Abstract

A tropical–polar connection and its seasonal dependence are examined using the real-time multivariate Madden–Julian oscillation (MJO) (RMM) index and daily indices for the annular modes, the Arctic Oscillation (AO) and the Antarctic Oscillation (AAO). On the intraseasonal time scale, the MJO appears to force the annular modes in both hemispheres. On this scale, during the cold season, the convection in the Indian Ocean precedes the increase of the AO/AAO. Interestingly, during the boreal winter (Southern Hemisphere warm season), strong MJOs in the Indian Ocean are related to a decrease of the AAO index, and AO/AAO tendencies are out of phase. On the longer time scales, a persistent AO/AAO anomaly appears to influence the convection in the tropical belt and impact the distribution of MJO-preferred phases. It is shown that this may be a result of the sea surface temperature (SST) changes related to a persistent AO, with cooling over the Indian Ocean and warming over Indonesia. In the Southern Hemisphere, the SST anomalies are to some extent also related to a persistent AAO pattern, but this relationship is much weaker and appears only during the Southern Hemisphere cold season. On the basis of these results, a mechanism involving the air–sea interaction in the tropics is suggested as a possible link between persistent AO and convective activity in the Indian Ocean and western Pacific.

1. Introduction

The Madden–Julian oscillation (MJO) is primarily a tropical phenomenon, with its influence extending to the extratropics through the teleconnection reaching the polar regions in both hemispheres. This interaction appears to be stronger in the Northern Hemisphere (NH), where the largest MJO amplitudes coincide with the winter circulation. There is some evidence that the MJO influences the southern as well as northern polar regions. An MJO influence on the Arctic Oscillation (AO) was noted by Miller et al. (2003), who found that positive AO months were associated with the MJO convection concentrated in the Indian Ocean. The opposite was true for the negative AO. In the Southern Hemisphere (SH), the influence of Antarctic Oscillation (AAO) was limited to the polar regions.

Zhou and Miller (2005) showed that the MJO influences the AO through the Rossby wave dispersion in the Pacific. L’Heureux and Higgins (2008) related the eastward progression of the MJO to the tendency of the AO index (Thompson and Wallace 1998) and noticed the similarities of the extratropical features of the MJO and the AO anomalies. Kim et al. (2006) examined a mechanism of the tropical–extratropical interaction during different phases of MJO and stressed the importance of the position of the tropical convection relative to the Asian–Pacific jet for the generation of midlatitude anomalies of divergent circulations. Contrary to the findings of Miller et al. (2003), Matthews and Meredith (2004) observed the connection between the MJO and polar circulation around 60°S. They showed that, during the SH winter, an increase of MJO convection in the Indian Ocean is followed by anomalous westerlies around 60°S. The stronger surface westerlies lead to an increased circumpolar oceanic transport around Antarctica.

Asymmetric modes, such as the North Atlantic Oscillation (NAO) and the Pacific–North America pattern (PNA), also exhibit phase locking with the MJO. As shown by Mori and Watanabe (2008), the positive PNA is associated with MJO convection over the Bay of Bengal through the development of the Rossby wave train. A similar mechanism is responsible for an increase of NAO amplitude following MJO convection in the warm pool of the Indian Ocean and western Pacific Ocean (Lin et al. 2009). As shown by Roundy et al. (2010), the strength of the response of polar circulation (NAO) to the phase of MJO is modulated by El Niño–Southern Oscillation (ENSO), with the strongest response observed during the cold ENSO phase.

Even though, by now, a great deal is known about MJO forcing of AO, it is not clear how persistent positive or negative AO influences the distribution of MJO convection in the tropics. The goal of this research is to answer this question by examining the seasonality of phase locking between the MJO and annular modes in both hemispheres. We investigate the dependence of this interaction on the time scale of annular mode variability and on the strength of the MJO and AO/AAO. This paper is organized as follows: section 2 describes the processing of the data, sections 3 and 4 examine the relationship between the MJO and annular modes, and section 5 discusses a conceptual model of the influence of annular modes on tropical convection.

2. Methodology for data analysis

The interaction between the MJO and annular modes is studied using the daily indices of the MJO (Wheeler and Hendon 2004) and AO and AAO indices in the 30-yr-long time series from 1979 to 2009. Wheeler and Hendon’s real-time multivariate MJO (RMM) index is based on the first two normal modes [empirical orthogonal functions (EOFs)] of 850-hPa and 200-hPa winds and outgoing longwave radiation (OLR) averaged in the 15°N–15°S belt. The projection of wind and OLR anomalies on these first two EOFs gives the two principal components (PCs) called RMM1 and RMM2. Plotted in the phase space, RMM1 and RMM2 constitute an index describing the propagating mode, where the index amplitude indicates the strength of the MJO and the phase shows the geographic location of the MJO-related convection. We use the index data calculated by Wheeler and published online by the Australian Bureau of Meteorology (http://cawcr.gov.au/staff/mwheeler/maproom/RMM/RMM1RMM2.74toRealtime.txt). The same index was used by L’Heureux and Higgins (2008) in an investigation of AO forcing by the MJO.

To assess the state of the polar circulation, we use the Climate Prediction Center’s AO and AAO indices, which are based on the daily projection of 1000-hPa and 700-hPa geopotential height anomalies poleward of 20°N and 20°S, respectively, on the leading EOF mode. We further filter the AO and AAO indices to retain the intraseasonal component (40–90 days; denoted as AOI and AAOI) and long-term component (120 days and longer; denoted as AOL and AAOL). We consider two seasons, May–October (warm season for AO and cold for AAO) and November–April (the opposite). The cold seasons in both hemispheres are more active periods for annular modes, with higher amplitude of the variance in daily 1000-hPa (NH) and 850-hPa (SH) geopotential height fields (Thompson and Wallace 2000). We calculate the distribution of MJO phases for positive and negative annular mode events for each season and for various amplitudes of MJO, AOI, AOL, AAOI, and AAOL. To emphasize the difference between the negative and positive phases of an annular mode, the results are normalized using the “background” distribution of MJO phases, that is, the distribution that includes both negative and positive signs of annular modes.

To be included in our calculations, the MJO episodes need to have amplitudes larger than 1; otherwise, the phase calculations are meaningless since no active MJO is usually observed (Wheeler and Hendon 2004). The amplitude and phase of the RMM index can be modified by tropical convectively coupled waves (Roundy et al. 2009). As shown by Roundy et al. (2009), filtering of the zonal wind and OLR before projecting them on RMM normal modes yields a much smoother index. However, in this paper we use the index as a measure of the equatorial convection and we do not attempt to separate the tropical modes that contribute to its value. In our calculations, very short episodes related to convectively coupled waves can be removed by imposing conditions on running means of MJO amplitude instead of its daily value. However, since this procedure does not appear to influence the results of our calculations, the conditions for individual days are used. Therefore, our requirements for MJO-related convection are less strict than those used in L’Heureux and Higgins (2008) and can include some signal from tropical waves, not only MJO. For annular modes, we include those cases in which the amplitude is larger than one standard deviation of the component we consider; for example, the positive AOI event should have the amplitude larger than the standard deviation of AOI. This normalization is done separately for warm and cold seasons because of the difference in variability of the annular modes. Lagged correlations between the MJO PCs, RMM1 and RMM2, and annular modes are also discussed.

Since we consider not only the influence of the MJO on annular modes but also the impact of the phase of these modes (especially AO) on tropical convection, we do not group the active MJO days into individual events, as was done by L’Heureux and Higgins (2008), but rather, we consider each day separately. This allows us to filter the data according to the strength of the MJO convection on each day. The statistical significance of the results is tested using the Monte Carlo approach. For each case, the number of days ND that satisfy the imposed criteria is determined, and 1000 subsets of ND/2 days are randomly drawn. MJO phase distributions for these subsets are used to evaluate the standard deviation of the fractions of days per MJO phase.

In addition, we calculate the MJO phase distribution treating the consecutive days with the same MJO phase as a single “event” instead of considering individual days. In this case, the results are not biased toward MJOs in which the strong convection persists for an extended time in one phase. The results of this calculation are shown in Table 1.

Table 1.

The distribution of MJO phases for the MJO amplitudes larger than 1.2 and normalized amplitudes of AO and AAO components larger than 1.2, calculated with the assumption that the consecutive days spent in a particular MJO phase are counted only once. The values in boldface show the increased convection either in the Indian Ocean (1–4) or Pacific Ocean (5–8) basins.

The distribution of MJO phases for the MJO amplitudes larger than 1.2 and normalized amplitudes of AO and AAO components larger than 1.2, calculated with the assumption that the consecutive days spent in a particular MJO phase are counted only once. The values in boldface show the increased convection either in the Indian Ocean (1–4) or Pacific Ocean (5–8) basins.
The distribution of MJO phases for the MJO amplitudes larger than 1.2 and normalized amplitudes of AO and AAO components larger than 1.2, calculated with the assumption that the consecutive days spent in a particular MJO phase are counted only once. The values in boldface show the increased convection either in the Indian Ocean (1–4) or Pacific Ocean (5–8) basins.

3. The relationship between the Madden–Julian oscillation and the Arctic Oscillation

First, we consider the dependence of MJO phase distribution on the sign of AO for warm and cold seasons and different strengths of the MJO and AO. Figures 1 and 2 show the results for relatively strong MJO and AO/AAO events, with MJO and normalized AO/AAO amplitudes exceeding 1.2. In addition, the 7-day lag is used when considering the intraseasonal component of the polar modes (AOI and AAOI following MJO).

Fig. 1.

(a) Distribution of MJO phases for the positive and negative states of the intraseasonal component of the AOI for the cold season (November–April). (b) As in (a), but for the warm season (May–October). (c) As in (a), but for the seasonal component AOL. (d) As in (c), but for the large (>2) MJO amplitudes. The number of days used in the calculation is shown for each case. The dashed lines indicate the standard deviation calculated using the Monte Carlo technique described in the paper.

Fig. 1.

(a) Distribution of MJO phases for the positive and negative states of the intraseasonal component of the AOI for the cold season (November–April). (b) As in (a), but for the warm season (May–October). (c) As in (a), but for the seasonal component AOL. (d) As in (c), but for the large (>2) MJO amplitudes. The number of days used in the calculation is shown for each case. The dashed lines indicate the standard deviation calculated using the Monte Carlo technique described in the paper.

Fig. 2.

(a) Distribution of MJO phases for the positive and negative states of the intraseasonal component of AAOI for the cold season (May–October). (b) As in (a), but for November–April. (c) As in (a), but for the seasonal component AAOL. (d) As in (b), but for the seasonal component AAOL.

Fig. 2.

(a) Distribution of MJO phases for the positive and negative states of the intraseasonal component of AAOI for the cold season (May–October). (b) As in (a), but for November–April. (c) As in (a), but for the seasonal component AAOL. (d) As in (b), but for the seasonal component AAOL.

The cold season results (Fig. 1a) show that, for the positive AOI, the MJO convection is concentrated in the Indian Ocean (MJO phases 1–4), while for the negative AOI the Pacific convection (MJO phases 5–8) dominates. For the positive AOI, the largest fraction of the MJO convection appears in phases 2 and 3, while the smallest fraction can be seen in phases 7 and 8. These results confirm what was observed by L’Heureux and Higgins (2008) for the winter months [December–February (DJF)]. Our calculations indicate that a similar MJO–AOI phase-locking pattern is observed during the warm period (Fig. 1b), although the asymmetry of the distribution is less pronounced and the errors are much larger. Therefore, even though the conditions for the teleconnection are less favorable during the warm season and the MJO and AO episodes are observed less frequently, some evidence of phase locking exists between the MJO and AO throughout the whole year. The AO–MJO connection is the weakest for the summer months [June–August (JJA), especially in June and July] and strongest in winter (DJF) and spring (March–May; not shown here). It appears that in all the seasons the difference between the positive and negative AO phase is more pronounced in the Indian Ocean than in the western Pacific.

As mentioned before, the MJO phase distribution was also calculated using the assumption that all consecutive days in one MJO phase constitute one event. These results are shown in Table 1 and are consistent with Fig. 1. For cold and warm seasons, the convection is stronger in the Indian Ocean for the positive AOI and in the western Pacific for negative AOI. The asymmetry between the warm and cold seasons observed for AAO, evident in the results in Table 1, agrees with the calculations presented in Fig. 2 and will be discussed in the next section. Since the consecutive days spent in one MJO phase are not really “independent events,” establishing the effective degrees of freedom for our calculation is not obvious. To account for that deficit, we recalculated the errors in Figs. 1 and 2, using the numbers from Table 1 as a measure of independent events, and used this number in Monte Carlo error calculations. In this approach, the error indicated in Figs. 1 and 2 increases by about 15%–20%.

In addition, an interesting relationship emerges between the intraseasonal and long time scales in terms of the MJO–AO phase locking (Fig. 1c). It appears that, during the cold season, the intraseasonal (i.e., AOI) and long-term (i.e., AOL) components of the AO variability are related to different distributions of the MJO convection. The positive sign of the AOL index is related to decreased Indian Ocean convection and increased Maritime Continent and western Pacific convection. The opposite is true for the negative AOL. This distribution asymmetry is weaker than what we observe for an intraseasonal component and is limited to the cold season. In the western Pacific, the standard deviation for distribution becomes quite large, and the clear separation between the positive and negative AOL develops only in the Indian Ocean. The difference between the Indian Ocean and western Pacific convection appears to increase for stronger MJOs. Figure 1d shows the phase distribution for MJO magnitudes exceeding 2. In this case, a clear difference between negative and positive AOL is apparent in both basins.

This relationship could provide the negative feedback in the MJO–AO interplay, with the persistent AO pattern influencing the magnitude of the MJO convection, which will be discussed later in this paper.

4. The relationship between the Madden–Julian oscillation and the Antarctic Oscillation

In the SH during the cold season (May–October), the MJO–AAOI relationship (Fig. 2a) looks somewhat similar to the Arctic cold season mode, although the maximum difference between positive and negative signs of the polar circulation is shifted slightly to the east (phases 1–2 instead of 2–3).

Figure 2 also shows that, unlike for the AOI, the MJO–AAOI relationship changes during the SH warm months (November–April) and the asymmetry between the Indian Ocean and western Pacific convection is quite clear. This is less apparent for smaller MJO and AAO amplitudes (not shown here). This is consistent with what was observed by Carvalho et al. (2005) for boreal winter (DJF). Our results contradict to some extent the findings of Pohl et al. (2010). Pohl et al. (2010) used the AAO and MJO indices and concluded that the connection between the MJO and AAO is very weak, even though the MJO clearly influences the polar circulations. According to Pohl et al., this influence does not exhibit the zonally symmetric pattern typical for AAO, but it has wavenumber-1 characteristics and therefore cannot impact annular modes. They state that a consideration of the seasonal patterns does not change their conclusion. In our calculations, however, the seasonal differences are quite evident. We suggest that the fact that there is a seasonal dependence of the AAO and MJO phase locking may explain the result of Pohl et al. (2010) that there is no significant relationship between the MJO and AAO variability at any frequency. We examine this problem further by analyzing the tendency of the AAOI, defined as a change of the index over 1 day. This approach was used by L’Heureux and Higgins (2008) to show that the positive tendency of AO is concentrated in MJO phases 1–3.

Figure 3 shows the comparison of the lag correlations of AOI and AAOI tendencies with the MJO PCs, RMM1 and RMM2 (Wheeler and Hendon 2004), for all the days with observed MJO. Since RMM1 and RMM2 describe the propagating phenomenon, the correlation coefficient is similar for both PCs, but there is a difference in the lag for the maximum correlation. The periodicity in the lagged correlation is related to periodicity of MJO and appears also in the lag correlation for unfiltered AO data (not shown), although the values of the correlations are smaller than what we calculated for the intraseasonal component (i.e., AOI). The correlation for the AAOI tendency is much smaller than what is observed for the AOI; this is probably the reason why Pohl et al. (2010) conclude that there is no significant influence of the MJO on AAO. However, the pattern changes when we consider the cold and warm months separately. Figure 3b shows the lag correlation between index tendencies and RMM2 only for the warm and cold seasons in both hemispheres. For the cold months the AAOI–MJO correlation looks similar to that of AOI–MJO, but for the warm months the correlation is slightly out of phase with what is observed for the SH cold months. The correlation coefficient increases for larger amplitudes of MJO (not shown). This increase is especially significant during the SH warm months, since this is the period when intense heat sources related to strong MJO events can develop. If we limit ourselves to very strong events with amplitudes larger than 3, the correlation coefficient between the AAOI tendency and RMM2 reaches the maximum value of about 0.4 (with 95% significance level) for the AAOI tendency, lagging the RMM2 by about 15 days (not shown). This implies that the strength of the MJO convection is an important factor in triggering the AAO changes, with only the strong episodes having an effect on polar circulations.

Fig. 3.

(a) The lag correlation between AOI (black) and AAOI (red) and the MJO normal vectors RMM1 (solid) and RMM2 (dashed). The positive values on the x axis indicate that MJO is leading AO/AAO. The symbols denote the correlations which are significant at the 95% level. Only the days with MJO amplitudes larger than 1 are considered. Both warm and cold seasons are included. (b) Seasonal lag correlation (days) of the RMM2 vector with AOI (solid) and AAOI (dashed) for each hemisphere. The black lines denote the cold season, while the red lines denote the warm season. The symbols denote the correlations that are significant at the 95% level.

Fig. 3.

(a) The lag correlation between AOI (black) and AAOI (red) and the MJO normal vectors RMM1 (solid) and RMM2 (dashed). The positive values on the x axis indicate that MJO is leading AO/AAO. The symbols denote the correlations which are significant at the 95% level. Only the days with MJO amplitudes larger than 1 are considered. Both warm and cold seasons are included. (b) Seasonal lag correlation (days) of the RMM2 vector with AOI (solid) and AAOI (dashed) for each hemisphere. The black lines denote the cold season, while the red lines denote the warm season. The symbols denote the correlations that are significant at the 95% level.

The seasonality of the RMM2–AAOI correlation and its dependence on the MJO strength are reflected in the pattern of the AAOI tendency. Figure 4 compares the distribution of the average tendency of an index for AOI and AAOI for the warm and cold months. In the cold months, for all MJO events with the magnitude larger than 1, the average positive tendency is equally distributed between phases 1–3 for AOI, but concentrated mostly in phase 1 for AAOI. This may be explained by a slightly different lag between the MJO and the strongest response observed for the AAO and AO (Fig. 3b); it apparently takes longer for the tropical signal to have an impact on the AAO tendency. It is worth noting that, even though the average tendency of the AOI and AAOI indices shows a clear pattern, the standard deviation for the calculated averages is rather large compared with the average tendencies, from about 0.05 for the warm seasons to about 0.08 for the cold seasons when the variability of the annular oscillations is larger. This implies that, while important, the MJO is not the only or even the primary factor influencing polar annular circulations. The seasonality of the MJO–AAO interaction can be explained by the seasonal changes in the annular mode itself. Barnes and Hartmann (2010) show that the 300-hPa flow has different characteristics during summer and winter, with a strong jet over the Indian Ocean in winter and zonal asymmetries of the flow that would enable interaction with the zonally asymmetric MJO circulation. Interestingly, the second EOF of the Antarctic 700-hPa variability (not shown here), which explains 10% of the variability, also shows the seasonal dependence on MJO phase. This mode consists of the intense midlatitude high in the southern Pacific to the west of the Drake Passage. The strength of this high is related to the western Pacific MJO convection during DJF. Other features of the circulation near the Drake Passage, such as the zonal surface winds, also exhibit a relationship with the MJO phase similar to that observed for AAO.

Fig. 4.

The average change (tendency) of the index over 1 day for (a) AOI and (b) AAOI. The cool colors denote the changes for the winter cold season, and the warm colors denote the changes for the warm season. The minimum MJO amplitude considered in the calculation of each average tendency is indicated on the figure.

Fig. 4.

The average change (tendency) of the index over 1 day for (a) AOI and (b) AAOI. The cool colors denote the changes for the winter cold season, and the warm colors denote the changes for the warm season. The minimum MJO amplitude considered in the calculation of each average tendency is indicated on the figure.

Very strong MJOs developing during the SH warm season (NH cold season) appear to influence the annular modes in both hemispheres. In Fig. 5, we show the tendencies of the AOI and AAOI for MJO episodes, with the magnitudes exceeding 3. According to the lag correlations, the strongest correlation between RMM2 and AOI develops with the 6-day lag and for AAOI with the 15-day lag; therefore, we shift the AOI tendency by 6 days and AAOI by 15 days relative to the MJO magnitude. The figure shows a series of individual episodes lasting between 4 and ~20 days, separated by the blue vertical lines. During these episodes, AO and AAO tendencies are rather large, with the index increasing or decreasing rapidly over a few days. The tendencies for AO and AAO are out of phase, with increasing AOI associated with decreasing AAOI and vice versa. This suggests that very strong MJO episodes could trigger a global teleconnection pattern, not limited to the winter hemisphere only.

Fig. 5.

The change of the AOI and AAOI indices over 1 day for the very strong MJO episodes that developed during the NH cold season. Only the days with the MJO index magnitude larger than 3 are considered. The solid line shows the tendency for the AO; the dashed line shows the tendency for the AAO. The vertical blue lines separate the individual episodes, which are accumulated in days for convenience and shown on the x axis. The numbers at the top indicate the year for which the episode was observed.

Fig. 5.

The change of the AOI and AAOI indices over 1 day for the very strong MJO episodes that developed during the NH cold season. Only the days with the MJO index magnitude larger than 3 are considered. The solid line shows the tendency for the AO; the dashed line shows the tendency for the AAO. The vertical blue lines separate the individual episodes, which are accumulated in days for convenience and shown on the x axis. The numbers at the top indicate the year for which the episode was observed.

The interhemispheric connection during the Antarctic summer is also evident in the EOFs of the global 700-hPa height. Figure 6 shows the first EOF of the global 700-hPa height anomaly calculated from the 30 years of monthly data for Antarctic summer (DJF) and winter (JJA). During DJF, the patterns on both hemispheres resemble the EOFs calculated separately for 20°–90°S and 20°–90°N, but with a smaller percentage of variance explained (15% versus 21% and 19%, respectively). The Antarctic winter pattern, with 21% of the variance explained, is very similar to the 20°–90°S all-season pattern and does not exhibit any correlation between the hemispheres. It is not clear what the reason is for this interhemispheric connection during the boreal winter. While the annular modes are an intrinsic mode of atmospheric variability and appear to be forced by eddy fluxes (Limpasuvan and Hartmann 2000), the impact of MJO on the link between these modes is an interesting possibility.

Fig. 6.

The global EOFs for 700-hPa geopotential height for (a) DJF and (b) JJA in (top) the Antarctic and (bottom) the Arctic. The percentage of variance explained by these patterns is 15% for DJF and 21% for JJA.

Fig. 6.

The global EOFs for 700-hPa geopotential height for (a) DJF and (b) JJA in (top) the Antarctic and (bottom) the Arctic. The percentage of variance explained by these patterns is 15% for DJF and 21% for JJA.

5. Discussion

The discrepancy of the MJO phase distribution between the intraseasonal and long-term components of AO suggests that, while the intraseasonal component of AO is forced to some extent by the MJO, the long-term component, that is, the persistent state of the AO, can in turn affect the MJO convection. We hypothesize that this influence could also be explained in terms of the air–sea interaction. One possible mechanism, suggested by the results of previous research, involves a cooling of the ocean by the tropical component of OLR anomalies related to AO (Miller et al. 2003). Miller et al. (2003) showed that there is a positive AO index signature in the OLR pattern in the tropical oceans, with the negative OLR anomaly in the Indian Ocean and positive anomaly in the western Pacific. This pattern, if persistent, could influence the sea surface temperature (SST) anomalies in this region, which in turn could impact the strength of MJO-related convection.

The study by Buermann et al. (2005) indicates that this might be the case. Buermann et al. showed an influence of the winter AO on the early season development of the Indian monsoon. In particular, winters with a prevalent negative AO are followed by warmer SSTs in the Arabian Sea region, contributing to intense rainfall in the monsoon region. Buermann et al. suggested that the negative AO is related to the suppressed cloudiness over North Africa, the Arabian Peninsula, and the northern Indian Ocean, which can result in a warming of the ocean and land surface in the months preceding the monsoon. Such an SST change could also influence the development of the MJO.

This conceptual model is consistent with the results of Gong et al. (2009), who examined the strontium content (a proxy for SST) in coral from the South China Sea. They found the relationship between the winter AO and SST anomaly in the following spring. They showed that January, with the positive AO, results in higher SSTs in the South China Sea in the following months (especially March). We hypothesize that such an SST anomaly could enhance convective anomalies in MJO phase 5 and enable transition of the MJO convection to the western Pacific, which can be seen in Fig. 1c. The fact that the strong correlation observed in the Gong et al. (2009) data occurs only in January may explain why the difference in the MJO phase distribution between AOI and AOL occurs only during the cold season. As seen in Figs. 1b and 1d, during the warm season there is no significant difference in the MJO phase distribution between AOI and AOL, and the increased Indian Ocean convection is related to the positive index value for both intraseasonal and long-term components of AO.

In the SH for strong MJO and AAO events, the reversal of the MJO distribution is also observed for AAOI and AAOL during SH cold season. It is similar to, but less pronounced than, what we have shown for AOI and AOL. According to Gong et al. (2009), the relationship between AAO and SST in the South China Sea is opposite to that observed for AO; that is, if there is a strong AAO episode in August, the South China Sea becomes colder in the following months. That would suggest a reduction of MJO phases 5–6 relative to negative AAOL; this is indeed the case, but the MJO convection in the Indian Ocean also seems to be reduced.

Our hypothesis that persistent AO could influence the tropical SST pattern agrees with the results of Wu (2010), who showed the relationship between the polar height anomalies representing AO and the global SST pattern. He used the lagged maximum covariance analysis (MCA) of the 500-hPa height and SST in the global tropics to show that the second mode of MCA represents the influence of AO on tropical SST anomalies. In particular, he showed that the positive AO pattern is related to the negative phase of the El Niño Modoki mode (Ashok et al. 2007). El Niño Modoki is defined as the second EOF of the tropical Pacific SST variability. It is characterized by a V-shaped warm anomaly in the central Pacific, with the center on the equator around the date line and the arms extending eastward and poleward. Cold anomalies are present in the eastern Pacific cold tongue and in the vicinity of the Maritime Continent. As shown by Ashok et al. (2007), the positive Modoki mode is associated with increased precipitation in the western and central Pacific and Indian Ocean, while the precipitation around the Maritime Continent from 100° to about 140°E is decreased. The negative Modoki, associated with positive AO influence, is characterized by the positive anomalies in the Indonesian area and therefore should contribute to increased MJO magnitude in phases 4–6.

To further analyze the SST variability in the Indian Ocean and Indonesian region, where the convection associated with the MJO is most active, we calculate the normal modes of the Kaplan SST monthly anomaly (Kaplan et al. 1998). The first mode (not shown) corresponds to El Niño and explains 25% of the variability in this region. The second mode, which explains 17% of the variability, is shown in Fig. 7 and is similar to the negative El Niño Modoki. It has a positive anomaly in the Indonesian region and a negative anomaly in the Indian Ocean. The SST anomaly related to this pattern should contribute to depressed convection in the Indian Ocean (i.e., during MJO phases 1–3) and increased convection around Maritime Continent (phases 4–6). We are trying to determine a time scale on which the interaction between AO and SST patterns may occur by looking at the correlations between the indices for various time-filtering scales. Figure 8 shows the time series of the monthly AO index and PC of the second EOF filtered on the 5–10-month time scale. The results indicate that the positive values of the second PC are correlated with the positive AO index and lag the index by about 1–2 months. The correlation is the largest (0.43 and 0.31) for the time scales of 5–10 and 4–10 months, respectively. For longer and shorter periods, the correlation decreases and approaches values of correlations of random time series filtered in a similar manner. This time scale is consistent with the observations of Buermann et al. (2005) and is slightly larger than the scale of about 3 months that can be inferred from the results of Gong et al. (2009). Interestingly, Fig. 8 suggests that the correlation between AO and the second PC of SST seems to be the lowest when the first PC is negative, that is, in 1988–90 and 1998–2000. These are the years in which a large La Niña was observed, and thus, the AO influence on SST was probably obscured by the large-scale ocean processes. The correlation between the second PC and the SST anomaly and the AAO coefficient is much weaker than for AO (0.13 versus 0.31, respectively), and it can be observed only during the SH winter months (JJA). Therefore, the forcing of the AAO on the SST anomaly pattern and the MJO convection seems to be weaker than what was observed for AO.

Fig. 7.

The second EOF (explaining 17% of the variability) of the 30-yr time series of the Kaplan monthly SST anomaly in the Indian Ocean and Indonesia. The color scale denotes the SST anomaly in degrees Celsius.

Fig. 7.

The second EOF (explaining 17% of the variability) of the 30-yr time series of the Kaplan monthly SST anomaly in the Indian Ocean and Indonesia. The color scale denotes the SST anomaly in degrees Celsius.

Fig. 8.

The time series of the monthly AO index (black) and the second PC (PC2) corresponding to the EOF (red). The indices are filtered to keep the time scales between 10 and 5 months, for which the largest correlation between the AO and the PC2 of the SST is observed. The blue curve denotes the PC of the first mode (PC1), which is related to the global El Niño mode.

Fig. 8.

The time series of the monthly AO index (black) and the second PC (PC2) corresponding to the EOF (red). The indices are filtered to keep the time scales between 10 and 5 months, for which the largest correlation between the AO and the PC2 of the SST is observed. The blue curve denotes the PC of the first mode (PC1), which is related to the global El Niño mode.

While in our conceptual model of AO the role of air–sea interaction in AO influence on tropical convection appears to agree with previous observations, it is highly speculative. The correlations of the time series shown in Fig. 8 are relatively low and appear only for the filtered time series. To fully investigate this process, a coupled dynamic model would be necessary, but it is beyond the scope of this paper.

The examination of the circulation anomalies related to AOI and AOL indicates that zonal wind anomalies can also contribute to an enhancement of convection in the western Pacific, following the strong and long-lasting AO positive phase. Figure 9 shows the composites of the 200-hPa zonal wind associated with the MJO and positive AOI and AOL (corresponding to the cases shown in Fig. 1a and listed in Table 1).

Fig. 9.

The composites of the 200-hPa zonal wind anomaly corresponding to the presence of the MJO and the positive (a) AOI and (b) AOL indices (as described in Table 1). The color scale denotes the zonal wind in meters per second.

Fig. 9.

The composites of the 200-hPa zonal wind anomaly corresponding to the presence of the MJO and the positive (a) AOI and (b) AOL indices (as described in Table 1). The color scale denotes the zonal wind in meters per second.

The 200-hPa wind anomaly for the positive AOI shows the wave train originating in the Indian Ocean, consistent with the forcing of AO changes by propagating MJO, as suggested by L’Heureux and Higgins (2008). The composite for the positive AOL exhibits much weaker anomalies in the Indian Ocean, but has a distinctive westward-tilting pattern in the Atlantic with a strong anomaly in the midlatitudes, a negative anomaly around 30°N, and a positive anomaly in the equatorial western Atlantic. This pattern closely resembles the regression map of 200-hPa zonal wind, lagged with respect to the NAO index by 3 pentads from Lin et al. (2009). Lin et al. argue that a further eastward propagation of this pattern leads to a tropical zonal wind anomaly very similar to the 200-hPa zonal wind distribution associated with phase 6 of the MJO. They suggest this extratropical forcing may be responsible for prevalent phases 6–7 of MJO that follow the strong episodes of positive NAO. The resemblance of the circulation features observed for long-lasting AO to the NAO features from Lin et al. (2009) suggests that the dynamical effect could enforce the air–sea interaction mechanism described earlier in this paper.

6. Summary

We have examined the relationship between the polar annular modes (AO and AAO) and the MJO phase distribution in all seasons for various magnitudes of the MJO and annular mode indices. Some of the results confirm what was observed previously (L’Heureux and Higgins 2008; Matthews and Meredith 2004), namely, that for the intraseasonal time scale during the cold season, positive indices are related to Indian Ocean convection (MJO phases 1–4) for both AO and AAO. We have shown that, for AO, this relationship also holds in the warm season, but the phase distribution for AAO changes during the warm months (Carvalho et al. 2005). Contrary to the conclusion of Pohl et al. (2010) that the MJO does not have a major impact on AAO, we find a significant contribution of the MJO to the AAO tendency on the intraseasonal scale, especially for strong MJO episodes.

We show that during the cold season the phase distribution of MJO for positive and negative annular modes depends on the time scale of these modes. For the positive phases of AOI and AAOI, the MJO convection is located mostly in the Indian Ocean, but the opposite is true for the negative phases of AOL and AAOL. We propose a conceptual model of the processes based on the air–sea interaction in the Indian Ocean and western Pacific. In this conceptual model, the persistent OLR anomalies related to annular mode contribute to SST modification and influence the distribution of MJO convection. This model is somewhat speculative, and positive correlations between AO and SST patterns are limited to the 4–10 time scales. However, it is consistent with observations of Miller et al. (2003), Gong et al. (2009), and Buermann et al. (2005). Since the correlation is not very strong, a coupled global circulation model should be used to further examine the role of air–sea interaction in the connection between polar circulations and the MJO.

Some questions remain unanswered: AAO’s influence on SST appears to be much weaker than what we see for AO. However, we can still observe the phase reversal between AAOI and AAOL in the MJO phase distribution. In addition, it is not clear why this interaction is observed only during the cold season for both AO and AAO. Since the amplitude of the MJO is usually smaller in boreal summers, we could expect weaker SST anomalies in May–November and less air–sea interaction impact. However, the phase-locking reversal can be observed for the AAO during this time. One possible answer is the combined effect of the air–sea interaction and dynamical influences similar to that observed for NAO (Lin et al. 2009), as shown in Fig. 9. Another possibility may be the impact of stratospheric processes on the MJO–AO (or AAO) phase-locking pattern. Kim and Flatau (2010) confirmed the strong influence of the winter Arctic vortex on the surface in terms of the AO index. Baldwin and Dunkerton (2001) show that the development of the strong polar vortex (positive AO) is much more likely with the westerly phase of quasi-biennial oscillation (QBO) than with its easterly phase. Observations indicate the interaction between the QBO and tropical convection, in which the westerly phase of QBO is related to stronger winter convection in the western Pacific (Collimore et al. 2003). This behavior is consistent with the predominant western Pacific convection (MJO phases 5–7) that coincides with positive AOL. Since the connection between stratospheric and tropospheric polar vortex is stronger in winter months (Baldwin et al. 2003), the stratospheric interaction could explain why the MJO–AO phase lock shift between intraseasonal and seasonal time scales is observed only during cold seasons. This is a subject of a further study.

Acknowledgments

The authors appreciate the support by the Office of Naval Research under Program Element 0601153N. The authors are grateful to the anonymous reviewers for the insightful and extensive comments that contributed to improvement of the manuscript. They would like to thank Piotr J. Flatau for his suggestions and help.

REFERENCES

REFERENCES
Ashok
,
K.
,
S. K.
Behera
,
S. A.
Rao
,
H.
Weng
, and
T.
Yamagata
,
2007
:
El Niño Modoki and its possible teleconnection
.
J. Geophys. Res.
,
112
,
C11007
,
doi:10.1029/2006JC003798
.
Baldwin
,
M. P.
, and
T. J.
Dunkerton
,
2001
:
Stratospheric harbingers of anomalous weather regimes
.
Science
,
294
,
581
584
.
Baldwin
,
M. P.
,
D. B.
Stephenson
,
D. W. J.
Thompson
,
T. J.
Dunkerton
,
A. J.
Charlton
, and
A.
O’Neill
,
2003
:
Stratospheric memory and skill of extended-range weather forecasts
.
Science
,
301
,
636
640
.
Barnes
,
E. A.
, and
D. L.
Hartmann
,
2010
:
Dynamical feedbacks of the southern annular mode in winter and summer
.
J. Atmos. Sci.
,
67
,
2320
2330
.
Buermann
,
W.
,
B.
Lintner
, and
C.
Bonfils
,
2005
:
A wintertime Arctic Oscillation signature on early-season Indian Ocean monsoon intensity
.
J. Climate
,
18
,
2247
2269
.
Carvalho
,
L. M. V.
,
C.
Jones
, and
T.
Ambrizzi
,
2005
:
Opposite phases of the Antarctic Oscillation and relationships with intraseasonal to interannual activity in the tropics during the austral summer
.
J. Climate
,
18
,
702
718
.
Collimore
,
C. C.
,
D. W.
Martin
,
M. H.
Hitchman
,
A.
Huesmann
, and
D. E.
Waliser
,
2003
:
On the relationship between the QBO and tropical deep convection
.
J. Climate
,
16
,
2552
2568
.
Gong
,
D.-Y.
,
S.-J.
Kim
, and
C.-H.
Ho
,
2009
:
Arctic and Antarctic Oscillation signatures in tropical coral proxies over the South China Sea
.
Ann. Geophys.
,
27
,
1979
1988
.
Kaplan
,
A.
,
M. A.
Cane
,
Y.
Kushnir
,
A. C.
Clement
,
M. B.
Blumenthal
, and
B.
Rajagopalan
,
1998
:
Analyses of global sea surface temperature 1856–1991
.
J. Geophys. Res.
,
103
,
18 567
18 589
.
Kim
,
B.-M.
,
G.-H.
Lim
, and
K.-Y.
Kim
,
2006
:
A new look at the midlatitude–MJO teleconnection in the Northern Hemisphere winter
.
Quart. J. Roy. Meteor. Soc.
,
132
,
585
503
.
Kim
,
Y.-J.
, and
M.
Flatau
,
2010
:
Hindcasting the January 2009 Arctic sudden stratospheric warming and its influence on the Arctic Oscillation with unified parameterization of orographic drag in NOGAPS. Part I: Extended-range stand-alone forecast
.
Wea. Forecasting
,
25
,
1628
1644
.
L’Heureux
,
M. L.
, and
R. W.
Higgins
,
2008
:
Boreal winter links between the Madden–Julian oscillation and the Arctic Oscillation
.
J. Climate
,
21
,
3040
3050
.
Limpasuvan
,
V.
, and
D. L.
Hartmann
,
2000
:
Wave-maintained annular modes of climate variability
.
J. Climate
,
13
,
4414
4429
.
Lin
,
H.
,
G.
Brunet
, and
J.
Derome
,
2009
:
An observed connection between the North Atlantic Oscillation and the Madden–Julian oscillation
.
J. Climate
,
22
,
364
380
.
Matthews
,
A. J.
, and
M. P.
Meredith
,
2004
:
Variability of Antarctic circumpolar transport and the southern annular mode associated with the Madden–Julian oscillation
.
Geophys. Res. Lett.
,
31
,
L24312
,
doi:10.1029/2004GL021666
.
Miller
,
A. J.
,
S.
Zhou
, and
S. K.
Yang
,
2003
:
Relationship of the Arctic and Antarctic Oscillations to the outgoing longwave radiation
.
J. Climate
,
16
,
1583
1592
.
Mori
,
M.
, and
M.
Watanabe
,
2008
:
Growth and triggering mechanisms of the PNA: A MJO-PNA coherence
.
J. Meteor. Soc. Japan
,
86
,
213
236
.
Pohl
,
B.
,
N.
Fauchereau
,
C. J. C.
Reason
, and
M.
Rouault
,
2010
:
Relationships between the Antarctic Oscillation, the Madden–Julian oscillation, and ENSO, and consequences for rainfall analysis
.
J. Climate
,
23
,
238
254
.
Roundy
,
P. E.
,
C. J.
Schreck
, and
M. A.
Janiga
,
2009
:
Contributions of convectively coupled equatorial Rossby waves and Kelvin waves to the real-time multivariate MJO indices
.
Mon. Wea. Rev.
,
137
,
469
478
.
Roundy
,
P. E.
,
K.
MacRitchie
,
J.
Asuma
, and
T.
Melino
,
2010
:
Modulation of the global atmospheric circulation by combined activity in the Madden–Julian oscillation and the El Niño–Southern Oscillation during boreal winter
.
J. Climate
,
23
,
4045
4059
.
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
.
Wheeler
,
M. C.
, and
H. H.
Hendon
,
2004
:
An all-season real-time multivariate MJO index: Development of an index for monitoring and prediction
.
Mon. Wea. Rev.
,
132
,
1917
1932
.
Wu
,
Q.
,
2010
:
Forcing of tropical SST anomalies by wintertime AO-like variability
.
J. Climate
,
23
,
2465
2472
.
Zhou
,
S. T.
, and
A. J.
Miller
,
2005
:
The interaction of the Madden–Julian oscillation and the Arctic Oscillation
.
J. Climate
,
18
,
143
159
.

Footnotes

*

Current affiliation: The Korea Institute of Atmospheric Prediction Systems, Seoul, South Korea.