Abstract

This study examines a mechanism of the interaction between the tropical Atlantic meridional and equatorial modes. To derive robust heat content (HC) variability, the ensemble-mean HC anomalies (HCA) of six state-of-the-art global ocean reanalyses for 1979–2007 are analyzed. Compared with previous studies, characteristic oceanic processes are distinguished through their dominant time scales. Using the ensemble empirical mode decomposition (EEMD) method, the HC fields are first decomposed into components with different time scales. The authors’ analysis shows that these components are associated with distinctive ocean dynamics. The high-frequency (first three) components can be characterized as the equatorial modes, whereas the low-frequency (the fifth and sixth) components are featured as the meridional modes. In between, the fourth component on the time scale of 3–4 yr demonstrates “mixed” characteristics of the meridional and equatorial modes because of an active transition from the predominant meridional to zonal structures on this time scale. Physically, this transition process is initiated by the discharge of the off-equatorial HCA, which is first accumulated as a part of the meridional mode, into the equatorial waveguide, which is triggered by the breakdown of the equilibrium between the cross-equatorial HC contrast and the overlying wind forcing, and results in a major heat transport through the equatorial waveguide into the southeastern tropical Atlantic. It is also shown that remote forcing from El Niño–Southern Oscillation (ENSO) exerts important influence on the transition from the equatorial to meridional mode and may partly dictate its time scale of 3–4 yr. Therefore, the authors’ results demonstrate another mechanism of the equatorial Atlantic response to the ENSO forcing.

1. Introduction

Previous studies have identified two distinct modes in the tropical Atlantic variability (TAV): the equatorial mode and the meridional mode. The equatorial mode has been recognized as resulting from dynamical air–sea interaction, in which the displacement of the equatorial thermocline and propagation of the oceanic waves play a vital role (e.g., Handoh and Bigg 2000), much like its counterpart, El Niño–Southern Oscillation (ENSO) in the Pacific (Zebiak 1993; Carton and Huang 1994; Huang and Shukla 1997; Xie and Carton 2004). The Atlantic meridional gradient mode is mainly associated with the SST meridional variability generated by thermodynamic air–sea feedback among the surface wind speed, evaporation, and SST (WES feedback; Chang et al. 1997; Xie 1999). Previous studies have also indicated that there are substantial off-equatorial subsurface heat content (HC) changes forced by anomalous wind curl associated with the meridional mode (Huang and Shukla 1997; Ruiz-Barradas et al. 2000; Joyce et al. 2004; Lee and Wang 2008; Doi et al. 2009, 2010). The role of the off-equatorial HC anomalies (HCA) in the TAV, however, is not yet clear. Some studies suggested that their effects on SST changes might be negligible (e.g., Carton et al. 1996; Seager et al. 2001).

Studies also found that the two modes do not evolve independently and are linked together. Based on observational indices and those from a forced oceanic general circulation model (OGCM) simulation, Servain et al. (1999, 2000) suggested that the meridional mode is linked to the equatorial mode almost simultaneously at both decadal and short-interannual (1–2 yr) time scales because both are related to the anomalous displacements of the ITCZ. On the other hand, the relationship between the two modes may be changed before and after the global-scale climate shift in late 1970s (Murtugudde et al. 2001). As an extension of Servain et al. (1999, 2000), Foltz and McPhaden (2010), by a theoretical study, further emphasized the importance of the delayed negative feedback from western boundary reflections of the Rossby waves generated by meridional mode-induced equatorial wind stress anomalies in the interaction between the meridional and equatorial modes. Analyzing observational data, Chiang et al. (2002) showed that the SST warming over the tropical North Atlantic induces southerly cross-equatorial winds and the associated easterly anomalies near the equator tend to force a cold event of Atlantic Niña. Enfield and Mayer (1997) also hinted at the similar sequence of development. A common assumption of the above studies is that the zonal mode is triggered by the equatorial zonal wind anomalies, which are due to the meridional shift of the ITCZ. Essentially, the link between the zonal and meridional modes is through the atmosphere. On the other hand, Huang et al. (1995) and Huang and Shukla (1997) proposed that the off-equatorial subsurface heat content changes generated by the off-equatorial wind curl may propagate into the equatorial waveguide, stimulate the equatorial mode, and thus generate SST anomalies (SSTAs) in the equatorial and southeastern Atlantic later on.

This study further explores the interaction between the meridional and equatorial modes by examining the role of ocean dynamics using observation-based ocean reanalyses. According to our previous study (Zhu et al. 2012), current ocean reanalysis systems contain considerable uncertainty in estimating the subsurface oceanic state in the tropical Atlantic Ocean, which prompted us to utilize an ensemble mean (EM) of multiple ocean reanalyses to estimate upper-ocean HCA (defined as the averaged temperature anomalies in the upper 300 m) over the tropical Atlantic. Taking advantage of the fact that the TAV is composed of variations on multiple time scales (Huang and Shukla 1997; Ruiz-Barradas et al. 2000; Ayina and Servain 2003; black solid and dashed curves in Fig. 1) and that the zonal processes, being centered within the equatorial waveguide, generally has shorter time scales than the meridional processes, we distinguish these different processes based on their time scales and analyze them separately. Therefore, the HC dataset is first decomposed objectively by a sophisticated technique of time–frequency analysis and examine the different components systematically. Our results will not only show the characteristics of the “pure” zonal and meridional modes but also demonstrate that there is a component with the 3–4-yr time scale, in which the Atlantic meridional and equatorial zonal modes coexist, and can interact with each other. The analysis procedure used by Plaut and Vautard (1994), Moron et al. (1998), and Huang et al. (2012a) is further applied to explore the mechanism behind the evolution within a typical life cycle of this particular component.

Fig. 1.

Power spectra for the PC1 time series of first six EEMD components and the PC1/PC2 time series of the raw EM HCA fields as shown in Fig. 4 of Zhu et al. (2012).

Fig. 1.

Power spectra for the PC1 time series of first six EEMD components and the PC1/PC2 time series of the raw EM HCA fields as shown in Fig. 4 of Zhu et al. (2012).

The paper is arranged as follows: The datasets and analysis methods are described in the next section. Section 3 describes the tropical Atlantic HC components with respect to their different characteristic time scales. Section 4 puts emphasis on a component with a 3–4-yr time scale, on which both the Atlantic meridional and equatorial zonal modes coexist. The summary and discussion are given in section 5.

2. Datasets and analysis methods

Zhu et al. (2012) found that current ocean reanalysis systems contain considerable uncertainty in estimating the subsurface oceanic state, particularly in the tropical Atlantic Ocean. To improve the signal-to-noise ratio, they made an ensemble average from all HCAs to produce an ensemble-mean dataset (referred to as the EM analysis) for the 29-yr period of 1979–2007 based on six ocean reanalysis products, including Ocean Reanalysis, version S3 (ORA-S3; Balmaseda et al. 2008) and a NEMOVAR (NV)-based ocean reanalysis for the FP7 Comprehensive Modelling of the Earth System for Better Climate Prediction and Projection (COMBINE) project (http://www.combine-project.eu/) COMBINE-NV (Balmaseda et al. 2010) from the European Center for Medium-Range Weather Forecasts (ECMWF); Global Ocean Data Assimilation System (GODAS; Behringer 2007) and the Climate Forecast System Reanalysis CFSR (Saha et al. 2010) from the National Centers for Environmental Prediction (NCEP); Simple Ocean Data Assimilation (SODA), jointly generated by University of Maryland and Texas A&M University (Carton and Giese 2008); and an ensemble coupled data assimilation (ECDA) (Zhang et al. 2007) from the Geophysical Fluid Dynamics Laboratory (GFDL).

The analysis by Zhu et al. (2012) further demonstrated that the leading modes of the empirical orthogonal functions (EOF) derived from the EM HCA analysis are closely associated with the well-known surface patterns of the meridional and zonal modes, much more consistent than those derived from the HCAs of individual reanalyses, which show large discrepancies among each other. Projecting the individual analyses on to the EM EOF modes, we found that these variations exist in most of the ocean reanalyses with comparable magnitudes, which confirms that they are most likely the signals injected by data insertion and surface forcing through the assimilation system. The quality of the EM analysis is further validated more favorably against Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO) altimetry sea level anomay (SLA) data and Prediction and Research Moored Array in the Tropical Atlantic (PIRATA) mooring station data. Based on these facts, the same EM HC data are used in study.

Considering that the TAV is composed of variations on multiple time scales (Huang and Shukla 1997; Ruiz-Barradas et al. 2000; Ayina and Servain 2003; black solid and dashed curves in Fig. 1), the EM HCA dataset is first decomposed into components based on their time scales. In this study, the ensemble empirical mode decomposition (EEMD; Wu and Huang 2009) method is applied. Unlike the Fourier transform-based time series analysis, which uses a priori global basis functions of rigid periods, the method of empirical mode decomposition (EMD) decomposes a complicated time series into a relatively small collection of empirically determined intrinsic mode functions (IMFs) based on the local characteristic time scale of the data, usually leading to easier physical interpretations (Huang et al. 1998). More recently, the EEMD technique adds multiple noise realizations to the single observed time series before the decomposition, mimicking a scenario of multiple experimental trials with inherent uncertainty. The ensemble average is used to derive the corresponding IMFs (Wu and Huang 2009). It has been shown that the EEMD extracts scale-consistent signals and makes the IMFs less dependent on the sample size. More detailed descriptions of the method can be found in Huang and Wu (2008) and Wu and Huang (2009).

Extending the EEMD technique to the spatiotemporal fields of the HCA, we follow the procedure of the pseudo-bi-dimensional EMD (pseudoBEMD) described in Wu et al. (2009). The pseudoBEMD conducts a one-dimensional EEMD with respect to time at each grid point independently and reconstructs the space–time EEMD components using the common IMFs from all spatial points. Wu et al. (2009) argue that, because the spatial structure is closely linked to the time scales for many physical processes, a priori assumption of spatial structures of the variability may not be essential. Instead, coherent spatial structures may emerge through the pseudoBEMD strategy and identify the physical processes that drive the evolving features on particular time scales. Using the SST cross section along 60°N in the North Atlantic as an example, Wu et al. (2009) showed that, in comparison with the EMD, the EEMD separates the IMF components more closely to their intrinsic time scales and coherent spatial structures do emerge naturally from the IMFs because of the shared physical characters among adjacent points. In this way, the pseudoBEMD method can reveal the time–space structures of coherent evolving features with characteristic time scales objectively selected by the EEMD. In practice, Coughlin and Tung (2004) have applied the EMD in a similar way to isolate the vertical distribution of the 11-yr solar cyclone in the atmospheric geopotential heights. More recently, Huang et al. (2012b) and Hu et al. (2012) also successfully applied it to isolate the signals of quasi-biennial oscillation from direct measurements and reanalysis datasets. In our case, the pseudoBEMD technique decomposes the EM HCA into different EEMD components, which have the same dimensions as the original HCA data, which are referred to as EEMD1, EEMD2, and so on. These components are ranked according to the lengths of their characteristic time scales in an increasing order. To further ensure spatial coherence, we use the leading EOF modes to identify the space–time patterns of the EEMD components, which only select signals that have strong covariance among grid points within the basin. As pointed out by several studies (e.g., Dommenget and Latif 2002; Hannachi et al. 2007), one has to exercise great care when trying to ascribe physical meaning to EOF modes. In this study, we will explain the derived HCA EOF modes in combination with other variables. They will also be subjected to a significance test following the criterion of North et al. (1982). More sophisticated techniques to perform such a test can be seen in Dommenget (2007) and Korres et al. (2000).

Among these components, EEMD4 (characterized by a 39-month cycle; see green curve in Fig. 1) is of particular interest in this study. As will be shown in following sections, the Atlantic meridional gradient and equatorial zonal modes coexist in this component. To further analyze its physical meanings, we applied the multichannel singular spectrum analysis (MSSA; see a review by Ghil et al. 2002) to explore the propagating features of the leading HCA oscillatory mode. Based on the characteristics revealed by the MSSA, a composite analysis is conducted to explore the mechanism of the interaction between the zonal and meridional processes in a complete cycle. The associated changes of the observed surface variables in the tropical Atlantic region are also examined, such as SST [optimum interpolation (OI) SST; Reynolds et al. 2002], wind stress [(NCEP–Department of Energy (DOE) Atmospheric Model Intercomparison Project (AMIP-II) reanalysis; Kanamitsu et al. 2002], and precipitation [provided by the National Oceanic and Atmospheric Administration (NOAA)/Climate Prediction Center and available online at http://ingrid.ldgo.columbia.edu/; Xie and Arkin 1997].

3. Time scales of the tropical Atlantic HC by EEMD

As outlined in last section, the anomalous HC fields from the EM dataset are first decomposed into different components by the EEMD technique. The first six components account for 7.3%, 10.5%, 12.3%, 14.2%, 11.6%, and 6.0% of total HCA variance over the tropical Atlantic, respectively (Table 1). The other components explain less than 1% of the total variance and will not be discussed further.

Table 1.

The explained variance by each EEMD component over the global and tropical Atlantic regions.

The explained variance by each EEMD component over the global and tropical Atlantic regions.
The explained variance by each EEMD component over the global and tropical Atlantic regions.

The leading EOF modes of these EEMD components are characterized by well-defined time scales and coherent spatial structures with clear physical meanings. In fact, the power spectra of their first principal components (PCs) are generally dominant by a single peak (Fig. 1). Therefore, the EEMD components can be categorized into the subannual, interannual, and low-frequency components based on their characteristic time scales. In particular, EEMD1 represents the highest-frequency variability, on a month-to-month scale. EEMD2 is characterized by a 10-month cycle, which, together with EEMD1, can be categorized as the subannual variations. EEMD3 and EEMD4 are characterized by 19 and 39 months, respectively, and can be categorized to the interannual component. EEMD5 and EEMD6 are characterized by 70 and 116 months, respectively, and can be categorized to the low-frequency component. It is interesting to note that the peaks of the power spectra from the first PCs of these components generally have their counterparts in the power spectra of either PC1 or PC2 of the total EM HCA fields (Fig. 1). A detailed description of the EOF modes from the total EM HCA fields can be found in Zhu et al. (2012). This consistency confirms that the characteristic time scales, as identified from the EEMD analysis, are intrinsic to the original fields.

The different EEMD components are also associated with distinctive physical processes. As an example, Fig. 2a presents the first EOF pattern of the HCA fields from the EEMD2 component, which is separated from the other EOFs according to the criterion of North et al. (1982). Its spatial structure is characterized by a zonal pattern symmetric to the equator, with warm HCAs appearing in the eastern equatorial basin and cold ones in the western Atlantic both south and north of the equator. The accompanying SSTA (Fig. 2b), derived by regression onto the HCA PC1, is characterized by warm anomalies in the mideastern and southeastern tropical Atlantic; the surface wind stress weakens to the west of anomalously warm water; and the associated rainfall tends to increase on the Guinea coast. All these changes are consistent with the characteristics of the equatorial zonal modes (Zebiak 1993; Carton and Huang 1994; Huang and Shukla 1997; Handoh and Bigg 2000; Ruiz-Barradas et al. 2000). As a matter of fact, the lagged linear regression analysis of HCA from the EEMD2 component against its PC1 (not shown here) does show zonally propagating signals along and off the equator, which are part of the adjustment by ocean waves due to the fluctuations in the equatorial winds, like ENSO in the tropical Pacific. The dominant spatial patterns for EEMD1 and EEMD3 are similar to those for EEMD2 (not shown), but with different meridional scales, increasing from a higher-frequency mode (EEMD1) to a lower-frequency one (EEMD3). Therefore, it can be concluded that the three EEMD components are all characterized by the pure equatorial Atlantic variability on intraseasonal to interannual time scales. The relationship between frequency and meridional scale is consistent with the characteristics of the equatorial zonal modes in Huang and Shukla (1997).

Fig. 2.

EOF1 mode of the HCA (a) EEMD2 and (c) EEMD5 and EEMD6 components, with their explained variance shown in parenthesis. The contour interval in (a),(c) is 0.025°C. (b),(d) The regression maps of the observed anomalous SST (contours, units: °C), precipitation (color shading, units: mm day−1) and wind stress (vectors, units: N m−2) onto HCA PC1 of the EEMD components. In (b),(d), only values above 95% significance test are shown, whereas wind stress vectors are shown if either meridional or zonal wind stress is above 95% significance test.

Fig. 2.

EOF1 mode of the HCA (a) EEMD2 and (c) EEMD5 and EEMD6 components, with their explained variance shown in parenthesis. The contour interval in (a),(c) is 0.025°C. (b),(d) The regression maps of the observed anomalous SST (contours, units: °C), precipitation (color shading, units: mm day−1) and wind stress (vectors, units: N m−2) onto HCA PC1 of the EEMD components. In (b),(d), only values above 95% significance test are shown, whereas wind stress vectors are shown if either meridional or zonal wind stress is above 95% significance test.

On the other hand, EEMD5 and EEMD6 are found to be characterized by the pure meridional mode. Figure 2c shows the first dominant EOF of the combined HCA EEMD5 and EEMD6 components, which is also separated from other EOFs according to the criterion of North et al. (1982). Its spatial structure shows an asymmetry with respect to the equator, with positive basinwide HCAs occupying almost all regions north of 10°S and small negative anomalies around 20°S and north of 20°N. Accompanying the HCAs (Fig. 2c; derived by regression onto its PC1), the SSTAs show a remarkable meridional gradient, with significant positive anomalies spreading from the North African coast into the northeastern tropical Atlantic and slight negative anomalies in the south. Moreover, the northeast trade winds are weakened and the precipitation field is dominated by a negative anomaly centered over northeast Brazil. All these variations are consistent with the characteristics of the meridional gradient mode (Moura and Shukla 1981; Nobre and Shukla 1996; Enfield and Mayer 1997; Huang and Shukla 1997; Ruiz-Barradas et al. 2000; Xie and Carton 2004) and may be explained by the thermodynamic WES feedback (Chang et al. 1997; Xie 1999). More analysis about these components with the pure equatorial or meridional mode is beyond the scope of this study.

In this study, we are most interested in the characteristics of the EEMD4 component, which is dominated by fluctuations around 3–4 yr. Among all EEMD components, EEMD4 explains the highest percentage of the total HCA variance over both the global and tropical Atlantic regions (Table 1), with the former a little higher, probably reflecting its close connections with ENSO, to be further analyzed later. More interestingly, the leading EOF modes of the EEMD4 show “mixed” characteristics of both equatorial zonal and meridional modes. Figures 3a,c show the spatial patterns of the first two EOFs, which account for about 16.0% and 13.1% of the total variance of the EEMD4 HCA, respectively. Clearly, the two patterns bear remarkable resemblance to the HCA pattern in the meridional and zonal modes, compared with the patterns shown in Figs. 2a,c, as well as EOF1 and EOF2 of the total HCA in Zhu et al. (2012, their Fig. 4). Particularly, EOF1 (Fig. 3a) is characterized by positive basinwide HCAs occupying almost all regions from the equator to 20°N, except for a negative belt at 10°N and a negative anomaly centered at 8°S. The HCAs are related to the Ekman pumping induced by simultaneous changes of the trade winds, especially north of the equator (Fig. 4). In EOF2, warm HCAs appear in the eastern equatorial basin and cold ones appear in the off-equatorial western Atlantic on both sides, with the northern center more pronounced. The corresponding changes of SST, surface wind stress, and precipitation (Figs. 3b,d) are also consistent with the characteristics of the meridional gradient and equatorial zonal modes, respectively. Actually, the atmospheric responses to EEMD4 may even be stronger than the atmospheric responses to the components of pure zonal and meridional modes. For example, the precipitation anomalies over northeast Brazil associated with EEMD4 (Fig. 3c) are much larger than those related to the combined EEMD5–EEMD6 component (Fig. 2d).

Fig. 3.

(a) EOF1 and (c) EOF2 modes of the HCA EEMD4 component. The contour interval in (a),(c) is 0.025°C the Regression maps of the observed anomalous SST (contours, units: °C), precipitation (color shading, units: mm day−1), and wind stress (vectors, units: N m−2) onto HCA (b) PC1 and (d) PC2 of the EEMD4 component. In (b),(d), only values above 95% significance test are shown, whereas wind stress vectors are shown if either meridional or zonal wind stress is above 95% significance test.

Fig. 3.

(a) EOF1 and (c) EOF2 modes of the HCA EEMD4 component. The contour interval in (a),(c) is 0.025°C the Regression maps of the observed anomalous SST (contours, units: °C), precipitation (color shading, units: mm day−1), and wind stress (vectors, units: N m−2) onto HCA (b) PC1 and (d) PC2 of the EEMD4 component. In (b),(d), only values above 95% significance test are shown, whereas wind stress vectors are shown if either meridional or zonal wind stress is above 95% significance test.

Fig. 4.

The wind stress curl (contours, units: N m−3) and Ekman pumping velocity (color shading, units: m s−1) derived from the surface wind stress changes in Fig. 3b. Note that wind stress curl and Ekman pumping velocity have been multiplied by 109 and 107, respectively for plotting.

Fig. 4.

The wind stress curl (contours, units: N m−3) and Ekman pumping velocity (color shading, units: m s−1) derived from the surface wind stress changes in Fig. 3b. Note that wind stress curl and Ekman pumping velocity have been multiplied by 109 and 107, respectively for plotting.

In addition to the same frequency band, the potential connection between the EOF1 and EOF2 of the EEMD4 HCA can also be seen from the relatively fixed phase relationship between their PCs (not shown). Moreover, because the amounts of variance explained by these two modes are relatively close [they are separated from each other very reluctantly according to the criterion of North et al. (1982)], there may be a certain degree of degeneracy between them; that is, they represent different aspects of a single oscillation that has the characteristics of both zonal and meridional modes. This point will be further pursued in the next section, with a statistical method more suitable for dealing with this kind of process. Moreover, it is important to understand the relationship between the meridional gradient and equatorial zonal modes on this time scale. How can a mixed mode exist? In particular, what determines the 3–4-yr time scale? What role does the ocean dynamics play? These questions will be explored in the next section.

4. The interaction between the meridional and equatorial modes

To further explore the potential oscillation in the 3–4-yr-frequency band, MSSA is applied to analyze the HCA EEMD4 component. MSSA derives eigenvalues of the lag-covariance matrix for the HCA EEMD4 components with a chosen maximum lag M. For a given eigenvalue, the corresponding eigenvector forms a lag sequence of spatial fields while the projections of the eigenvector onto the dataset constitute its time series. Even though it is similar to the EOF analysis in formulation, which corresponds to M = 0, MSSA allows the identification of propagating oscillations as degenerate pairs of eigenmodes, having close eigenvalues and similar spatiotemporal structures, except for an orthogonality in phase.

Our analysis generally follows the procedure used by Plaut and Vautard (1994), Moron et al. (1998), and Huang et al. (2012a). In our case, M = 40 months is chosen for the MSSA analysis, based on the time scale derived from the EEMD analysis. The propagating oscillatory mode is identified from the pair of the first and second MSSA eigenmodes, which together account for 26.5% (13.5% plus 13.0%) of the total variance of the HCA EEMD4 component. The closeness of their eigenvalues suggests that the two eigenmodes form a degenerate pair. Following the procedure of Moron et al. (1998), the oscillation mode is reconstructed based on these two eigenmodes as the filtered space–time series with the same dimensions as the original HCA data. In this way, the characteristics of the oscillation can then be represented economically by the two leading EOF modes of the filtered data, as shown in Figs. 5a,b with their respective time series in Fig. 5c, as the solid and dashed black curves. These two EOF modes explain 55.9% and 43.6% of the total variance of the filtered fields, respectively.

Fig. 5.

The (a) EOF1 and (b) EOF2 modes of reconstructed HCA EEMD4 component based on the degenerate pair of the first and second MSSA eigenmodes, which together explain about 26.5% of total variance for the EEMD4 component. (c) Their respective time series and Niño-3.4 index [from the Extended Reconstructed SST version 3 (ERSST v3)] are shown. The contour interval in (a),(b) is 0.025°C.

Fig. 5.

The (a) EOF1 and (b) EOF2 modes of reconstructed HCA EEMD4 component based on the degenerate pair of the first and second MSSA eigenmodes, which together explain about 26.5% of total variance for the EEMD4 component. (c) Their respective time series and Niño-3.4 index [from the Extended Reconstructed SST version 3 (ERSST v3)] are shown. The contour interval in (a),(b) is 0.025°C.

Comparing the spatial patterns of the two EOF modes with those in the unfiltered HCA EEMD4 fields (Fig. 3a versus Fig. 5a; Fig. 3b versus Fig. 5b), only a very slightly difference can be seen, which is also true when comparing their respective PCs. As a matter of fact, the associated SST, surface wind stress, and precipitation changes (derived by regression onto their PCs in Fig. 5c; figures not shown here) are also almost the same as in Figs. 3b,d. All these characteristics further prove the robustness of the existence of both the meridional and equatorial modes in the HCA EEMD4 component. In addition, the PCs of these two EOFs show similar variations and have a clear 90° phase shift, as expected from an oscillation (Fig. 5c). This oscillation has a dominant period of 3–4 yr and is intermittent in nature. Their magnitudes are large in the 1980s and the late 1990s but weak in the 2000s and even weaker in the early 1990s.

To examine the evolution of the oscillation, as well as relationship between the meridional and equatorial modes within this evolution, we further applied the phase-compositing technique described by Plaut and Vautard (1994) and Moron et al. (1998). An instantaneous phase index is calculated from the PC1 (black solid curve in Fig. 5c) of the reconstructed space–time series and its time derivative (both normalized), which by definition varies between 0 and 2π and describes the life cycles of the oscillation. In constructing a composite life cycle of the oscillation based on its phase, the spatiotemporal evolution of any variables can be concisely and objectively presented by their averages within predetermined π/4 sectors of the instantaneous phase index throughout the considered time period. As a result, there are in total eight consecutive composites to describe a complete mean life cycle, which is referred to here as the eight phases of the life cycle. Because the period of the oscillation is around 39 months, each of these phase composites roughly represents an averaging period of 5 months. Note that here the composites for all variables are based on the original datasets, without any filtering processes. Composing the raw data instead of the EEMD4 component provides a more robust result, which, however, may also bring some uncorrelated signals into the composite maps.

Figure 6 shows the eight phases of the HCA evolution in a complete cycle. The corresponding evolutions of the SSTA and the anomalies of the wind stress and precipitation are presented in Fig. 7, for the composite life cycle. In phase 1, the anomalies of all variables, including HC, SST, wind stress, and precipitation, are structured with the meridional mode characteristics. The SST exhibits warm anomalies in the northern tropical Atlantic, with slightly cool anomalies shown in the Southern Hemisphere (contours in Fig. 7a). The trade wind is significantly weaker than average in the Northern Hemisphere but has smaller changes in the Southern Hemisphere (vectors in Fig. 7a). The rainfall in the northeast Brazil is apparently below average (shading in Fig. 7a). Correspondingly, HCA is characterized as positive basinwide anomalies occupying almost all regions from the equator to 20°N, except for a small negative belt at 10°N, whereas negative anomalies are centered at 8°S (Fig. 6a). The relationship between the wind and SST anomalies in the northern tropical Atlantic is consistent with the WES mechanism (Chang et al. 1997; Xie 1999). Associated with the WES-induced wind anomalies, a negative wind curl is formed between the equator and 10°N, which is consistent with the deepened thermocline in the western tropical Atlantic (Fig. 5a). The northern band of the warm HCA is possibly due to the convergence of the southward Ekman transport by the weakened northeast trades farther north.

Fig. 6.

The composite life cycle of the HCA oscillations at a period of around 39 months based on the raw EM dataset for phases (a) 1, (b) 2, (c) 3, (d) 4, (e) 5, (f) 6, (g) 7, and (h) 8. The contour interval is 0.05°C.

Fig. 6.

The composite life cycle of the HCA oscillations at a period of around 39 months based on the raw EM dataset for phases (a) 1, (b) 2, (c) 3, (d) 4, (e) 5, (f) 6, (g) 7, and (h) 8. The contour interval is 0.05°C.

Fig. 7.

As in Fig. 6, but for the composite of the observed anomalous SST (contours, units: °C), precipitation (color shading, units: mm day−1), and wind stress (vectors, units: N m−2). All figures are based on the raw unfiltered data.

Fig. 7.

As in Fig. 6, but for the composite of the observed anomalous SST (contours, units: °C), precipitation (color shading, units: mm day−1), and wind stress (vectors, units: N m−2). All figures are based on the raw unfiltered data.

In the next phases, the HCAs accumulated in the northwestern tropical Atlantic redistribute and trigger the movement of surface anomalies from the northern tropical Atlantic toward the southeastern Atlantic Ocean. In phase 2, a part of the warm HCAs in the northwestern Atlantic has been discharged into the equatorial waveguide and propagates along the equator into the eastern Atlantic (Fig. 6b), possibly because of the weakening of the wind (wind curl) anomalies between the equator and 10°N (vectors in Fig. 7b). The warm SSTAs in the northern tropical Atlantic are weakened while the warm SSTAs started to appear in the eastern equatorial Atlantic and replaced the negative anomalies in the Southern Hemisphere (contours in Fig. 7b). The drought condition in northeast Brazil also becomes less severe (shading in Fig. 7b).

In phase 3, the pattern of the equatorial mode is established. In the HCA fields (Fig. 6c), warm oceanic anomalies appear in the eastern equatorial basin, whereas cold anomalies appear in the off-equatorial western Atlantic, with the northern center more pronounced. The SSTA field is featured by a warm anomaly in the central and eastern equatorial Atlantic Ocean (contours in Fig. 7c). As a result, more precipitation occurs along the equator, including the Gulf of Guinea coast (shading in Fig. 7c), which is accompanied by anomaly wind convergence into this region (vectors in Fig. 7c). In the western Atlantic, consistent with the warm phase of the equatorial mode, there are westerly wind stress anomalies.

In phase 4 (Figs. 6d, 7d), the characteristics of the equatorial zonal mode become clearer, with the corresponding changes enhanced significantly. Particularly, the positive SST anomaly in the central and eastern equatorial Atlantic increases from around 0.2°C in phase 3 to above 0.5°C in phase 4. The changes in the related wind stress and precipitation are also twice of those in phase 3. The enhancement may be explained by the positive Bjerknes-type feedback among the equatorial wind, thermocline, and SST (Bjerknes 1969). For phases 5–8 in Figs. 6 and 7, the similar evolution processes can be identified, but with a generally opposite polarity.

Although the whole cycle takes about 3–4 yr to complete, the transition from the meridional to the zonal mode can be relatively fast. In particular, the propagation of the northern tropical HCA into the equatorial waveguide and to the eastern Atlantic happens swiftly and repeatedly. The actual process can be seen more clearly in specific anomalous events. Figure 8 shows the month-to-month evolution of HCA and other surface variables for 1987. In this year, the tropical Atlantic experienced a transition from the meridional mode in May to a weak equatorial mode in September. If we follow the HCA evolution from May to September, it took 3 months, from June to August, for the warm HCA to propagate from the northwestern Atlantic to the Gulf of Guinea. In particular, the front of the HCA propagation, as an equatorial Kelvin wave signal, may take less than 1 month to cross the basin, given the narrow width of the tropical Atlantic Ocean. It is interesting to note that the evolution in HCA from dominantly meridional pattern in boreal spring to the dominantly zonal pattern in boreal summer appears to be similar to the seasonal changes of these anomalous patterns from spring to summer based on SSTA and other surface variables (Bates 2010).

Fig. 8.

The time evolution of some physical fields in the tropical Atlantic during spring–summer 1987. (a)–(e) HC (color bar is the top one). (f)–(j) SST (contours; values less than 0 shown in blue dashed and are otherwise shown in red solid; units: °C), surface wind stress (vectors; units: N m−2), and precipitation (colors; color bar is the bottom one; units: mm day−1). (a),(f) May, (b),(g) June, (c),(h) July, (d),(i) August, and (e),(j) September 1987.

Fig. 8.

The time evolution of some physical fields in the tropical Atlantic during spring–summer 1987. (a)–(e) HC (color bar is the top one). (f)–(j) SST (contours; values less than 0 shown in blue dashed and are otherwise shown in red solid; units: °C), surface wind stress (vectors; units: N m−2), and precipitation (colors; color bar is the bottom one; units: mm day−1). (a),(f) May, (b),(g) June, (c),(h) July, (d),(i) August, and (e),(j) September 1987.

The characteristics of the evolution based on the MSSA can also be identified by more basic statistical methods, such as the lagged regression onto the PC1 of the HCA EEMD4 component (not shown here). The initial discharge of the off-equatorial HCA into the equatorial waveguide, which triggers the subsequent development, is possibly resulted from the meridional gradient of the oceanic potential energies, built up by the anomalous heat accumulation in the off-equatorial northwestern Atlantic. In the meridional mode’s peak phase, the north–south contrast is in equilibrium with the overlying wind forcing. Once the wind starts to weaken, the equilibrium cannot be sustained and the warm water has the tendency to propagate southward. Because of the dynamical constraint of the equatorial beta-plane effect, the natural route is to propagate along the equatorial waveguide.

Our results suggest that the transition is mainly from the meridional to zonal, although, in principle, the opposite should also be possible because a change in the equatorial SST can also change the ITCZ position, which can trigger the WES feedback in the northern tropical Atlantic. The transition from the former to the latter is more effective when taking remote forcing into account, because the WES feedback in the northern tropical Atlantic can be more effectively generated by remote forcing. As a matter of fact, previous studies have shown that the meridional mode, especially its Northern Hemisphere component [i.e., the north tropical Atlantic (NTA)], is significantly influenced by external forcing, such as ENSO and NAO (Czaja et al. 2002). Xie and Tanimoto (1998) also demonstrated how remote extratropical wind forcing induces the meridional mode. Particularly for ENSO, many observational (Enfield and Mayer 1997) and modeling (Saravanan and Chang 2000; Alexander and Scott 2002; Huang et al. 2002) studies have shown that the ENSO influence on the tropical Atlantic is strongest in the NTA region, with Atlantic warming occurring 4–5 months after the mature phases of Pacific warm events [see the review by Xie and Carton (2004) for more comprehensive summary about the influence of ENSO on TAV], which has prompted ENSO to be highly valued in the statistical TAV forecasts (Penland and Matrosova 1998). Hindcast experiments by dynamical models have also found that models are particularly skillful in forecasting the SST anomaly in the NTA region, which largely results from the remote influence of ENSO (Chang et al. 2003; Hu and Huang 2007).

Considering these, we are shifting in favor of the external influence from ENSO to understand the accumulation of the HCA in the northwestern tropical Atlantic in the first place. If we simply compare the Niño-3.4 index with the time series related to the meridional mode in the HCA EEMD4 component (Fig. 5c), it can be seen that the Niño-3.4 index is clearly correlated to the appearance of the meridional mode with a lead of a few months. This is quantitatively confirmed by lagged correlation analysis between them (Fig. 9), which exhibits highest correlation (above 0.6) when ENSO leads the meridional mode by 4–5 months. The lag time is also consistent with the studies mentioned above (Enfield and Mayer 1997; Saravanan and Chang 2000; Alexander and Scott 2002; Huang et al. 2002; Xie and Carton 2004; and others), where the mechanism of the Atlantic response to ENSO can also be found. Actually, the remote influence of ENSO, which is characterized by a 3–5-yr period (Philander 1990), can also explain the 3–4-yr time scale for the mixed mode.

Fig. 9.

The lagged correlation between Niño-3.4 index and the PC1 of reconstructed EEMD4 HCA, which are both shown in Fig. 5c.

Fig. 9.

The lagged correlation between Niño-3.4 index and the PC1 of reconstructed EEMD4 HCA, which are both shown in Fig. 5c.

To sum up, the mixed mode on the 3–4-yr time scale in the tropical Atlantic can be explained by the following schematic diagram: Four to five months after the El Niño in the Pacific, usually peaking in boreal winter, a basinwide warming is provoked in the NTA region by mechanisms including WES (Chang et al. 1997; Xie 1999). Accompanying the surface meridional gradient structure, the subsurface north Atlantic is also occupied by positive basinwide HC anomalies. When the meridional mode starts to decay, the off-equatorial HCA will be discharged into the equatorial waveguide and propagate to eastern equatorial Atlantic, stimulating the equatorial mode about 10–15 months later.

5. Conclusions and discussion

This study provides observational evidence for a mechanism of interaction between the meridional gradient mode and the equatorial zonal mode in the tropical Atlantic. It suggests that the HC changes generated in the meridional mode can propagate into the equatorial waveguide and stimulate the equatorial mode, as proposed by Huang et al. (1995) and Huang and Shukla (1997). To derive more reliable HC variability, the study utilized a HC dataset estimated through an ensemble method based on six state-of-the-art global ocean reanalysis products (Zhu et al. 2012). Considering that the TAV is composed of variations on multiple time scales, the sophisticated EEMD technique is first applied to decompose the total HC fields into ones with different time scales. Our analysis about these decomposed components shows that the high-frequency modes (including EEMD1, EEMD2, and EEMD3) are characterized as the “pure” equatorial mode, whereas the low-frequency modes (including EEMD5 and EEMD6) are featured as the pure meridional mode.

In this study, we are more interested in a component with a mixed mode, EEMD4, which is on the 3–4-yr time scale. Our analysis, based on MSSA and a phase-compositing technique used by Plaut and Vautard (1994), Moron et al. (1998), and Huang et al. (2012a), indicates that the meridional and equatorial modes coexist and interact with each other in EEMD4. Particularly, the transition from the meridional to the equatorial mode involves the redistribution of the off-equatorial subsurface heat content, which may result from the breakdown of the equilibrium between the oceanic potential energy contrast of the Northern Hemisphere versus the Southern Hemisphere and the overlying wind forcing. The appearance of the meridional mode may largely be explained by the remote forcing, ENSO, which may also play a selective role in the time scale of the mixed mode with 3–4 yr of fluctuation. The mechanism proposed here for the transition from the meridional to equatorial zonal variations is complementary to those described in previous studies (Servain et al. 1999, 2000; Chiang et al. 2002; Foltz and McPhaden 2010). These previous studies emphasized the processes within the equatorial waveguide, whereas the HCA redistribution in our study reflects a connection between the equatorial and off-equatorial oceans.

In addition, previous investigations have demonstrated that the relationship between ENSO and the Atlantic equatorial zonal mode is fragile (Zebiak 1993), which is arguably a result of “destructive interference” between atmospheric and oceanic processes in response to ENSO (Chang et al. 2006). This study, on the other hand, presents evidence for the influence of ENSO on the Atlantic equatorial zonal mode on the 3–4-yr time scale with a different mechanism. In the mechanism, ENSO and the equatorial mode are bridged through the meridional gradient mode by an oceanic subsurface channel. However, in reality the influence of ENSO on the Atlantic equatorial zonal mode is significantly complicated by the essence of TAV with multiple time scales, which made it hard to detect the relationship by a simple correlation analysis of SST indices in raw data.

Meanwhile, a recent study also revealed that the intrinsic ocean dynamics of the deep equatorial Atlantic, in the forms of vertically alternating deep zonal jets of short vertical wavelength, can affect SST, wind, and rainfall in the tropical Atlantic region and constitutes a 4.5-yr climate cycle (Brandt et al. 2011). This study indicates that, at a similar (or slightly shorter) time scale (~39 months), these climatic variables in the equatorial Atlantic Ocean can be altered through another alternative mechanism: that is, the horizontal propagation of upper layer heat content from the off-equatorial region to the equatorial region.

Acknowledgments

Funding for this study is provided by the NOAA/CVP program (NA07OAR4310310). ZW is supported by the National Science Foundation project AGS-1139479. The authors thank Drs. J. Shukla and J. Kinter for their guidance and support of this project. We are grateful to Drs. Zeng-Zhen Hu, Bin Wang, and the reviewers for their suggestions and comments. We would also like to thank Dr. V. Krishnamurthy for his comments on an earlier version of the manuscript.

REFERENCES

REFERENCES
Alexander
,
M.
, and
J.
Scott
,
2002
:
The influence of ENSO on air-sea interaction in the Atlantic
.
Geophys. Res. Lett.
,
29
,
1701
,
doi:10.1029/2001GL014347
.
Ayina
,
L.-H.
, and
J.
Servain
,
2003
:
Spatial-temporal evolution of the low frequency climate variability in the tropical Atlantic. Interhemispheric Water Exchange in the Atlantic Ocean, G. J. Goni and P. M. Rizzoli, Eds., Oceanography Series, Vol. 68, Elsevier, 475–495
.
Balmaseda
,
M. A.
,
A.
Vidard
, and
D. L. T.
Anderson
,
2008
:
The ECMWF ocean analysis system: ORA-S3
.
Mon. Wea. Rev.
,
136
,
3018
3034
.
Balmaseda
,
M.
,
K.
Mogensen
,
F.
Molteni
, and
A.
Weaver
,
2010
.
The NEMOVAR-COMBINE ocean re-analysis. COMBINE Tech. Rep. 1, 11 pp. [Available online at http://www.combine-project.eu/fileadmin/user_upload/combine/tech_report/COMBINE_TECH_REP_n01.pdf.]
Bates
,
S. C.
,
2010
:
Seasonal influences on coupled ocean-atmosphere variability in the tropical Atlantic Ocean
.
J. Climate
,
23
,
582
604
.
Behringer
,
D. W.
,
2007
:
The Global Ocean Data Assimilation System (GODAS) at NCEP. Preprints, 11th Symp. on Integrated Observing and Assimilation Systems for the Atmosphere, Oceans, and Land Surface (IOAS-AOLS), San Antonio, TX, Amer. Meteor. Soc., 3.3. [Available online at http://ams.confex.com/ams/pdfpapers/119541.pdf.]
Bjerknes
,
J.
,
1969
:
Atmospheric teleconnections from the equatorial Pacific
.
Mon. Wea. Rev.
,
97
,
163
172
.
Brandt
,
P.
,
A.
Funk
,
V.
Hormann
,
M.
Dengler
,
R. J.
Greatbatch
, and
J. M.
Toole
,
2011
:
Interannual atmospheric variability forced by the deep equatorial Atlantic Ocean
.
Nature
,
473
,
497
500
,
doi:10.1038/nature10013
.
Carton
,
J. A.
, and
B.
Huang
,
1994
:
Warm events in the tropical Atlantic
.
J. Phys. Oceanogr.
,
24
,
888
903
.
Carton
,
J. A.
, and
B. S.
Giese
,
2008
:
A reanalysis of ocean climate using Simple Ocean Data Assimilation (SODA)
.
Mon. Wea. Rev.
,
146
,
2999
3017
.
Carton
,
J. A.
,
X.
Cao
,
B. S.
Giese
, and
A. M.
da Silva
,
1996
:
Decadal and interannual SST variability in the tropical Atlantic Ocean
.
J. Phys. Oceanogr.
,
26
,
1165
1175
.
Chang
,
P.
,
L.
Ji
, and
H.
Li
,
1997
:
A decadal climate variation in the tropical Atlantic Ocean from thermodynamic air-sea interactions
.
Nature
,
385
,
516
518
.
Chang
,
P.
,
R.
Saravanan
, and
L.
Ji
,
2003
:
Tropical Atlantic seasonal predictability: The roles of El Niño remote influence and thermodynamic air-sea feedback
.
Geophys. Res. Lett.
,
30
,
1501
1504
.
Chang
,
P.
,
Y.
Fang
,
R.
Saravanan
,
L.
Ji
, and
H.
Seidel
,
2006
:
The cause of the fragile relationship between the Pacific El Niño and the Atlantic Niño
.
Nature
,
443
,
324
328
.
Chiang
,
J. C. H.
,
Y.
Kushnir
, and
A.
Giannini
,
2002
:
Deconstructing Atlantic Intertropical Convergence Zone variability: Influence of the local cross-equatorial SST gradient, and remote forcing from the eastern equatorial Pacific
.
J. Geophys. Res.
,
107
,
4004
,
doi:10.1029/2000JD000307
.
Coughlin
,
K.
, and
K. K.
Tung
,
2004
:
Eleven-year solar cycle signal throughout the lower atmosphere
.
J. Geophys. Res.
,
109
,
D21105
,
doi:10.1029/2004JD004873
.
Czaja
,
A.
,
P.
van der Vaart
, and
J.
Marshall
,
2002
:
A diagnostic study of the role of remote forcing in tropical Atlantic variability
.
J. Climate
,
15
,
3280
3290
.
Doi
,
T.
,
T.
Tozuka
, and
T.
Yamagata
,
2009
:
Interannual variability of the Guinea Dome and its possible link with the Atlantic meridional mode
.
Climate Dyn.
,
33
,
985
998
.
Doi
,
T.
,
T.
Tozuka
, and
T.
Yamagata
,
2010
:
The Atlantic meridional mode and its coupled variability with the Guinea Dome
.
J. Climate
,
23
,
455
475
.
Dommenget
,
D.
,
2007
:
Evaluating EOF modes against a stochastic null hypothesis
.
Climate Dyn.
,
28
,
517
531
.
Dommenget
,
D.
, and
M.
Latif
,
2002
:
A cautionary note on the interpretation of EOFs
.
J. Climate
,
15
,
216
225
.
Enfield
,
D. B.
, and
D. A.
Mayer
,
1997
:
Tropical Atlantic SST variability and its relation to El Niño-Southern Oscillation
.
J. Geophys. Res.
,
102
,
929
945
.
Foltz
,
G. R.
, and
M. J.
McPhaden
,
2010
:
Interaction between the Atlantic meridional and Niño modes
.
Geophys. Res. Lett.
,
37
,
L18604
,
doi:10.1029/2010GL044001
.
Ghil, M
., and
Coauthors
,
2002
:
Advanced spectral methods for climatic time series
.
Rev. Geophys.
,
40
,
1003
,
doi:10.1029/2000RG000092
.
Handoh
,
I. C.
, and
G. R.
Bigg
,
2000
:
A self-sustaining climate mode in the tropical Atlantic, 1995-1997: Observations and modelling
.
Quart. J. Roy. Meteor. Soc.
,
126
,
807
821
.
Hannachi
,
A.
,
I. T.
Jolliffe
, and
D. B.
Stephenson
,
2007
:
Empirical orthogonal functions and related techniques in atmospheric science: A review
.
Int. J. Climatol.
,
27
,
1119
1152
.
Hu
,
Z.-Z.
, and
B.
Huang
,
2007
:
The predictive skill and the most predictable pattern in the tropical Atlantic: The effect of ENSO
.
Mon. Wea. Rev.
,
135
,
1786
1806
.
Hu
,
Z.-Z.
,
B.
Huang
,
J. L.
Kinter
III
,
Z.
Wu
, and
A.
Kumar
,
2012
:
Connection of stratospheric QBO with global atmospheric general circulation and tropical SST. Part II: Interdecadal variations
.
Climate Dyn.
,
38
,
25
43
,
doi:10.1007/s00382-011-1073-6
.
Huang
,
B.
, and
J.
Shukla
,
1997
:
Characteristics of the interannual and decadal variability in a general circulation model of the tropical Atlantic Ocean
.
J. Phys. Oceanogr.
,
27
,
1693
1712
.
Huang
,
B.
,
J. A.
Carton
, and
J.
Shukla
,
1995
:
A numerical simulation of the variability in the tropical Atlantic Ocean, 1980–88
.
J. Phys. Oceanogr.
,
25
,
835
854
.
Huang
,
B.
,
P. S.
Schopf
, and
Z.
Pan
,
2002
:
The ENSO effect on the tropical Atlantic variability: A regionally coupled model study
.
Geophys. Res. Lett.
,
29
,
2039
,
doi:10.1029/2002GL014872
.
Huang
,
B.
,
Z.-Z.
Hu
,
E. K.
Schneider
,
Z.
Wu
,
Y.
Xue
, and
B.
Klinger
,
2012a
:
Influences of subtropical air-sea interaction on the multidecadal AMOC variability in the NCEP climate forecast system
.
Climate Dyn.
,
doi:10.1007/s00382-011-1258-z, in press
.
Huang
,
B.
,
Z.-Z.
Hu
,
J. L.
Kinter
,
Z.
Wu
, and
A.
Kumar
,
2012b
:
Connection of stratospheric QBO with global atmospheric general circulation and tropical SST. Part I: Methodology and composite life cycle
.
Climate Dyn.
,
32
,
1
23
.
Huang
,
N. E.
, and
Z.
Wu
,
2008
:
A review on Hilbert-Huang transform: Method and its applications on geophysical studies
.
Rev. Geophys.
,
46
,
RG2006
,
doi:10.1029/2007RG000228
.
Huang
,
N. E.
, and
Coauthors
,
1998
:
The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis
.
Proc. Roy. Soc. London
,
454A
,
903
995
.
Joyce
,
T. M.
,
C.
Frankignoul
,
J.
Yang
, and
H. E.
Phillips
,
2004
:
Ocean response and feedback to the SST dipole in the tropical Atlantic
.
J. Phys. Oceanogr.
,
34
,
2525
2540
.
Kanamitsu
,
M.
,
W.
Ebisuzaki
,
J.
Woollen
,
S.-K.
Yang
,
J. J.
Hnilo
,
M.
Fiorino
, and
G. L.
Potter
,
2002
:
NCEP–DOE AMIP-II Reanalysis (R-2)
.
Bull. Amer. Meteor. Soc.
,
83
,
1631
1643
.
Korres
,
G.
,
N.
Pinardi
, and
A.
Lascaratos
,
2000
:
The ocean response to low-frequency interannual atmospheric variability in the Mediterranean Sea. Part II: Empirical orthogonal function analysis
.
J. Climate
,
13
,
732
745
.
Lee
,
S.-K.
, and
C.
Wang
,
2008
:
Tropical Atlantic decadal oscillation and its potential impact on the equatorial atmosphere–ocean dynamics: A simple model study
.
J. Phys. Oceanogr.
,
38
,
193
212
.
Moron
,
V.
,
R.
Vautard
, and
M.
Ghil
,
1998
:
Trends, interdecadal and interannual oscillations in global sea-surface temperatures
.
Climate Dyn.
,
14
,
545
569
.
Moura
,
A. D.
, and
J.
Shukla
,
1981
:
On the dynamics of the droughts in northeast Brazil: Observations, theory and numerical experiments with a general circulation model
.
J. Atmos. Sci.
,
38
,
2653
2675
.
Murtugudde
,
R. G.
,
J.
Ballabrea-Poy
,
J.
Beauchamp
, and
A. J.
Busalacchi
,
2001
:
Relationship between zonal and meridional modes in the tropical Atlantic
.
Geophys. Res. Lett.
,
28
,
4463
4466
.
Nobre
,
P.
, and
J.
Shukla
,
1996
:
Variations of sea surface temperature, wind stress, and rainfall over the tropical Atlantic and South America
.
J. Climate
,
9
,
2464
2479
.
North
,
G. R.
,
T. L.
Bell
,
R. F.
Cahalan
, and
F. J.
Moeng
,
1982
:
Sampling errors in the estimation of empirical orthogonal functions
.
Mon. Wea. Rev.
,
110
,
699
706
.
Penland
,
C.
, and
L.
Matrosova
,
1998
:
Prediction of tropical Atlantic sea surface temperatures using linear inverse modeling
.
J. Climate
,
11
,
483
496
.
Philander
,
S. G. H.
,
1990
:
El Niño, La Niña, and the Southern Oscillation. International Geophysics Series, Vol. 46, Academic Press, 293 pp
.
Plaut
,
G.
, and
R.
Vautard
,
1994
:
Spells of low-frequency oscillations and weather regimes in the Northern Hemisphere
.
J. Atmos. Sci.
,
51
,
210
236
.
Reynolds
,
R. W.
,
N. A.
Rayner
,
T. M.
Smith
,
D. C.
Stokes
, and
W. Q.
Wang
,
2002
:
An improved in situ and satellite SST analysis for climate
.
J. Climate
,
15
,
1609
1625
.
Ruiz-Barradas
,
A.
,
J. A.
Carton
, and
S.
Nigam
,
2000
:
Structure of interannual-to-decadal climate variability in the tropical Atlantic sector
.
J. Climate
,
13
,
3285
3297
.
Saha
,
S.
, and
Coauthors
,
2010
:
The NCEP Climate Forecast System Reanalysis
.
Bull. Amer. Meteor. Soc.
,
91
,
1015
1057
.
Saravanan
,
R.
, and
P.
Chang
,
2000
:
Interactions between the Pacific ENSO and tropical Atlantic climate variability
.
J. Climate
,
13
,
2177
2194
.
Seager
,
R.
,
Y.
Kushnir
,
P.
Chang
,
N.
Naik
,
J.
Miller
, and
W.
Hazeleger
,
2001
:
Looking for the role of the ocean in tropical Atlantic decadal climate variability
.
J. Climate
,
14
,
638
655
.
Servain
,
J.
,
I.
Wainer
,
J. P.
McCreary
, and
A.
Dessier
,
1999
:
Relationship between the Equatorial and meridional modes of climatic variability in the tropical Atlantic
.
Geophys. Res. Lett.
,
26
,
485
488
.
Servain
,
J.
,
I.
Wainer
,
H. L.
Ayina
, and
H.
Roquet
,
2000
:
The relationship between the simulated climatic variability modes of the tropical Atlantic
.
Int. J. Climatol.
,
20
,
939
953
.
Wu
,
Z.
, and
N. E.
Huang
,
2009
:
Ensemble empirical mode decomposition: A noise-assisted data analysis method
.
Adv. Adapt. Data Anal.
,
1
,
1
41
.
Wu
,
Z.
,
N. E.
Huang
, and
X.
Chen
,
2009
:
The multi-dimensional ensemble empirical mode decomposition method
.
Adv. Adapt. Data Anal.
,
1
,
339
372
.
Xie
,
P.
, and
P.
Arkin
,
1997
:
Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs
.
Bull. Amer. Meteor. Soc.
,
78
,
2539
2558
.
Xie
,
S.-P.
,
1999
:
A dynamic ocean–atmosphere model of the tropical Atlantic decadal variability
.
J. Climate
,
12
,
64
70
.
Xie
,
S.-P.
, and
Y.
Tanimoto
,
1998
:
A pan-Atlantic decadal climate oscillation
.
Geophys. Res. Lett.
,
25
,
2185
2188
.
Xie
,
S.-P.
, and
J. A.
Carton
,
2004
:
Tropical Atlantic variability: Patterns, mechanisms, and impacts. Earth’s Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 121–142
.
Zebiak
,
S. E.
,
1993
:
Air–sea interaction in the equatorial Atlantic region
.
J. Climate
,
8
,
1567
1586
.
Zhang
,
S.
,
M. J.
Harrison
,
A.
Rosati
, and
A.
Wittenberg
,
2007
:
System design and evaluation of coupled ensemble data assimilation for global oceanic climate studies
.
Mon. Wea. Rev.
,
135
,
3541
3564
.
Zhu
,
J.
,
B.
Huang
, and
M.
Balmaseda
,
2012
:
An ensemble estimation of the variability of upper-ocean heat content over the tropical Atlantic Ocean with multi-ocean reanalysis products
.
Climate Dyn.
,
doi:10.1007/s00382-011-1189-8
,
in press
.