Abstract

Rectification of (Madden–Julian oscillation) MJO-induced wind speed and latent heat flux variations across the tropical Indian and western Pacific Oceans is estimated using 51 yr of NCEP–NCAR reanalysis. The rectified wind speed anomaly is calculated from the difference in wind speed based on 30- and 90-day low-pass-filtered winds. During periods when the MJO is active, the wind speed is typically enhanced by about 1 m s−1 south of the equator in the western Pacific. The largest rectified latent heat flux occurred during the large MJO event of March 1997 in the western Pacific warm pool. The magnitude of the rectification is found to depend strongly on the mean wind speed, and this affects the temporal and spatial variations of the rectification.

1. Introduction

Prior to the onset of El Niño in mid-1997, two episodes of large-scale convection associated with the Madden–Julian oscillation (MJO; Madden and Julian 1972) passed eastward across the tropical Indian and western Pacific Oceans (e.g., McPhaden 1999). The far western equatorial Pacific was observed to cool and oceanic Kelvin waves were generated, which may have initiated the eastward displacement of the warm pool in the western Pacific (Bergman et al. 2001). The unusually rapid development of this El Niño, which was not well forecast, has been attributed to these MJOs (e.g., Kessler and Kleeman 2000, hereafter KK00; McPhaden and Yu 1999).

The MJO may possibly affect lower-frequency phenomena such as ENSO by producing a mean rectified surface forcing of the warm pool. KK00 explored this possibility using idealized intraseasonally varying winds as forcing in a ocean general circulation model (OGCM). In their study, active episodes of the MJO acted to increase the wind speed across the warm pool, because both positive and negative intraseasonally varying zonal wind anomalies increase the wind speed in this region of near-zero mean wind speed. For reasonable magnitudes of the imposed intraseasonally varying winds, the far western Pacific cooled by about 0.4°C over a period of 4 months and the central Pacific warmed by about 0.1°C. The cooling resulted primarily from increased evaporation. The warming in the central Pacific resulted from zonal heat advection associated with a mean westerly equatorial current that was spun up. KK00 speculate that this cooling in the western Pacific and warming in the central Pacific may act to reverse the zonal SST gradient on the equator, which then can induce an additional low-frequency westerly wind anomaly as a result of the low-level zonal pressure gradient.

Although KK00 demonstrated the potential importance of rectification of intraseasonally varying forcing associated with the MJO, the idealized winds and, thus, the resulting latent heat flux used in their study are different than observed for typical MJO events. For instance, the largest latent heat flux associated with the MJO typically occurs when the wind is anomalously westerly (Hendon and Glick 1997). In KK00, their idealized latent heat flux is most enhanced during the easterly phase (Fig. 5 in KK00).

The purpose of this paper is to quantify the observed rectified wind forcing and latent heat flux caused by the MJO and to describe their spatial and temporal variations using 51 yr of the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis winds (Kalnay et al. 1996). This paper is organized as follows: section 2 describes the method to calculate the rectified component of wind speed from the data. In section 3, spatial and temporal variations of rectified wind speed are described. In section 4, the dependence of the rectified wind speed on the mean wind is discussed. Finally, conclusions and discussion are provided in section 5.

2. Calculation of rectified wind forcing and latent heat flux

Surface winds from the NCEP–NCAR reanalysis (Kalnay et al. 1996) are used to calculate wind forcing and latent heat flux. Daily mean winds for the period from 1948 to 1998 on a Gaussian grid of T62 resolution are analyzed. The quality of the NCEP–NCAR reanalysis surface wind data will be discussed in section 5. In order to isolate the rectified component of wind speed due only to the intraseasonal wind variation, a method is developed that removes effects due to interannual variations unrelated to changes in MJO activity. This method is demonstrated by considering the daily area average zonal wind in the western Pacific warm pool during 1996/97 winter (Fig. 1a). Two episodes of large-scale convection associated with the MJO passed over the western Pacific warm pool in December 1996 and March 1997 (e.g., Bergman et al. 2001). The westerly wind episodes associated with this active convection are evident in the daily wind data.

Fig. 1.

(a) Time series of daily mean (open circle) and 90-day low-pass-filtered (thick line) zonal winds averaged over 15°S–5°N, 130°E–180°. (b) Time series of 30- (open circle) and 90-day low-pass-filtered (thick line) zonal winds averaged over 15°S–5°N, 130°E–180°. (c) Time series of wind speed calculated from daily mean winds (open circle) and 90-day low-pass-filtered winds (thick line). (d) Time series of wind speed calculated from 30- (open circle) and 90-day low-pass-filtered winds (thick line)

Fig. 1.

(a) Time series of daily mean (open circle) and 90-day low-pass-filtered (thick line) zonal winds averaged over 15°S–5°N, 130°E–180°. (b) Time series of 30- (open circle) and 90-day low-pass-filtered (thick line) zonal winds averaged over 15°S–5°N, 130°E–180°. (c) Time series of wind speed calculated from daily mean winds (open circle) and 90-day low-pass-filtered winds (thick line). (d) Time series of wind speed calculated from 30- (open circle) and 90-day low-pass-filtered winds (thick line)

In order to isolate the contribution to wind speed by intraseasonally varying winds, the time series of zonal and meridional winds are first low-pass filtered (via Fourier transform) for periods longer than 30 (U30, V30) and 90 days (U90, V90). The time series of U30 and U90 during the 1996/97 winter are shown in Fig. 1b. The U30 includes the intraseasonal westerly winds associated with the MJO, while these westerly wind episodes are removed in U90. The U90 and V90 represent the variation of low-frequency winds that is longer than intraseasonal timescale. This low-frequency component is referred to as the “mean” winds hereafter. In this definition of the mean, it is assumed that the intraseasonal zonal wind anomaly induced by the MJO has no influence on the mean zonal wind. Wind speed is then calculated from these filtered winds:

 
formula

Figure 1d shows the time series of WS30 and WS90. Strong westerly winds associated with the MJO are well captured by 30-day low-pass-filtered time series. The mean zonal wind during the December 1996 event is nearly zero and is weak westerly during the March 1997 event (Fig. 1b). During the easterly phase of MJO (February 1997), WS30 is close to the mean wind speed (WS90) because the weak mean westerly wind and the intraseasonal easterly anomaly offset each other. During the westerly phase (March 1997), WS30 is much larger than WS90 because the mean wind (U90) and the intraseasonal anomaly (U30U90) are both westerly and thus the wind speed is enhanced. Hence, the wind speed over one cycle of the MJO is increased over the mean for this period. Also shown is the wind speed calculated from unfiltered daily zonal and meridional winds (Fig. 1c). During the MJO periods, it is slightly higher than the WS30 since shorter period wind fluctuations (less than 30 days) also contribute to the enhancement of the wind speed. During non-MJO periods (September–November 1996), the wind speed is also significantly enhanced due to the shorter timescale wind fluctuations that are not associated with the MJO.

The difference of two curves (WS30 − WS90; Fig. 1d) is the increase in wind speed that is produced by the intraseasonal variation of winds (Fig. 2a), and it is referred to as “increased wind speed” in the following. The increased wind speed is regarded as the rectified component of wind speed caused by the intraseasonal wind variation.

Fig. 2.

(a) Time series of the latent heat flux calculated from the increased wind speed. (b) Time series of the increased wind speed (open circle) and PC12 + PC22 (closed circle) from the EOF analysis of zonal winds. A 30-day running mean is applied to PC12 + PC22

Fig. 2.

(a) Time series of the latent heat flux calculated from the increased wind speed. (b) Time series of the increased wind speed (open circle) and PC12 + PC22 (closed circle) from the EOF analysis of zonal winds. A 30-day running mean is applied to PC12 + PC22

The mechanism of the wind speed enhancement during the MJO diagnosed above is basically the same as what KK00 incorporated in their experiments. KK00 compared the wind speed calculated from the smoothed annual cycle of zonal winds with the wind speed calculated including idealized intraseasonal zonal winds. Because of the nonlinearity, the wind speed is increased due to the MJO-induced zonal wind anomaly. The 90-day low-pass-filtered wind in this study corresponds to the annual cycle wind of KK00's study and the 30-day low-pass-filtered wind corresponds to their idealized intraseasonal wind. Here we have used 90-day low-pass-filtered winds instead of the annual cycle, in order to remove the interannual variation.

The enhanced latent heat flux δQe produced by the enhanced wind speed is estimated using a standard bulk formula:

 
δQe = ρaL(WS30 − WS90)Ceδq,

where ρa is the air density, Ce is the exchange coefficient, L is the latent heat of vaporization, and δq is the difference between air specific humidity and the saturation specific humidity at the sea surface. We have used Ce = 1.3 × 10−3, ρa = 1.2 kg m−3, L = 2.441 × 106 J kg−1, and δq = 6 g kg−1. Here, it is assumed that the rectified latent heat flux is primarily caused by enhanced wind speed. On intraseasonal timescales, variations of latent heat flux across the warm pool are primarily driven by variations of wind speed rather than by changes in the humidity difference δq (e.g., Hendon and Liebmann 1990; Shinoda et al. 1998). The increased latent heat flux is calculated in the same fashion as the increased wind speed and is shown in Fig. 2a for the period September 1996–May 1997.

Periods of active MJO are identified using EOFs of zonal winds in the domain (15°N–15°S, 60°E–90°W). The EOFs are calculated from zonal winds that are intraseasonally filtered (via spectral transform) to periods of 30–90 days. The first two EOFs capture 17% and 11% of the filtered variance, respectively. The principal components of these two EOFs correlate at 0.63 at a lag of 12 days. Thus, taken together these two EOFs describe a zonally propagating disturbance with a near-50-day period, and are taken here to depict the MJO (cf. Zhang and Hendon 1997).

An index of the level of MJO activity is defined by PC12 + PC22, where PC1 and PC2 are the principal components of first and second mode of EOFs, respectively. An example of the index during the 1996/97 winter is shown in Fig. 2b. Two maxima of the index correspond to the period when large-scale convection associated with the MJO passed over the western Pacific.

Periods of active MJO are defined when the index exceeds three standard deviations. A 50-day running mean is applied to the index in order to smooth over individual events. In this study, the threshold value is chosen to select only large events such as the 1996/97 winter events that could be potentially important for the onset of El Niño. Thus, 6.4% of the entire record is identified as having active MJOs (the index is a positive quantity). However, after 1974, 10.3% of the record is identified as being active. This difference in level of activity, possibility related to the advent of satellite observations in the mid-1970s, is discussed further in section 5. Periods of nonactive MJO are also defined when the index is less than 0.32 standard deviations, which yields the same percentage of the record (i.e., 6.4%) inactive as active.

3. Temporal and spatial variations of rectification

The relation between the level of MJO activity and the area-averaged increased wind speed in the western Pacific warm pool is displayed in Fig. 3. A 50-day running mean is applied to the time series of the increased wind speed in order to emphasize increases of wind speed averaged over the life cycle of the MJO. Increased wind speed is correlated with the MJO index at 0.45, which is significant at the 99% level.

Fig. 3.

Relation between PC12 + PC22 and the increased wind speed averaged over 15°S–5°N, 130°E–180°. The sampling interval is 5 days

Fig. 3.

Relation between PC12 + PC22 and the increased wind speed averaged over 15°S–5°N, 130°E–180°. The sampling interval is 5 days

Time series of the area average increased wind speed in the tropical Indian, western Pacific, and eastern Pacific Oceans are shown in Fig. 4. Red points indicate active MJO periods based on the MJO index. Large values of the increased wind speed often occur during MJO events in the western Pacific and Indian Oceans (Figs. 4a,b). The largest increase of wind speed occurred during the large MJO event in March 1997 in the western Pacific (Fig. 4a). The area average increased wind speed often exceeds 1 m s−1 during large MJO events. This increased wind speed corresponds to an increased latent heat flux of about 23 W m−2. In the central and eastern Pacific, on the other hand, rectification (the increased wind speed) is small (Fig. 4c). Also, it does not correlate well with the level of MJO activity.

Fig. 4.

Time series of the increased wind speed averaged over (a) 15°S–5°N, 130°E–180°; (b) 15°S–5°N, 50°–100°E; and (c) 15°S–5°N, 160°–110°W. Red points indicate PC12 + PC22 > 164.3 (std dev × 3), and blue points indicate PC12 + PC22 > 18.0. The 50-day running mean is applied to the time series. The time interval of each point is 5 days

Fig. 4.

Time series of the increased wind speed averaged over (a) 15°S–5°N, 130°E–180°; (b) 15°S–5°N, 50°–100°E; and (c) 15°S–5°N, 160°–110°W. Red points indicate PC12 + PC22 > 164.3 (std dev × 3), and blue points indicate PC12 + PC22 > 18.0. The 50-day running mean is applied to the time series. The time interval of each point is 5 days

Blue points in Fig. 4 indicate periods when the MJO is not active. In the western Pacific and Indian Oceans, the increased wind speed is low during most of the nonactive MJO periods. Increased wind speed is sometimes large in the western Pacific and Indian Oceans even if the MJO index is not large. The spatial scale of the intraseasonal wind variation during these periods may not be large enough to be identified as the MJO periods by the index. For instance, the large increased wind speed in the western Pacific is evident during March 1991 when the MJO index is not large. During this period, there is almost no intraseasonal wind signal in the Indian Ocean (not shown).

Figure 5a shows the average increased wind speed for the entire 51 yr. The maximum rectified wind speed occurs around 10°–12°S between 160°E and the date line in the tropical Pacific. Similar values of increased wind speed are seen in the eastern tropical Indian Ocean. The average increased wind speed exceeds 0.3 m s−1 in these regions, which corresponds to about 7 W m−2 evaporative cooling. Since the annual mean surface heat flux is small in the western Pacific warm pool (e.g., ≈20 W m−2 estimated by Esbensen and Kushnir 1981), the above estimate of rectified cooling suggests that the intraseasonal wind variations contribute significantly to the mean surface heat flux in the warm pool. During large MJO events, the increased wind speed is much higher than the average over most of the area of the western Pacific and Indian Oceans (Fig. 5b). The magnitude of the increased wind speed exceeds 1 m s−1 in the western Pacific south of the equator and the eastern Indian Ocean near Sumatra.

Fig. 5.

(a) Time mean increased wind speed (m s−1). (b) Time mean increased wind speed (m s−1) during the period PC12 + PC22 > 164.3

Fig. 5.

(a) Time mean increased wind speed (m s−1). (b) Time mean increased wind speed (m s−1) during the period PC12 + PC22 > 164.3

In order to examine if winds during the MJO are actually stronger than climatological winds, the average total wind speed anomaly at each location during the active MJO periods is calculated. Daily wind speed is calculated from the daily winds and anomalies are formed by removing the annual cycle of daily wind speed. Then 50-day running mean is applied to the wind speed anomalies. Figure 6 shows the average wind speed anomaly during the active MJO period determined by the MJO index. In most of the areas of the western Pacific and Indian Ocean warm pool, winds are stronger than the annual cycle during the active MJO period. The large wind speed anomaly is found south of the equator in the western Pacific and eastern Indian Oceans, which is consistent with the spatial variation of the increased wind speed in Fig. 5b. Wind speed anomalies are smaller than the increased wind speed since the annual cycle of wind speed includes a rectified component of wind speed caused by the MJO. There are also other differences between the two figures. For instance, there is a zonal band of maximum north of the equator in the Pacific in Fig. 6 that is absent in Fig. 5b. These differences could be due to the fact that interannual variations of winds that are unrelated to the MJO activity, such as the low-frequency variation associated with ENSO, are also included in the wind speed anomaly.

Fig. 6.

Average wind speed anomaly during the period when the MJO is active. The active MJO period is identified by the MJO index PC12 + PC22 > 164.3 (std dev × 3) from the EOF analysis of zonal winds. Anomalies are calculated with respect to the first three harmonics from 51 yr (1948–98) of data

Fig. 6.

Average wind speed anomaly during the period when the MJO is active. The active MJO period is identified by the MJO index PC12 + PC22 > 164.3 (std dev × 3) from the EOF analysis of zonal winds. Anomalies are calculated with respect to the first three harmonics from 51 yr (1948–98) of data

Large intraseasonal wind variations tend to cause large rectification of wind speed. In order to examine whether the geographical distribution of rectified wind speed in Fig. 5b is primarily determined by the strength of MJO activities at each location, the average intraseasonal deviation of wind speed is calculated and it is compared with the spatial variation of the increased wind speed. The intraseasonal deviation is computed from 30- and 90-day low-pass-filtered time series of zonal and meridional winds. For each 50-day segment of time series, average wind speed is calculated from daily values of (U30, V30) and (U90, V90), respectively. The difference of the average wind speed calculated from (U30, V30) and (U90, V90) is defined as the intraseasonal deviation during the 50-day period.

The geographical distribution of the average intraseasonal deviation of wind speed (Fig. 7) is similar to that of the rectified wind (Figs. 5a,b). Maximum deviation is located south of the equator in the western Pacific and eastern Indian Ocean. However, there are significant differences. For instance, the increased wind speed around 10°–15°N, near the date line is very small, while the intraseasonal deviation is not small in this region. The difference is caused by spatial variations in the mean wind speed and is discussed in the next section.

Fig. 7.

Intraseasonal deviation of wind speed (m s−1)

Fig. 7.

Intraseasonal deviation of wind speed (m s−1)

4. Dependence on mean winds

The mechanism of rectification proposed by KK00 depends on the mean winds being weak. The observed mean winds in the warm pool undergo a seasonal cycle as well as being affected by significant interannual variations. To understand this impact on rectification of the intraseasonally varying winds, we first examine the dependence on the mean winds for the idealized case similar to the KK00.

Suppose the total zonal wind (Utot) is described by a combination of a sinusoidal intraseasonal fluctuation and a lower-frequency zonal wind (MU):

 
Utot = A sin(ωt) + MU,

where A (>0) is the amplitude of intraseasonal variation, t is time, and ω is the frequency. Here it is assumed that the meridional wind component is zero, and MU and A are constant for the simplicity. The rectified wind speed is defined by

 
|
Utot
| − |MU|,

where |Utot| is the time mean of |Utot| over one cycle of the intraseasonally varying winds. If |MU| > A, |Utot| is same as the background mean wind speed |MU|, and the rectified wind speed defined above is equal to zero. Hence, intraseasonally varying winds do not cause any rectification if mean winds are strong enough.

For this simple case, the rectified wind speed has maximum value (2A/π) when MU = 0. If 0 < |MU| < A, the rectified wind speed is between 0 and the maximum value (2A/π). For example, if A = 3 m s−1 and MU = 1 m s−1, which are typical in the western Pacific, the rectified wind speed is 1.02 m s−1. This is similar to that estimated in the warm pool (Fig. 4a). However, in the warm pool the background mean wind (MU) and the amplitude of intraseasonal fluctuations (A) vary seasonally and interannually. Also A could be a function of MU. The spatial variation of MU and A is also large. Thus, further details of the effect of the background winds cannot be quantified from this idealized case.

Figure 8a shows the relation between the mean wind speed (WS90) and the increased wind speed due to rectification (WS30 − WS90). Large rectification occurs when the mean wind is weak. However, the MJO is more active during periods of weak mean winds, and increased MJO activity enhances the rectification. For instance, rectification greater than 1 m s−1 mostly occurs when the mean wind speed is less than 3.5 m s−1 and the MJO is active (red points in the figure). Hence, the large rectification under low mean winds is partly caused by larger MJO activities. Further analyses is necessary to isolate the effect of the mean wind on the rectification.

Fig. 8.

(a) Relation between area average (15°S–5°N, 130°E–180°) increased wind speed and wind speed calculated from 90-day low-pass-filtered winds. Red points indicate PC12 + PC22 > 164.3, and blue points indicate PC12 + PC22 < 18.0. (b) Same as (a) except the intraseasonal variation is normalized

Fig. 8.

(a) Relation between area average (15°S–5°N, 130°E–180°) increased wind speed and wind speed calculated from 90-day low-pass-filtered winds. Red points indicate PC12 + PC22 > 164.3, and blue points indicate PC12 + PC22 < 18.0. (b) Same as (a) except the intraseasonal variation is normalized

In order to remove the effect of the magnitude of the intraseasonal wind variation on the rectification, the difference between 30- and 90-day low-pass-filtered daily zonal wind (U30U90) is first normalized by the standard deviation of (U30U90) during each 50-day period. The meridional wind fluctuation is also normalized in the same way as the zonal wind. These normalized deviations are referred to as UN30 − UN90 and VN30 − VN90. Then the normalized increased wind speed (WSN) is calculated from the normalized difference of zonal and meridional winds as follows: WSN = (UN30 − UN90)2 + (VN30 − VN90)2.

A 50-day running mean is applied to the time series of WSN. In this manner, the effect of the magnitude of the intraseasonal wind variation on the rectification is removed. Figure 8b shows the increased wind speed calculated from the normalized time series of winds. A linear relationship between the increased wind speed and the background mean wind is evident. There is almost no rectification when the mean wind speed is more than 5–6 m s−1.

Figure 9a shows the map of the increased wind speed calculated from normalized time series. The maximum is located near the equator in the western Pacific. This geographical distribution is similar to the mean wind (WS90; Fig. 9b). The increased wind speed (normalized) near the equator around 150°E is about 0.35 m s−1, while it is 0.05–0.1 m s−1 at 10°–15°N around the date line (Fig. 9a). This difference is significant compared to the magnitude of rectification (calculated without normalization; see Fig. 5a), and it is caused entirely by the difference of mean winds. The mean wind difference in the two locations is about 4.5 m s−1. The rectification around 10°–15°N in the western and central Pacific is largely reduced due to the strong mean wind of about 7 m s−1.

Fig. 9.

(a) Increased wind speed calculated from normalized anomalies (see text for details). (b) Wind speed calculated from 90-day low-pass-filtered winds

Fig. 9.

(a) Increased wind speed calculated from normalized anomalies (see text for details). (b) Wind speed calculated from 90-day low-pass-filtered winds

Since the MJO-induced zonal wind anomaly is generally stronger than the meridional wind anomaly, the zonal component of mean wind might affect the rectification more than the meridional component. In order to understand the effect of each component of the mean wind on the rectification, the relation between the mean zonal wind and the normalized increased wind speed is also examined (not shown). The correlation coefficient between the mean zonal wind and the increased wind speed (normalized) is 0.27, which is lower than the correlation between the total mean wind speed and the increased wind speed (−0.43). This suggests that the meridional component of the MJO-induced wind anomaly is often significant, and that the meridional component of mean winds also affects the rectification.

5. Conclusions and discussion

Enhanced wind speed and increased latent heat flux during the life cycle of the MJO are examined using 51 yr of wind data. The rectified wind speed is estimated using the difference in wind speed based on 30- and 90-day low-pass-filtered winds.

Averaged over the life cycle of strong MJO events, the mean wind speed is enhanced by about 1 m s−1 across much of the western Pacific from the equator to 15°S. This increased wind speed corresponds to an increase in latent heat flux of about 23 W m−2. This amount of evaporative cooling could have a strong impact on air–sea coupled processes in the western Pacific warm pool since the annual mean net surface heat flux in this region could be as small as 20 W m−2 (Esbensen and Kushnir 1981).

Large increases in rectified wind speed are often associated with large MJO events in the Indian Ocean and western Pacific. For instance, the large MJO event during March 1997 produced the largest increase of evaporative cooling in the western Pacific in the 51-yr record used here. The magnitude of the rectification also strongly depends on the mean wind speed. When the mean wind speed is more than 5–6 m s−1, almost no rectification occurs even if individual intraseasonal wind variations are large. Hence, the mean wind variation significantly affects the temporal and spatial variations of the rectification.

In idealized model experiments, KK00 found an SST cooling of about 0.4°C in the western Pacific, which resulted primarily from rectified evaporative cooling. They hypothesized that this cooling will act to flatten the background zonal SST gradient, which then may induce additional mean westerly wind anomalies. Such lower-frequency (i.e., frequencies lower than the MJO itself) anomalies may affect the onset and evolution of El Niño.

To explore the possibility that MJO-induced rectification in the western Pacific systematically precedes the development of El Niño, we correlated the time series of increased wind speed in the western Pacific (smoothed with a 50-day running mean to remove intraseasonal fluctuations) with a similarly smoothed time series of Niño-3 SST for the period 1950–98. No significant correlations were found at any lags from −12 to +12 months.

We further examined whether cooling in the western Pacific is observed to follow occurrences of large rectification. Weekly SST analyses from November 1981 to 1998 (Reynolds and Smith 1994) are first interpolated to daily values, and then SST anomalies from the annual cycle are calculated. A 50-day running mean is applied to time series of SST tendency anomalies and the increased wind speed to remove the intraseasonal fluctuations. The correlation coefficient of SST tendency anomalies and the increased wind speed in the western Pacific is calculated. Although the average value of correlation coefficient for the entire period is essentially zero (≈−0.07 for the area average 15°S–5°N, 130°–180°E), the correlation varies substantially for each year. For instance, a large correlation (−0.58) between SST tendency and increased wind speed in the western Pacific warm pool (15°S–5°N, 130°–180°E) is found during September 1994–May 1995.

Since the mixed layer is shallow in the warm pool (e.g., Lukas and Lindstrom 1991), the SST in this region is sensitive to a variety of processes such as shortwave radiation and entrainment. The relative importance of each process may be different for each MJO event (e.g., Bergman et al. 2001). When the wind speed is enhanced during large MJOs, the entrainment rate is also changed. Since the mixed layer depth is a function of u3 (where u is a friction velocity) in simple bulk mixed layer physics (e.g., Kraus and Turner 1967), the net entrainment heat flux during one cycle of the MJO could be very sensitive to the rectified wind speed. However, the entrainment heat flux strongly depends on the temperature gradient below the mixed layer. The thermohaline structure below the mixed layer is controlled by various processes in the warm pool (e.g., Lukas and Lindstrom 1991; Shinoda and Lukas 1995; Vialard and Delecluse 1998) and thus the temperature gradient below the mixed layer can vary widely from event to event. Entrainment can even cause upper-ocean warming that may compensate the enhanced evaporative cooling (e.g., Shinoda and Hendon 1998, 2001). Further studies that estimate the complete upper-ocean heat budget during the MJO are necessary to fully understand the net effect of the MJO on the lower-frequency SST variation in the warm pool.

The NCEP–NCAR reanalysis surface winds are used in this study since they are the longest daily records available. Deficiencies in their quality are acknowledged, and these deficiencies could affect the subtle changes in mean winds that are diagnosed here. Shinoda et al. (1999) show that the wind stress and latent heat flux from the NCEP reanalysis agree reasonably well with direct measurements and other independent estimates. However, they examined data only for the period of 1986–93. The quality of the NCEP reanalysis winds for the entire 51-yr period cannot be easily evaluated. Reliability of the data in earlier periods may be different since satellite observations and Tropical Ocean Global Atmosphere (TOGA) Tropical Atmosphere Ocean (TAO) data have been included in recent years. In order to confirm the results of the analysis based on the 51 yr of records, the rectified wind speed is also calculated from the data for the period 1979–98 (not shown). The average increased wind speed during the MJO periods from the 1979–98 data agrees well with the one from the 1948–98 data, and thus major conclusions in this paper do not depend on the period of the data. Figure 5 shows that active MJO periods are less frequent during 1948–74 compared to recent periods. It is uncertain whether decreased MJO activity in earlier periods is due to true interdecadal variation or differences of data quality. Further comparisons of different surface wind products is necessary to address this issue.

Acknowledgments

We thank Klaus Weickmann for his valuable suggestions. We are also grateful for the constructive comments of the reviewers. Support for this work was provided by CLIVAR-Pacific Grants from NOAA's office of Global Programs.

REFERENCES

REFERENCES
Bergman
,
J. W.
,
H. H.
Hendon
, and
K. M.
Weickmann
,
2001
:
Intraseasonal variation of west Pacific convection at the onset of the 1997/98 El Niño.
J. Climate
,
14
,
1702
1719
.
Esbensen
,
S. K.
, and
Y.
Kushnir
,
1981
:
The heat budget of the global ocean: An atlas based on estimates from surface marine observations.
Climate Research Institute Rep. 29, Oregon State University, Covalis, OR, 27 pp
.
Hendon
,
H. H.
, and
B.
Leibmann
,
1990
:
The intraseasonal (30–50 day) oscillation of the Australian summer monsoon.
J. Atmos. Sci.
,
47
,
2909
2923
.
Hendon
,
H. H.
, and
J.
Glick
,
1997
:
Intraseasonal air–sea interaction in the tropical Indian and Pacific Oceans.
J. Climate
,
10
,
647
661
.
Kalnay
,
E.
,
Coauthors
.
1996
:
The NCEP/NCAR 40-Year Reanalysis Project.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
Kessler
,
W. S.
, and
R.
Kleeman
,
2000
:
Rectification of the Madden–Julian oscillation into the ENSO cycle.
J. Climate
,
13
,
3560
3575
.
Kraus
,
E. B.
, and
J. S.
Turner
,
1967
:
A one-dimensional model of the seasonal thermocline. II: The general theory and its consequences.
Tellus
,
19
,
98
106
.
Lukas
,
R.
, and
E.
Lindstrom
,
1991
:
The mixed layer of the western equatorial Pacific Ocean.
J. Geophys. Res.
,
96
,
(Suppl.),
.
3343
3357
.
Madden
,
R. A.
, and
P. R.
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.
, and
X.
Yu
,
1999
:
Equatorial waves and the 1997–98 El Niño.
Geophys. Res. Lett.
,
26
,
2961
2964
.
Reynolds
,
R. W.
, and
T. M.
Smith
,
1994
:
Improved global sea surface temperature analyses using optimum interpolation.
J. Climate
,
7
,
929
948
.
Shinoda
,
T.
, and
R.
Lukas
,
1995
:
Lagrangian mixed layer modeling of the western equatorial Pacific.
J. Geophys. Res.
,
100
,
2523
2541
.
Shinoda
,
T.
, and
H. H.
Hendon
,
1998
:
Mixed layer modeling of intraseasonal variability in the tropical western Pacific and Indian Oceans.
J. Climate
,
11
,
2668
2685
.
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.
,
H. H.
Hendon
, and
J.
Glick
,
1998
:
Intraseasonal variability of surface fluxes and sea surface temperature in the tropical western Pacific and Indian Ocean.
J. Climate
,
11
,
1685
1702
.
Shinoda
,
T.
,
H. H.
Hendon
, and
J.
Glick
,
1999
:
Intraseasonal surface fluxes in the tropical western Pacific and Indian Ocean from NCEP reanalyses.
Mon. Wea. Rev.
,
127
,
678
693
.
Vialard
,
J.
, and
P.
Delecluse
,
1998
:
An OGCM study for the TOGA decade. Part II: Barrier layer formation and variability.
J. Phys. Oceanogr.
,
28
,
1089
1106
.
Zhang
,
C.
, and
H. H.
Hendon
,
1997
:
Propagating and standing components of the intraseasonal oscillation in tropical convection.
J. Atmos. Sci.
,
54
,
741
752
.

Footnotes

Corresponding author address: Dr. Toshiaki Shinoda, Climate Diagnostic Center, 325 Broadway, Boulder, CO 80303. Email: ts@cdc.noaa.gov