Atmospheric circulation changes during boreal winter of the second half of the twentieth century exhibit a trend toward the positive polarity of both the Northern Hemisphere annular mode (NAM) and the Southern Hemisphere annular mode (SAM). This has occurred in concert with other trends in the climate system, most notably a warming of the Indian Ocean. This study explores whether the tropical Indian Ocean warming played a role in forcing these annular trends. Five different atmospheric general circulation models (AGCMs) are forced with an idealized, transient warming of Indian Ocean sea surface temperature anomalies (SSTA); the results of this indicate that the warming contributed to the annular trend in the NH but offset the annular trend in SH. The latter result implies that the Indian Ocean warming may have partly cancelled the influence of the stratospheric ozone depletion over the southern polar area, which itself forced a trend toward the positive phase of the SAM. Diagnosis of the physical mechanisms for the annular responses indicates that the direct impact of the diabatic heating induced by the Indian Ocean warming does not account for the annular response in the extratropics. Instead, interactions between the forced stationary wave anomalies and transient eddies is key for the formation of annular structures.
The atmospheric circulation of both the Northern Hemisphere (NH) and Southern Hemisphere (SH) in the boreal winter exhibited a trend toward positive geopotential height anomalies at midlatitudes and negative geopotential height anomalies over the polar areas during the second half of the twentieth century (e.g., Thompson and Solomon 2002; Gillett and Thompson 2003; Hoerling et al. 2004; Solomon et al. 2007). The spatial structure of these trends projects onto the positive phase of the Northern Hemisphere annular mode (NAM) and the Southern Hemisphere annular mode (SAM; Thompson and Wallace 1998, 2000; Gong and Wang 1999). The NAM (SAM) is the leading mode of variability of NH (SH) extratropical circulation. Previous studies demonstrated that these modes account for a substantial part of the observed circulation change in both the NH and SH (Hurrell 1995; Thompson and Wallace 2001; Gillett et al. 2006); understanding their cause is of great interest.
Concurrent with these extratropical circulation changes, significant warming trends occurred in the global tropical ocean, which were more or less attributed to anthropogenic increases of greenhouse gases (Knutson et al. 1999; 2006; Hoerling et al. 2004). Although warming trends over the tropical Atlantic and west Pacific Oceans are modulated by substantial variability on decadal time scales, the tropical Indian Ocean (Fig. 1) is the epicenter for a detectable anthropogenic change because of its large monotonic sea surface temperature (SST) increase relative to its moderate decadal variability (Knutson et al. 2006). Previous atmospheric general circulation model (AGCM) and coupled air–sea model experiments reveal that the progressive tropical Indian Ocean warming has accounted for a significant fraction of the NH annular-like circulation trend (Hoerling et al. 2001, 2004; Li et al. 2006a). This was further confirmed by a recent study by Cassou (2008), which shows a close relationship between the North Atlantic Oscillation (NAO) and the Madden–Julian oscillation (MJO) at the subseasonal time scale. In particular, when there is enhanced convection in the Indian Ocean, the NAO tends to be in the positive phase. This link is verified by the coherence between the reversed NAO trend (Overland and Wang 2005) and the weakened Indian Ocean SST trend after the mid-1990s. However, whether the warming has also contributed to the annular circulation trend in the SH is unclear.
Significant depletion of stratospheric ozone over the southern polar area during austral spring was observed in the recent decades. This ozone decrease reduces the absorption of shortwave solar radiation and cools the polar stratosphere, thus increasing the meridional temperature gradient (Perlwitz et al. 2008). The accompanying dynamical response resulted in a 1-month delay of the winter polar vortex breakdown in the stratosphere and a shift of the summertime SAM toward its positive phase (Thompson and Solomon 2002; Gillett and Thompson 2003; Perlwitz et al. 2008).
The coexistence of the two climate forcings, where each of which is attributable to human influences, ozone depletion, and tropical oceanic warming, makes it difficult to separate their individual role in forming the observed southern climate trend with the short historical instrumental records at hand. There is therefore a need for investigating the sole effect of the tropical ocean warming. One AGCM study (Grassi et al. 2005) suggests that the tropical SST changes significantly contributed to the formation of the SAM trend together with the ozone depletion. Deser and Phillips (2009) illustrated, however, that tropical SST changes from 1950 to 1999 cause a shift of the SAM toward its negative phase. Similarly, Hu and Fu (2009) forced an AGCM with observed SSTs and reproduced the warming trend in the Antarctic stratosphere in recent decades. Other studies suggest that tropical Pacific Ocean warming (cooling) resulting from El Niño (La Niña) leads to a negative (positive) phase of the SAM (Zhou and Yu 2004; L’Heureux and Thompson 2006). The goal of the present study is to compare NH and SH circulation sensitivities to progressive tropical Indian Ocean warming using experiments with AGCMs and to understand plausible mechanisms using diagnostic models.
The paper is organized as follows: Section 2 describes the methodology employed in this study. It also introduces the five AGCMs and the linear baroclinic model (LBM) utilized. Section 3 describes the results, discussing the circulation response in one AGCM in more detail. Section 4 explores the physical processes for the modeled annular response by diagnosing transient experiments with two AGCMs as well as with a steady-state LBM. Summary and discussion are given in section 5.
a. Sensitivity experiments
A series of idealized transient SST warming ensemble experiments were conducted using five AGCMs, the details of which are given in Table 1. Ensemble size for individual AGCMs varies from 5 to 11. For each simulation, the model is integrated over 12 yr (from year 0 to 11), during which Indian Ocean SSTs are subjected to a linear rate of SST increase. Parts of the experiments were performed previously and used to explore the impact of Indian Ocean warming on East Asian summer monsoon (Li et al. 2008) or on the Southern Hemispheric stratospheric polar vortex (Li 2009).
The progressive Indian Ocean warming is represented by forcing the ACGM with a linearly increasing SST anomaly (SSTA) within the tropical Indian Ocean basin domain as indicated in Fig. 1. The SSTA maximizes over the equatorial Indian Ocean from 5°S to 5°N and from 40° to 110°E and gradually decreases in amplitude to zero over 30°N and 30°S. Repeating annual cycle climatological SSTs are prescribed in the remaining ocean domain. The pattern of Indian Ocean warmth is kept fixed during the integration, whereas the amplitude increases linearly. For January of the first model year (noted as model year 0), the maximum SSTA over the equatorial areas is near zero and then increases with the constant rate of 0.2 K yr−1, reaching 2.2°C at the end of the model integration (December of model year 11). Model year 0 is discarded to allow for model spinup. The mean of the SSTA over the equatorial Indian Ocean is 0.5 K for the first 5-yr period (model years 1–5) and 1.5°C for the second 5-yr period (model years 6–10). Thus, the SSTA difference between the second and first 5-yr periods has a value of 1.0 K, which is similar to the observed warming that occurred over the Indian Ocean during the second half of the twentieth century (Hoerling et al. 2004; Fig. 1).
b. Diagnostic experiments
A second set of experiments is carried out using two of the AGCMs [National Center for Atmospheric Research (NCAR) Climate Community Model version 3 (CCM3) and National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS)] to diagnose physical processes leading to the formation of extratropical circulation responses. We investigate the transient atmospheric adjustment to a sudden switch on of the idealized 1-K Indian Ocean SSTA pattern. For each model, we carried out an ensemble of 60 runs with the anomalous SST forcing. Each run starts from randomly selected 1 January atmospheric initial conditions and is integrated over 45 days when the atmospheric anomalies were equilibrated. The 60-member ensemble is compared with a control ensemble of the same size forced only with climatological SST. The daily evolving differences in atmospheric states between these parallel ensemble averages are analyzed based on 5-day means, and the statistical significance of the differences are estimated using a Student’s t test. Such an experimental design had been used previously in Hoerling et al. (2004) using CCM3; here, we repeat that analysis using a different model.
An LBM is used to investigate the direct atmospheric response to a tropical forcing. The LBM is the same as used by Peng and Whitaker (1999). It is a time-dependent spectral model with a horizontal resolution of T21 and 10 equally spaced pressure levels. No topography is prescribed. The linearization is about a three-dimensional time-mean flow. The model treats the diabatic heating and the transient-eddy flux of vorticity as forcings. The basic state is calculated from the ensemble winter [December–January (DJF)] or summer [June–August (JJA)] mean of the 10-yr detrended data in each CCM3 equilibrium experiment. Rayleigh friction and Newtonian damping have rates of (1 day)−1 at the lowest level and (7 days)−1 at other levels. Biharmonic horizontal diffusion with a coefficient of 2 × 1016 m4 s−1 is applied everywhere, and Fickian thermal diffusion with a coefficient of 2 × 106 m2 s−1 is included to represent the heat fluxes by transient eddies. With these values for dissipation and diffusion, a steady response to a forcing is reached after about 40 days. Averages over the final 5 days of the 40-day integrations are used to approximate the steady linear responses.
3. Results of sensitivity experiments
The nature of the annular mode responses occurring in the transient Indian Ocean warming experiments of the five AGCMs is first described. Because the changes are qualitatively similar among the five AGCMs, we present the results in detail only for the CCM3 and give a more general description for the other four models.
To illustrate the structure of the annular modes in the models, Fig. 2 presents the leading empirical orthogonal function (EOF) of NH and SH 500-hPa height in CCM3. The EOF analysis is conducted for detrended time series of monthly 500-hPa heights for the NH (20°–90°N) and the SH (20°–90°S). The time series of the detrended height fields for all 10 of the CCM3 runs are combined to form one time series of 1200 months.
Figure 2 shows the regression patterns of the first EOF. In the NH and SH, these modes explain 25% and 34% of total variance, respectively. The spatial pattern depicts a height-anomaly seesaw between mid and high latitudes with anomalies north of 65°N (south of 65°S) closely related to anomalies of the opposite sign between 30° and 50°N (S). The spatial pattern strongly resembles the NAM (SAM) of observations for which previous papers have documented coherent vertical structures extending from the surface to the stratosphere (e.g., Thompson and Wallace 2000).
Two methods are used to illustrate the models’ annular responses. One uses the temporal projection coefficient of monthly 500-hPa heights on these leading NH/SH EOFs to represent the NAM/SAM index. The other uses an NH (SH) polar cap height index determined as area average of height anomalies from 65° to 90°N (S), which represents one main center of action of the NAM (SAM). That the latter is a reasonable proxy of NAM (SAM) behavior is confirmed by the strong temporal correlation of −0.79 (−0.92) between the model’s 500-hPa NH (SH) polar cap height index and the NAM (SAM) index.
Figure 3 summarizes the seasonal cycle of changes in the annular mode indices and polar cap height indices, derived by computing the difference between the second and first 5-yr periods of the transient simulations (see section 2a for detailed description). Figure 3 also shows changes in the normalized 500-hPa NAO index,1 which is closely related to the NAM but whose centers of action are more confined to the North Atlantic region. During the boreal cold season (December–April), the progressive Indian Ocean warming causes a significant shift of the NAM/NAO index toward its positive polarity (Figs. 3a,c). This result is consistent with the previous study by Hoerling et al. (2004). During May–November, tropical Indian Ocean warming does not induce annular mode–like changes as indicated by the very small response in the model’s NAM index.
The SH also exhibits annular mode sensitivity to Indian Ocean warming, but it interestingly has polarity opposite to the NH response. The SH response is dominated by a shift of the SAM index toward its negative phase (Fig. 3b) and, as with the NH, exhibits greatest sensitivity during December–April (Fig. 3e). During austral winter (June–August), the indices of the annular mode (polar cap height index and SAM index) do not show any significant response. Note also that, for both hemispheres, the equivalent barotropic polar cap height responses (Figs. 3d,e) indicate that annular responses extend through the whole troposphere.
The regional and seasonal structures in the extratropical circulation response can be illustrated further by presenting the pattern of 200-hPa height changes for DJF and JJA (Fig. 4). Clearly, during DJF both the NH (SH) height fields show negative (positive) changes over high latitudes and positive (negative) height anomalies over midlatitudes consistent with the NAM and SAM changes. During JJA, the NH change pattern is dominated by positive changes over most of the hemisphere with no pronounced annular structure. The SH change pattern is dominated by a well-organized midlatitude wave train. We speculate that the SAM response occurs mostly in the summer season because of sensitivity to the spatial structure of the background flow and the occurrence of a single eddy-driven jet.
The previously described annular responses to Indian Ocean warming are further quantified by calculating the frequency of occurrence of NAM/SAM index values during the boreal winter months (DJF) for the two 5-yr periods (Fig. 5). For the NH, the number of months with a positive NAM index increases in the second 5-yr period relative to the first 5-yr period. For the SH, the situation is opposite, and the number of months with negative values increases. There is also an apparent increase in skewness of the SAM distribution as the Indian Ocean warms, with a substantial increase in extreme negative SAM states. It is unclear whether these extreme event changes are statistically significant, and we do not have a theory to understand such sensitivity. Little evidence for skewness change is seen in the NH.
In summary, the CCM3 results reveal that the SH annular mode sensitivity to tropical Indian Ocean warming is very different from its NH counterpart. First, the polarity is opposite. Second, the phasing with the seasonal cycle is also opposite, with the SAM sensitivity greatest in austral summer but the NAM sensitivity greatest in boreal winter. It is plausible that model biases in simulation of the climatological seasonal cycle and/or that its sensitivity to Indian Ocean forcing could render these results unique to CCM3. However, that these findings are not an artifact of a single model is confirmed by a multimodel analysis (Fig. 6 and Table 2), which demonstrates identical opposing behaviors of annular responses between the hemispheres. Somewhat less reproducible is the distinct seasonality of responses seen in CCM3, with more complicated monthly varying sensitivity occurring especially in the Max Planck Institute (MPI) ECHAM5 and Geophysical Fluid Dynamics Laboratory (GFDL) models.
Two additional features of the sensitivity to Indian Ocean warming are noteworthy, because they are key not only to understanding the structure of atmospheric responses but also to provide mechanisms for their proximate cause. Figure 7a shows the DJF zonal-mean zonal wind changes, and Fig. 8 shows the DJF precipitation changes. It is well known that the positive annular modes coincides with strong midlatitude westerlies (Thompson and Wallace 2000), and Fig. 7a shows strengthening (weakening) in the climatological westerlies of the NH (SH). The patterns can also be interpreted as a poleward (equatorward) shift of the westerlies in the NH (SH) that is consistent with a shift of the NAM (SAM) toward positive (negative) polarity.
Precipitation changes provide a picture for the principal characteristic of the tropical sensitivity to Indian Ocean warming. Figure 8 shows a marked increase in rainfall over the warming Indian Ocean, which tends to shift meridionally with the seasonal cycle of that region’s monsoonal climate and the changing position of intertropical convergence zone (ITCZ). Both the spatial extent and intensity of the increased rainfall are largest in DJF, despite the fact that identical patterns of SST change occur in all seasons.
4. Results of diagnostic experiments
In this section, we investigate possible processes that may contribute to the modeled annular-like responses and their polarity. We will investigate so-called transient adjustment runs, diagnosing the development of atmospheric anomalies prior to equilibration, to gain insights on possible causes. The suitability of studying such additional experiments to understand the equilibrium results of section 3 is supported by the resemblance of the zonal-mean zonal wind response at quasi equilibrium (roughly day 45) to the DJF seasonal mean response in the prior runs (cf. Figs. 7a,b). We also further investigate the role of diabatic heating associated with the tropical rainfall sensitivity in driving the extratropical circulation response (section 4c).
a. Evolution of transient response
The transient atmospheric adjustment to a tropical SSTA, for instance, as studied in Jin and Hoskins (1995), can be broadly viewed as occurring on a 2–4-week time scale before equilibration. In the first week, tropical heating anomalies develop and mature while the circulation response is mostly local and tropically confined. Linear wave dynamics and a global stationary wave response mature within the second to third weeks. The latter can be viewed as direct linear response to a tropical heating forcing modifying the basic state resulting in the reorganization of transient eddies, which induce anomalous transient-eddy flux that further interact with the evolving large-scale flow. This latter feedback process is important in determining the equilibrated solution (e.g., Held et al. 1989), whose ultimate adjustment time scale is on the order of a few weeks for steady forcing. The different phases of this adjustment can be monitored in output of experiments where the SST forcing of 1°C Indian Ocean warmth is suddenly switched on and the subsequent daily evolution of model states is examined. These states are based on a large 60-member ensemble to isolate the coherent part of the forcing–response relation (see also Hoerling et al. 2004).
For the NH (Fig. 9, left), the 200-hPa responses are less than half their equilibrium amplitudes at week 2, and there is virtually no indication of annularity, especially for the NH. Around week 2, the NH response consists of two wave train–like structures, one negative-phase Pacific–North America (PNA)-like pattern emerging from the tropical central east Pacific and the other emerging from the tropical Indian Ocean and propagating downstream toward East Asia (see the thick dashed line in the NH days 11–15 panel in Fig. 9). Such a response is in agreement with linear theory about the extratropical atmospheric responses to a tropical heating (e.g., Trenberth et al. 1998). The PNA response emerging from the central east Pacific may be associated with the Walker cell anomaly linked to the Indian Ocean forcing. This result agrees with the study by Mori and Watanabe (2008), which highlights a close relationship between the negative PNA phase and Indian Ocean convection associated with the MJO. After week 2, the responses intensify and become increasingly annular in the NH. At week 4, a positive NAO-like response is seen over the North Atlantic, which is in agreement with Cassou (2008), and a northern annular response is distinguishable. The progressive buildup process of the NAM is clearly seen from the evolution of the spatial correlation coefficients of the hemispheric responses with the NAM spatial pattern (Table 3).
For the SH (Fig. 9, right), from the beginning to week 2, the response is weak and exhibits a zonally propagating chain along 55°S. Two wave trains can be seen (see the thick dashed line in the SH days 11–15 panel in Fig. 9), one Pacific–South America (PSA)-like pattern emerging from the tropical central east Pacific and the other arched chain emerging from the tropical Indian Ocean and propagating downstream toward Australia (e.g., Wang 2005). The PSA response may share the same source with the NH PNA response. From week 2 to week 4, the responses over the southern Indian and Pacific Oceans double in strength and alter their spatial structure. However, the strongest change is over the southern Atlantic with values switching from weak positive to significant negative anomalies, and an overall midlatitude negative height pattern emerges, whereas heights within the polar cap become predominately positive. This progressive process of the SAM response buildup can also be seen in Table 3. For both the NH and SH, the significant difference in responses averaged from weeks 1–2 versus weeks 4–5 suggests that the transient feedback is critical. A similar transient adjustment can be seen in the GFS experiments indicating the robustness of the anomaly evolution (Fig. 10).
Figure 11 displays the estimated probability distribution functions (PDFs) of annular mode indices of 41–45-day mean 200-hPa heights in the transient CCM3 experiments. In the forced experiments, 70% of the ensemble members yield a positive NAM response, whereas 55% of the members yield a negative SAM response. The NAM index has greater spread relative to the control ensemble, indicating an intensification of amplitude as well. For the SAM, the index is more concentrated, suggesting that a shift to negative phase is dominant. Further calculations (not shown) reveal that the largest differences of the simulated responses from annularity locate in the region from the East Siberia Uplands to the Bering Sea for the Northern Hemisphere and in the southern Indian Ocean off the Antarctica for the Southern Hemisphere.
b. Role of transient momentum forcing
As indicated previously, the atmospheric responses have attained quasi equilibrium with the SSTA forcing by 45 days, in so far as the zonal-mean zonal wind anomalies at day 45 are very similar to the response in the equilibrium experiments (Fig. 7a). Now, we diagnose the role of transient-eddy feedback in forming the annular responses using budget balance calculations. Dynamically, the annular mode is a transient phenomenon with a time scale of about 10 days (Luo et al. 2007). The importance of transient eddies can be further elaborated by budget diagnostics of zonal-mean zonal wind. The equation for the zonal-mean zonal wind can be written as follows (Seager et al. 2003):
Here, the zonal-mean variations are divided into contributions from the zonal-mean circulation (denoted as brackets), stationary waves (denoted as asterisks), and transient eddies (denoted as primes). The variable u is the zonal wind, υ is the meridional wind, a is the radius of the earth, ϕ is latitude, p is pressure, f is the Coriolis parameter, D〈u〉 is a damping, and the upper bar denotes the time mean. The first term on the right side is the advection of the zonal-mean wind by the mean meridional circulation, the second term is the Coriolis torque, the third and fourth terms are the forcing by the stationary waves through inducing momentum flux convergence, and the fifth and sixth terms are the forcing by the transient eddies through inducing momentum flux convergence.
We calculate the zonal-mean momentum balance for the quasi-equilibrium period of the CCM3 transient responses from day 31 to day 45. The stationary wave fluxes are calculated using 9-day running means, whereas the transient eddies are computed from the daily original value minus the 9-day running mean. The balance in the extratropics is described by a tendency of the stationary wave and transient-eddy terms to induce a tendency for westerly acceleration (deceleration) near ∼60°N (∼60°S) that balances the Coriolis torque term (not shown). A similar role of these terms in maintaining the ENSO-associated zonal-mean zonal wind is seen in previous studies (Seager et al. 2003; L’Heureux and Thompson 2006).
In so far as the Coriolis torque can be interpreted as a response to the stationary and transient momentum fluxes, further diagnosis of time-evolving transient-eddy and stationary-eddy fluxes is desirable. Figure 12 displays the daily evolution of 300-hPa stationary zonal-mean zonal wind anomaly (u), anomalous stationary wave momentum forcing, and synoptic transient momentum forcing. From Fig. 12a, prior to day 10, (u) is weak, especially over the extratropics, which is consistent with the fact that transient feedback is not involved at this stage of adjustment. From day 5 to day 15, positive (u) anomalies begin to develop in the SH subtropics, and become larger (Fig. 12a). At the same time, negative anomalies develop between 30° and 50°S. These anomalies reach −0.5 m s−1 starting day 20 and increase up to about 1 m s−1 by the end of the 45-day period while the region of negative anomalies shifts southward. These features are consistent with a southward shift of the jet. The development in the NH is characterized by an expansion of positive anomalies toward the equator reaching amplitudes of 3 m s−1 at 5°N. Together with the equatorial positive wind anomaly, a decrease of westerly winds around 30°N (up to −2.5 m s−1) and an increase near 60°N (up to 3.5 m s−1) can be found. The fact that the positive zonal-mean zonal wind anomaly extends over a wider domain in the SH tropics than in the NH is related to the location of the climatological heating during this season. The diabatic heating anomaly as seen from the rainfall responses (Fig. 8a) shifts toward the south with a maximum around 10°S.
Figures 12b,c illustrate the relative contributions of momentum flux convergence induced by stationary and transient eddies. For the NH, momentum flux convergence induced by transient eddy is evidently greater than that induced by stationary waves. Furthermore, the evolution of −d(u′υ′)/dy bears more resemblances to the zonal-mean zonal wind than −d(u*υ*)/dy, suggesting that the synoptic eddies are more important in inducing the northern annular response. For the SH, however, the two terms are more close to each other in magnitude, suggesting that they jointly contribute to the formation of the negative-phase SAM trend. This is particularly clear after day 35. Thus, the relative role of stationary waves and synoptic scale eddies in inducing the annular response is different for the two hemispheres. This difference was also found in Seager et al. (2003), where the momentum budget of hemispherically symmetric zonal-mean zonal wind anomalies resulting from El Niño was studied. Nonetheless, the present analysis suggests that the atmospheric direct response to the tropical Indian Ocean heating is not annular. An annular response cannot be induced without transient-eddy feedback.
c. Role of diabatic heating
The sensitivity of the annular responses to the direct tropical heating is illustrated by utilizing LBM experiments (e.g., Peng et al. 2003; Li et al. 2006b, 2007). An idealized heating is given, as shown in Fig. 13. It is largely based on the anomalous rainfall induced by the Indian Ocean SSTA (Fig. 8a).
Figures 14a,b display the LBM response to the idealized heating under the DJF basic state. The NH response consists of two wave train–like circulation anomalies: one PNA-like pattern emerging from the tropical central Pacific and the other directly emerging from the tropical Indian Ocean and propagating downstream. These wave train–like responses are not only similar to the transient responses through days 11–15 (Figs. 9, 10) but also correspond to linear theory studies (Trenberth et al. 1998).
For the SH, the LBM response is somewhat complicated, but two kinds of wave train are still discernable: an arched Rossby wave train emerging from the southern tropical Indian Ocean propagating downstream eastward to South Indian Ocean and reaching Australia and a second wave train evolving from the central-eastern Pacific and propagating downstream eastward to South America, reminiscent of the PSA-like pattern (e.g., Wang 2005).
These responses can be explained by the linear Rossby waves excited by the tropical heating. The PNA and PSA responses may be associated with the wave source over the central-east Pacific linked to the Walker cell. The tropical Indian Ocean triggering anomalous wave activity to influence the SH extratropics was also observed in previous studies (e.g., Quintanar and Mechoso 1995). The LBM responses in the SH are overall similar to the 5-day mean through days 16–20 in the transient runs (Figs. 9, 10). These results support the point that the atmospheric linear response to tropical Indian Ocean forcing is not annular. High-frequency transience, however, cannot fully explain the formation of the annular response from the PNA- and PSA-like patterns, because it is most typical that low-frequency eddies excite higher-frequency transient eddies in a manner that reinforces the low-frequency eddies. One possible mechanism is that the Indian Ocean convection changes the refractive index for propagating waves, as discussed by Seager et al. (2003). Nonetheless, annularity does not emerge without feedback from transient eddies, which has been seen from Fig. 12, albeit discussed previously (e.g., Li et al. 2006b, 2007).
The same LBM experiments are also performed under a JJA basic state with the same idealized heating, except for the maximum location shift to the Bay of Bengal, with intriguing results. For the NH (Fig. 14c), the JJA response is considerably weak. For the SH (Fig. 14d), the response is evidently much stronger and its magnitude is 2 or 3 times larger than during DJF. This suggests that the atmospheric direct response in the NH to a tropical Indian Ocean heating is more seasonally dependent, in agreement with the stronger seasonality of the NH response revealed in the trend experiments (e.g., Fig. 6).
5. Summary and discussion
During the second half of the twentieth century, significant tropical Indian Ocean warming occurred along with a shift of SAM and NAM indices toward their positive phase. We investigated the influence of the tropical Indian Ocean warming on extratropical circulation by carrying out sensitivity studies with five AGCMs. The results of this study reveal the following:
Indian Ocean warming induces an annular-like atmospheric circulation response in both hemispheres, but of opposite polarity: positive in the NH and negative in the SH.
The annular responses maximize in different seasons in the hemispheres: boreal winter in the NH and austral summer in the SH.
The simulated NH responses to Indian Ocean warming are in agreement with both the observed NAM trends and the results of previously published AGCM or coupled model studies (e.g., Hoerling et al. 2001, 2004; Li et al. 2006a). This suggests that the Indian Ocean warming may have materially contributed to the observed NH trend. However, the simulated responses are opposite to the observed SAM trend. In so far as this was a robust sensitivity in all AGCMs, the results suggest that other forcings are dominating the observed SH trend. Stratospheric ozone depletion as described in section 1 has been previously shown to be a key factor. Our result suggests, therefore, that the impact of stratospheric ozone depletion on the SH tropospheric circulation during DJF is partly offset by the impact of Indian Ocean warming and that a stronger than observed SAM trend toward its positive phase would have resulted in the absence of Indian Ocean warming.
The SAM results of the present study are in agreement with a recent study by Deser and Phillips (2009) on the relative contribution of observed circulation changes to both direct radiative forcings and to sea surface temperatures to circulation changes from 1950 to 1999. For DJF, they showed that changes in the direct atmospheric forcing (including well-mixed greenhouse gases and ozone changes) cause a shift toward positive SAM index polarity, whereas SST changes (mainly of tropical origin) cause a shift toward its negative phase.
During the 1951–99 period, tropical SST changes were not only characterized by Indian Ocean warming but also by a general warming in the tropical eastern Pacific (e.g., Fig. 9f in Deser and Phillips 2009). Thus, changes in the tropical Indian Ocean cannot be considered in isolation from changes in the tropical Pacific. On the interannual time scale, there is a significant negative correlation between the ENSO index (measured by the cold-tongue index) and the summertime SAM index (L’Heureux and Thompson 2006; Zhou and Yu 2004). However, indices of the NAM and ENSO are not significantly correlated during any season (L’Heureux and Thompson 2006). Deser and Phillips (2009) determined that the tropical warming causes positive trends in SAM index and North Atlantic Oscillation (NAO). The study by L’Heureux and Thompson (2006) and our study provide additional insight into the results by Deser and Phillips (2009) by suggesting that warming in both tropical Indian Ocean and tropical Pacific contributed to a negative SAM trend, whereas Indian Ocean warming is the main tropical SST cause for the positive NAO trend. Whether possible nonlinear interaction between SST changes in the tropical Pacific and the Indian Ocean and recent changes in the trend behavior of tropical SST (cooling in tropical Pacific, leveling off of tropical Indian Ocean warming) plays a role in the near-neutral or negative phase behavior of the NAM since the mid-1990s (Overland and Wang 2005) requires further investigation.
Our findings raise two additional questions: 1) Why do the two hemispheres yield opposite polarity annular responses to identical SST forcing that is equatorially centered and axially symmetric? 2) Why do the two hemispheres’ responses maximize in opposite phases of their annual cycles? The transient adjustment experiments with two AGCMs together with the idealized heating experiments with an LBM indicate that these two questions are closely related and that more studies are necessary to answer them completely. First of all, the transient adjustment experiments indicate that the annular response was initiated by a thermally driven wave-like response followed by midlatitude transient-eddy feedback with the mean flow. A lack of an NH summer impact on the extratropical circulation can be attributed to the lack of any significant wave train–like response (Fig. 13c) and the fact that NH stationary and transient-eddy activity during that season is weak. We note that, in the SH, transient-eddy activity has no pronounced seasonal cycle. The spinup experiments suggest that the differences in the polarity of the NH and SH responses result from the differences between the NH and SH climatological basic states of zonal-mean zonal wind. However, further analysis and additional experiments have to be carried out for an in-depth investigation about the interhemispheric difference of climatological basic flow, transience eddy activity, and their feedbacks, as well as sensitivity to the strength and location of the tropical heating anomalies.
The results of the present study have implications for the attribution of extratropical circulation changes, the role of SST changes, and the need for the investigation of the sensitivity to SST changes in individual ocean basins. An open question is how a further warming of Indian Ocean SST combined with a possible recovery of the Antarctic ozone hole (e.g., Perlwitz et al. 2008) will reshape SH circulation changes during the next decades.
Finally, there are several limitations in this study linked especially to the idealized nature of our experiments. The prescribed rate of Indian Ocean warming is, for instance, much swifter than observed. To what extent this will influence the atmospheric trend behavior toward the annular structure is unclear. The present study may also miss possible impacts of stratospheric climate change and its impact on the extratropical troposphere. For example, none of the models analyzed included a proper simulation of stratospheric dynamics. Scaife et al. (2005) argued that realistic changes in the NAO can only be simulated when including realistic changes in the stratospheric circulation. A recent study by Lin et al. (2009) showed that increased wave forcing of stratospheric circulation in boreal spring caused a warming of the polar lower stratosphere resulting from an increase of the Brewer–Dobson circulation and has partly offset the cooling of the stratosphere by ozone depletion. The implications on tropospheric climate change have not been studied to date.
The authors thank anonymous reviewers for constructive suggestions, which led to a significant improvement of the manuscript, and acknowledge Dr. Walter Robinson for helpful discussion and Dr. Tore Furevik for commenting on an early version of the manuscript. The experiments with the GFS, CCM3, CAM3, and ECHAM5 models were carried out by Xiaowei Quan, Gary Bates, Adam Phillips, and Kaiming Hu, respectively. This study was jointly supported by the Innovation Key Program (Grant KZCX2-YW-BR-14) of the Chinese Academy of Sciences, the National Basic Research Program (Grant 2010CB428602), and the National Natural Science Foundation of China (Grant 40775053). The contribution of JP was supported by the NOAA/Climate Program Office.
Corresponding author address: Dr. Shuanglin Li, Nansen-Zhu International Research Centre, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China. Email: email@example.com