Abstract

Observations of the development of recent El Niño events suggest a pivotal role for the Madden–Julian oscillation (MJO). Previous attempts to uncover a systematic relationship between MJO activity and the El Niño–Southern Oscillation (ENSO), however, have yielded conflicting results. In this study the MJO–ENSO relationship is stratified by season, and the focus is on MJO activity in the equatorial western Pacific. The results demonstrate that MJO activity in late boreal spring leads El Niño in the subsequent autumn–winter for the period 1979–2005. Spring is the season when MJO activity is least asymmetric with respect to the equator and displays the most sensitivity to SST variations at the eastern edge of the warm pool. Enhanced MJO activity in the western Pacific in spring is associated with an eastward-expanded warm pool and low-frequency westerly surface zonal wind anomalies. These sustained westerly anomalies in the western Pacific are hypothesized to project favorably onto a developing El Niño in spring.

1. Introduction

Development of recent El Niño events suggests a pivotal role for the Madden–Julian oscillation (MJO): enhanced MJO activity across the western Pacific has been observed to precede the peak of El Niño by a few months (e.g., Weickmann 1991; Kessler et al. 1995; McPhaden 1999; Zhang and Gottschalck 2002; Lau 2005). Although intraseasonal variations of surface winds and convection over a range of time and space scales have also been observed in the western Pacific in the lead up to El Niño (e.g., Luther et al. 1983; Gutzler 1991; Vecchi and Harrison 2000), the component of this variability associated with the MJO is of most importance for the interaction with El Niño–Southern Oscillation (ENSO) because of the large spatial and temporal coherence of the surface fluxes of heat and momentum (e.g., Moore and Kleeman 1999; Zhang and Gottschalck 2002; Zavala-Garay et al. 2005). The MJO is thought to promote El Niño because its eastward-propagating, westerly stress anomalies efficiently excite downwelling Kelvin waves in the equatorial Pacific (e.g., Kessler et al. 1995; Hendon et al. 1998), which remotely act to warm the eastern Pacific (e.g., Zhang and Gottschalck 2002). Investigation of the onset of the strong 1997–98 El Niño (Bergman et al. 2001; Lengaigne et al. 2002) suggests, however, that another critical region for interaction of the MJO and ENSO is in the western Pacific warm pool. There, the MJO cools the far western equatorial Pacific by increasing the ocean–atmosphere heat flux and upper-ocean mixing (e.g., Kessler and Kleeman 2000; Shinoda and Hendon 2002). In addition, the MJO tends to warm the eastern edge of the warm pool by spinning up westerly surface currents that advect the mean east–west temperature gradient eastward (e.g., Kessler and Kleeman 2000; Shinoda and Hendon 2001). Together these mechanisms act to reduce the east–west temperature gradient on the eastern edge of warm pool, which promotes anomalous surface westerlies in the western Pacific that help shift the warm pool eastward at the onset of El Niño (e.g., Picaut et al. 2001; Lengaigne et al. 2003). Similarly, reduced MJO activity has been suggested to be associated with anomalous easterlies and a contracted warm pool in the western Pacific, thus promoting development of La Niña (Lau 2005).

Attempts to quantify the observed relationship of the MJO with the evolution of ENSO, however, have yielded conflicting results. On the one hand, Zhang and Gottschalck (2002) found a robust leading relationship between MJO-forced Kelvin wave activity in the western equatorial Pacific and strength of El Niño 6 to 12 months later (see also Batstone and Hendon 2005). A similar leading relationship has also been observed for westerly wind bursts (e.g., Vecchi and Harrison 2000). However, those winds bursts associated with the MJO are of most importance for interaction with the ocean ENSO, as a large oceanic response is only driven by wind events that are spatially and temporally coherent (e.g., Kessler et al. 1995; Shinoda and Hendon 2001; Zhang and Gottschalck 2002).

On the other hand, little relationship between global MJO activity and ENSO has been observed (Hendon et al. 1999; Slingo et al. 1999). However, these studies focused on the contemporaneous relationship with ENSO and on behavior during northern winter, when the MJO is strongest, or used data from all months, which biases the results to winter when the MJO is strongest. Though weaker than in winter, the eastward-propagating MJO does still occur in spring and summer (Salby and Hendon 1994; Zhang and Dong 2004), which appear to be the critical seasons for interaction with a developing El Niño (Fedorov 2002). Thus, the previous studies that were biased to wintertime MJO activity may not have revealed an important seasonal dependence of the lagged association of the MJO with ENSO.

Although little relationship of the global level of MJO activity and ENSO has been revealed, an eastward shift of MJO activity into the central Pacific does accompany the eastward shift of the warm pool as El Niño events mature (e.g., Fink and Speth 1997; Hendon et al. 1999). Kessler (2001) argues that this shift increases the fetch of MJO winds across the equatorial Pacific, thus increasing the oceanic response to the MJO at these times. However, Kessler (2001) did not emphasize a pronounced leading relationship between western Pacific MJO activity and the development of El Niño. Thus, the eastward shift of MJO activity may be nothing more than a reactive modulation of MJO behavior once El Niño has commenced and the highest SST shifts toward the date line (e.g., Yu et al. 2003; Batstone and Hendon 2005; Eisenman et al. 2005). However, Kessler’s (2001) analysis was based on MJO variance computed in a running 1-yr window, which precludes detection of any leading relationship with El Niño that is tightly tied to the annual cycle. A seasonally varying relationship between MJO activity and the ENSO cycle might be anticipated both because MJO activity varies seasonally and development of El Niño is tightly coupled to the annual cycle (e.g., Rasmusson and Carpenter 1982; Tziperman et al. 1997).

The current study will explore the lagged relationship between MJO activity and ENSO, when seasonal variations are resolved. Seasonal stratification will overcome the problem of domination by the winter season when the MJO is strongest, as measured globally, but is apparently unrelated to the state of ENSO. It will also allow us to relate the interaction of the MJO and ENSO to the annual variations of the mean state upon which MJO activity and ENSO evolve.

Some of the seasonal characteristics of near-equatorial MJO activity in the Pacific, which is the region of most relevance to the interaction with ENSO, are reviewed in section 2. We will then demonstrate in section 3 a robust leading relationship between enhanced western Pacific MJO activity in spring–early summer and subsequent El Niño strength in autumn–winter. This finding bolsters the hypothesis for a pivotal role of the MJO in the development of El Niño. The analysis indicates a linear relationship with La Niña events as well, suggesting that reduced MJO activity in spring promotes stronger La Niña events in winter. A mechanism for the two-way interaction of the MJO with ENSO is hypothesized in section 4. Conclusions are provided in section 5.

2. Seasonal independence of the equatorial MJO activity

Previous studies have emphasized the pronounced seasonal variation of MJO activity, with strongest activity shifting south of the equator in late southern summer and a secondary maximum occurring north of the equator in northern summer. Maximum activity near the equator tends to occur near the equinoxes (e.g., Salby and Hendon 1994; Zhang and Dong 2004). Furthermore, contrasting propagation characteristics in the two respective summer seasons have been emphasized, with activity in southern summer exhibiting predominantly eastward propagation along and to the south of the equator and that in northern summer exhibiting both eastward and poleward propagation into the Asian summer monsoon (e.g., Wang and Rui 1990). Recently, however, Wheeler and Hendon (2004) emphasized the seasonal invariance of the near-equatorial structure of the MJO. In particular, they showed that near-equatorial behavior of the MJO displays the classical structure of the eastward-propagating, zonal-overturning circulation along the equator, as originally depicted by Madden and Julian (1972), during both northern and southern summers.

Space–time spectra of equatorially averaged (5°S–5°N) surface zonal wind (Usfc; Fig. 1) emphasize this seasonally invariant zonal and temporal structure of the near-equatorial component of the MJO. The daily surface zonal wind was obtained from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalyses (Kalnay et al. 1996). Spectra were computed in 180-day segments centered on 3-month seasons (which are 1 month later than traditional seasons in order to coincide with the peaks in activity that occur in January–March south of the equator and in July–September north of the equator; Salby and Hendon 1994). Prior to computing power, the seasonal cycle, time mean, and linear trend in each 180-day segment were removed, and then a cosine tapered window was applied over the first and last 45 days. Power was then computed each year during 1979–2005 (data end in April 2005) and then averaged.

Fig. 1.

Space–time spectra of surface zonal wind anomalies averaged over 5°S–5°N for the period 1979–2005. Spectra were computed in 180-day segments centered on (a) January–March, (b) April–June, (c) July–September, and (d) October–December each year and then averaged for all years. Prior to computing power, the time mean and linear trend in each segment were removed and the first and last 45 days tapered with a cosine-squared window. Frequency bandwidth is 1/180 cycles per day. The abscissa is linear in frequency but labeled with equivalent period in days. The contour interval is 7.5 × 10−3 m2 s−2 per unit frequency per wavenumber.

Fig. 1.

Space–time spectra of surface zonal wind anomalies averaged over 5°S–5°N for the period 1979–2005. Spectra were computed in 180-day segments centered on (a) January–March, (b) April–June, (c) July–September, and (d) October–December each year and then averaged for all years. Prior to computing power, the time mean and linear trend in each segment were removed and the first and last 45 days tapered with a cosine-squared window. Frequency bandwidth is 1/180 cycles per day. The abscissa is linear in frequency but labeled with equivalent period in days. The contour interval is 7.5 × 10−3 m2 s−2 per unit frequency per wavenumber.

In all seasons, a pronounced spectral peak at eastward wavenumber 1 with a period of 30–90 days is observed (Fig. 1). Power is relatively uniform from October through June, while only July–September exhibits a pronounced reduction. However, the eastward MJO, as evidenced by the spectral peak at eastward wavenumber 1 with periods of 30–90 days, predominates year-round. A similar picture emerges for near-equatorial outgoing longwave radiation (OLR; not shown), which is widely used to depict the convective component of the MJO, though the spectral peak is spread into higher wavenumbers (2–5) due to the localization of the convective component of the MJO to the warm pool of the Indian and western Pacific Oceans (e.g., Salby and Hendon 1994).

The space–time spectra in Fig. 1 emphasize the tropical-wide characteristics of the near-equatorial behavior of the MJO. Activity in the equatorial Pacific, however, is of more direct relevance to interaction with El Niño. Figure 2a shows the annual cycle of the amplitude of the MJO in Usfc for the Pacific basin. The MJO amplitude was computed by taking the square root of MJO-filtered Usfc variance at each longitude. The MJO filter retains eastward wavenumbers 1–5 with periods of 30–95 days, which captures the bulk of the MJO signal (e.g., Wheeler and Kiladis 1999) as well as interannual longitudinal variations of the MJO associated with ENSO (e.g., Hendon et al. 1999). After filtering, the data were averaged over 5°S–5°N. At each grid point in longitude the root-mean-square amplitude was computed in a 90-day moving window and then sampled monthly. Averaging all like months for the period 1979–2004 formed the annual cycle of MJO amplitude (Fig. 2a). Also shown in Fig. 2b is the annual cycle of total SST averaged over 5°S–5°N, based on analyses of Reynolds and Smith (1994) for the period November 1981 through April 2005. For clarity 1.5 annual cycles are shown.

Fig. 2.

Annual cycle of equatorially averaged (5°S–5°N) (a) MJO-filtered, surface zonal wind amplitude and (b) SST for the Pacific basin. One and one-half annual cycles are shown for clarity. Contour interval (CI) in (a) is 0.06 m s−1 and in (b) is 0.75°C.

Fig. 2.

Annual cycle of equatorially averaged (5°S–5°N) (a) MJO-filtered, surface zonal wind amplitude and (b) SST for the Pacific basin. One and one-half annual cycles are shown for clarity. Contour interval (CI) in (a) is 0.06 m s−1 and in (b) is 0.75°C.

MJO amplitude in the western equatorial Pacific is relatively constant between about November and May and reaches a distinct minimum in July–September (Fig. 2a). This minimum coincides both with the cold season in the eastern equatorial Pacific (Fig. 2b) and with the strong shift of MJO activity away from the equator into the Northern Hemisphere. Interestingly, the annual variation of MJO activity in the western Pacific bears no clear relationship with the annual variation of SST in the equatorial warm pool (i.e., the region in the western Pacific and Indian Oceans where SST is greater than 28.5°C), which is pronouncedly small (e.g., Picaut et al. 2001).

In summary, near-equatorial MJO activity in the western Pacific exhibits relatively uniform intensity from November through May and a distinct minimum in July–September. This extended duration of maximum near-equatorial activity in the western Pacific from November to May is longer than that implied by the annual variation of globally averaged MJO activity, which maximizes during boreal summer (e.g., Salby and Hendon 1994). From the global perspective, MJO activity peaks markedly in January–March, but activity is then shifted well into the Southern Hemisphere. The extended duration of activity in the west equatorial Pacific into early northern summer could be important for interaction with ENSO because this is the time of year when El Niño tends to develop.

3. Lagged relation of MJO activity with ENSO

El Niño tends to develop in northern spring and peak in the subsequent winter season (e.g., Rasmusson and Carpenter 1982). To explore the relationship between MJO activity and ENSO while taking account of this phase locking to the seasonal cycle, we compute the lag regression of MJO activity onto the 3-month mean Niño-3.4 SST index centered on January. We obtained the monthly Niño-3.4 SST time series from the U.S. Weather Service’s Climate Prediction Center Web site (http://www.cpc.ncep.noaa.gov/data/indices/).

MJO activity is diagnosed using MJO-filtered, equatorially averaged (5°S–5°N) Usfc and OLR (Liebmann and Smith 1996). As in section 2, the MJO filter retains eastward wavenumbers 1–5 for periods of 30–95 days. MJO amplitude is computed by taking the square root of the variance of MJO-filtered Usfc and OLR computed in a 3-month running window. Hence, for example the MJO amplitude in February 1979 is the square root of the mean variance of MJO-filtered Usfc and OLR computed over January–March 1979. We will refer to these amplitude time series, which we sample monthly, as MJOOLR and MJOUsfc

The lagged regression of MJOOLR and MJOUsfc as functions of longitude onto Niño-3.4(Jan) is shown in Fig. 3. Also shaded in Fig. 3 are regions where the correlation coefficient has magnitude greater than 0.4, which are judged to be significantly different than zero at the 95% level assuming Gaussian statistics with 25 independent samples. The zero lag regression [Jan(1)] depicts the eastward displacement of MJOOLR into the eastern Pacific at the peak of El Niño and the concurrent reduction of activity in the Indian Ocean (e.g., Fink and Speth 1997; Hendon et al. 1999; Kessler 2001). A similar picture emerges with MJOUsfc, though the eastward shift of enhanced activity into the central Pacific is less pronounced, while the suppression of activity in the Indian Ocean is more marked. The eastward displacement of MJOOLR into the central Pacific at the peak of El Niño can be tracked backward in time to the western Pacific, as discussed by Kessler (2001), Yu et al. (2003), and Lau (2005). This systematic eastward progression as El Niño matures is not as apparent in MJOUsfc. However, a more continuous eastward shift is evident if MJOUsfc is computed over a wider range of latitudes (e.g., 15°S–15°N; not shown). Nonetheless, both MJOOLR and MJOUsfc show a dramatic increased amplitude in the western Pacific some 8 months before El Niño peaks [i.e., April–May (0)]. In the case of MJOUsfc the enhanced amplitude at this time also extends into the eastern Pacific, while for MJOOLR it extends into the eastern Indian Ocean. This zonally extensive region of enhanced amplitude some 7 months before the peak of El Niño implies an increase of global-mean MJO activity at this time as well (cf. Table 1). This strong positive correlation of MJO activity in the far western Pacific (locally exceeding 0.7 in the vicinity of 120°–150°E) some 8 months before El Niño peaks was not uncovered by Kessler’s (2001) analysis. Zhang and Gottschalck (2002), on the other hand, found a similar 8-month lag between Niño-3.4 and MJO-forced Kelvin activity in the far western Pacific.

Fig. 3.

Monthly regression (contours) and correlation (shading) of equatorially averaged (a) MJOOLR and (b) MJOUsfc as function of longitude onto Niño-3.4(Jan). Lag is in months, with Jan(1) indicating zero lag (simultaneous) regression. Zero subscripts indicate Niño-3.4 lagging. Anomalies are scaled for a 2 standard anomaly of Niño-3.4(Jan). Contour interval is (a) 1 W m−2 and (b) 60 × 10−3 m s−1. The magnitude of the correlation coefficient is shaded (shading level 0.1) with first level at 0.4.

Fig. 3.

Monthly regression (contours) and correlation (shading) of equatorially averaged (a) MJOOLR and (b) MJOUsfc as function of longitude onto Niño-3.4(Jan). Lag is in months, with Jan(1) indicating zero lag (simultaneous) regression. Zero subscripts indicate Niño-3.4 lagging. Anomalies are scaled for a 2 standard anomaly of Niño-3.4(Jan). Contour interval is (a) 1 W m−2 and (b) 60 × 10−3 m s−1. The magnitude of the correlation coefficient is shaded (shading level 0.1) with first level at 0.4.

Table 1.

Correlation of MJO indices in May with Niño-3.4 SST in the following December. Correlations exceeding 0.35 are deemed to differ significantly from zero at the 95% level using a one-tailed t test assuming 25 degrees of freedom.

Correlation of MJO indices in May with Niño-3.4 SST in the following December. Correlations exceeding 0.35 are deemed to differ significantly from zero at the 95% level using a one-tailed t test assuming 25 degrees of freedom.
Correlation of MJO indices in May with Niño-3.4 SST in the following December. Correlations exceeding 0.35 are deemed to differ significantly from zero at the 95% level using a one-tailed t test assuming 25 degrees of freedom.

To further explore this sensitivity of El Niño to preceding MJO activity in the western Pacific, we form amplitude indices of western Pacific MJO activity by taking the square root of the mean variance of MJO-filtered Usfc and OLR over a 3-month running window and averaged (5°S–5°N, 120°E–180°). This is the region in Fig. 3 that shows the strongest leading correlation with respect to Niño-3.4[Jan(1)], however the results are not significantly affected using a larger Pacific domain. We refer to these monthly indexes as WPacMJOOLR and WPacMJOUsfc, respectively.

The lag correlation of Niño-3.4 (smoothed with a 3-month running mean to be consistent with the MJO indices) with respect to WPacMJOOLR and WPacMJOUsfc, as functions of start month, is shown in Figs. 4a and 5a, respectively. WPacMJOOLR shows significant simultaneous correlation with Niño-3.4 from about May through November (Fig. 4a), which reflects the eastward shift of activity in conjunction with El Niño (as also inferred from Fig. 3a). However, the most outstanding feature in Figs. 4a and 5a is the strong lagged correlation between WPacMJO activity in spring–early summer and Niño-3.4 up to 8–10 months later. For instance, Fig. 4a shows that WPacMJOOLR in June is correlated with Niño-3.4 in the following December at about 0.8. Similar behavior is displayed for MJOUsfc (Fig. 5a), although its maximum correlation is slightly reduced, occurs 1 month earlier in May, and persists over a longer lag (5–9 months). When all months are considered together, the maximum lagged correlation of Niño-3.4 with respect to WPacMJOOLR and WPacMJOUsfc is reduced to ∼0.4 (Niño-3.4 lagging by 3 months) and ∼0.3 (Niño-3.4 lagging by 8 months), respectively, which further emphasizes the strong seasonality of the MJO–ENSO relationship.

Fig. 4.

(a) Lagged correlation of Niño-3.4 with respect to WPacMJOOLR as function of start month. The abscissa indicates the start month for WPacMJOOLR. The lag of Niño-3.4 is indicated in months on the ordinate. The contour interval is 0.1 (negative dashed) with first contour at ±0.3. (b) Same as in (a), except for the partial correlation of Niño-3.4 at lag with respect to WPacMJOOLR, where the linear relationship with Niño-3.4 at zero lag is first removed.

Fig. 4.

(a) Lagged correlation of Niño-3.4 with respect to WPacMJOOLR as function of start month. The abscissa indicates the start month for WPacMJOOLR. The lag of Niño-3.4 is indicated in months on the ordinate. The contour interval is 0.1 (negative dashed) with first contour at ±0.3. (b) Same as in (a), except for the partial correlation of Niño-3.4 at lag with respect to WPacMJOOLR, where the linear relationship with Niño-3.4 at zero lag is first removed.

Fig. 5.

Same as in Fig. 4, except for (a) lagged correlation of Niño-3.4 with respect to WPacMJOUsfc and (b) partial correlation of Niño-3.4 with respect to WPacMJOUsfc where the linear relationship with Niño-3.4 at zero lag is first removed.

Fig. 5.

Same as in Fig. 4, except for (a) lagged correlation of Niño-3.4 with respect to WPacMJOUsfc and (b) partial correlation of Niño-3.4 with respect to WPacMJOUsfc where the linear relationship with Niño-3.4 at zero lag is first removed.

The persistence of the correlation of western Pacific MJO activity in spring with Niño-3.4 well into autumn–winter might partly stem from their joint correlation with Niño-3.4 in spring–summer. That is, WPacMJOOLR in June might be automatically correlated with Niño-3.4 in the following December because WPacMJOOLR in June is correlated simultaneously with Niño-3.4 in June, and Niño-3.4 in June is correlated with itself in December due to strong persistence at these times of year. To explore this possibility, we remove the linear relationship with Niño-3.4 at zero lag prior to computing the lagged correlations. The resulting partial correlations are shown in Figs. 4b and 5b. The robust lagged correlation of WPacMJO in spring/summer with Niño-3.4 into winter survives, suggesting that there is a direct relationship between the two.

Another area of concern is the possibility that the robust lag relationship comes about primarily from the strongest El Niño event in the record (1997), which was preceded by very strong MJO activity in spring of 1997 (e.g., Bergman et al. 2001). Figure 6 displays the scatterplot of WPacMJOOLR in May compared with Niño-3.4 in the following December. The 1997 El Niño event clearly contributes to the strong positive correlation, but removal of 1997 from the record only reduces the correlation from 0.79 to 0.75. A more significant reduction occurs when WPacMJOUsfc is considered (Table 1), with the May to December correlation dropping from 0.72 to 0.66 when 1997 is removed. Nonetheless, highly significant and physically relevant (i.e., explained variances exceeding 40%) correlations remain even in the absence of the unusual 1997 event. Further more, La Niña behavior appears to linearly contribute to the positive correlation as well, with reduced WPacMJO in spring/early summer preceding cold conditions 7 months later.

Fig. 6.

Scatterplot of WPacMJOOLR in May with Niño-3.4 in the following December. Units are std devs. The 1997 value is indicated by “97.” The correlation including (excluding) 1997 is 0.79 (0.75).

Fig. 6.

Scatterplot of WPacMJOOLR in May with Niño-3.4 in the following December. Units are std devs. The 1997 value is indicated by “97.” The correlation including (excluding) 1997 is 0.79 (0.75).

Figure 3 indicates that, although the peak correlation of Niño-3.4[Jan(1)] with MJO activity in the preceding spring occurs in the western Pacific, the positive correlation at this time extends into the eastern Pacific for MJOUsfc and into the Indian Ocean for MJOOLR. As mentioned above, this suggests a relationship between globally averaged MJO activity in spring and subsequent development of El Niño. This possibility is explored by forming some global measures of MJO activity. These include averaging MJOOLR and MJOUsfc in a global domain rather than restricting them to the western Pacific. We also employ the Realtime Multivariate MJO (RMM) Indices developed by Wheeler and Hendon (2004). They showed that by combining equatorially averaged upper- and lower-tropospheric zonal winds with OLR, the eastward-propagating MJO is readily returned as a leading pair of empirical orthogonal functions (EOFs). Importantly, Wheeler and Hendon showed that the associated daily principal component time series, RMM1 and RMM2, could be used to efficiently diagnose the MJO in all seasons. Here, we form a global measure of MJO amplitude by summing the squares of RMM1 and RMM2, averaging in a 90-day running window, and then taking the square root and sampling monthly. We refer to this amplitude time series as the RMM index, which can be obtained from the Australian Bureau of Meteorology Web site (http://www.bom.gov.au/bmrc/clfor/cfstaff/matw/maproom/RMM).

Table 1 lists correlations using these various global MJO indices in May with Niño-3.4 in the following December. While significant correlations between Niño-3.4(Dec) and global MJOOLR and MJOUsfc, and, to a lesser degree, RMM in the preceding May are found, they are in general weaker than the correlations based on indices of western Pacific MJO activity. This is despite the fact that the global measures of MJO activity correlate highly with the west Pacific indices of MJO activity (e.g., the simultaneous correlation between global MJOUsfc and WPacMJOUsfc using all months is ∼0.85). The only modest reduction in correlation using the global index of MJOOLR as compared to WPacMJOOLR stems from the fact that the MJO signal in OLR is primarily confined to the western Pacific and Indian Oceans. The more dramatic drop in correlation using the global index of MJOUsfc over that of the western Pacific index occurs because the zonal wind component of the MJO is more globally distributed (e.g., Salby and Hendon 1994), and thus variations outside of the western Pacific that contribute to variability of this global index are not necessarily of relevance to El Niño variability. The weaker correlation of Niño-3.4 with RMM, which uses zonal winds at 200 and 850 hPa and OLR, also stems from the fact that variations outside of the western Pacific contribute to variability of the RMM index. However, the reduction is also contributed to by the fact that RMM is computed by zonally averaging winds and OLR in a greater range of latitudes (15°S–15°N) than the range (5°S–5°N) used for MJOOLR and MJOUsfc, and it only includes 2 spatial degrees of freedom (i.e., interannual longitudinal displacements of MJO activity are not well captured by the RMM index). Nonetheless, a modest but significant relationship between global MJO activity in spring and subsequent El Niño variability in winter has been uncovered, but this analysis points to the specific details of the MJO variability in the western equatorial Pacific as being of most importance for the interaction with ENSO (e.g., Kessler 2001).

4. Hypothesis for mechanism of interaction

We have identified a strong relationship between ENSO in autumn–winter and preceding MJO activity in spring–early summer. Important issues are raised by this analysis: what is the mechanism by which the MJO interacts with ENSO, why does the interaction occur in spring, or is the relationship nothing more than just a deterministic response of the MJO to the SST anomalies that develop during El Niño?

Several studies have suggested that the MJO interacts with ENSO because enhanced MJO activity results in anomalous westerly surface winds in the western Pacific (e.g., Kessler and Kleeman 2000; Bergman et al. 2001; Lengaigne et al. 2003; Batstone and Hendon 2005; Zavala-Garay et al. 2005), which then project efficiently onto the developing ENSO mode. For example, detailed examination of the interaction of the MJO with the development of the 1997/98 El Niño (Bergman et al. 2001; Lengaigne et al. 2002) indicates that a series of MJOs in late winter/spring 1997 induced cooling in the far western Pacific together with an eastward displacement (i.e., local warming) of the eastern edge of the warm pool. This lead to sustained westerly surface wind anomalies in the western Pacific (Lengaigne et al. 2003) that then further promoted the eastward expansion of the warm pool at the onset of El Niño (e.g., Kessler et al. 1995; Picaut et al. 2001). The low-frequency westerly anomalies associated with MJO activity have also been interpreted as the direct low-frequency tail of the MJO itself due to its episodic and impulsive character (e.g., Zavala-Garay et al. 2005).

The association of enhanced MJO activity with an expanded western Pacific warm pool and anomalous surface zonal wind anomalies is demonstrated by considering the simultaneous relationship of WPacMJO variability with equatorially averaged SST (Fig. 7a) and Usfc (Fig. 7b) as a function of month. Again, 1.5 annual cycles are shown for clarity. From about October to June enhanced WPacMJO is associated with anomalously warm SST centered on about the date line (i.e., an eastward-expanded warm pool) together with anomalous surface westerlies across the western Pacific. This relationship is strongest in March–June. While the magnitude of the regression coefficient with SST is largest in the east Pacific in November, the correlation then is only marginally significant. There is also some indication of an association of enhanced western Pacific MJO activity with cooler SST in the far western Pacific (e.g., Hendon et al. 1999; Bergman et al. 2001; Lengaigne et al. 2003), but only in winter.

Fig. 7.

Annual cycle of simultaneous regression (contours) and correlation (shading) of equatorially averaged (a) SST and (b) Usfc onto WPacMJOUsfc as function of month. One and one-half annual cycles are shown for clarity. Anomalies are shown for a two standard anomalies of WPacMJOUsfc. The contour interval is (a) 0.25°C and (b) 0.5 m s−1. The magnitude of the correlation coefficient is shaded (shading level 0.1 with first level at 0.4).

Fig. 7.

Annual cycle of simultaneous regression (contours) and correlation (shading) of equatorially averaged (a) SST and (b) Usfc onto WPacMJOUsfc as function of month. One and one-half annual cycles are shown for clarity. Anomalies are shown for a two standard anomalies of WPacMJOUsfc. The contour interval is (a) 0.25°C and (b) 0.5 m s−1. The magnitude of the correlation coefficient is shaded (shading level 0.1 with first level at 0.4).

The sensitivity of MJO activity to SST on the eastern edge of the warm pool in spring (Fig. 7a) and, to a lesser degree, autumn reflects that then the MJO is most equatorially focused (e.g., Salby and Hendon 1994) and is most sensitive to zonal expansion and contraction of the warm pool (e.g., Lau 2005). This is also the time of year that the mean zonal SST gradient is flattest (i.e., when warmest conditions occur in the east Pacific; Fig. 2b) and anomalous convection and surface westerlies are most sensitive to SST anomalies on the edge of the warm pool (e.g., Spencer 2004). The general lack of sensitivity of western Pacific MJO activity to warm pool SST (and surface zonal wind) in both northern and southern summers presumably stems from the pronounced poleward shift of MJO activity into the monsoons at these times.

The mechanism for interaction of the MJO with ENSO, then, is hypothesized to occur via the westerly anomalies in the western Pacific that are associated with enhanced MJO activity in northern spring (Fig. 7b). Such westerly anomalies in the western Pacific are well known to precede the development of El Niño (e.g., Clarke and Shu 2000), as can be seen from the regression of equatorial SST and surface zonal wind onto the Niño-3.4 index in January (Fig. 8). El Niño typically begins with a rapid expansion of the warm pool into the eastern Pacific beginning in about April, referred to as April(0) (Fig. 8a), which coincides with the normal seasonal redevelopment of the cold tongue (Fig. 2b; e.g., Rasmusson and Carpenter 1982). This initial eastward expansion of the warm pool is accompanied by westerly anomalies in the western Pacific (Fig. 8b) that are evident as far back as the preceding winter (e.g., Clarke and Shu 2000). Thus, the westerly anomalies associated with enhanced MJO activity (Fig. 7b) would project strongly onto these westerlies that accompany development of El Niño in spring.

Fig. 8.

Monthly regression (contours) of equatorially averaged (a) SST and (b) Usfc onto Niño-3.4 [Jan(1)]. Contour interval in (a) is 0.3°C and in (b) is 0.5 m s−1 (zero contour not drawn). Anomalies are shown for a two std devs anomaly of Niño-3.4 (Jan). The magnitude of the correlation coefficient is shaded (shading level 0.1) with first level at 0.4.

Fig. 8.

Monthly regression (contours) of equatorially averaged (a) SST and (b) Usfc onto Niño-3.4 [Jan(1)]. Contour interval in (a) is 0.3°C and in (b) is 0.5 m s−1 (zero contour not drawn). Anomalies are shown for a two std devs anomaly of Niño-3.4 (Jan). The magnitude of the correlation coefficient is shaded (shading level 0.1) with first level at 0.4.

Furthermore, a positive feedback between the MJO and El Niño development is implied in spring. Figure 9 shows the regression (and correlation) of equatorial SST onto WPacMJOUsfc in May. The zero lag–correlation is now indicated at May(0) so that the time evolution can be compared directly with Fig. 8a. Careful inspection of Fig. 9 reveals that WPacMJO in May(0) is most strongly correlated with SST in the west Pacific 1–2 months earlier. That is, anomalously warm SST on the eastern edge of the warm pool in March and April, which typically precedes development of El Niño (Fig. 8a), promotes enhanced WPacMJO in May, which then promotes enhanced surface westerlies in the western Pacific, which are highly conducive to El Niño conditions 6–8 months later.

Fig. 9.

Monthly regression (contours) and correlation (shading) of equatorially averaged SST onto WPacMJOUsfc in May(0). Anomalies are shown for a two std devs anomaly of WPacMJOUsfc. The contour interval is 0.3°C. The magnitude of the correlation coefficient is shaded (shading level 0.1 with first level at 0.4).

Fig. 9.

Monthly regression (contours) and correlation (shading) of equatorially averaged SST onto WPacMJOUsfc in May(0). Anomalies are shown for a two std devs anomaly of WPacMJOUsfc. The contour interval is 0.3°C. The magnitude of the correlation coefficient is shaded (shading level 0.1 with first level at 0.4).

It is also interesting to note that WPacMJO in May explains more variance of SST in the equatorial eastern Pacific in the following January (i.e., more than 50% of the variance in the vicinity of 150°W) than does Niño-3.4 from the preceding May. Indeed, Niño-3.4 exhibits a pronounced springtime persistence barrier, with the lag–correlation of Niño-3.4 from May to the following January dropping to below 0.4 (e.g., McPhaden 2003). In this regard, springtime MJO activity in the western Pacific may be used as a predictor to overcome it.

5. Conclusions

Based on analyses for the period 1979–2005, western Pacific MJO activity in spring is associated with subsequent El Niño strength in autumn and winter. A modest leading relationship for global-mean MJO activity is also observed, but the present analysis emphasizes that it is MJO activity in the western Pacific that more directly relates to the subsequent evolution of El Niño. An interesting question is what the relationship between global and western Pacific MJO activity is, especially in spring, and we will pursue this in a subsequent study.

We have argued that the causative relationship between enhanced spring/early summer MJO activity and El Niño in autumn/winter stems from the association of enhanced MJO activity with anomalous surface westerlies in the western Pacific, both of which promote and are promoted by warm SSTs on the eastern edge of the Pacific warm pool. Spring is when the mean equatorial zonal SST gradient in the Pacific is weakest and MJO activity is most centered on the equator. The opposite relationship with the development of La Niña tends to hold as well (i.e., decreased western Pacific MJO activity in spring is associated with easterly anomalies in the western Pacific and a contracted warm pool), indicating some linearity of this relationship that has received only limited emphasis (e.g., Lau 2005). Presumably, normal MJO activity contributes a westerly component to the mean state in the western Pacific. Years with less-than-normal MJO activity exhibit an easterly anomaly while years with above-normal MJO activity exhibit a westerly anomaly.

The reliance of MJO activity in the western Pacific on the occurrence of mean (or low frequency) surface westerly winds has previously been emphasized (e.g., Inness and Slingo 2003; Zhang and Dong 2004). That is, MJO activity in the western Pacific is facilitated or promoted by the occurrence of mean surface westerlies, which reflect a convecting basic state driven by the warm SST in the western Pacific. Here, we hypothesize a positive feedback, whereby MJO activity in the western Pacific promotes sustained westerly surface winds, which themselves promote and are promoted by warm SST and atmospheric convection on the eastern edge of the warm pool. This feedback is most pronounced in spring, when convection and westerlies are most sensitive to SST anomalies on the edge of the warm pool (e.g., Spencer 2004).

We have argued that the MJO interacts with El Niño because the low-frequency westerly anomalies that are associated with enhanced MJO activity project efficiently onto the El Niño mode in spring (e.g., Kessler and Kleeman 2000; Zavala-Garay et al. 2005). This interpretation is different, though not entirely inconsistent, from Zhang and Gottschalck (2002), who suggest that the connection between the two is via excitation of intraseasonal oceanic Kelvin waves that then propagate into the eastern Pacific where they interact with El Niño. Their analysis, however, shows that the strongest lagged relationship, at least with El Niño–related SST variations in the central Pacific (i.e., as indicated by correlation with the Niño-3.4 or Niño-3 SST index), is with MJO-forced Kelvin wave activity at the western boundary of the Pacific basin. A consistent interpretation, then, is that the apparent relationship between MJO-forced, intraseasonal Kelvin waves and development of El Niño stems from MJO activity and associated low-frequency effects directly in the western Pacific, which then affect El Niño development in the central Pacific. On the other hand, intraseasonal Kelvin waves that are excited by the MJO in the western Pacific may play a more primary role in the evolution of El Niño–related SST anomalies in the far eastern Pacific (e.g., as monitored by the Niño-1 and -2 SST indexes).

Finally, we cannot dismiss the possibility that the lagged relationship developed here is nothing more than a diagnosis of the deterministic evolution of the atmospheric noise through the El Niño cycle (e.g., Yu et al. 2003; Batstone and Hendon 2005; Vecchi et al. 2006). That is, SST anomalies associated with the initial development of El Niño may provide favorable conditions on the eastern edge of the Pacific warm pool for enhancement of MJO activity in late spring–summer, but this enhancement of MJO activity is not the cause of the preceding SST anomalies and thus does not necessarily feed back onto El Niño. The resolution of these issues requires a better understanding of the role of the MJO in the maintenance and variability of the mean state of the tropical Pacific atmosphere and of the mechanisms of interaction of the MJO with the ocean. This understanding is also required so that the influences of the MJO on El Niño can be properly simulated and assessed in coupled forecast and climate models.

Acknowledgments

Javier Zavala-Garay, Klaus Weickmann, and two anonymous reviewers provided insightful comments on earlier versions of the manuscript. CZ thanks BOM for supporting his visit to BMRC, during which this research was undertaken, and acknowledges support by NOAA’s Office of Global Programs through awards under Cooperative Agreement NA67RJO149 to CIMAS.

REFERENCES

REFERENCES
Batstone
,
C.
, and
H. H.
Hendon
,
2005
:
Stochastic variability associated with ENSO and the role of the MJO.
J. Climate
,
18
,
1773
1789
.
Bergman
,
J. W.
,
H. H.
Hendon
, and
K. M.
Weickmann
,
2001
:
Intraseasonal air–sea interactions at the onset of El Niño.
J. Climate
,
14
,
1702
1719
.
Clarke
,
A. J.
, and
L.
Shu
,
2000
:
Quasi-biennial winds in the far western equatorial Pacific phase-locking El Niño to the seasonal cycle.
Geophys. Res. Lett.
,
27
,
771
774
.
Eisenman
,
I.
,
L.
Yu
, and
E.
Tziperman
,
2005
:
Westerly wind bursts: ENSO’s tail rather than the dog?
J. Climate
,
18
,
5224
5238
.
Fedorov
,
A. V.
,
2002
:
The response of the coupled tropical ocean-atmosphere to westerly wind bursts.
Quart. J. Roy. Meteor. Soc.
,
128
,
1
23
.
Fink
,
A.
, and
P.
Speth
,
1997
:
Some potential forcing mechanisms of the year-to-year variability of tropical convection and its intraseasonal (25–70 day) variability.
Int. J. Climatol.
,
17
,
1513
1534
.
Gutzler
,
D. S.
,
1991
:
Interannual fluctuations of intraseasonal variance of near-equatorial zonal winds.
J. Geophys. Res.
,
96
,
3173
3185
.
Hendon
,
H. H.
,
B.
Liebmann
, and
J. D.
Glick
,
1998
:
Oceanic Kelvin waves and the Madden–Julian Oscillation.
J. Atmos. Sci.
,
55
,
88
101
.
Hendon
,
H. H.
,
C.
Zhang
, and
J.
Glick
,
1999
:
Interannual variation of the Madden–Julian oscillation during austral summer.
J. Climate
,
12
,
2538
2550
.
Inness
,
P.
, and
J. M.
Slingo
,
2003
:
Simulation of the Madden–Julian oscillation in a coupled general circulation model. Part II: The role of the basic state.
J. Climate
,
16
,
365
382
.
Kalnay
,
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
Kessler
,
W. S.
,
2001
:
EOF representations of the Madden–Julian oscillation and its connection with ENSO.
J. Climate
,
14
,
3055
3061
.
Kessler
,
W. S.
, and
R.
Kleeman
,
2000
:
Rectification of the Madden–Julian oscillation into the ENSO cycle.
J. Climate
,
13
,
3560
3575
.
Kessler
,
W. S.
,
M. J.
McPhaden
, and
K. M.
Weickmann
,
1995
:
Forcing of intraseasonal Kelvin waves in the equatorial Pacific.
J. Geophys. Res.
,
100
,
10613
10631
.
Lau
,
W. K. M.
,
2005
:
El Niño Southern Oscillation connection.
Intraseasonal Variability in the Atmosphere-Ocean Climate System, W. K. M. Lau and D. E. Waliser, Eds., Praxis Publishing, 271–300
.
Lengaigne
,
M.
,
J-P.
Boulanger
,
C.
Menkes
,
S.
Masson
,
G.
Madec
, and
P.
Delecluse
,
2002
:
Ocean response to the March 1997 Westerly Wind Event.
J. Geophys. Res.
,
107
.
8015, doi:10.1029/2001JC000841
.
Lengaigne
,
M.
,
J-P.
Boulanger
,
C.
Menkes
,
G.
Madec
,
P.
Delecluse
,
E.
Guilyardi
, and
J.
Slingo
,
2003
:
The March 1997 westerly wind event and the onset of the 1997/98 El Niño: Understanding the role of the atmospheric response.
J. Climate
,
16
,
3330
3343
.
Liebmann
,
B.
, and
C. A.
Smith
,
1996
:
Description of a complete (interpolated) outgoing longwave radiation dataset.
Bull. Amer. Meteor. Soc.
,
77
,
1275
1277
.
Luther
,
D. S.
,
D. E.
Harrison
, and
R. A.
Knox
,
1983
:
Zonal winds in the central equatorial Pacific and El Niño.
Science
,
222
,
327
330
.
Madden
,
R.
, and
P.
Julian
,
1972
:
Description of global-scale circulation cells in the tropics with a 40–50 day period.
J. Atmos. Sci.
,
29
,
1109
1123
.
McPhaden
,
M. J.
,
1999
:
Genesis and evolution of the 1997–98 El Niño.
Science
,
283
,
950
954
.
McPhaden
,
M. J.
,
2003
:
Tropical Pacific Ocean heat content variations and ENSO persistence barriers.
Geophys. Res. Lett.
,
30
.
1480, doi:10.1029/2003GL016872
.
Moore
,
A. M.
, and
R.
Kleeman
,
1999
:
Stochastic forcing of ENSO by the intraseasonal oscillation.
J. Climate
,
12
,
1199
1220
.
Picaut
,
J.
,
M.
Ioualalen
,
T.
Delcroix
,
F.
Masia
,
R.
Murtugudde
, and
J.
Vialard
,
2001
:
The oceanic zone of convergence on the eastern edge of the Pacific warm pool: A synthesis of results and implications for El Niño-Southern Oscillation and biogeochemical phenomena.
J. Geophys. Res.
,
106
,
2363
2386
.
Rasmusson
,
E.
, and
T.
Carpenter
,
1982
:
Variations in the tropical sea surface temperature and surface wind fields associated with the Southern Oscillation/El Niño.
Mon. Wea. Rev.
,
110
,
354
384
.
Reynolds
,
R. W.
, and
T. M.
Smith
,
1994
:
Improved global sea surface temperature analyses using optimum interpolation.
J. Climate
,
7
,
929
948
.
Salby
,
M. L.
, and
H. H.
Hendon
,
1994
:
Intraseasonal behavior of clouds, temperature, and motion in the Tropics.
J. Atmos. Sci.
,
51
,
2207
2224
.
Shinoda
,
T.
, and
H. H.
Hendon
,
2001
:
Upper-ocean heat budget in response to the Madden–Julian oscillation in the western equatorial Pacific.
J. Climate
,
14
,
4147
4165
.
Shinoda
,
T.
, and
H. H.
Hendon
,
2002
:
Rectified wind forcing and latent heat flux produced by the Madden–Julian oscillation.
J. Climate
,
15
,
3500
3508
.
Slingo
,
J. M.
,
D. P.
Rowel
, and
K. R.
Sperber
,
1999
:
On the predictability of the interannual behavior of the Madden-Julian Oscillation and its relationship with El Niño.
Quart. J. Roy. Meteor. Soc.
,
125
,
583
609
.
Spencer
,
H.
,
2004
:
Role of the atmosphere in seasonal phase locking of El Niño.
Geophys. Res. Lett.
,
31
.
L24104, doi:10.1029/2004GL021619
.
Tziperman
,
E.
,
S. E.
Zebiak
, and
M. A.
Cane
,
1997
:
Mechanisms of seasonal − ENSO interaction.
J. Atmos. Sci.
,
54
,
61
71
.
Vecchi
,
G. A.
, and
D. E.
Harrison
,
2000
:
Tropical Pacific sea surface temperature anomalies, El Niño, and equatorial westerly wind events.
J. Climate
,
13
,
1814
1830
.
Vecchi
,
G. A.
,
A. T.
Wittenberg
, and
A.
Rosati
,
2006
:
Reasssesing the role of stochastic forcing in the 1997–1998 El Niño.
Geophys. Res. Lett.
,
33
.
L0176, doi:10.1029/2005GL024738
.
Wang
,
B.
, and
H.
Rui
,
1990
:
Synoptic climatology of transient tropical intraseasonal convection anomalies: 1975–1985.
Meteor. Atmos. Phys.
,
44
,
43
61
.
Weickmann
,
K. M.
,
1991
:
El Niño Southern Oscillation and the Madden-Julian (30–60 day) oscillation during 1981–82.
J. Geophys. Res.
,
96
,
3187
3196
.
Wheeler
,
M. C.
, and
G. N.
Kiladis
,
1999
:
Convectively coupled equatorial waves: Analysis of clouds and temperature in the wavenumber–frequency domain.
J. Atmos. Sci.
,
56
,
374
399
.
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
.
Yu
,
L.
,
R. A.
Weller
, and
W. T.
Liu
,
2003
:
Case analysis of a role of ENSO in regulating the generation of westerly wind bursts in the Western Equatorial Pacific.
J. Geophys. Res.
,
108
.
3128, doi:10.1029/2002JC001498
.
Zavala-Garay
,
J.
,
C.
Zhang
,
A. M.
Moore
, and
R.
Kleeman
,
2005
:
The linear response of ENSO to the Madden–Julian oscillation.
J. Climate
,
18
,
2441
2459
.
Zhang
,
C.
, and
J.
Gottschalck
,
2002
:
SST Anomalies of ENSO and the Madden–Julian oscillation in the equatorial Pacific.
J. Climate
,
15
,
2429
2445
.
Zhang
,
C.
, and
M.
Dong
,
2004
:
Seasonality in the Madden–Julian oscillation.
J. Climate
,
17
,
3169
3180
.

Footnotes

Corresponding author address: Harry H. Hendon, BMRC, GPO Box 1289, Melbourne 3001, Australia. Email: hhh@bom.gov.au