pattern, and the Antarctic Circumpolar Wave in the CSIRO coupled model. J. Climate, 12 , 3087–3104.
Cai, W., P. G. Baines, and H. B. Gordon, . 2001: Reply. J. Climate, 14 , 1332–1334.
Deser, C., and M. L. Blackmon, 1995: On the relationship between tropical and North Pacific sea surface temperature variations. J. Climate, 8 , 1677–1680.
Farrara, J. D., M. Ghil, C. R. Mechoso, and K. C. Mo, 1989: Empirical orthogonal functions and multiple flow regimes in the Southern Hemisphere winter. J. Atmos. Sci., 46 , 3219–3223.
Frederiksen, C. S., and J. S. Frederiksen, 1996: A theoretical model of Australian northwest cloudband disturbances and Southern Hemisphere storm tracks: The role of SST anomalies. J. Atmos. Sci., 53 , 1410–1432.
Frederiksen, J. S., and P. J. Webster, 1988: Alternative theories of atmospheric teleconnections and low-frequency fluctuations. Rev. Geophys., 26 , 459–494.
Garreaud, R. D., and D. S. Battisti, 1999: Interannual (ENSO) and interdecadal (ENSO-like) variability in the Southern Hemisphere tropospheric circulation. J. Climate, 12 , 2113–2123.
Gent, P. R., and J. C. McWilliams, 1990: Isopycnal mixing in ocean circulation models. J. Phys. Oceanogr., 20 , 150–155.
Ghil, M., and K. Mo, 1991: Intraseasonal oscillations in the global atmosphere. Part II: Southern Hemisphere. J. Atmos. Sci., 48 , 780–790.
Gordon, H. B., and S. P. O'Farrell, 1997: Transient climate change in the CSIRO coupled model with dynamic sea ice. Mon. Wea. Rev., 125 , 875–907.
Graham, N. E., T. P. Barnett, R. Wilde, M. Ponater, and S. Schubert, 1994: On the roles of tropical and mid latitude SSTs in forcing interannual and interdecadal variability in the winter Northern Hemisphere circulation. J. Climate, 7 , 1416–1441.
Hirst, A. C., S. P. O'Farrell, and H. B. Gordon, 2000: Comparison of a coupled ocean–atmosphere model with and without oceanic eddy-induced advection. Part I: Ocean spinup and control integrations. J. Climate, 13 , 139–163.
Horel, J. D., and J. M. Wallace, 1981: Planetary-scale atmospheric phenomena associated with the Southern Oscillation. Mon. Wea. Rev., 109 , 813–829.
Hoskins, B. J., and D. J. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci., 38 , 1179–1196.
Karoly, D., 1989: Southern Hemisphere circulation features associated with El Niño–Southern Oscillation events. J. Climate, 2 , 1239–1252.
Karoly, D., . 1990: The role of transient eddies in low-frequency zonal variations of the Southern Hemisphere circulation. Tellus, 42 , 41–50.
Karoly, D., . 1995: Observed variability of the Southern Hemisphere atmospheric circulation. Natural Climate Variability on Decadal-to-Century Time Scales, National Research Council, 111–118.
Karoly, D., R. A. Plumb, and M. Ting, 1989: Examples of the horizontal propagation of quasi-stationary waves. J. Atmos. Sci., 46 , 2802–2811.
Kidson, J. W., 1986: Index cycles in the Southern Hemisphere during the Global Weather Experiment. Mon. Wea. Rev., 114 , 1654–1663.
Kidson, J. W., . 1988a: Indices of the Southern Hemisphere zonal wind. J. Climate, 1 , 183–194.
Kidson, J. W., . 1988b: Interannual variations in the Southern Hemisphere circulation. J. Climate, 1 , 1177–1198.
Kidson, J. W., . 1991: Intraseasonal variations in the Southern Hemisphere circulation. J. Climate, 4 , 939–953.
Kidson, J. W., and M. R. Sinclair, 1995: The influence of persistent anomalies on Southern Hemisphere storm tracks. J. Climate, 8 , 1938–1950.
Kiladis, G. N., and K. M. Weickmann, 1997: Horizontal structure and seasonality of large-scale circulation associated with submonthly tropical convection. Mon. Wea. Rev., 125 , 1997–2013.
Kiladis, G. N., and K. C. Mo, 1999: Interannual and intraseasonal variability in the Southern Hemisphere. Meteorology of the Southern Hemisphere, D. J. Karoly and D. G. Vincent, Eds., Amer. Meteor. Soc., 307–336.
Lau, N. C., 1997: Interactions between global SST anomalies and the midlatitude atmospheric circulation. Bull. Amer. Meteor. Soc., 78 , 21–33.
Lau, N. C., and J. Nath, 1994: A modeling study of the relative role of tropical and extratropical SST anomalies in the variability of the global atmosphere–ocean system. J. Climate, 7 , 1184–1200.
Limpasuvan, V., and D. L. Hartmann, 2000: Wave-maintained annular modes of climate variability. J. Climate, 13 , 4414–4429.
Madden, R., and P. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122 , 814–837.
Mo, K. C., 1986: Quasi-stationary states in the Southern Hemisphere. Mon. Wea. Rev., 114 , 808–823.
Mo, K. C., and H. van Loon, 1984: Some aspects of the interannual variation of mean monthly sea level pressure on Southern Hemisphere. J. Geophys. Res., 89 , 9541–9546.
Mo, K. C., and C. H. White, 1985: Teleconnections in the Southern Hemisphere. Mon. Wea. Rev., 113 , 22–37.
Mo, K. C., and M. Ghil, 1987: Statistics and dynamics of persistent anomalies. J. Atmos. Sci., 44 , 877–901.
Mo, K. C., and R. W. Higgins, 1997: Planetary waves in the Southern Hemisphere and linkages to the Tropics. Harry van Loon symposium, studies in climate, Part II. NCAR Tech. Note TN433+Proc., 290 pp.
Mo, K. C., and R. W. Higgins, . 1998: The Pacific–South American modes and the tropical convection during the Southern Hemisphere winter. Mon. Wea. Rev., 126 , 1581–1596.
North, G. R., T. L. Bell, and R. F. Cahalan, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110 , 699–706.
Philander, S. G. H., 1990: El Niño, La Niña, and the Southern Oscillation. International Geophysics Series, Vol. 46, Academic Press, 289 pp.
Pittock, A. B., 1973: Global meridional interactions in stratosphere and troposphere. Quart. J. Roy. Meteor. Soc., 99 , 424–437.
Rodwell, M. J., D. P. Rowell, and C. K. Folland, 1999: Oceanic forcing of the wintertime North Atlantic and European climate. Nature, 398 , 320–323.
Rogers, J. C., and H. van Loon, 1982: Spatial variability of sea level pressure and 500 mb height anomalies over the Southern Hemisphere. Mon. Wea. Rev., 110 , 1375–1392.
Sardeshmukh, P. D., and B. J. Hoskins, 1988: The generation of global rotational flow by steady idealized tropical divergence. J. Atmos. Sci., 45 , 1228–1251.
Simmons, A. J., 1982: The forcing of stationary wave motion by tropical diabatic heating. Quart. J. Roy. Meteor. Soc., 108 , 503–534.
Thompson, D. W. J., and J. M. Wallace, 1998: The Arctic Oscillation signature in the wintertime geopotential height and temperature fields. Geophys. Res. Lett., 25 , 1297–1300.
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-mean flow in the Southern Hemisphere. Mon. Wea. Rev., 107 , 1515–1524.
Trenberth, K. E., . 1981: Observed Southern Hemisphere eddy statistics at 500 mb: Frequency and spatial dependence. J. Atmos. Sci., 38 , 2585–2605.
Trenberth, K. E., and D. A. Paolino, 1981: Characteristic patterns of variability of sea level pressure in the Northern Hemisphere. Mon. Wea. Rev., 109 , 1169–1189.
Wallace, J. M., and D. S. Gutzler, 1981: Teleconnections in the geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev., 109 , 784–812.
Wallace, J. M., and H. H. Hsu, 1983: Ultra-long waves and the two-dimensional Rossby waves. J. Atmos. Sci., 40 , 2211–2219.
Walland, D. J., S. B. Power, and A. C. Hirst, 2000: Decadal climate variability simulated in a coupled general circulation model. Climate Dyn., 16 , 201–211.
Watterson, I. G., 2000: Southern midlatitude zonal wind vacillation and its interaction with the ocean in GCM simulations. J. Climate, 13 , 562–578.
Watterson, I. G., . 2001: Zonal wind vacillation and its interaction with the ocean: Implication for interannual variability and predictability. J. Geophys. Res., 106 , 23965–23975.
Watterson, I. G., . 2002: Wave–mean flow feedback and the persistence of simulated zonal flow vacillation. J. Atmos. Sci., 59 , 1274–1288.
Yu, J-Y., and D. L. Hartmann, 1993: Zonal flow vacillation and eddy forcing in a simple GCM of the atmosphere. J. Atmos. Sci., 50 , 3244–3259.
Zhang, Y., J. M. Wallace, and N. Iwasaka, 1996: Is climate variability over the North Pacific a linear response to ENSO? J. Climate, 9 , 1468–1478.
Patterns of the first three EOFs of the 500-hPa geopotential height anomalies. (left) Modes from annual mean anomalies of the coupled experiment. They account for 36.8%, 9.8%, 7.6%, respectively, of the total variance. (middle) Modes from monthly anomalies of the coupled experiment, accounting for 33.8%, 10.1%, and 7.1%, respectively, of the total variance. (right) Modes from monthly anomalies of the NCEP–NCAR reanalysis. They account for 36.5%, 11.7%, and 8.3% of the total variance, respectively. Note that EOF3 from the NCEP–NCAR reanalysis is plotted in the second row of the right panels
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Patterns of the first three EOFs of the 500-hPa geopotential height anomalies. (left) Modes from annual mean anomalies of the coupled experiment. They account for 36.8%, 9.8%, 7.6%, respectively, of the total variance. (middle) Modes from monthly anomalies of the coupled experiment, accounting for 33.8%, 10.1%, and 7.1%, respectively, of the total variance. (right) Modes from monthly anomalies of the NCEP–NCAR reanalysis. They account for 36.5%, 11.7%, and 8.3% of the total variance, respectively. Note that EOF3 from the NCEP–NCAR reanalysis is plotted in the second row of the right panels
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Patterns of the first three EOFs of the 500-hPa geopotential height anomalies. (left) Modes from annual mean anomalies of the coupled experiment. They account for 36.8%, 9.8%, 7.6%, respectively, of the total variance. (middle) Modes from monthly anomalies of the coupled experiment, accounting for 33.8%, 10.1%, and 7.1%, respectively, of the total variance. (right) Modes from monthly anomalies of the NCEP–NCAR reanalysis. They account for 36.5%, 11.7%, and 8.3% of the total variance, respectively. Note that EOF3 from the NCEP–NCAR reanalysis is plotted in the second row of the right panels
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The first three EOFs of 500-hPa geopotential height from seasonal mean anomalies of the coupled experiment. (top) For the DJF season, (bottom) for JJA. Shown are anomalies associated with a 1 std dev of each mode. The contours (in m), and shading patterns are as shown in the label bar
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The first three EOFs of 500-hPa geopotential height from seasonal mean anomalies of the coupled experiment. (top) For the DJF season, (bottom) for JJA. Shown are anomalies associated with a 1 std dev of each mode. The contours (in m), and shading patterns are as shown in the label bar
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The first three EOFs of 500-hPa geopotential height from seasonal mean anomalies of the coupled experiment. (top) For the DJF season, (bottom) for JJA. Shown are anomalies associated with a 1 std dev of each mode. The contours (in m), and shading patterns are as shown in the label bar
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
(left) The 500-hPa geopotential height anomalies (in m) and (right) surface temperature anomalies (in °C) associated with the HLM mode, in the (top row) coupled experiment, (second row) mixed-layer experiment, (third row) climatology experiment, and (bottom row) in the coupled experiment but with anomalies associated with the model ENSO removed. The 500-hPa geopotential anomalies are obtained by multiplying the pattern of weights of the HLM with a 1 std dev of the HLM time series. The surface temperature anomalies are obtained by multiplying the regression coefficient (regressing gridpoint surface temperature anomalies upon the time series of the HLM) with a 1 std dev of the HLM time series
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
(left) The 500-hPa geopotential height anomalies (in m) and (right) surface temperature anomalies (in °C) associated with the HLM mode, in the (top row) coupled experiment, (second row) mixed-layer experiment, (third row) climatology experiment, and (bottom row) in the coupled experiment but with anomalies associated with the model ENSO removed. The 500-hPa geopotential anomalies are obtained by multiplying the pattern of weights of the HLM with a 1 std dev of the HLM time series. The surface temperature anomalies are obtained by multiplying the regression coefficient (regressing gridpoint surface temperature anomalies upon the time series of the HLM) with a 1 std dev of the HLM time series
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
(left) The 500-hPa geopotential height anomalies (in m) and (right) surface temperature anomalies (in °C) associated with the HLM mode, in the (top row) coupled experiment, (second row) mixed-layer experiment, (third row) climatology experiment, and (bottom row) in the coupled experiment but with anomalies associated with the model ENSO removed. The 500-hPa geopotential anomalies are obtained by multiplying the pattern of weights of the HLM with a 1 std dev of the HLM time series. The surface temperature anomalies are obtained by multiplying the regression coefficient (regressing gridpoint surface temperature anomalies upon the time series of the HLM) with a 1 std dev of the HLM time series
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The same as Fig. 3 but for the wavenumber-3 mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The same as Fig. 3 but for the wavenumber-3 mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The same as Fig. 3 but for the wavenumber-3 mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The same as Fig. 3 but for the PSA mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The same as Fig. 3 but for the PSA mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The same as Fig. 3 but for the PSA mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Maps of the correlation coefficients between gridpoint rainfall anomalies and the time series of the first three EOFs of Z500 of the climatology experiment: (a) EOF1, (b) EOF2, and (c) EOF3
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Maps of the correlation coefficients between gridpoint rainfall anomalies and the time series of the first three EOFs of Z500 of the climatology experiment: (a) EOF1, (b) EOF2, and (c) EOF3
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Maps of the correlation coefficients between gridpoint rainfall anomalies and the time series of the first three EOFs of Z500 of the climatology experiment: (a) EOF1, (b) EOF2, and (c) EOF3
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Anomalies associated with the model ENSO mode in the coupled experiment. (a) Surface temperature (in °C), (b) 500-hPa geopotential height (in m), and (c) rainfall (in mm day−1). The ENSO mode is obtained from an EOF analysis of SST anomalies in the tropical Pacific domain of 30°S–30°N. The associated anomalies are obtained by regressing gridpoint anomalies with the time series of the model ENSO mode and then multiplying the regression coefficients with the value of 1 std dev of the time series of the ENSO mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Anomalies associated with the model ENSO mode in the coupled experiment. (a) Surface temperature (in °C), (b) 500-hPa geopotential height (in m), and (c) rainfall (in mm day−1). The ENSO mode is obtained from an EOF analysis of SST anomalies in the tropical Pacific domain of 30°S–30°N. The associated anomalies are obtained by regressing gridpoint anomalies with the time series of the model ENSO mode and then multiplying the regression coefficients with the value of 1 std dev of the time series of the ENSO mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Anomalies associated with the model ENSO mode in the coupled experiment. (a) Surface temperature (in °C), (b) 500-hPa geopotential height (in m), and (c) rainfall (in mm day−1). The ENSO mode is obtained from an EOF analysis of SST anomalies in the tropical Pacific domain of 30°S–30°N. The associated anomalies are obtained by regressing gridpoint anomalies with the time series of the model ENSO mode and then multiplying the regression coefficients with the value of 1 std dev of the time series of the ENSO mode
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Power spectra of the time series of the total PSA mode (solid curve) and the time series of the internal PSA mode, i.e., the PSA mode obtained after removing ENSO-induced anomalies (dark dotted curve). The maximum time lag was 0.1 of the series length. A Blackman–Harris smoothing was used. Confidence intervals of 2.5% and 97.5% for a “red noise” or Markov fit to the series are also shown (dashed curves). The ticks on the x axis indicate cpy and the values on the ticks are 0.5, 0.4, 0.3, 0.2, and 0.1, and so on, corresponding to periods of 2, 2.5, 3.3, 5, and 10 yr, respectively
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Power spectra of the time series of the total PSA mode (solid curve) and the time series of the internal PSA mode, i.e., the PSA mode obtained after removing ENSO-induced anomalies (dark dotted curve). The maximum time lag was 0.1 of the series length. A Blackman–Harris smoothing was used. Confidence intervals of 2.5% and 97.5% for a “red noise” or Markov fit to the series are also shown (dashed curves). The ticks on the x axis indicate cpy and the values on the ticks are 0.5, 0.4, 0.3, 0.2, and 0.1, and so on, corresponding to periods of 2, 2.5, 3.3, 5, and 10 yr, respectively
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Power spectra of the time series of the total PSA mode (solid curve) and the time series of the internal PSA mode, i.e., the PSA mode obtained after removing ENSO-induced anomalies (dark dotted curve). The maximum time lag was 0.1 of the series length. A Blackman–Harris smoothing was used. Confidence intervals of 2.5% and 97.5% for a “red noise” or Markov fit to the series are also shown (dashed curves). The ticks on the x axis indicate cpy and the values on the ticks are 0.5, 0.4, 0.3, 0.2, and 0.1, and so on, corresponding to periods of 2, 2.5, 3.3, 5, and 10 yr, respectively
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Time series of the correlation coefficient using a 51-yr sliding window between the time series of the ENSO mode and the time series of the total PSA mode. The correlation coefficient between the two time series using all samples (600) is plotted as a straight line (thin solid line)
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Time series of the correlation coefficient using a 51-yr sliding window between the time series of the ENSO mode and the time series of the total PSA mode. The correlation coefficient between the two time series using all samples (600) is plotted as a straight line (thin solid line)
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Time series of the correlation coefficient using a 51-yr sliding window between the time series of the ENSO mode and the time series of the total PSA mode. The correlation coefficient between the two time series using all samples (600) is plotted as a straight line (thin solid line)
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
(a) Pattern of SST EOF2, accounting for 6.8% of the total variance. (b) Pattern of EOF1 of the oceanic meridional overturning circulation, accounting for 51.1% of the total variance
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
(a) Pattern of SST EOF2, accounting for 6.8% of the total variance. (b) Pattern of EOF1 of the oceanic meridional overturning circulation, accounting for 51.1% of the total variance
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
(a) Pattern of SST EOF2, accounting for 6.8% of the total variance. (b) Pattern of EOF1 of the oceanic meridional overturning circulation, accounting for 51.1% of the total variance
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Pattern of the second EOF of 500-hPa geopotential height anomalies normalized by the std dev at each grid point. Before normalization the ENSO-induced anomalies are removed. The first four modes account for 21.7%, 16.0%, 6.3%, and 4.6%, respectively, of the total variance
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Pattern of the second EOF of 500-hPa geopotential height anomalies normalized by the std dev at each grid point. Before normalization the ENSO-induced anomalies are removed. The first four modes account for 21.7%, 16.0%, 6.3%, and 4.6%, respectively, of the total variance
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
Pattern of the second EOF of 500-hPa geopotential height anomalies normalized by the std dev at each grid point. Before normalization the ENSO-induced anomalies are removed. The first four modes account for 21.7%, 16.0%, 6.3%, and 4.6%, respectively, of the total variance
Citation: Journal of Climate 15, 10; 10.1175/1520-0442(2002)015<1159:MOIVOT>2.0.CO;2
The std dev (m) of the time series of the EOF modes for the coupled experiment (annual and monthly mean) and the “observed” NCEP–NCAR reanalysis (monthly mean). The HLM, wavenumber-3, and PSA mode are EOFs 1, 2, and 3, respectively, of the model Z500 variability
The std dev (m) of the time series of the EOF modes of interannual variability for the coupled, mixed-layer, climatology and coupled except ENSO removed experiments
