The observational analysis reveals two distinct precipitation modes, the zonal dipole (DP) mode and the monopole (MP) mode, in the tropical Indian Ocean (TIO) during the El Niño mature winter, even though sea surface temperature anomalies (SSTAs) have a similar basinwide warming pattern [referred to as the Indian Ocean basin mode (IOBM)]. The formation of the two precipitation modes depends on the distinct evolutions of the SSTA in the tropical Pacific and Indian Ocean. Both of the precipitation modes are preceded by an Indian Ocean dipole (IOD). The IOD associated with the DP mode developed in late summer and was triggered by Pacific El Niño through a “Sumatra–Philippine pattern.” The IOD associated with the MP mode developed in early summer when the Pacific SSTAs were still normal. The different IOD onset time leads to salient differences in subsequent evolution including the transfer of a dipole SST pattern to a basinwide pattern. As a result, in the boreal winter, the zonal SSTA gradient associated with the DP mode is much stronger than that associated with the MP mode. The strong SSTA zonal gradient associated with the DP mode drives an anomalous Walker circulation in the TIO, while the nearly uniform warm SSTA associated with the MP mode forces a basin-scale upward motion. The two modes have opposite impacts on the zonal wind over the equatorial western Pacific, with anomalous westerly (easterly) occurring during the DP (MP) mode, and thus they may have distinct impacts on El Niño evolution.
During ENSO mature winter, the sea surface temperature anomalies (SSTAs) in the tropical Indian Ocean (TIO) evolve to a basin-scale uniform mode, referred to as Indian Ocean basin mode (IOBM) (Klein et al. 1999; Huang and Kinter 2002; Krishnamurthy and Kirtman 2003). The IOBM is generally seen as a response to ENSO remote forcing (Klein et al. 1999; Venzke et al. 2000; Alexander et al. 2002; Lau and Nath 2003). The warm SSTA in the equatorial central-eastern Pacific stimulate an anomalous Walker circulation whose descending branch would suppress the convection over the eastern TIO and thus influence the atmospheric circulation over the TIO. The suppressed convection would increase the downward shortwave radiative flux, and the low-level wind anomalies would decrease the upward latent heat flux (Klein et al. 1999). Meanwhile, the anticyclonic wind stress anomalies would excite oceanic Rossby waves (Huang and Kinter 2002; Yu et al. 2005). When the Rossby waves propagate westward to the southwestern TIO, where the climatological thermocline is shallow, it would warm the SSTA there (Xie et al. 2002; Baquero-Bernal and Latif 2005). Therefore, both surface thermal forcing and oceanic subsurface dynamics contribute to the formation of the IOBM (Hong et al. 2010). In addition, the variation of the oceanic mixed layer depth plays a role in the TIO warming (Lau and Nath 2003).
The Indian Ocean dipole mode (IOD), which usually develops in boreal summer and reaches a peak phase in boreal fall (Saji et al. 1999; Webster et al. 1999), is another air–sea coupled mode that may influence the occurrence of the IOBM (Li et al. 2003; Hong et al. 2010). During summer and fall, the IOD, especially its eastern pole, develops and maintains through a positive “wind–evaporation–SST” feedback— the cold SSTAs suppress local convection and thus stimulate an anomalous anticyclone over the southern TIO; in turn, the southeasterly anomalies to the northeastern flank of the anticyclone, which is in the same direction with the background mean wind, further cool the SSTA by enhancing local evaporation, coastal upwelling and mixing (Li et al. 2002, 2003; Shinoda et al. 2004). The operation of the positive feedback relies on the maintenance of the mean southeasterly wind. However, with the onset of the Asian winter monsoon, the mean northwesterly gradually replaces the southeasterly and controls the coast of Sumatra. The change in the mean wind causes the positive feedback transfer to a negative feedback that damps the cold SSTA quickly (Li et al. 2003; Tokinaga and Tanimoto 2004).
On the other hand, during boreal summer and fall, the easterly anomalies over the equatorial Indian Ocean, excited by the negative heating over the southeastern TIO would stimulate oceanic downwelling Rossby waves propagating westward (Yu et al. 2005). The Rossby waves reflect to equatorial downwelling Kelvin waves in the western boundary (Yuan and Liu 2009). After the Kelvin waves reach the eastern TIO, it would help the negative SSTA to damp and transfer to the positive SSTA (Li et al. 2003; Yuan and Liu 2009). The studies above suggested that the evolution from IOD to IOBM is attributed to the combined effects of the local negative feedback and the oceanic wave dynamics.
As a dominant mode in the TIO, the IOBM may exert a great impact on the variability in the tropical Pacific. During El Niño decaying summer, with the decay of warm SSTA in the central-eastern Pacific, the IOBM plays an important role in the maintenance of the western North Pacific (WNP) anomalous anticyclone (WNPAC) (Yang et al. 2007; Li et al. 2008; Xie et al. 2009; Wu et al. 2009a, 2010; Huang et al. 2010), and thus has a strong impact on the East Asian summer monsoon.
However, whether the IOBM has a great effect on the tropical Pacific during El Niño mature winter is still a controversial issue because of the presence of strong El Niño remote forcing. Using a linearized model forced by a basin-wide warm SSTA pattern, Watanabe and Jin (2002) showed that the IOBM may induce an anomalous anticyclone in the WNP through induced basin-wide convection anomaly over the TIO. A similar experiment but with a different model was done by Annamalai et al. (2005a), who emphasized the role of SSTA over the southwestern TIO. The easterly wind stress anomalies to the southern flank of the anomalous anticyclone induced by the IOBM may stimulate oceanic upwelling Kelvin waves and favor a fast termination of the El Niño (Weisberg and Wang 1997a,b; Kim and Lau 2001; Lau and Wu 2001; Kug and Kang 2006). In contrast, Wu et al. (2009a) noted, based on observational data analyses, that during El Niño mature winter, the convection over the eastern TIO is suppressed by the descending branch of the El Niño–induced anomalous Walker circulation. As a result, the warm SSTA in the eastern TIO is a passive response to El Niño remote forcing. On the other hand, warm SSTA in the western TIO plays an active role in forcing anomalous ascending motion. As a consequence of the dipole precipitation anomaly pattern in the TIO, a westerly wind anomaly rather than an eastern anomaly should appear over the equatorial western Pacific.
A related question is whether or not the TIO variability can significantly contribute to the ENSO evolution. The clarification of the question may advance our understanding of the interactive nature of the TIO and tropical Pacific. As shown in section 3, the TIO has two distinct modes of precipitation anomalies in boreal winter. Though both the modes are associated with basin-wide warming patterns, they have opposite impacts on the zonal wind over the tropical western Pacific. In this study we will analyze the two anomalous precipitation modes, especially their formation mechanisms. We will also explore their teleconnection patterns with the tropical Pacific variability.
The rest of the paper is organized as follow. Datasets, analysis methods, and model are described in section 2. In section 3, we obtain two dominant modes of the TIO precipitation during boreal winter through an EOF analysis. The differences in the precipitation, atmospheric circulation, and SST anomalies between the two dominant modes are investigated. In section 4, we present possible mechanisms responsible for the formation of the two distinct modes. The different impacts of the two TIO precipitation modes on the atmospheric circulation over the tropical western Pacific are investigated in section 5. Section 6 summarizes the major findings.
2. Datasets, methods, and model
The datasets used in the present study consist of 1) precipitation data from the Global Precipitation Climatology Project (GPCP) (Adler et al. 2003); 2) SST data from the Met Office (UKMO) Hadley Center’s sea ice and SST dataset (HadISST) (Rayner et al. 2003); 3) atmospheric circulation and surface heat flux from the National Centers for Environment Prediction (NCEP)–Department of Energy Atmospheric Model Intercomparison Project II reanalysis (NCEP2) (Kanamitsu et al. 2002); 4) ocean temperature and circulation from the NCEP Global Ocean Data Assimilation System (GODAS), which is forced by the momentum flux, heat flux, and freshwater flux from the NCEP2 (Behringer and Xue 2004). All the datasets cover the period of 1979–2008, except for the GODAS data, which are available over the period of 1980–2008.
In addition, the following datasets are used for the comparison with the above datasets: 1) National Oceanic and Atmospheric Administration (NOAA) interpolated outgoing longwave radiation (OLR) (Liebmann and Smith 1996); 2) the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) data (Xie and Arkin 1997); 3) the surface heat flux in the reanalysis data from the Japanese long-term reanalysis cooperative research project (JRA-25) carried out by the Japan Meteorological Agency and the Central Research Institute of Electric Power Industry (Onogi et al. 2007); 4) ocean temperature and current derived from the Simple Ocean Data Assimilation (SODA) (Carton et al. 2000); and 5) the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data (Kalnay et al. 1996).
Following previous studies, IOD and Niño-3.4 indexes are used to depict the IOD and ENSO. The IOD index is defined as the difference of the area-averaged SSTA between western (10°S–10°N, 50°–70°E) and eastern TIO (10°S–0°, 90°–110°E) in fall [September–November (SON)], and the Niño-3.4 index is defined as the area-averaged SSTA in the equatorial central-eastern Pacific (5°S–5°N, 170°–120°W) in winter [December–February (DJF)].
Because anomalous precipitation and associated diabatic heating, rather than the SSTA, have direct impacts on the atmospheric circulation, we first analyze the two dominant EOF modes of the DJF mean precipitation anomalies over the TIO (15°S–15°N, 40°–110°E). Corresponding atmospheric and oceanic anomalies are then obtained through regressions against the time series of normalized principle components (PCs). It is worth mentioning that the two distinct dominant patterns cannot be derived from the DJF SSTA field in the TIO.
The robustness of the EOF results is carefully examined. First, we test the credibility of the precipitation data by applying the same EOF analysis to the OLR data or CMAP precipitation data. The obtained EOF spatial patterns are generally similar to that derived from the GPCP data, indicating that the results are not sensitive to the data used. Second, we apply EOF analysis to 500-hPa omega data for the period of 1950–2009 (NCEP–NCAR reanalysis data). The obtained EOF spatial pattern is also analogous to that derived from the GPCP data, suggesting that the results are also insensitive to the analysis period. Third, we test the credibility of the EOF analysis method by comparing with a composite analysis. The years, during which the normalized PC values are greater than 1.0 or less than −1.0 standard deviations, are selected to make the composite analysis. The spatial pattern derived from the composite is generally consistent with that derived from the EOF analysis, indicating that the two dominant modes are realistic. Because our focus is on the interannual variability, variances longer than 8 years are filtered out through a Lanczos filter prior to the analysis (Duchon 1979).
In the study, a dry version of the Princeton AGCM (Held and Suarez 1994) is used to assess the atmospheric responses to specific heating over the TIO. Following previous studies (Ting and Yu 1998; Jiang and Li 2005; Li 2006), the model is linearized by specified realistic 3D winter-mean (December–February) wind and temperature fields. Based on the linearization, we can test the model response to specific anomalous heating under the realistic basic mean state. The model use a sigma vertical coordinate with 5 evenly spaced levels from the top of the atmosphere (σ = 0) to the surface (σ = 1). It has a horizontal resolution of T42, equivalent to 2.8° × 2.8°. A more detailed description of the model can be found in Jiang and Li (2005).
3. Two dominant winter precipitation patterns and associated large-scale circulations
The EOF analysis is applied to the DJF precipitation anomaly field. The Rule N test based on the Monte Carlo experiments is used to select the meaningful EOF mode (Preisendorfer 1988; Li et al. 2000). The first three modes pass the 5% significance level. The first and second modes account for 25.2% and 14.4% of total variances, respectively. According to the North role (North et al. 1982), the two modes are well distinguished from each other and from other modes.
The spatial patterns of the first two leading modes are shown in Figs. 1a,b. The TIO precipitation anomalies in the EOF1 exhibit a dipole pattern, with a negative center in the southeastern TIO and a positive center in the western TIO. In contrast, the TIO precipitation anomalies in the EOF2 generally exhibit a monopole pattern. According to their distinct spatial patterns, the EOF1 and EOF2 are referred to as the dipole precipitation (DP) mode and the monopole precipitation (MP) mode, respectively.
The time series of principal components (PCs) of the two modes are shown in Fig. 1e. The power spectrum densities and red noises of the PCs are shown in Fig. 2. The PC1 has a major spectral peak in 2–3 yr and a secondary peak in 4–5 yr, while the PC2 has a major spectral peak in 5 yr and a secondary peak in 2–3 yr.
Though the precipitation anomalies in the TIO are distinct, both the modes correspond to a similar basinwide warming in the TIO at a first glance (Figs. 1c,d). However, in fact, the spatial structures of the IOBM in the two modes also have differences. For the MP mode, the warm SSTA in the TIO is quite uniform, while the SSTA associated with the DP mode shows a strong zonal gradient, with the warm SSTA in the western TIO being much stronger than that in the eastern TIO.
The different SSTA patterns correspond to different zonal atmospheric circulation anomalies. Figure 3 shows 200-hPa velocity potential. For the DP mode, a divergence over the tropical central-eastern Pacific connects a divergence to its west, whose center is located over the eastern TIO, Maritime Continent and tropical western Pacific (Fig. 3a). Almost the entire TIO is covered by the convergence, except for the far western TIO, indicating that the vertical motion over the TIO is suppressed by the descending branch of the anomalous Walker circulation excited by the anomalous ascent motion over the equatorial central-eastern Pacific (Chou 2004; Ham et al. 2007; Wu et al. 2009a). For the MP mode, a well-organized zonal tripole structure is seen over the tropical Pacific and TIO (Fig. 3b). Two divergences over the TIO and central-eastern Pacific grip a convergence over the tropical western Pacific and Maritime Continent. The divergence over the TIO is stronger and more significant than that over the central Pacific, implying that the subsidence over the tropical western Pacific links to the ascending motion over the TIO more tightly.
The main difference of 200-hPa velocity potential between the two modes is that the DP mode corresponds to the upper-tropospheric convergence over the TIO and divergence over the central Pacific, while the MP mode corresponds to the divergence over the TIO and convergence over the western Pacific. The difference implies a different Indian Ocean–Pacific relationship between the two modes, which will be further explored in section 5.
4. Formation of the DP and MP modes
In the section, mechanisms responsible for the formations of the two distinct precipitation modes are explored. As shown in the previous section, both of the modes correspond to the IOBM, which may be caused by ENSO remote forcing, or evolved from preceding IOD event as presented in the introduction. To explore the relationships between the two modes and IOD or ENSO, the correlations between PCs and IOD (Niño-3.4) indexes are calculated. The PC1 is more correlated with Niño-3.4 than the PC2, with the coefficients being 0.59 and 0.38, while the situation is opposite for the correlations with IOD index (coefficients being 0.44 and 0.62). The correlation analysis suggests that DP (MP) mode is associated with ENSO (IOD) more tightly.
a. Early evolution
To explore the formation processes of the two modes, we first investigate their early evolution characteristics through a lag-regression analysis. The complete evolutions of the SST, 925-hPa wind, and precipitations for the two modes from preceding June to December are shown in Fig. 4 and Fig. 5.
The difference is evident in the early summer. For the DP mode, canonical El Niño–like warm SSTAs are apparent in June, while no significant cold SSTAs are seen in the TIO (Fig. 4a). On the contrary, for the MP mode, dipole SSTAs have developed in the TIO, with cold SSTAs located in the coast of Sumatra and warm SSTAs in the eastern coast of Africa, while El Niño has not established yet (Fig. 4h). In July, the El Niño associated with the DP mode and the IOD associated with the MP mode gradually develop (Figs. 4b,i). Meanwhile, for both the modes, the positive precipitation anomalies over the WNP, the negative precipitation anomalies over the southeastern TIO, and the northward cross-equatorial low-level wind linking the two opposite precipitation anomalies form a classical antisymmetric Gill pattern, referred to as the “Sumatra–Philippine pattern” (Figs. 5b,i, Li et al. 2002, 2006; Wu et al. 2009b).
The driving forces to the Sumatra–Philippine patterns associated with the DP mode are different from that associated with the MP mode. The former is attributed to the anomalous heating over the WNP monsoon region associated with the El Niño warming. The latter is attributed to the anomalous heating over the southeastern TIO associated with the IOD. Because of that, their spatial structures present some subtle differences. For example, for the DP mode, the negative precipitation anomalies cover the entire equatorial Indian Ocean, Maritime Continent, and equatorial western Pacific in June and gradually move southward in July (Figs. 5a,b). In contrast, for the MP mode, the negative precipitation anomaly is centered at about 10°S, consistent with the area of the underlying negative SSTA (Fig. 5h). It originates from the southeastern TIO and gradually extends eastward (Figs. 5h,i).
Specific processes associated with the Sumatra–Philippine patterns are discussed as follows. For the DP mode, the pattern is primarily attributed to the warm SSTA in the equatorial central-eastern Pacific (Figs. 4a,b). The warm SSTA enhances the local convection, and thus stimulates twin Rossby wave–like anomalous cyclones symmetric about the equator to its west and an anomalous Walker circulation. The anomalous cyclone over Northern Hemisphere enhances the WNP summer monsoon and thus increases the precipitation there (Wu et al. 2009b). Meanwhile, the subsidence branch of the anomalous Walker circulation suppressed the convection over the southeastern TIO and Maritime Continent (Figs. 5a,b). In contrast, the Sumatra–Philippine pattern associated with the MP mode mainly links to the cold SSTA in the eastern TIO (Figs. 4h,i), which significantly suppresses local convection (Figs. 5h,i). The suppressed convection stimulates a Kelvin wave–like westerly anomaly to its east (Annamalai et al. 2005b) and enhances the convection over the central-western Pacific through an anomalous Walker circulation (Li et al. 2006) or Kelvin wave–induced boundary layer divergence (Wu et al. 2010).
In August and September (AS), the differences of the precipitation, SST, and circulation between the two modes become more salient. For the MP mode, the IOD further intensifies (Figs. 4j,k). The dipole SSTA stimulates dipole precipitation anomalies and equatorial easterly anomalies (Figs. 5j,k). Compared with the MP mode, the TIO air–sea coupled system associated with the DP mode develops much slower. The cold SSTA in the southeastern TIO just slightly strengthens and SSTA is hardly seen in the western TIO (Figs. 4c,d). Though the negative precipitation anomalies in the eastern equatorial Indian Ocean extend westward with the enhancement of the El Niño remote forcing, the low-level wind anomalies over the TIO are still scattered (Figs. 5c,d).
In October, the IOD associated with the MP mode reaches peak phase, especially the warm SSTA of the western pole being much stronger than the preceding AS (Fig. 4l). The warm SSTA center is located in the southwestern TIO, where the climatological thermocline is shallow at the time. The SSTA results from the combined effects of the oceanic Rossby wave (Xie et al. 2002; Du et al. 2009) and surface latent heat flux (Wu and Yeh 2010). Correspondingly, the anomalous precipitation, equatorial easterly, and southern TIO anomalous anticyclone (SIOAC) are much stronger than that seen in the AS (Fig. 5l). For the DP mode, with the warm SSTA developing in the far western TIO, an IOD-like SSTA pattern establishes (Fig. 4e). Correspondingly, the dipole precipitation pattern, equatorial easterly anomalies and SIOAC also form (Fig. 5e). However, it is worth noting that the spatial structure of the IOD associated with the DP mode is different from that associated with the MP mode. For the DP mode, the warm SSTA of the western pole is mainly located in the far northwestern TIO and the coast of Africa, so that the zonal distance between the two poles of the IOD is much larger than that associated with the MP mode. Correspondingly, the positive precipitation anomalies in the western TIO are shifted westward, and the negative precipitation anomalies in the eastern TIO extend westward. The equatorial easterly anomalies extend westward 20° farther than that associated with the MP mode, and the SIOAC is also shifted westward.
The differences in the southwestern TIO warm SSTA between the two modes are associated with oceanic dynamic processes. Figure 6 shows the time–latitude distributions of the 1000-hPa wind anomaly averaged over the 5°N–5°S, and the sea surface height (SSH) anomaly averaged over the 6°–8°S. For the MP mode, the equatorial easterly anomalies over the TIO are seen in the early summer. Anticyclonic vorticity anomalies associated with the easterly anomalies stimulate strong oceanic downwelling Rossby waves propagating westward. When the Rossby waves reach the southwestern TIO where the climatological thermocline is shallow all year around, they cause the SST warm through deepening the thermocline (Xie et al. 2002). Compared with the MP mode, the equatorial easterly anomaly associated with the DP mode forms much later, due to the later establishment of the IOD. Correspondingly, the oceanic Rossby waves onset two-months later, and reach maximum magnitude one or two months later than that associated with the MP mode. During the fall (October–November), the oceanic thermocline anomaly associated with the DP mode, represented by the SSH anomaly, is much weaker than that associated with the MP mode, so that the former SSTA in the southwestern TIO from 60°–90°E is much weaker than the latter (Figs. 4e,f,l,m).
During the early evolution stage, the two modes show salient differences, but how these differences determine the establishment of the distinct precipitation anomalies in DJF is still unknown. In the next subsection, we will resolve this issue.
b. Transition from IOD to IOBM
In this subsection, we analyze the difference in the transition processes from IOD to IOBM between the two modes. Figure 7 shows the evolution of the SSTA in the eastern TIO (20°S–0°, 90°–110°E). It indicates that the cold SSTA associated with the MP mode decays one month earlier than that associated with the DP mode, with the former starting in October, but the latter starting in November. The prior warming of the eastern TIO SSTA causes the IOBM to establish earlier and the wintertime zonal gradients of the TIO SSTA to be weaker for the MP mode.
October–November (ON) is a key stage for the occurrence of the different evolutions of the eastern TIO SSTA between the DP and MP mode, with the latter starting to decay and the former continuing to develop. To understand the opposite SSTA tendency, we diagnose the oceanic mixed layer heat budget in the ON for the two modes. Following previous studies, the mixed layer temperature tendency equation is written as
where T denotes the mixed layer temperature; V and w denote horizontal and vertical velocity of the ocean current, which are defined as the vertical average of the entire mixed layer; ∇ and denote the horizontal and vertical gradient operator, respectively; prime and bar in the equation denote climatology and anomaly variables, respectively; QSW, QLW, QLH, and QSH denote net downward shortwave radiative flux, downward longwave radiative flux, latent heat flux, and sensible heat flux, respectively; and ρ (1026 kg m−3), Cp (3986 J kg K−1), and H denote density of seawater, specific heat of water, and mixed layer depth, respectively. The mixed layer depth is defined as the depth where the temperature deviation from the surface temperature is less than 0.1°C. The equation means that the temperature tendency of the mixed layer is associated with the horizontal advection, vertical advection, and surface heat flux. Both the horizontal and vertical advection terms are the sum of the linear (first two terms in each bracket) and nonlinear terms (last term in each bracket). The results show that the mixed layer temperature tendencies are highly consistent with the sum of the temperature advection and net surface heat flux (Figs. 8a,b), suggesting that the analysis is valid here.
Budget analysis indicates that although the opposite SSTA tendencies between the two modes are seen in both the equatorial (8°S–0°, 90°–110°E) and off-equatorial southeastern Indian Ocean (20°–8°S, 90°–110°E) (Figs. 8a,b), they are caused by different mechanisms.
At the equatorial zone, the difference in the temperature tendency between the two modes is primarily attributed to the zonal temperature advection. Thus the zonal advection has a strong warming effect on the SSTA associated with the MP mode, but it is generally negligible for the DP mode (Fig. 8c). The difference in the zonal advection is primarily contributed by .
As noted in section 4a, the location of the IOD warm pole has remarkable difference between the two modes. For the MP mode, the warm pole center is located in the central-western TIO, about 75°E, and thus forms a strong temperature gradient with the cold pole (Fig. 9b). The mean South Equatorial Countercurrent and South Java current efficiently transport the anomalous warm water to the cold pole and damp it. In contrast, for the DP mode, the IOD warm pole is shifted much westward, with center located in the coast of the eastern Africa (Fig. 9a). Thus the anomalous temperature gradient in the central-eastern TIO is quite weak, so that the mean current has little impact on the eastern TIO SSTA.
For the off-equatorial region, the difference in the temperature tendency is primarily caused by the surface heat flux anomalies (Fig. 8d). The difference in the surface heat flux is attributed to the latent heat flux. For the MP mode, the off-equatorial southeastern Indian Ocean is controlled by the positive downward latent heat flux anomalies (Fig. 10d), which tend to warm the SSTA in situ, while the latent heat flux associated with the DP mode is very weak there (Fig. 10c).
The latent heat flux is associated with the surface wind speed and the difference between the surface air specific humidity and saturated specific humidity at SST. Here the remarkable difference in the latent heat flux is primarily attributed to the difference in the surface wind speed anomalies. As noted in section 4a, the SIOAC associated with the DP mode is located to the west of that associated with the MP mode. Correspondingly, the northwesterly anomaly in the southern flank of the anomalous anticyclone is shifted more westward, and thus has little impact on the off-equatorial southeastern Indian Ocean. In the ON, the off-equatorial southeastern Indian Ocean is dominated by the mean southwesterly (Figure not shown). For the MP mode, the northwesterly anomaly significantly reduces the wind speed (Fig. 10b) and thus enhances the downward latent heat flux (Fig. 10d). On the contrary, for the DP mode, both the wind speed and the latent heat flux anomalies are very week (Figs. 10a,c).
Above analysis shows that both the processes responsible for the opposite SSTA tendencies in the southeastern TIO are closely associated with differences in the spatial structures of the SSTA and corresponding anomalous wind fields. It indicates that the distinct early evolutions of the IOD play an essential role in the formation of the two types of the IOBM.
5. Opposite effects of the two modes on the equatorial western Pacific wind
The low-level wind anomalies over the equatorial western Pacific associated with the two modes have salient differences (Fig. 11). The low-level easterly anomalies to the east of the DP mode are restricted to the west of 145°E, while to the east of the MP mode they expand eastward to 160°E. Meanwhile, the WNPAC associated with the MP mode also greatly shifts westward. These differences can be partially explained by the distinct TIO anomalous precipitation patterns in the two modes. Previous studies noted that the WNPAC and the easterly anomalies to its southern flank are primarily maintained by air–sea interaction in the WNP and El Niño remote forcing (Wang et al. 2000; Li et al. 2006; among many others). However, in terms of the classical Gill response, the negative heating in the eastern TIO (DP mode) tends to reduce the easterly anomalies to its east, while the positive heating (MP mode) tends to enhance them.
To intentionally exclude other impacts and clearly demonstrate the effect of the TIO forcing, we conduct a series of idealized numerical experiments using the Geophysical Fluid Dynamics Laboratory (GFDL) dry anomaly model.
Two anomalous heating fields, which have an idealized vertical profile with a maximum at 300 hPa, are constructed according to the horizontal distributions of the two TIO precipitation patterns (shown in Fig. 1). Two experiments, RUN1 and RUN2, are conducted by forcing the model with the two specified heating fields, respectively. The heating rate at the 300 hPa and low-level wind response of the model are shown in Fig. 12. The model reproduces the SIOAC and the strong easterly anomaly over the equatorial Indian Ocean in the RUN1. It also reproduces the southward cross-equatorial winds in the western TIO and the intersection of the easterly and westerly in the equatorial central Indian Ocean in the RUN2. The reasonable simulations of the model in the TIO give us confidence to further examine the response of the equatorial western Pacific to the specified heating fields.
As expected, a westerly extending from the Maritime Continent to the central Pacific is seen in RUN1, while an easterly blowing from the central Pacific to the central TIO is seen in RUN2. The nearly opposite model responses indicate that the two dominant precipitation modes in the TIO, although being in association with a similar basinwide SSTA warming pattern, exert significantly different impacts on the wind over the equatorial western Pacific, and thus may modulate the El Niño evolution differently.
6. Conclusions and discussion
In the study, we obtain two distinct modes of precipitation anomalies over the TIO during boreal winter. They account for 25.2% and 14.4% of total variance, respectively. Though both the modes correspond to underlying basinwide warming, they have opposite impacts on the atmospheric circulation over the equatorial western Pacific. The mechanisms responsible for the formation of the two modes are explored. The major findings are summarized below.
The first mode of the TIO precipitation anomalies is a zonal dipole mode (DP mode), that is, the precipitation anomalies over the southeastern TIO are opposite to the western TIO. In contrast, the second mode is a monopole mode (MP mode), that is, precipitation anomalies are quite uniform over the entire TIO.
The formation of the two distinct modes is closely related to the underlying SSTA pattern. The TIO SSTA zonal gradient associated with the DP mode is much large than that associated with the MP mode during boreal winter. The former corresponds to an anomalous Walker circulation in the TIO, with an upward motion in the west and a downward motion in the east, while the latter forces a basin-scale upward motion anomaly.
The difference in the TIO SSTA patterns between the two modes is closely associated with their distinct early evolutions. During the preceding summer and fall, though the El Niño associated with the DP mode is much stronger than that associated with the MP mode, the IOD event associated with the MP mode establishes much earlier. Correspondingly, the well-organized easterly anomalies over the equatorial Indian Ocean establish much earlier for the MP mode. The anticyclonic voriticy anomalies associated with the easterly anomalies stimulate oceanic Rossby waves, which propagate westward and warm the SST in the southwestern TIO, where climatological thermocline is shallow. As a result, during ON, the key stage for the formations of the two modes, the warm pole center of the IOD preceding the MP mode is located in the southwestern TIO, while that preceding the DP mode is shifted to the far western TIO and the coast of the East Africa. Correspondingly, the SIOAC associated with the DP mode is shifted westward relative to that associated with the MP mode.
During ON, the difference in the IOD spatial pattern results in an opposite SSTA tendencies in the southeastern TIO, with the negative SSTA associated with the MP mode starting to decay, while that associated with the DP mode still enhancing. Two mechanisms responsible for the difference operate in the equatorial and off-equatorial zones, respectively. For the equatorial zone, since the warm pole center of the IOD associated with the MP mode is located to the east of that associated with the DP mode, the former zonal temperature gradient in the central-eastern TIO is much larger than latter. Therefore, the mean South Equatorial Countercurrent transports the warm water eastward much more effectively for the MP mode. In the off-equatorial zone, for the MP mode, the northeasterly anomalies to the southern flank of the SIOAC reduce the wind speed and thus suppress the upward latent heat flux anomalies. However, for the DP mode, due to the westward shift of the anticyclone, the latent heat flux in the southeastern TIO is normal.
The distinct TIO precipitation anomalies have opposite impacts on the atmospheric circulation anomalies over the equatorial western Pacific during ENSO mature winter. Idealized numerical experiments show that the dipole (monopole) precipitation pattern tends to force westerly (easterly) anomalies over the equatorial western Pacific. The result suggests that caution is needed in interpreting the IOBM effect on the anomalous wind over the equatorial western Pacific. Both the SST and precipitation anomaly patterns need to be carefully examined to determine the active or passive role of the SSTA.
In this study, some interesting aspects of TIO–Pacific interactions are noted. To what extent the Pacific SSTA affect TIO variability and the TIO feeds back to the Pacific deserve further investigations.
The differences in the surface wind over the equatorial western Pacific between the two modes may further influence SSTA evolution in the equatorial central-eastern Pacific. For the MP mode, the IOD-related SSTA establishes prior to the SSTA in the equatorial central-eastern Pacific (Figs. 4h,i). The negative pole of the IOD suppresses local convection and thus stimulates Kelvin wave–like westerly anomalies to its east, extending to the equatorial central Pacific. The westerly anomalies would contribute to the onset and development of ENSO, suggesting that IOD may be one of triggering factors of ENSO, if it established prior to ENSO (Saji and Yamagata 2003).
The DP and MP modes would have opposite contributions to ENSO phase transition through the opposite impacts on the low-level wind anomalies over the equatorial western Pacific during boreal winter. In terms of a lag-regression analysis, the SSTA in the equatorial central-eastern Pacific associated with the MP mode tend to transform to an opposite phase earlier than that associated with the DP mode (Figure not shown). The stronger easterly wind anomalies to the east of the MP mode may stimulate stronger oceanic upwelling Kelvin waves, which propagate eastward and accelerate El Niño decaying.
Though the EOF and lead–lag analyses indicate that the TIO variability associated with the MP mode tends to influence the low-level wind variability over the tropical Pacific, caution is needed to interpret the relationship between the MP mode and ENSO. In addition to remote forcing from the Indian Ocean, ENSO is primarily modulated by a variety of air–sea interaction processes in the Pacific. For example, three typical MP mode years, 1980, 1982, and 1994, have different TIO–Pacific relationships. They all have a strong IOD event, but only 1982 has a strong ENSO event. 1980 is a normal year, while 1994 is a weak El Niño year (Meyers et al. 2007). This is consistent with the fact that the MP mode only has weak correlation with Niño-3.4 index (section 4a).
We found that the MP modes are closely linked to the preceding IOD. However, the initiation of the IOD is still unknown. Our preliminary analysis indicates that the IOD associated with the MP mode is trigged by a shoaling of the thermocline in the eastern equatorial Indian Ocean. In the preceding year of the IOD initiation, Rossby waves in the southern off-equatorial Indian Ocean propagate westward and reflect to Kelvin waves in the East African coast, which propagate eastward and shoal the thermocline in the eastern tropical Indian Ocean (Rao et al. 2009). The mechanism responsible for the IOD initiation deserves further studies.
Major conclusions from this study are obtained from the observational analysis. They should be further tested and explored by numerical experiments. We analyzed twentieth-century experiments (20C3M) of the some coupled general circulation models that participated in phase 3 of the Coupled Model Intercomparison Project (CMIP3). All runs are integrated from the 1850 to the present. It is found that the UKMO Hadley Centre general circulation model version 1 (UKMO-HADGEM1) reproduces the DP and MP modes analogous to the observation (figure not shown). Since the 150-yr model output provides much larger sample sets, we plan to analyze the long integration result to reveal statistically significant features. We will also diagnose models that failed to reproduce the two distinct precipitation modes. In addition, idealized numerical experiments may be conducted to investigate specific processes responsible for the remote El Niño impact on the TIO and the effect of the TIO SSTA on the Pacific SST and wind evolution.
This work was supported by NSFC Grant 41005040, National Program on Key Basic Research Project (2010CB951904), and National High-Tech Research and Development Plan of China (2010AA012302). TL was supported by ONR Grants N000140810256 and N000141010774 and by the International Pacific Research Center that is sponsored by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC), NASA (NNX07AG53G), and NOAA (NA17RJ1230).
School of Ocean and Earth Science and Technology Contribution Number 8523 and the International Pacific Research Center Contribution Number 831.