Indian Ocean Dipolelike Variability in the CSIRO Mark 3 Coupled Climate Model

Wenju Cai CSIRO Atmospheric Research, Aspendale, Victoria, Australia

Search for other papers by Wenju Cai in
Current site
Google Scholar
PubMed
Close
,
Harry H. Hendon BMRC, Melbourne, Australia

Search for other papers by Harry H. Hendon in
Current site
Google Scholar
PubMed
Close
, and
Gary Meyers CSIRO Marine Research, Hobart, Tasmania, Australia

Search for other papers by Gary Meyers in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Coupled ocean–atmosphere variability in the tropical Indian Ocean is explored with a multicentury integration of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Mark 3 climate model, which runs without flux adjustment. Despite the presence of some common deficiencies in this type of coupled model, zonal dipolelike variability is produced. During July through November, the dominant mode of variability of sea surface temperature resembles the observed zonal dipole and has out-of-phase rainfall variations across the Indian Ocean basin, which are as large as those associated with the model El Niño–Southern Oscillation (ENSO). In the positive dipole phase, cold SST anomaly and suppressed rainfall south of the equator on the Sumatra–Java coast drives an anticyclonic circulation anomaly that is consistent with the steady response (Gill model) to a heat sink displaced south of the equator. The northwest–southeast tilting Sumatra–Java coast results in cold sea surface temperature (SST) centered south of the equator, which forces anticylonic winds that are southeasterly along the coast, which thus produces local upwelling, cool SSTs, and promotes more anticylonic winds; on the equator, the easterlies raise the thermocline to the east via upwelling Kelvin waves and deepen the off-equatorial thermocline to the west via off-equatorial downwelling Rossby waves. The model dipole mode exhibits little contemporaneous relationship with the model ENSO; however, this does not imply that it is independent of ENSO. The model dipole often (but not always) develops in the year following El Niño. It is triggered by an unrealistic transmission of the model’s ENSO discharge phase through the Indonesian passages. In the model, the ENSO discharge Rossby waves arrive at the Sumatra–Java coast some 6 to 9 months after an El Niño peaks, causing the majority of model dipole events to peak in the year after an ENSO warm event. In the observed ENSO discharge, Rossby waves arrive at the Australian northwest coast. Thus the model Indian Ocean dipolelike variability is triggered by an unrealistic mechanism. The result highlights the importance of properly representing the transmission of Pacific Rossby waves and Indonesian throughflow in the complex topography of the Indonesian region in coupled climate models.

Corresponding author address: Dr. Wenju Cai, CSIRO Atmospheric Research, PMB1, Aspendale, Victoria 3195, Australia. Email: Wenju.Cai@csiro.au

Abstract

Coupled ocean–atmosphere variability in the tropical Indian Ocean is explored with a multicentury integration of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Mark 3 climate model, which runs without flux adjustment. Despite the presence of some common deficiencies in this type of coupled model, zonal dipolelike variability is produced. During July through November, the dominant mode of variability of sea surface temperature resembles the observed zonal dipole and has out-of-phase rainfall variations across the Indian Ocean basin, which are as large as those associated with the model El Niño–Southern Oscillation (ENSO). In the positive dipole phase, cold SST anomaly and suppressed rainfall south of the equator on the Sumatra–Java coast drives an anticyclonic circulation anomaly that is consistent with the steady response (Gill model) to a heat sink displaced south of the equator. The northwest–southeast tilting Sumatra–Java coast results in cold sea surface temperature (SST) centered south of the equator, which forces anticylonic winds that are southeasterly along the coast, which thus produces local upwelling, cool SSTs, and promotes more anticylonic winds; on the equator, the easterlies raise the thermocline to the east via upwelling Kelvin waves and deepen the off-equatorial thermocline to the west via off-equatorial downwelling Rossby waves. The model dipole mode exhibits little contemporaneous relationship with the model ENSO; however, this does not imply that it is independent of ENSO. The model dipole often (but not always) develops in the year following El Niño. It is triggered by an unrealistic transmission of the model’s ENSO discharge phase through the Indonesian passages. In the model, the ENSO discharge Rossby waves arrive at the Sumatra–Java coast some 6 to 9 months after an El Niño peaks, causing the majority of model dipole events to peak in the year after an ENSO warm event. In the observed ENSO discharge, Rossby waves arrive at the Australian northwest coast. Thus the model Indian Ocean dipolelike variability is triggered by an unrealistic mechanism. The result highlights the importance of properly representing the transmission of Pacific Rossby waves and Indonesian throughflow in the complex topography of the Indonesian region in coupled climate models.

Corresponding author address: Dr. Wenju Cai, CSIRO Atmospheric Research, PMB1, Aspendale, Victoria 3195, Australia. Email: Wenju.Cai@csiro.au

1. Introduction

The Indian Ocean dipole (IOD) refers to the episodic occurrence of an anomalous zonal gradient in sea surface temperature (SST) across the equatorial Indian Ocean and related changes in the topography of the thermocline (e.g., Saji et al. 1999; Webster et al. 1999; Yu and Rienecker 1999; Murtugudde and Busalacchi 1999; Murtugudde et al. 2000; Udea and Matsumoto 2000; Xie et al. 2002; Rao et al. 2002; Feng et al. 2001; Feng and Meyers 2003). An IOD episode usually begins with anomalous cooling in the tropical eastern Indian Ocean during May–June, when enhanced surface easterlies generate an anomalously shallow thermocline, enhanced latent heat flux, and upwelling off the Sumatra–Java coast. The cold anomaly usually peaks in September–October, by which time the western Indian Ocean has warmed as a result of increased insolation, reduced evaporation, and deepened thermocline. The cold anomaly in the east typically disappears rapidly with the monsoon transition by December–January, often yielding a basinscale warm anomaly by the following boreal spring. This seasonality in the evolution of the IOD is often referred to as the seasonal phase-locking feature.

Development of the IOD (with cold anomalies off the Sumatra–Java coast) typically, but not always, accompanies development of El Niño conditions in the Pacific, when atmospheric convection shifts eastward along the equator toward the date line. This induces suppressed rainfall and easterly anomalies in the equatorial eastern Indian Ocean (e.g., Lau and Nath 2004; Shinoda et al. 2004a). Because the climatological mean winds are easterly south of the equator during June–August, the induced easterlies act to increase the wind speed and the latent and sensible heat fluxes. Shoaling of the thermocline and upwelling of colder subsurface water are also promoted off the Sumatra–Java coast. Increased upwelling and increased latent and sensible heat flux overcome the increased shortwave radiation associated with reduced rainfall, and the southeastern Indian Ocean initially cools. In the western portion of the basin, a warm anomaly develops as the equatorial thermocline deepens in response to easterly anomalies. An off-equatorial depression of the thermocline forced by Ekman pumping develops and translates westward as an oceanic forced Rossby wave. These processes enhance the warming in the west (Chambers et al. 1999; Webster et al. 1999; Xie et al. 2002; Feng and Meyers 2003). In the east–west direction, as an anomalous zonal temperature gradient sets up, the easterly anomalies strengthen. The thermocline responds with further shoaling in the east and deepening in the west. The process resembles the Bjerknes (1969) feedback for the equatorial Pacific Ocean, in the sense that it grows as a consequence of a feedback linking anomalies of SST, rainfall, wind, and depth of the thermocline, although it develops in the southern tropical Indian Ocean and involves extraequatorial processes. The demise of the IOD often occurs soon after the Australian summer monsoon commences in December, when the mean winds become westerly in the equatorial eastern Indian Ocean. The induced easterly anomalies then act to reduce the wind speed. Reduced latent heat flux along with increased surface shortwave radiation act to warm the eastern Indian Ocean, yielding a basinscale warm anomaly. The basinscale anomaly has been traditionally described as the response of the Indian Ocean to ENSO (e.g., Klein et al. 1999).

The cold anomaly off the Sumatra–Java coast and the anomalous easterlies in the equatorial southeastern tropical Indian Ocean do not always develop in unison with an El Niño event (e.g., Reverdin et al. 1986). Consequently, there are IOD events that develop without clear ENSO forcing. This suggests that the ocean–atmosphere coupled process in the Indian Ocean sector has the capability to generate its own variability in response to a variety of triggering mechanisms, challenging the view that the Indian Ocean is merely a slave of the Pacific Ocean (e.g., Latif and Barnett 1995). There is also evidence (Saji et al. 1999; Feng and Meyers 2003) that a delayed feedback process operates in the tropical Indian Ocean during some periods similar to that suggested for the Pacific Ocean (Battisti and Hirst 1989) and that the delay is responsible for the biennial signal of the IOD. Similarly, not all biennial events have a corresponding signal in the Pacific (Meehl and Arblaster 2002), again providing support for the capability of the Indian Ocean to develop modes of coupled variability of its own. However, over the last several decades many IOD events have occurred in conjunction with development of El Niño. Consequently, El Niño is recognized as the dominant, but not the only, triggering mechanism.

The occurrence and dynamics of the IOD are of interest because significant rainfall anomalies both around the Indian Ocean basin and globally are associated with development of the dipole (e.g., Saji et al. 1999; Birkett, et al. 1999; Hendon 2003; Saji and Yamagata 2003; Yamagata et al. 2004). Because the IOD often coevolves with ENSO, it is difficult to attribute the cause of the rainfall anomalies to SSTs in one particular region. Thus, in order to understand the role of IOD events for global climate variability, we must first develop model systems that are capable of simulating IOD events with realistic structure and evolution. Much progress has been made in simulating the relevant ocean dynamics and thermodynamics (Murtugudde and Busalacchi 1999; Murtugudde et al. 2000; Behera et al. 2000) and the ENSO-induced atmospheric anomalies across the Indian Ocean (Lau and Nath 2003; Shinoda et al. 2004a) using stand-alone ocean or atmosphere models. Recently, fully coupled models have been successful in reproducing IOD-like events (e.g., Iizuka et al. 2000; Gualdi et al. 2003; Lau and Nath 2004), which allow complete two-way interaction of the atmosphere and ocean in all basins. In the present study, we explore tropical Indian Ocean dipole variability and its relationship to ENSO in the latest version (Mark 3) of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) coupled ocean–atmosphere climate model. We identify both realistic and unrealistic features of the Mark 3 dipolelike variability and identify the topography of the Indonesian region as a critical feature that should be considered carefully in model development and intercomparison studies.

The remainder of this paper is arranged as follows. In section 2, we briefly describe the CSIRO Mark 3 coupled model. In section 3, we apply empirical orthogonal function (EOF) analysis to seasonal SST anomalies and show that the dipole mode is dominant, similar to observations, in the June–July–August (JJA) and (September–October–November) SON seasons. Counter to the observed dipole, we show that it peaks two–three seasons after the peak of El Niño in the preceding DJF (December–January–February) season. We then examine in section 4 the evolution and the dynamics of the model IOD and assess its interaction with model ENSO. Section 5 provides discussion, and conclusions are presented in section 6.

2. Model and model run

The CSIRO coupled model used in this study is denoted as the Mark 3 version. It is based on the Mark 2 coupled model as described in Gordon and O’Farrell (1997). The Mark 3 model has been significantly improved relative to the Mark 2 model (Gordon et al. 2002); some brief details are provided below. The Mark 3 model is run without the use of flux adjustments and has only a moderate amount of (initial) climate drift. The horizontal resolution of the atmospheric model is spectral T63 (approximately 1.875° latitude × 1.875° longitude) with 18 vertical levels (hybrid sigma-pressure vertical coordinate). The atmospheric model has been upgraded to include a comprehensive cloud microphysical parameterization (Rotstayn et al. 2000), and the convection parameterization is based on that used in the Hadley Centre model (Gregory and Rowntree 1990). This convection parameterization is linked to the cloud microphysics scheme via the detrainment of liquid and frozen water at the cloud top. Atmospheric moisture advection (vapor, liquid, and frozen) is carried out by the semi-Lagrangian method (McGregor 1993). A simple treatment of the direct radiative effect of sulfate, which entails a perturbation of the surface albedo (Mitchell et al. 1995), is included. The land surface scheme (six layers of moisture and temperature) with a vegetation canopy (Kowalczyk et al. 1991, 1994) includes a three-layer snow model. Multiple soil types and vegetation types are included. A dynamic–thermodynamic polar ice model is incorporated, which includes a variable fraction of leads (O’Farrell 1998).

The Mark 3 ocean model is based upon the Modular Ocean Model (MOM2.2) version of the Geophysical Fluid Dynamics Laboratoy (GFDL) model. The oceanic component has horizontal resolution matching that of the atmospheric model’s grid in the east–west direction and twice that in the north–south direction. Thus the grid spacing is 0.9375° latitude × 1.875° longitude (approximately; latitude is on a Gaussian grid). Because there are two ocean grid points per atmospheric grid point in the meridional direction, the atmosphere model and ocean model subcomponents have identical land–sea masks. There are 31 levels in the vertical, with the spacing of the levels gradually increasing with depth, from 10 m at the surface to 400 m in the deep ocean. A parameterization of mixing of tracers based on the formation of Griffies et al. (1998) and Griffies (1998) is included.

We analyze the output of a 260-yr integration. The first 20 yr is discarded as a slight drift was detected. Throughout this study, monthly (seasonal) anomaly fields are formed by subtracting the monthly (seasonal) mean climatology.

The model produces strong ENSO variability (e.g., Cai et al. 2003). The standard deviation of Niño-3.4 (SST anomaly averaged over 5°S–5°N and 120°–170°W) is greater than that of the observed over the last 130 yr (0.94 versus 0.78). The model also produces a stronger biennial spectral peak than observed (Cai et al. 2003). Nonetheless, the model still suffers from a common “cold tongue bias” in the Pacific, with the equatorial cold tongue extending too far west. As a result, the eastern Indian Ocean is also too cold, possibly affecting the variability and interbasin interactions.

On the other hand, the basic features of climatological SST and surface winds in the Indian Ocean are reasonable, given that the model is run without flux adjustment. Southeasterly trade winds blow throughout the year in the southern tropical Indian Ocean, while the northern Indian Ocean is dominated by monsoonal winds. It is emphasized that aspects of the climatology that are relevant to the dipole dynamics are reasonably simulated: southeasterlies along the Java–Sumatra coast are strongest in JJA and SON, and the flow reverses in DJF associated with onset of the Australian monsoon. The model produces Wyrkti jets in boreal spring and fall with a stronger jet in boreal spring. The model thermocline in the equatorial Indian Ocean slopes slightly upward toward the west, as in the observations. The annual mean depth of 20°C isotherm (D20) exhibits a pronounced zonal ridge along 10°S, although it is generally too deep, extends too far to the east, and is shallowest around 80°E instead of the observed 60°E (Feng and Meyers 2003). This ridge geostrophically balances the South Equatorial Current, which flows westward, and the South Equatorial Countercurrent, which flows eastward, parallel to the ridge.

3. Modes of surface and subsurface variability

a. EOF patterns of SST anomalies

The dominant modes of interannual variability in the model’s Indian Ocean are identified using EOF analysis on seasonally stratified anomalous SST and D20 data. As in the observations, the leading mode of SST variability varies strongly with seasons; hence EOF analysis on seasonally stratified data helps prevent modal mixing. The leading EOF of SST in the Indian Ocean (hereafter IO1; Fig. 1) accounts for 31.4%, 39.4%, 51.3%, and 41.5% of variance in DJF, March–April–May (MAM), JJA, and SON, respectively. Figure 1 shows that there is a dipole structure in JJA and SON that resembles the observed in that both the modeled and the observed have strong loadings along the Java–Sumatra coast and oppositely signed loadings in the off-equatorial western Indian Ocean. As in observations, the cold anomaly develops first off the southern coast of Java in the far eastern Indian Ocean and strengthens as it moves equatorward (Susanto et al. 2001) in the following seasons.

From JJA to SON, the cold anomaly grows and expands westward. We have constructed time series of east−west SST gradient from averages over an eastern (0°–10°S, 100°–110°E) and western (0°–10°S, 50°–70°E) box. The standard deviations of the SST gradient for the SON, JJA, MAM, and DJF seasons are 1.13°, 0.78°, 0.37°, and 0.45°C, respectively. The most pronounced zonal gradient occurs in SON as in the observations (e.g., Baquero-Bernal et al. 2002), when the cold SST anomalies off the Java–Sumatra coast peak. During DJF and MAM, IO1 is more zonally uniform, with largest loadings south of the equator. As in observations, there is little evidence of a dipole structure in these seasons.

b. Relationship with model ENSO

There has been a vigorous debate as to the relationship between IOD and ENSO (Saji et al. 1999; Allan et al. 2001; Udea and Matsumoto 2000; Annamalai et al. 2003; Loschnigg et al. 2003; Clark et al. 2003; Lau and Nath 2003; Krishnamurthy and Kirtman 2003). We explore this by computing correlations between IO1 time series and the model Niño-3.4 index. For DJF and MAM, the correlation coefficient is strong (r = 0.60 for DJF and r = 0.54 for MAM). The correlation is largely due to an atmospheric teleconnection from the Pacific to Indian Ocean SST. But for JJA and SON, when the dipole is growing and most prominent, the correlation is small (r = 0.20 for both JJA and SON). This would appear to suggest that the model is generating a dipole mode that has an internal dynamics not forced by ENSO. However, ENSO often does initiate the dipole mode in this model but with a significant delay, which is explained later.

Close inspection of the time series of IO1 and Niño-3.4 for SON (Fig. 2a) reveals that most dipole events peak in the year following an El Niño. Over the 240 yr of model integration (discarding the first 20 yr of the 260-yr integration), there are 74 El Niño events, but there are only 34 IOD events. The majority of model IOD events, 23 out of 34, peak 1 yr after an El Niño event peaks. Occasionally, a dipole event coincides with ENSO (e.g., years 35, 62, and 66), but throughout the whole 240-yr integration, only 11 (out of 34) such events occur. To further illustrate this lag, the time series of IO1 for SON is shifted backward 1 yr (Fig. 2b). The correlation is large when Niño-3.4 leads by 1 yr (r = 0.75 for JJA (not shown) and r = 0.61 for SON). The majority of dipole events occur after an El Niño event, but not every El Niño event is followed by a dipole event, especially not those El Niño events with shorter (i.e., biennial) time scales.

Two additional features are worth noting. First, occasionally (years 109 and 159), a relatively strong IOD event takes place even though the El Niño event is quite weak. This suggests that sometimes an initial gentle kick is enough to generate the model IOD. Second, most of the model IODs are characterized by a substantial cooling that emerges in two consecutive SON seasons. Although in observations there are such events, for example, in 1982 and 1983, they do not occur as frequently as they do in Mark 3, and the mechanism is different. As will be clear, in our model the cooling in the first SON is triggered by El Niño–related wind anomalies, and the cooling in the SON of the following year is triggered by oceanic Rossby waves from the Pacific. The latter is an unrealistic feature of the model.

The significance of these SST dipole events to the climate of the surrounding regions is emphasized by the associated rainfall variability (Fig. 3a) for SON when the model IOD reaches its mature phase. For comparison, the ENSO rainfall teleconnection pattern for SON is also plotted (Fig. 3b). While the pattern associated with IO1 in SON is similar to that coincident with El Niño (Fig. 3b), with a basinwide dipole pattern in rainfall reflecting the dipole SST structure, the correlations with IO1 are more contrasting in both the zonal (in the 0°–10°S latitude band) and the meridional directions. Note that the strong positive rainfall anomaly north of the equatorial eastern Indian Ocean may be unrealistic (akin to a split ITCZ), and it helps promote the alongshore winds in the east. In observations, this feature of the IOD appears farther north over the Asian continent in at least one IOD event (Behera et al. 1999).

c. Atmospheric circulation anomalies

The anomaly patterns of wind and rainfall associated with a one-standard-deviation value of the IO1 time series for each season are shown in Fig. 4. The patterns are constructed by regressing gridpoint anomaly fields onto the IO1 time series and multiplying the regression coefficient by a one-standard-deviation value of the IO1 time series. During DJF, easterly and southerly wind anomalies (first and third columns of Fig. 4a) occur in the eastern Indian Ocean and off the Sumatra–Java coast (associated with mature El Niño conditions in the Pacific); however, there is little SST anomaly (Fig. 1a). During MAM, easterlies (first column of Fig. 4a) are present over the eastern Indian Ocean off Java, and stronger easterlies cover most of the northern tropical Indian Ocean; although there is little cold anomaly off the Sumatra–Java coast, warm anomalies have developed in the southern tropical western Indian Ocean, accompanied by marked positive rainfall anomalies (second column of Fig. 4b).

By JJA, the cold pole in the eastern Indian Ocean has become substantial with an increasing zonal SST gradient (Fig. 1c; from 0.37°C in MAM to 0.78°C in JJA). This is accompanied by a strengthening zonal wind anomaly and decreasing local rainfall (Fig. 4), and both the zonal wind and rainfall anomalies expand westward from the Sumatra–Java coast. There is also a significant north–south rainfall gradient (second column of Fig. 4c) and a significant southerly anomaly at around 100°E (third column of Fig. 4c). The cold anomaly in the east reaches a maximum in SON (Fig. 1d, when the zonal SST is strongest) and, by this time, has expanded farther west, as do the easterly wind and the negative rainfall anomalies. The southerly anomaly, which contributes to the alongshore wind anomalies along the Sumatra–Java coast, is strongest in this season (third column of Fig. 4c).

The growth and westward expansion of easterly anomalies, zonal SST gradient, and rainfall gradient suggest that an ocean–atmosphere feedback involves wind, SST, and rainfall anomalies. To explore this further, we have formed a time series of zonal wind stress anomalies averaged over a box bounded by 0°–10°S and 70°–100°E and zonal rainfall gradient from the two boxes for the zonal SST gradient (see section 3a) for each season. Correlation coefficients between the zonal SST gradient and the zonal wind for SON, JJA, MAM, and DJF are 0.86, 0.73, 0.56, and 0.50, respectively; and the correlation coefficients between the zonal wind and the zonal rainfall gradient for SON, JJA, MAM, and DJF are 0.85, 0.44, 0.59, and 0.58, respectively.

Likewise, the strong meridional wind stress anomalies over the Sumatra–Java region in JJA and SON also show strong correlations with meridional SST and rainfall gradient. Correlation coefficients between meridional wind stress anomaly averaged over a 0°–10°S, 8°S–4°N box and meridional SST gradient constructed from two boxes (0°–10°S, 100°–115°E; 0°–10°N, 95°–110°E) for SON and JJA are 0.90 and 0.79, respectively, while the correlation coefficients between meridional wind stress and meridional rainfall gradient (similarly constructed) for SON and JJA are 0.83 and 0.81, respectively. Thus correlation between wind and rainfall gradients and between wind and SST gradients is comparable.

The picture that emerges is that there are two parts to the mechanism indicating the ocean–atmosphere feedback: 1) the wind anomaly responds to the SST gradient, and convergent wind at the end of the gradient favors uplift and rain; 2) uplift and rain over the warmest SST draws the wind into the convergent region. The SST, wind, and rain are all part of the consistent mechanism that operates during the JJA to SON season and should not be separated. The atmosphere–ocean forcing necessarily involves a response in the thermocline, which is discussed in the next section. The complete feedback loop is not simply between zonal wind on the equator and SST along the Java–Sumatra coast. Rather, the feedback involves alongshore winds that promote upwelling on the Java–Sumatra coast and near-equatorial zonal wind that raises the thermocline to the east and deepens it to the west. In the positive dipole phase, a cold SST anomaly and suppressed rainfall south of the equator on the Java–Sumatra coast drive an anticyclonic circulation anomaly that is consistent with the steady response (Gill model) to a heat sink displaced south of the equator. This anticyclonic structure is similar to the observed (e.g., Saji et al. 1999; Shinoda et al. 2004b). Because the coast is tilted from the southeast to the northwest, upwelling is promoted on the eastern side of the anticyclone, while the near-equatorial easterly anomalies also act to elevate the thermocline to the east. Off-equatorial Rossby waves act to deepen the thermocline to the west. Variations in the model’s depth of the thermocline are similar to observed variation of IOD based on expendable bathythermographic (XBT) and altimetric data (Feng and Meyers 2003; Wijffels and Meyers 2004). The positive feedback between anomalous SST, rainfall, zonal and meridional winds, and depth of the thermocline in the east only occurs in June–October when the mean thermocline in the east is shallow and seasonal upwelling occurs along the coast (Fig. 7).

d. EOF patterns of thermocline anomalies

In contrast to the strong seasonality in the structure of the leading EOF of SST (Fig. 1), the leading EOF of D20 (hereafter D1) is not strongly seasonal (Fig. 5). These EOFs account for some 13%–17% of the total variance (the low percentage is due to the presence of large variance in the western Indian Ocean that is not related to IOD but contributes to the total variance). In each season, the thermocline topography shows a dipole structure and patterns associated with Rossby wave propagation. The Rossby wave is particularly large along 10°–15°S. Correlation coefficients among time series of these EOFs range from 0.68 to 0.92, supporting the finding of Huang and Kinter (2002) and Rao et al. (2002) that the thermocline variability in a season evolves from the previous season.

Correlation coefficients between time series of D1 and IO1 for the same season are very high, in the range of 0.80–0.86. This indicates that the subsurface dipole can exist throughout the year in the subsurface thermocline structure. The highest correlation of D1 with Niño-3.4 occurs simultaneously in DJF and MAM, while during JJA and SON it occurs with D1 lagging Niño-3.4 by 1 yr. This suggests that in each season the leading modes of variability at the surface and subsurface are to a large extent driven by the same forcing, with the exception of a lag relative to Niño-3.4.

e. Thermocline feedback

Xie et al. (2002) and Huang and Kinter (2002) have discussed the necessary conditions for a thermocline and SST coupling feedback, whereby a change in the thermocline depth can lead to an SST anomaly. These conditions include the presence of upwelling and a shallow thermocline that is lifted into the depth of upwelling. In general, coupling and decoupling of SST and thermocline can be expected for a variety of reasons. For example, an easterly anomaly off Java when superimposing on a climatological westerly wind will produce a warm anomaly in the surface, but the same easterly anomaly will generate an anomalously shallow thermocline, giving negative local correlations. If the anomalous wind is not strong enough to reverse the climatological westerly winds, there is no upwelling, and the thermocline, although rising, is deep, and hence unable to affect the surface temperature. Thus, both surface and subsurface anomalies are generated by the same forcing, but they are not dynamically coupled. This is generally the situation in the southern equatorial eastern Indian Ocean for DJF and MAM, as illustrated in Figs. 6a,b, which show a negative local correlation coefficient between D20 and SST because the anomalous easterly reduces latent heat loss. In the same region in the JJA and SON seasons, the mean winds are easterly and there is an easterly anomaly; together they have lifted the thermocline to the upwelling depth, contributing to surface cooling by upwelling colder water. Hence the local correlation coefficient between D20 and SST is positive.

Another region where thermocline and SST feedback can operate is the western tropical southern Indian Ocean. Xie et al. (2002) and Huang and Kinter (2002) suggest that in this region, the southeasterly trades are present most time of the year. With the presence of westerlies to the north, this forms a pattern of wind stress that generates upward Ekman pumping and formation of the zonal ridge in the thermocline along 10°S. As a downwelling Rossby wave propagates through, upwelling decreases, leading to a surface warming. In our model, this coupling appears to operate (Fig. 6). As discussed in section 2, the model ridge extends too far to the east relative to the observed. As a consequence, the zonal extent of this coupling also extends too far east (to 90°E) (Figs. 6a,b). Overall, the evolution from Fig. 6a to Fig. 6c displays the influence of the westward propagation of a Rossby wave and its subsequent reflection as an equatorial Kelvin wave. The strongest coupling in 60°–90°E in MAM is due to suppression of the strong mean upwelling as the downwelling Rossby wave propagates through the region (see Fig. 7c).

f. Seasonal phase locking

The seasons in which cold SST anomalies in the equatorial eastern Indian Ocean emerge and intensify are the seasons (JJA and SON) when the mean SST is cold (Fig. 7a), in conjunction with persistent easterlies (Fig. 7b), upwelling (Fig. 7c), an upward tilt toward the east of the thermocline (Fig. 7d), and a decreased rainfall (Fig. 7e). The climatological variations of these fields support the notion that the thermocline has to be located above a threshold determined by the maximum depth of surface-forced upwelling in order for cold SST anomalies to be generated in the presence of southeasterly wind anomalies (Susanto et al. 2001). Upwelling off the Sumatra–Java coast is mainly confined to the upper 100 m. The top of the thermocline off the Sumatra–Java coast is below this depth in DJF but near this depth in SON (not shown). Thus, in JJA and SON anomalous shoaling of the thermocline and anomalous strength of upwelling driven by local winds are conducive to the generation of cold SST anomalies, while in DJF and MAM shoaling of the thermocline does not influence the surface in a way that is conducive to the generation of cold SST anomalies. Seen in this way, the appearance of a dipole structure in the Indian Ocean may be regarded as the consequence of year-to-year fluctuations of the seasonal cycle, which may be driven by any forcing, local or remote, that is capable of generating a change in the upwelling regime.

4. Evolution and dynamics of the surface dipole

a. Anomaly pattern associated with IO1

Since the model dipole is most prominent in SON (Fig. 1), in what follows, we will use IO1 in SON as a base season and compute lag regression in order to illustrate the development and evolution of the model dipole. Figure 8 shows maps of one-standard-deviation anomaly pattern of wind stress (plotted as vectors) and SST (shading) associated with the SON at various lags. These patterns are obtained by first regressing gridpoint anomalies onto the SON IO1 time series at various lags and then multiplying the regression coefficient with a one-standard-deviation value of the SON IO1 time series. Figure 9 shows similar maps of regression patterns between IO1 and D20 (without the winds). To explore the coherent vertical structure of anomalies, Fig. 10 plots maps of correlation coefficients at various lags for upper-ocean temperature and vertical velocity (positive means upwelling) averaged between 8° and 12°S, a latitude band along which Pacific Rossby waves propagate into the Sumatra–Java coast. As will be clear, arrival of a strong signal from the Pacific on the Sumatra–Java coast is an unrealistic feature of the model, which is not found in observed variability of subsurface thermal structure (Wijffels and Meyers 2004). In Fig. 10, we use correlation coefficients because the amplitude of anomaly decreases rapidly with depth). The sequence starts from lag −4 seasons—1 yr prior to the peak of the dipole—and runs through lag +4 seasons—1 yr after the peak of the dipole.

As discussed in section 3b, lag −4 (Figs. 8a and 9a) effectively reflects the simultaneous correlation with El Niño because of the 1-yr lag of IO1 with El Niño (Fig. 2). In association with ENSO conditions in the Pacific (warm SST and surface westerlies), weak easterly–southeasterly anomalies are evident across the equatorial central and eastern Indian Ocean. The SST anomaly has a weak dipole structure, with cold anomalies underlying southeasterlies along the Sumatra–Java coast. The cold SST anomaly is associated with weak shallowing of the thermocline (Fig. 9a) and increased upward anomalous velocity (Fig. 10a). A strong, mature dipole does not develop in SON in association with El Niño because the thermocline in the Java–Sumatra upwelling region was too deep during the preceding two seasons. Off the equator, in both hemispheres, deepening of the thermocline (Fig. 9a) takes place as a result of Ekman mass transport driven by the surface equatorial easterlies. To the south of the equator, deepening of the thermocline occurs along the latitude of about 8°–12°S, where subsurface warming and downwelling anomalies (right column of Fig. 10a) are seen. Some similarities with the observed dipole are apparent (Saji et al. 1999; Webster et al. 1999), but the SST anomalies are weak, especially in the west.

Despite the persistence of El Niño conditions in the Pacific and the associated easterly anomalies across the Indian Ocean into the following two seasons [lag −3 (DJF) and lag −2 (MAM); Figs. 8b,c], the weak cold anomalies in the eastern Indian Ocean disappear. However, the subsurface features continue to develop from lag −4 to lag −2, in association with the discharge phase of the preceding El Niño. The thermocline continues to shoal off Java–Sumatra (Figs. 9b,c), and the deep thermocline west of 90°E propagates westward as a downwelling Rossby wave, reaching the western boundary by lag −2 (MAM; Fig. 9c). Meanwhile, the subsurface temperature anomalies remain in the region off the Sumatra–Java coast, seemingly awaiting a favorable time to interact with the surface in the upwelling season. These anomalies do not affect the surface in these seasons because the mean thermocline is deep in association with mean westerly wind and downwelling (Fig. 7) at this time of year.

By lag −2, substantial positive SST anomalies have developed in the southwestern portion of the basin, stemming from the downwelling Rossby wave (Figs. 8c and 9c). An equatorial downwelling Kelvin wave, stemming from reflection of the downwelling Rossby wave off of the western boundary, is evident at lag −2 (Fig. 9c) and is prominent east of 75°E at lag −1 (JJA; Fig. 9d). An analysis using monthly data (not presented) clearly shows that the reflection commences at May and reaches the eastern Indian Ocean by June and July.

Upon reaching the coast (lag 0), the downwelling Kelvin wave is not seen to propagate along the Sumatra–Java coast as a coastal downwelling Kelvin wave. A referee asks whether the model is able to resolve coastal Kelvin dynamics. Evidence that the model resolves coastal Kelvin wave dynamics appears at other times and places. In the Indian Ocean, we see what appears to be an upwelling coastal Kelvin wave propagating northward into the Bay of Bengal (lags 4, −3, and −2 in Fig. 9), and we see this upwelling Kelvin wave propagate southward onto the Java–Sumatra coast. Also, in the next section, equatorial Pacific Kelvin waves after hitting the eastern boundary propagate poleward as coastal Kelvin waves (Fig. 13). But, what we do not see at the time in question is the reflected downwelling Kelvin wave reaching the Java–Sumatra coast at lag 1 (Fig. 9). We suggest below that the downwelling Kelvin wave is offset by superposition with a stronger upwelling Rossby wave radiating from the coast. The upwelling Rossby wave is dominant; consequently, cold anomaly there grows in JJA and SON (Figs. 9d,e) in response to increased surface southeasterlies and elevation of the mean thermocline combined with the anomalous shallow thermocline south of Java.

In observations, the downwelling Kelvin waves will propagate into the Sumatra–Java coast as coastal Kelvin waves and destroy the negative SST anomalies (Gualdi et al. 2003; Huang and Kinter 2002) off the Sumatra–Java coast by deepening the thermocline. Often they initiate the development of warm anomalies, hence contributing to the biennial signals of the IOD (Feng and Meyers 2003).

The ocean–atmosphere physics at lag 0 requires special attention because they are controlling the growth of the models erroneous IOD 1 yr after ENSO. Understanding what is happening here may give a clue to correcting models. The upwelling anomalies that have developed from lag −4 to lag −2 appear to be transmitted from the western Pacific (Figs. 10a–c). These upwelling anomalies are associated with the westward-propagating Rossby waves generated in the western Pacific during the maturing of El Niño in the previous year. Its propagation through the Indonesian throughflow region and into the coastal waveguide south of Java is an unrealistic feature of the model, which will be discussed further in section 5.

As the upwelling Rossby wave propagates to the Java region during MAM (lag −2) and JJA (lag −1), negative SST anomalies grow rapidly (Fig. 10). This rapid development takes place in the presence of strong positive SST anomalies in the central and western Indian Ocean. This anomalous zonal SST gradient and the associated zonal rainfall gradient further induce easterly anomalies (Figs. 4c,d). These, together with the increasing southerly component induced by strengthening meridional rainfall and SST gradients, in turn generate further shoaling of the thermocline and further upwelling and cooling in the eastern Indian Ocean. This positive ocean–atmosphere feedback loop brings the model IOD to its maturity in SON (lag 0).

The intensification of the easterlies at lag −2 and lag −1 (Figs. 8c,d) generates another downwelling Rossby wave in the southern Indian Ocean. Compared to the initial Rossby wave at lag −3, this wave is slightly poleward and produces stronger deepening (Fig. 9d,e). Slower westward propagation is also evident (Figs. 9d–f), consistent with the lower phase speed of the higher-latitude Rossby wave. The excitation of the downwelling Rossby wave is again associated with Ekman pumping induced by the easterly anomalies, which are themselves a consequence of the positive ocean–atmosphere feedback. This time around, there are no Rossby wave signals in the north Indian Ocean. The subsequent positive feedback in the southern tropical Indian Ocean further reinforces this interhemispheric asymmetry. The Rossby wave reflects at the eastern boundary and forms an eastward-propagating, forced downwelling Kelvin wave on the equator, which accumulates at the eastern boundary at lag +4 (Fig. 9i). The Rossby–Kelvin wave dynamics and ocean–atmosphere feedback have many similarities to observed variability of the IOD (Feng and Meyers 2003).

To the extent that the anomalous SST and rainfall gradient, wind stress, and thermocline depth reinforce each other, ocean–atmosphere interactions play an important role in the development of the subsurface structure and have some influence on SST dipole through thermocline coupling discussed in section 3e, but ocean dynamics appear to play a dominant role for driving the SST anomalies. In our model, the role of ocean dynamics is unrealistically strong because of the transmission to the Sumatra–Java coast of the upwelling Rossby wave from the Pacific. Previous studies have shown that surface heat flux variations are also important (e.g., Shinoda et al. 2004a). Examinations of the relationship between heat flux and SST anomalies off the Sumatra–Java coast reveal that in the model, heat flux anomalies act to dampen the development of SST anomalies in the southern tropical eastern Indian Ocean, especially in SON (Fig. 11e). This is also an observed characteristic in the Java upwelling zone (Ansell 2002), although it is in contrast to observed behavior on a large scale of the eastern Indian Ocean (Hendon 2003), in which latent heat flux anomalies act to reinforce SST anomalies in JJA and SON.

In the model, positive heat flux anomalies in the eastern Indian Ocean are associated with generation of negative SST anomalies (Figs. 11d,e). The positive heat flux anomaly results from both latent and shortwave fluxes: first, as SST cools in JJA and SON, less evaporative heat is lost to the atmosphere; second as SST cools, convection and cloudiness are reduced and more shortwave radiative flux penetrates to the surface. Both processes contribute to positive flux into the ocean. Thus, in the model, negative SST anomalies develop despite the damping by the positive surface heat flux. By contrast, in observations (e.g., Behera et al. 1999; Hendon 2003), the easterly anomalies in JJA and SON act to increase the wind speed, thereby increasing the latent heat flux. This increased latent heat flux, together with increased upwelling, overcome increased shortwave radiation, and large-scale surface cooling results.

By lag +1 (DJF), the cold SST anomalies in the east abruptly disappear (Fig. 8f). Along the equator, this is partly explained by the arrival of a second downwelling Kelvin wave (Figs. 9e–g), associated with reflection of the second Rossby wave off the western boundary. However, south of the equator, the subsurface is still cold in the east (Fig. 10f). Hence, rapid surface warming is partly attributed to the positive heat flux that peaks in SON. It is also attributed to mean downwelling and deepening of the mean thermocline in DJF (Fig. 7), which are unfavorable for the development of cool SST anomalies off the Sumatra–Java coast at this time of year. Interestingly, long after the surface easterlies have abated, the signature of the forced downwelling Rossby wave in the western Indian Ocean persists (Figs. 9g–i).

b. Anomaly patterns associated with ENSO

We now compare the evolution of anomalies associated with IO1 with those associated with the typical ENSO cycle in the model. We do so by computing a one-standard-deviation anomaly pattern at various lags as in Figs. 8 and 9 but use Niño-3.4 in SON as the reference time series. Here, we show lags running from a simultaneous phase with Niño-3.4 (lag 0) out to lag +4 seasons for SST, surface winds (Fig. 12), and D20 (Fig. 13). Because of the 1-yr lag between IO1 and Niño-3.4, the simultaneous pattern associated with Niño-3.4 in SON (lag 0; Figs. 12a and 13a) is comparable to the lag −4 pattern associated with IO1 (Figs. 8a and 9a). Similarly, the lag +4 pattern associated with Niño-3.4 (Figs. 12d and 13d) is comparable to the simultaneous correlation with IO1 (Figs. 8d and 9d). Very similar evolution is depicted in both cases. For instance, in the SON following the peak of ENSO (Fig. 8e), the strongest SST dipole is evident in the Indian Ocean; furthermore, a weak dipole structure is evident in the SON season just prior to the peak of ENSO (Fig. 8a), but the warm anomalies in the western Indian Ocean are weak, especially south of the equator.

Comparing Figs. 8d,e with Figs. 12d,e, we see that the SST anomalies are larger in the Indian Ocean in Figs. 8d,e. Off the Sumatra–Java coast, SST anomalies associated with the IO1 are about twice as large as those associated with Niño-3.4. This indicates that ENSO alone cannot generate the full magnitude of the IOD SST signal and local ocean–atmosphere interaction has contributed to the full IOD response.

Another feature that is emphasized in Fig. 12 is the clear biennial component of ENSO in the equatorial Pacific, with SST anomalies of opposing polarity developing 1 yr after El Niño peaks (Figs. 12a,e). This is in sharp contrast to the evolution associated with IO1 in the eastern Indian Ocean, where there is no biennial signal.

Comparing the evolution of the thermocline associated with ENSO (Figs. 13a–e) and that with IO1 (Figs. 8a–e), we again see strong resemblance if a 1-yr lag is taken into account. The formation of the first Ekman ridge in both the northern and southern Indian Ocean, the associated westward propagation in the form of downwelling Rossby wave, and the subsequent reflection as a Kelvin wave are all present in Figs. 13a–e.

In the Pacific, the transfer of warm water (greater than 20°C) from the western Pacific to the eastern Pacific associated with the development of El Niño is clear at lag 0 and lag 1 (Figs. 13a–b). In the mean time, shoaling of the thermocline that starts in the western Pacific continues (Figs. 13b–d); by lag +3 and lag +4, the thermocline is anomalously shallow throughout the entire tropical Pacific, showing the mature phase of the El Niño discharge phase. Along the eastern boundary, poleward propagation of Kelvin waves is seen.

Shoaling of the equatorial Pacific thermocline as ENSO matures is accompanied by shoaling of the thermocline off the Sumatra–Java coast. The Pacific shoaling tends to precede development of the SST dipole in the Indian Ocean by two–three seasons. In the model, the influence of the Pacific shoaling on the eastern Indian Ocean is through Rossby wave propagation to the Java coast. In the next section, we explore this model feature, compare it to earlier observational and modeling studies, and discuss how it is related to the 1-yr delay in development of the model IOD in response to ENSO.

5. Discussion

The aim of the present study is to examine the Mark 3 climate model simulation of dipolelike variability in the Indian Ocean and to explore its dynamics, its associated rainfall teleconnection pattern, and its relationship with ENSO cycles. The main aim is to identify deficiencies in the model simulation and develop ideas on how to correct them if possible. We have shown that the CSIRO Mark 3 coupled model reproduces some realistic features of IOD, including the strong seasonal phase locking, the rainfall–IOD teleconnection pattern, the propagation of oceanic Rossby waves and reflections at the western boundary, and the ocean–atmosphere interaction in the development and evolution of the IOD. There are, however, several unrealistic features between the observed and modeled dipole.

Over the 240 yr of model integration (discarding the first 20 yr of the 260-yr integration), there are 74 El Niño events, while there are only 34 IOD events. The majority of model IOD events, 23 out of 34, peak 1 yr after an El Niño event peaks, and some 11 events take place simultaneously with an El Niño event. The 1-yr delay in model IOD events is an unrealistic feature that we need to understand. Furthermore, the IOD has no biennial cycles despite the fact that the model ENSO has a large number of biennial events. These features are in sharp contrast to those observed over the past several decades, which show significant biennial signals (e.g., Saji et al. 1999; Feng and Meyers 2003), and to the significant number of observed dipole events taking place simultaneously with ENSO events (e.g., Hendon 2003). In this section, we explore the possible causes.

a. The 1-yr lag

There are several possible factors that may contribute to the 1-yr lag with respect to ENSO. First, consistent with the dominant role of the oceanic dynamics in the generation of the dipole in this model, atmospheric heat flux acts to dampen the SST anomalies off the Sumatra–Java coast as the model dipole develops. The damping by the surface heat fluxes means that it will take a longer time for the dipole to peak and may preclude a simultaneous peak with ENSO. In the model, the damping is partially a result of a parameterization of latent heat flux that, although responsive to SST anomalies, does not vary with wind speed when the wind speed is below a threshold value. By contrast, in the observations, latent heat flux helps the development of the large-scale SST anomalies (Hendon 2003) with easterly anomalies acting on climatological easterly winds in JJA and SON, enhancing the evaporative heat loss. Although shortwave radiative flux into the ocean increases as a result of reduction in cloudiness, enhanced latent heat anomalies in Hendon (2003) dominate in the observed eastern pole. Hence, in observations, heat flux anomalies may act to reinforce the development of cold SST anomalies off the Sumatra–Java coast, but this does not occur in the model.

Second, we have shown that two–three seasons after an El Niño event peaks, the thermocline off the Sumatra–Java coast is still in the process of shoaling as a consequence of propagation of upwelling Rossby waves from the Pacific into the region (right column of Fig. 10). Two–three seasons after an El Niño peaks, the Pacific thermocline is at its shallowest depth, as a result of heat discharge. The model discharge in the Pacific involves Ekman pumping, Rossby wave propagation, and associated western boundary reflections. The upwelling wave that reaches the Java coast in the model is part of the discharge process at the western boundary and its interaction with the equatorial and coastal waveguide in the Indonesian throughflow region. In observations, the discharge signal from the Pacific appears primarily on the western coast of Australia, with only a weak signal on the coast of Java (Wijffels and Meyers 2004). This unrealistic feature in the model may account for the delayed IOD, which is further explored below.

b. Modeled ENSO discharge and the variability off the Sumatra–Java coast

To further identify the post–El Niño discharge in the model and its relationship to the elevated thermocline in the Indian Ocean in the year after El Niño peaks, we have applied EOF analysis to the model D20 anomalies in the tropical Pacific and Indian Ocean domain of 10°S–10°N for the SON season. The first two EOFs are plotted in Fig. 14 and the associated time series are plotted in Fig. 15. The first two EOFs account for 34.2% and 20% of the total variance, respectively.

The recharge–discharge paradigm of ENSO (Jin 1997a, b) and observational supporting evidence (Meinen and McPhaden 2000) suggest that EOF1 and EOF2 of D20 reflect different phases of the ENSO cycle. The first mode has an out-of-phase structure across the equatorial Pacific basin associated with the west-to-east transfer of warm water that accompanies mature ENSO conditions and is highly correlated with the Niño-3.4 index as seen in Fig. 15a (for the first 80 yr). The negative weights in the eastern Indian Ocean indicate shallowing of the thermocline driven by the ENSO-induced surface easterly anomalies over the eastern Indian Ocean (atmospheric teleconnection). The second mode represents zonal-mean shoaling of the equatorial Pacific thermocline, that is, the discharge of the zonal-mean equatorial Pacific heat content, which usually peaks some three seasons after an El Niño event matures (Fig. 15b). Upwelling Rossby waves and their western boundary reflection compensate for the mass transport to the east and play an important role in the discharge process. The strong negative weights off the Sumatra–Java coast reflect the effect of propagation of the upwelling Rossby wave from the Pacific into the equatorial Indian Ocean waveguide (oceanic teleconnection).

Shoaling of the model thermocline in the eastern Indian Ocean is at any time the combined effect of both atmospheric teleconnection (via remotely forced wind stress anomalies) and oceanic teleconnection (via transmission of Rossby waves through Indonesian passages). The shallow thermocline in the eastern Indian Ocean during peak El Niño (Fig. 9, lag −4) is generated by a thermocline response to El Niño wind anomalies (atmospheric teleconnection). The shallow thermocline in the eastern Indian Ocean a year later (Fig. 9, lag 0) is in part generated by Rossby wave propagation (oceanic teleconnection). As SST begins to cool in JJA (Fig. 8, lag −1), the feedback process linking SST, rainfall, wind, and thermocline results in strengthening easterly wind anomalies over the southern tropical Indian Ocean and strengthening alongshore wind along the Sumatra–Java coast.

At lag −4 relative to IO1, there are few thermocline anomalies in the western Pacific (Fig. 9a), but there are already significant thermocline anomalies in the eastern Indian Ocean, suggesting that these anomalies are mainly generated by wind anomalies (atmospheric teleconnection). However, from lag −3 (Fig. 9b) to lag 0 (Fig. 9e), the influence of an upwelling Rossby wave propagating from the Pacific Ocean (oceanic teleconnection) is evident, and the Rossby wave appears to be reinforced by the easterly anomalies (atmospheric teleconnection).

c. Observed relationship between ENSO discharge and variability in the eastern Indian Ocean

The Rossby wave propagation through the Indonesian throughflow region and into the Indian Ocean in observations has been documented by Wijffels and Meyers (2004). They find that the response to wind forcing over the Pacific is transmitted to the Indian Ocean primarily through the coastal waveguide of the Australian west coast. The different behavior of the model transmitting the discharge to the Java–Sumatra region is a significant discrepancy between model and observations. To investigate this further, we examine the upper-ocean temperature analyses compiled by Smith (1995a, b)). They contain in situ observations from XBT and ocean moorings. From about the mid-1990s, the analysis between 5°S and 5°N has included observations from the Tropical Atmosphere Ocean (TAO) array in the Pacific (Smith and Meyers 1996). These analyses are available on a 1° latitude × 2° longitude grid for each month from 1980 to the present down to 500 m on 14 standard levels. In the Indian Ocean, because no such gridded data are available, we employ a frequently repeated XBT section, IX1, which extends from Fremantle, Australia, to Sunda Strait (between Sumatra and Java; see Fig. 1 of Feng and Meyers 2003). The IX1 XBT section covers the period from 1982 onward. We first conduct EOF analysis on monthly D20 anomalies of the Pacific Ocean (not shown). This yields two EOF modes with patterns similar to Fig. 3 of Meinen and McPhaden (2000): EOF1 is the quasi-equilibrium mode showing east–west sloshing of warm water and is highly correlated with Niño-3.4 (r = 0.95); EOF2 is the zonally uniform discharge–recharge mode. The pattern shows positive weights in the tropical Pacific, so that the time series is most negative when the tropical Pacific is discharged. We then correlate the associated time series with monthly temperature anomalies of the IX1 section in the Indian Ocean. The correlation maps for EOF1 and EOF2 are shown in Figs. 16a and 16b, respectively.

Focusing on Fig. 16a, which reflects the simultaneous correlation with mature ENSO conditions in the Pacific, we see that near the Sumatra–Java region (10°S) negative correlations exist, indicating the role of atmospheric teleconnection (Fig. 13a). In Wijffels and Meyers (2004), the shallow thermocline is generated primarily by wind over the central equatorial Indian Ocean and transmitted to the eastern boundary by Kelvin waves. The strengthened easterly wind also generates a stronger South Equatorial Current in the upper 150 m. Positive correlations indicate a deep countercurrent in the coastal waveguide off Java, below strengthened upwelling. Strong negative correlations also exist in the latitude band 20°–25°S. This is the beginning of the discharge transmitted from the Pacific on the western Pacific trajectory identified by Wijffels and Meyers (2004).

Figure 16b, which reflects the simultaneous correlation with the discharge phase of ENSO, reveals that the Java coast (10°S) does not feel the impact of the Pacific discharge. This is in sharp contrast to the model results. In some models, the Java coast is screened by Timor, which is a separate island located some 200 km south of the Java coast, keeping the Rossby waves off the coast (e.g., Potemra 2001). Instead, the Rossby waves emanate from the Australian coastal waveguide in the latitude band 17°–25°S. These Rossby waves propagate very slowly and play no role in elevating the thermocline off the Java coast. The fully developed El Niño Pacific discharge appears in the eastern Indian Ocean primarily off the coast of western Australia (Fig. 15b). (This discharge appears in the EOF time series as a negative value so it is indicated by positive correlation.) Overall, these results indicate that while Rossby waves do come into the Indian Ocean, they end up off the northwestern Australian coast and that the pathway to the Java coast and into the near-equatorial waveguide transmits a weak signal at most (R. Murtugudde 2002, personal communication).

In the model, Java and Timor are connected and represented as a zonal east–west coast. It may be this unrealistic representation of the model geometry that forms a pathway for Pacific upwelling Rossby waves to reach the model Java coast. Some two–three seasons after an El Niño event peaks, the tropical Pacific is fully discharged. In the process, upwelling Rossby waves have lifted the thermocline off the Java coast to a shallow depth in the model. Coupled with the strong seasonal cycle of the mean upwelling and thermocline in the eastern Indian Ocean, this allows the surface dipole to peak in the following JJA and SON. Although the discharge Rossby wave reaches the Java coast as early as MAM (i.e., just after the peak of ENSO in DJF), the seasonal phase locking of upwelling onset will “round off” the time so that the lag between ENSO and the surface dipole is nearly a year.

An important question arises as to why the upwelling Rossby waves, upon arrival at the Sumatra–Java coast, do not propagate farther westward. Examination of this subsequent behavior requires application of techniques such as that of Chambers et al. (1999) and is beyond the scope of the present study. However, previous studies (e.g., McCreary and Anderson 1984) have shown that an upwelling Rossby wave will stall when it propagates into a region with easterly anomalies and downward Ekman pumping. The easterlies have the effect of depressing the thermocline to the west and off-setting the upwelling.

d. Contemporaneous IOD events

As mentioned earlier, there are some IOD events that take place simultaneously with El Niño events. What makes these events distinct from the rest? To address this, we composite SST, D20, and surface wind stress anomalies for the 11 simultaneous El Niño events (not shown). It is found that anomalies associated with these El Niño events develop early; for example, at lag −4 relative to peak El Niño (SON in the preceding year), substantial westerly anomalies are already seen in the western Pacific. By MAM (lag −2), there are already substantial SST anomalies in both the Indian and Pacific Ocean. By JJA (lag −1), there are substantial cold anomalies in the eastern Indian Ocean and strong anomalous shoaling of the thermocline in the eastern Indian Ocean. The positive air–sea interaction brings the IOD to its peak in SON, by which time the SST and wind anomalies in the western Pacific region have already lasted for some three–four seasons. Thus both the atmospheric and oceanic teleconnections conspire to ensure that the dipole peaks in the same year as ENSO. These El Niño events also decay faster and flip to a La Niña phase by lag 4 (four seasons after El Niño peaks). For El Niño events with lagged dipole events, the anomalies generally start late and grow slowly.

e. The lack of biennial IOD signals

The 1-yr lag with respect to ENSO is in part a cause for the lack of biennial signal in model IOD. As discussed briefly in section 3, the arrival at Sumatra–Java coast of upwelling Rossby waves from the Pacific Ocean after an El Niño peaks makes reflected downwelling Kelvin waves unable to initiate development of positive SST anomalies off the Sumatra–Java coast. In observations, this is a major forcing for generating biennial signals of the IOD, besides the biennial ENSO forcing.

Another factor that contributes to the lack of biennial IOD signals in the model is that the biennial ENSO anomaly patterns are quite different from those associated with ENSO cycles on longer time scales. Lag correlation analysis between the Niño-3.4 index and gridpoint SST and surface wind anomalies (figure not shown) reveals that the biennial ENSO anomalies are confined in the central and eastern Pacific region, with weak signals in the eastern Indian Ocean in the DJF and MAM seasons and few signals in the JJA and SON. Thus there is a lack of triggering by atmospheric teleconnection on this time scale.

6. Conclusions

We have explored surface dipole variability in the tropical Indian Ocean in the CSIRO coupled climate model. Several observed features of the dipole are reproduced. These include the seasonal phase locking with peak amplitude in the SON season; positive air–sea feedback involving anomalies of SST, rainfall gradient, wind, and upwelling along the Sumatra–Java coast; variation in thermocline depth; and finally, the role of oceanic Rossby waves in generation of positive SST anomalies in the western Indian Ocean through thermocline feedback.

The relationship of the dipole with ENSO is a complex one. On the one hand, the strong seasonal phase locking suggests that the model dipole results from year-to-year fluctuations of the JJA and SON climate and can be generated during years when El Niño does not occur. On the other hand, although not every model El Niño event induces an IOD event, the IOD events in the model are linked with either a simultaneous or a preceding ENSO warm event. The majority of the IOD events lag El Niño events by two–three seasons, and the IOD shows little biennial signal. These features are in sharp contrast to the observed IOD events, which display strong biennial signals and often peak in the same year as an El Niño event, and some of which develop independently of ENSO. We found that the two–three season lag is caused by an unrealistic oceanic teleconnection, that is, propagation of upwelling Rossby waves from the western Pacific into the Java coast, allowing the thermocline water there to participate in the heat discharge process during ENSO. Analysis of observations indicates that there is no pathway for discharge Rossby waves to reach the Java coast. The results highlight the importance of proper topographic and geographic representation of the Indonesian region in modeling variability of the Indian Ocean. Experiments with more realistic topography and geometry are being conducted and will be reported in a separate study.

Ocean–atmosphere interaction in the growth of model positive IOD events involves alongshore winds that promote upwelling on the Java–Sumatra coast and near-equatorial zonal wind that raises the thermocline to the east and deepens it to the west. In the positive dipole phase, a cold SST anomaly and suppressed rainfall south of the equator off the Java–Sumatra coast drives an anticyclonic circulation anomaly that is consistent with the steady response (Gill model) to a heat sink displaced south of the equator. Because the coast is tilted from the southeast to the northwest, upwelling is promoted on the eastern side of the anticyclone, while the near-equatorial easterly anomalies also act to elevate the thermocline to the east, enhancing the cooling of SST. Off-equatorial Rossby waves act to deepen the thermocline to the west, enhancing the warming of SST anomalies. The basinwide increasing SST gradients and associated rainfall patterns act to increase the southeasterly wind anomalies during the growth of IOD. The positive feedback between anomalous SST, rainfall, zonal and meridional winds, and depth of the thermocline in the east only occurs in June–October when the mean thermocline in the east is shallow and seasonal upwelling occurs along the coast. This scenario of growth is similar to the one put forward by Gualdi et al. (2003). In the Mark 3 model, the growth is initiated by a different mechanism than in their model. This indicates that the growth mode might be triggered by a number of different mechanisms.

Acknowledgments

The Australian Greenhouse Office and CSIRO Wealth from the Oceans Flagship supported this work. The efforts of members of the Climate Model and Application Program are gratefully acknowledged. We thank Mark Collier for arranging model outputs in an easily accessible format and Paul Durack for reviewing the paper before submission.

REFERENCES

  • Allan, R., and Coauthors, 2001: Is there an Indian Ocean Dipole, and is it independent of El Niño–Southern Oscillation? CLIVAR Exchanges, No. 3, International CLIVAR Project Office, Southampton, United Kingdom, 18–21.

  • Annamalai, H., R. Murtugudde, J. Potemra, S. P. Xie, P. Liu, and B. Wang, 2003: Coupled dynamics over the Indian Ocean: Spring initiation of the zonal dipole. Deep-Sea Res., 50 , 23052330.

    • Search Google Scholar
    • Export Citation
  • Ansell, T., 2002: Air–sea interaction in the tropical Indian Ocean and links with Australian rainfall variability. Ph.D. dissertation, University of Melbourne, 272 pp.

  • Baquero-Bernal, A., M. Latif, and S. Legutke, 2002: On dipolelike variability in the tropical Indian Ocean. J. Climate, 15 , 13581369.

    • Search Google Scholar
    • Export Citation
  • Battisti, D. S., and A. C. Hirst, 1989: Interannual variability in a tropical atmosphere–ocean model: Influence of the basic state, ocean geometry, and nonlinearity. J. Atmos. Sci., 46 , 16871712.

    • Search Google Scholar
    • Export Citation
  • Behera, S. K., R. Krishnan, and T. Yamagata, 1999: Unusual ocean–atmosphere conditions in the tropical Indian Ocean during 1994. Geophys. Res. Lett., 26 , 30013004.

    • Search Google Scholar
    • Export Citation
  • Behera, S. K., P. S. Salvekar, and T. Yamagata, 2000: Simulation of interannual SST variability in the tropical Indian Ocean. J. Climate, 13 , 34873499.

    • Search Google Scholar
    • Export Citation
  • Birkett, C., R. Murtugudde, and T. Allan, 1999: Indian Ocean climate event brings floods to East Africa’s lakes and the Sudd Marsh. Geophys. Res. Lett., 26 , 10311034.

    • Search Google Scholar
    • Export Citation
  • Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev., 97 , 163172.

  • Cai, W., M. A. Collier, H. B. Gordon, and L. J. Waterman, 2003: Strong ENSO variability and a super-ENSO pair in the CSIRO coupled climate model. Mon. Wea. Rev., 131 , 11891210.

    • Search Google Scholar
    • Export Citation
  • Chambers, D. P., B. D. Tapley, and R. H. Stewart, 1999: Anomalous warming in the Indian Ocean coincident with El Niño. J. Geophys. Res., 104 , 30353047.

    • Search Google Scholar
    • Export Citation
  • Clark, C. O., P. J. Webster, and J. E. Cole, 2003: Interdecadal variability of the relationship between Indian Ocean zonal mode and east African coastal rainfall anomalies. J. Climate, 16 , 548554.

    • Search Google Scholar
    • Export Citation
  • Feng, M., and G. Meyers, 2003: Interannual variability in the tropical Indian Ocean: A two-year time scale of Indian Ocean Dipole. Deep-Sea Res., 50 , 22632284.

    • Search Google Scholar
    • Export Citation
  • Feng, M., G. Meyers, and S. Wijffels, 2001: Interannual upper ocean variability in the tropical Indian Ocean. Geophys. Res. Lett., 28 , 41514154.

    • Search Google Scholar
    • Export Citation
  • Gordon, H. B., and S. P. O’Farrell, 1997: Transient climate change in the CSIRO coupled model with dynamic sea ice. Mon. Wea. Rev., 125 , 875907.

    • Search Google Scholar
    • Export Citation
  • Gordon, H. B., and Coauthors, 2002: CSIRO Mark 3 Climate System Model. CSIRO Atmospheric Research Tech. Paper 60, 130 pp.

  • Gregory, D., and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability dependent closure. Mon. Wea. Rev., 118 , 14831506.

    • Search Google Scholar
    • Export Citation
  • Griffies, S. M., 1998: The Gent–McWilliams skew flux. J. Phys. Oceanogr., 28 , 831841.

  • Griffies, S. M., A. Gnanadesikan, R. C. Pacanowski, V. D. Larichev, J. K. Dukowicz, and R. D. Smith, 1998: Isoneutral diffusion in a z-coordinate ocean model. J. Phys. Oceanogr., 28 , 805830.

    • Search Google Scholar
    • Export Citation
  • Gualdi, S., E. Guilyardi, A. Navarra, S. Masina, and P. Delecluse, 2003: The interannual variability in the tropical Indian Ocean as simulated by a CGCM. Climate Dyn., 10 , 567582.

    • Search Google Scholar
    • Export Citation
  • Hendon, H. H., 2003: Indonesian rainfall variability: Impacts of ENSO and local air–sea interaction. J. Climate, 16 , 17751790.

  • Huang, B., and J. L. Kinter III, 2002: Interannual variability in the tropical Indian Ocean. J. Geophys. Res., 107 .3199, doi:10.1029/2001JC001278.

    • Search Google Scholar
    • Export Citation
  • Iizuka, S., T. Matsuura, and T. Yamagata, 2000: The Indian Ocean SST dipole simulated in a coupled general circulation model. Geophys. Res. Lett., 27 , 33693372.

    • Search Google Scholar
    • Export Citation
  • Jin, F-F., 1997a: An equatorial ocean recharge paradigm for ENSO. Part I: Conceptual model. J. Atmos. Sci., 54 , 811829.

  • Jin, F-F., 1997b: An equatorial ocean recharge paradigm for ENSO. Part II: Stripped-down coupled model. J. Atmos. Sci., 54 , 830847.

  • Klein, S. A., B. J. Soden, and N-C. Lau, 1999: Evidence for a tropical atmospheric bridge. J. Climate, 12 , 917932.

  • Kowalczyk, E. A., J. R. Garratt, and P. B. Krummel, 1991: A soil-canopy scheme for use in a numerical model of the atmosphere—1D stand alone model. CSIRO Division of Atmospheric Research Tech. Paper 23, 60 pp.

  • Kowalczyk, E. A., J. R. Garratt, and P. B. Krummel, 1994: Implementation of a soil-canopy scheme into the CSIRO GCM—Regional aspects of the model response. CSIRO Division of Atmospheric Research Tech. Paper 32, 59 pp.

  • Krishnamurthy, V., and B. P. Kirtman, 2003: Variability of the Indian Ocean: Relation to monsoon and ENSO. Quart. J. Roy. Meteor. Soc., 129 , 16231646.

    • Search Google Scholar
    • Export Citation
  • Latif, M., and T. P. Barnett, 1995: Interactions of tropical oceans. J. Climate, 8 , 952964.

  • Lau, N-C., and M. J. Nath, 2003: Atmosphere–ocean variations in the Indo-Pacific sector during ENSO episodes. J. Climate, 16 , 320.

  • Lau, N-C., and M. J. Nath, 2004: Coupled GCM simulation of atmosphere–ocean variability associated with zonally asymmetric SST changes in the tropical Indian Ocean. J. Climate, 17 , 245265.

    • Search Google Scholar
    • Export Citation
  • Loschnigg, J., G. A. Meehl, P. J. Webster, J. M. Arblaster, and G. P. Compo, 2003: The Asian monsoon, the tropospheric biennial oscillation, and the Indian Ocean zonal mode in NCAR CSM. J. Climate, 16 , 16171641.

    • Search Google Scholar
    • Export Citation
  • McCreary, J. P., and D. L. T. Anderson, 1984: A simple model of El Niño and the Southern Oscillation. Mon. Wea. Rev., 112 , 934946.

  • McGregor, J. L., 1993: Economical determination of departure points for semi-Lagrangian models. Mon. Wea. Rev., 121 , 221230.

  • Meehl, G. A., and J. M. Arblaster, 2002: The tropical biennial oscillation and Asian–Australian monsoon rainfall. J. Climate, 15 , 722744.

    • Search Google Scholar
    • Export Citation
  • Meinen, C. S., and M. J. McPhaden, 2000: Observations of warm water volume changes in the equatorial Pacific and their relationship to El Niño and La Niña. J. Climate, 13 , 35513559.

    • Search Google Scholar
    • Export Citation
  • Mitchell, J. F. B., R. A. Davis, W. J. Ingram, and C. A. Senior, 1995: On surface temperature, greenhouse gases, and aerosols: Models and observations. J. Climate, 8 , 23642386.

    • Search Google Scholar
    • Export Citation
  • Murtugudde, R., and A. J. Busalacchi, 1999: Interannual variability in the Indian Ocean. J. Climate, 12 , 23002326.

  • Murtugudde, R., J. P. McCreary, and A. J. Busalacchi, 2000: Oceanic processes associated with anomalous events in the Indian Ocean with relevance to 1997–1998. J. Geophys. Res., 105 , 32953306.

    • Search Google Scholar
    • Export Citation
  • O’Farrell, S. P., 1998: Investigation of the dynamic sea ice component of a coupled atmosphere sea ice general circulation model. J. Geophys. Res., 103 , C8,. 1575115782.

    • Search Google Scholar
    • Export Citation
  • Potemra, J. T., 2001: Contribution of equatorial Pacific winds to southern tropical Indian Ocean Rossby waves. J. Geophys. Res., 106 , 24072422.

    • Search Google Scholar
    • Export Citation
  • Rao, S. A., S. K. Behera, Y. Masumoto, and T. Yamagata, 2002: Interannual subsurface variability in the tropical Indian Ocean with a special emphasis on the Indian Ocean Dipole. Deep-Sea Res., 49 , 15491572.

    • Search Google Scholar
    • Export Citation
  • Reverdin, G., D. Cadel, and D. Gutzler, 1986: Interannual displacement of convection and surface circulation over the equatorial Indian Ocean. Quart. J. Roy. Meteor. Soc., 112 , 4367.

    • Search Google Scholar
    • Export Citation
  • Rotstayn, L. D., B. F. Ryan, and J. J. Katzfey, 2000: A scheme for calculation of the liquid fraction in mixed-phase stratiform clouds in large-scale models. Mon. Wea. Rev., 128 , 10701088.

    • Search Google Scholar
    • Export Citation
  • Saji, N. H., and T. Yamagata, 2003: Possible impacts of Indian Ocean Dipole Mode events on global climate. Climate Res., 25 , 151169.

  • Saji, N. H., B. N. Goswami, P. N. Vinayachanfran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401 , 360363.

    • Search Google Scholar
    • Export Citation
  • Shinoda, T., M. A. Alexander, and H. H. Hendon, 2004a: Remote response of the Indian Ocean to SST variations in the tropical Pacific. J. Climate, 17 , 362372.

    • Search Google Scholar
    • Export Citation
  • Shinoda, T., H. H. Hendon, and M. A. Alexander, 2004b: Surface and subsurface dipole variability in the Indian Ocean and its relation with ENSO. Deep-Sea Res., 51 , 619635.

    • Search Google Scholar
    • Export Citation
  • Smith, N. R., 1995a: The BMRC ocean thermal analysis system. Aust. Meteor. Mag., 44 , 93100.

  • Smith, N. R., 1995b: An improved system for tropical ocean subsurface temperature analyses. J. Atmos. Oceanic Technol., 12 , 850870.

  • Smith, N. R., and G. Meyers, 1996: An evaluation of expendable bathythermograph and Tropical Atmosphere–Ocean Array data for monitoring tropical ocean variability. J. Geophys. Res., 101 , 2848928501.

    • Search Google Scholar
    • Export Citation
  • Susanto, R. D., A. L. Gordon, and Q. Zheng, 2001: Upwelling along the coast of Java and Sumatra and its relation to ENSO. Geophys. Res. Lett., 28 , 15991602.

    • Search Google Scholar
    • Export Citation
  • Udea, H., and J. Matsumoto, 2000: A possible triggering process of east–west asymmetric anomalies over the Indian Ocean in relation to 1997/1998 El Niño. J. Meteor. Soc. Japan, 78 , 803818.

    • Search Google Scholar
    • Export Citation
  • Webster, P. J., A. Moore, J. P. Loschnigg, and R. R. Leben, 1999: Coupled oceanic–atmospheric dynamics in the Indian Ocean during 1997–1998. Nature, 401 , 356360.

    • Search Google Scholar
    • Export Citation
  • Wijffels, S., and G. Meyers, 2004: An intersection of oceanic waveguides: Variability in the Indonesian Throughflow region. J. Phys. Oceanogr., 34 , 12321253.

    • Search Google Scholar
    • Export Citation
  • Xie, P., F. A. Schoot, and J. P. McCreary Jr., 2002: Structure and mechanism of South Indian Ocean climate variability. J. Climate, 15 , 864878.

    • Search Google Scholar
    • Export Citation
  • Yamagata, T., S. K. Behera, J-J. Luo, S. Masson, M. Jury, and S. A. Rao, 2004: Coupled ocean–atmosphere variability in the tropical Indian Ocean. Earth’s Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., No. 147, Amer. Geophys. Union, 189–212.

  • Yu, L. S., and M. M. Rienecker, 1999: Mechanisms for the Indian Ocean warming during the 1997–1998 El Niño. Geophys. Res. Lett., 26 , 735738.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Pattern of the first EOF of Indian Ocean SST anomalies (°C) for the (a) DJF, (b) MAM, (c) JJA, and (d) SON seasons. The EOFs have been scaled by a one-standard-deviation anomaly of the principal component.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 2.
Fig. 2.

(a) Time series of the IO1 (thick) and Niño-3.4 (thin) for SON, normalized by their respective standard deviations. (b) Same as in (a), but the IO1 is shifted backward by 1 yr.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 3.
Fig. 3.

Maps of correlation coefficient (a) between gridpoint SON rainfall and IO1, and (b) between gridpoint SON rainfall and Niño-3.4.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 4.
Fig. 4.

(first column) Anomaly patterns of zonal wind stress, (second column) rainfall, and (third column) meridional wind stress associated with a one-standard-deviation value of the time series of SST EOF1 for each season. These patterns are obtained by regressing gridpoint anomalies onto the time series and then multiplying the regression coefficient with the one-standard-deviation value of the time series.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 5.
Fig. 5.

Same as in Fig. 1, but for anomalies of 20°C isotherm depth (m).

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 6.
Fig. 6.

Maps of correlation coefficient between local SST and 20°C isotherm depth anomalies for each season.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 7.
Fig. 7.

Seasonal cycle of (a) SST (°C), (b) surface zonal wind stress (N m−2), (c) upwelling at 70-m depth (×10−3 m s−1), (d) 20° isotherm depth (m), and (e) rainfall, averaged over the latitude band between 10°S and 0°.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 8.
Fig. 8.

Anomaly patterns of SST (shaded, contour intervals are 0.2°C) and wind stress (vectors) associated with IO1 in SON at lags of −4 seasons to +4 seasons. These are obtained by regressing gridpoint anomalies of SST and surface wind stress in SON onto the time series of IO1 in SON and then multiplying the regression coefficient by a one-standard-deviation value of the time series. Positive lags indicate that IO1 leads.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 9.
Fig. 9.

Same as in Fig. 8, except for anomaly patterns of D20. Positive lags indicate that IO1 leads.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 10.
Fig. 10.

Correlation between IO1 in SON and anomalies of upper-ocean temperature and vertical velocity. Positive lags indicate that IO1 leads.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 11.
Fig. 11.

Same as in Fig. 9, except for oceanic surface heat flux anomaly. Positive lags indicate that IO1 leads.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 12.
Fig. 12.

Anomaly patterns of SST (shaded, contour intervals are 0.2°C) and wind stress (vectors) associated with Niño-3.4 in SON at lags of 0 seasons to +4 seasons. These are obtained by regressing gridpoint anomalies of SST and surface wind stress in SON onto the time series Niño-3.4 in SON and then multiplying the regression coefficient by a one-standard-deviation value of the Niño-3.4 time series. Positive lags indicate that the Niño-3.4 index leads.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 13.
Fig. 13.

Same as in Fig. 12, except for anomaly patterns of D20. Positive lags indicate that the Niño-3.4 index leads.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 14.
Fig. 14.

Pattern of (a) first and (b) second EOF of 20°C isotherm depth anomalies in the tropical Pacific and Indian Ocean domain. Shown is the one-standard-deviation anomaly (m) at each grid point obtained by multiplying the one-standard-deviation value of the associated time series with the weight at each grid point.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 15.
Fig. 15.

(a) Time series of the D20 EOF1 in the Indo-Pacific domain (solid curve) and the SON Niño-3.4 index (dashed curve). (b) Time series of D20 EOF2 in the Indo-Pacific domain and the SON IOD index.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Fig. 16.
Fig. 16.

Correlation (a) between observed D20 EOF1 time series and upper temperature anomalies along the IX1 XBT line (see Fig. 1 of Feng and Meyers 2003) and (b) between observed D20 EOF2 and the upper temperature anomalies along the same XBT line. D20 EOF1 is in quasi-equilibrium with Niño-3.4 representing the transferring of warm water (with temperature greater than 20°C) to the central eastern Pacific. EOF2 is the discharge–recharge mode with positive weights in the equatorial Pacific, so that when the time series is at its most negative, the tropical Pacific is discharged. These two patterns are similar to Fig. 3 of Meinen and McPhaden (2000).

Citation: Journal of Climate 18, 10; 10.1175/JCLI3332.1

Save
  • Allan, R., and Coauthors, 2001: Is there an Indian Ocean Dipole, and is it independent of El Niño–Southern Oscillation? CLIVAR Exchanges, No. 3, International CLIVAR Project Office, Southampton, United Kingdom, 18–21.

  • Annamalai, H., R. Murtugudde, J. Potemra, S. P. Xie, P. Liu, and B. Wang, 2003: Coupled dynamics over the Indian Ocean: Spring initiation of the zonal dipole. Deep-Sea Res., 50 , 23052330.

    • Search Google Scholar
    • Export Citation
  • Ansell, T., 2002: Air–sea interaction in the tropical Indian Ocean and links with Australian rainfall variability. Ph.D. dissertation, University of Melbourne, 272 pp.

  • Baquero-Bernal, A., M. Latif, and S. Legutke, 2002: On dipolelike variability in the tropical Indian Ocean. J. Climate, 15 , 13581369.

    • Search Google Scholar
    • Export Citation
  • Battisti, D. S., and A. C. Hirst, 1989: Interannual variability in a tropical atmosphere–ocean model: Influence of the basic state, ocean geometry, and nonlinearity. J. Atmos. Sci., 46 , 16871712.

    • Search Google Scholar
    • Export Citation
  • Behera, S. K., R. Krishnan, and T. Yamagata, 1999: Unusual ocean–atmosphere conditions in the tropical Indian Ocean during 1994. Geophys. Res. Lett., 26 , 30013004.

    • Search Google Scholar
    • Export Citation
  • Behera, S. K., P. S. Salvekar, and T. Yamagata, 2000: Simulation of interannual SST variability in the tropical Indian Ocean. J. Climate, 13 , 34873499.

    • Search Google Scholar
    • Export Citation
  • Birkett, C., R. Murtugudde, and T. Allan, 1999: Indian Ocean climate event brings floods to East Africa’s lakes and the Sudd Marsh. Geophys. Res. Lett., 26 , 10311034.

    • Search Google Scholar
    • Export Citation
  • Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev., 97 , 163172.

  • Cai, W., M. A. Collier, H. B. Gordon, and L. J. Waterman, 2003: Strong ENSO variability and a super-ENSO pair in the CSIRO coupled climate model. Mon. Wea. Rev., 131 , 11891210.

    • Search Google Scholar
    • Export Citation
  • Chambers, D. P., B. D. Tapley, and R. H. Stewart, 1999: Anomalous warming in the Indian Ocean coincident with El Niño. J. Geophys. Res., 104 , 30353047.

    • Search Google Scholar
    • Export Citation
  • Clark, C. O., P. J. Webster, and J. E. Cole, 2003: Interdecadal variability of the relationship between Indian Ocean zonal mode and east African coastal rainfall anomalies. J. Climate, 16 , 548554.

    • Search Google Scholar
    • Export Citation
  • Feng, M., and G. Meyers, 2003: Interannual variability in the tropical Indian Ocean: A two-year time scale of Indian Ocean Dipole. Deep-Sea Res., 50 , 22632284.

    • Search Google Scholar
    • Export Citation
  • Feng, M., G. Meyers, and S. Wijffels, 2001: Interannual upper ocean variability in the tropical Indian Ocean. Geophys. Res. Lett., 28 , 41514154.

    • Search Google Scholar
    • Export Citation
  • Gordon, H. B., and S. P. O’Farrell, 1997: Transient climate change in the CSIRO coupled model with dynamic sea ice. Mon. Wea. Rev., 125 , 875907.

    • Search Google Scholar
    • Export Citation
  • Gordon, H. B., and Coauthors, 2002: CSIRO Mark 3 Climate System Model. CSIRO Atmospheric Research Tech. Paper 60, 130 pp.

  • Gregory, D., and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability dependent closure. Mon. Wea. Rev., 118 , 14831506.

    • Search Google Scholar
    • Export Citation
  • Griffies, S. M., 1998: The Gent–McWilliams skew flux. J. Phys. Oceanogr., 28 , 831841.

  • Griffies, S. M., A. Gnanadesikan, R. C. Pacanowski, V. D. Larichev, J. K. Dukowicz, and R. D. Smith, 1998: Isoneutral diffusion in a z-coordinate ocean model. J. Phys. Oceanogr., 28 , 805830.

    • Search Google Scholar
    • Export Citation
  • Gualdi, S., E. Guilyardi, A. Navarra, S. Masina, and P. Delecluse, 2003: The interannual variability in the tropical Indian Ocean as simulated by a CGCM. Climate Dyn., 10 , 567582.

    • Search Google Scholar
    • Export Citation
  • Hendon, H. H., 2003: Indonesian rainfall variability: Impacts of ENSO and local air–sea interaction. J. Climate, 16 , 17751790.

  • Huang, B., and J. L. Kinter III, 2002: Interannual variability in the tropical Indian Ocean. J. Geophys. Res., 107 .3199, doi:10.1029/2001JC001278.

    • Search Google Scholar
    • Export Citation
  • Iizuka, S., T. Matsuura, and T. Yamagata, 2000: The Indian Ocean SST dipole simulated in a coupled general circulation model. Geophys. Res. Lett., 27 , 33693372.

    • Search Google Scholar
    • Export Citation
  • Jin, F-F., 1997a: An equatorial ocean recharge paradigm for ENSO. Part I: Conceptual model. J. Atmos. Sci., 54 , 811829.

  • Jin, F-F., 1997b: An equatorial ocean recharge paradigm for ENSO. Part II: Stripped-down coupled model. J. Atmos. Sci., 54 , 830847.

  • Klein, S. A., B. J. Soden, and N-C. Lau, 1999: Evidence for a tropical atmospheric bridge. J. Climate, 12 , 917932.

  • Kowalczyk, E. A., J. R. Garratt, and P. B. Krummel, 1991: A soil-canopy scheme for use in a numerical model of the atmosphere—1D stand alone model. CSIRO Division of Atmospheric Research Tech. Paper 23, 60 pp.

  • Kowalczyk, E. A., J. R. Garratt, and P. B. Krummel, 1994: Implementation of a soil-canopy scheme into the CSIRO GCM—Regional aspects of the model response. CSIRO Division of Atmospheric Research Tech. Paper 32, 59 pp.

  • Krishnamurthy, V., and B. P. Kirtman, 2003: Variability of the Indian Ocean: Relation to monsoon and ENSO. Quart. J. Roy. Meteor. Soc., 129 , 16231646.

    • Search Google Scholar
    • Export Citation
  • Latif, M., and T. P. Barnett, 1995: Interactions of tropical oceans. J. Climate, 8 , 952964.

  • Lau, N-C., and M. J. Nath, 2003: Atmosphere–ocean variations in the Indo-Pacific sector during ENSO episodes. J. Climate, 16 , 320.

  • Lau, N-C., and M. J. Nath, 2004: Coupled GCM simulation of atmosphere–ocean variability associated with zonally asymmetric SST changes in the tropical Indian Ocean. J. Climate, 17 , 245265.

    • Search Google Scholar
    • Export Citation
  • Loschnigg, J., G. A. Meehl, P. J. Webster, J. M. Arblaster, and G. P. Compo, 2003: The Asian monsoon, the tropospheric biennial oscillation, and the Indian Ocean zonal mode in NCAR CSM. J. Climate, 16 , 16171641.

    • Search Google Scholar
    • Export Citation
  • McCreary, J. P., and D. L. T. Anderson, 1984: A simple model of El Niño and the Southern Oscillation. Mon. Wea. Rev., 112 , 934946.

  • McGregor, J. L., 1993: Economical determination of departure points for semi-Lagrangian models. Mon. Wea. Rev., 121 , 221230.

  • Meehl, G. A., and J. M. Arblaster, 2002: The tropical biennial oscillation and Asian–Australian monsoon rainfall. J. Climate, 15 , 722744.

    • Search Google Scholar
    • Export Citation
  • Meinen, C. S., and M. J. McPhaden, 2000: Observations of warm water volume changes in the equatorial Pacific and their relationship to El Niño and La Niña. J. Climate, 13 , 35513559.

    • Search Google Scholar
    • Export Citation
  • Mitchell, J. F. B., R. A. Davis, W. J. Ingram, and C. A. Senior, 1995: On surface temperature, greenhouse gases, and aerosols: Models and observations. J. Climate, 8 , 23642386.

    • Search Google Scholar
    • Export Citation
  • Murtugudde, R., and A. J. Busalacchi, 1999: Interannual variability in the Indian Ocean. J. Climate, 12 , 23002326.

  • Murtugudde, R., J. P. McCreary, and A. J. Busalacchi, 2000: Oceanic processes associated with anomalous events in the Indian Ocean with relevance to 1997–1998. J. Geophys. Res., 105 , 32953306.

    • Search Google Scholar
    • Export Citation
  • O’Farrell, S. P., 1998: Investigation of the dynamic sea ice component of a coupled atmosphere sea ice general circulation model. J. Geophys. Res., 103 , C8,. 1575115782.

    • Search Google Scholar
    • Export Citation
  • Potemra, J. T., 2001: Contribution of equatorial Pacific winds to southern tropical Indian Ocean Rossby waves. J. Geophys. Res., 106 , 24072422.

    • Search Google Scholar
    • Export Citation
  • Rao, S. A., S. K. Behera, Y. Masumoto, and T. Yamagata, 2002: Interannual subsurface variability in the tropical Indian Ocean with a special emphasis on the Indian Ocean Dipole. Deep-Sea Res., 49 , 15491572.

    • Search Google Scholar
    • Export Citation
  • Reverdin, G., D. Cadel, and D. Gutzler, 1986: Interannual displacement of convection and surface circulation over the equatorial Indian Ocean. Quart. J. Roy. Meteor. Soc., 112 , 4367.

    • Search Google Scholar
    • Export Citation
  • Rotstayn, L. D., B. F. Ryan, and J. J. Katzfey, 2000: A scheme for calculation of the liquid fraction in mixed-phase stratiform clouds in large-scale models. Mon. Wea. Rev., 128 , 10701088.

    • Search Google Scholar
    • Export Citation
  • Saji, N. H., and T. Yamagata, 2003: Possible impacts of Indian Ocean Dipole Mode events on global climate. Climate Res., 25 , 151169.

  • Saji, N. H., B. N. Goswami, P. N. Vinayachanfran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401 , 360363.

    • Search Google Scholar
    • Export Citation
  • Shinoda, T., M. A. Alexander, and H. H. Hendon, 2004a: Remote response of the Indian Ocean to SST variations in the tropical Pacific. J. Climate, 17 , 362372.

    • Search Google Scholar
    • Export Citation
  • Shinoda, T., H. H. Hendon, and M. A. Alexander, 2004b: Surface and subsurface dipole variability in the Indian Ocean and its relation with ENSO. Deep-Sea Res., 51 , 619635.

    • Search Google Scholar
    • Export Citation
  • Smith, N. R., 1995a: The BMRC ocean thermal analysis system. Aust. Meteor. Mag., 44 , 93100.

  • Smith, N. R., 1995b: An improved system for tropical ocean subsurface temperature analyses. J. Atmos. Oceanic Technol., 12 , 850870.

  • Smith, N. R., and G. Meyers, 1996: An evaluation of expendable bathythermograph and Tropical Atmosphere–Ocean Array data for monitoring tropical ocean variability. J. Geophys. Res., 101 , 2848928501.

    • Search Google Scholar
    • Export Citation
  • Susanto, R. D., A. L. Gordon, and Q. Zheng, 2001: Upwelling along the coast of Java and Sumatra and its relation to ENSO. Geophys. Res. Lett., 28 , 15991602.

    • Search Google Scholar
    • Export Citation
  • Udea, H., and J. Matsumoto, 2000: A possible triggering process of east–west asymmetric anomalies over the Indian Ocean in relation to 1997/1998 El Niño. J. Meteor. Soc. Japan, 78 , 803818.

    • Search Google Scholar
    • Export Citation
  • Webster, P. J., A. Moore, J. P. Loschnigg, and R. R. Leben, 1999: Coupled oceanic–atmospheric dynamics in the Indian Ocean during 1997–1998. Nature, 401 , 356360.

    • Search Google Scholar
    • Export Citation
  • Wijffels, S., and G. Meyers, 2004: An intersection of oceanic waveguides: Variability in the Indonesian Throughflow region. J. Phys. Oceanogr., 34</