Abstract

Using the newly available decade-long Argo data for the period 2004–13, a detailed study is carried out on deriving four-dimensional (4D) modality of sea temperature in the upper ocean with emphasis on its interannual variability in terms of amplitude, phase, and periodicity. Three principal modes with central periodicities at 19.2, 33.8, and 50.3 months have been identified, and their relationship with El Niño–Southern Oscillation (ENSO) is investigated, yielding a number of useful results and conclusions: 1) A striking tick-shaped pipe-like feature of interannual variability maxima, which is named the “Niño pipe” in this paper, has been revealed within the 10°S–10°N upper Pacific Ocean. 2) The pipe core extends downward from ~50 m at 130°E to ~250 m near the date line before tilting upward to the sea surface at about 275°E, coinciding nicely with the pathway of the Pacific equatorial undercurrent (EUC). 3) The double-peak zonal modality pattern of the Niño pipe in the upper Pacific is echoed in the subsurface Atlantic and Indian Oceans through Walker circulation, while its single-peak meridional modality pattern is mirrored in the subsurface North and South Pacific through Hadley circulation. 4) A coherent three-peak modal structure implies a strong coupling between sea level variability at the surface and sea temperature variability around the thermocline. Accumulating evidence suggests that Rossby/Kelvin wave dynamics in tandem with EUC-based thermocline dynamics are the main mechanisms of the three-mode Niño pipe in ENSO cycles.

1. Introduction

In addition to seasonal cycles (including annual and semiannual), interannual modes are probably the second strongest regime in global oceanic variability. In the tropical oceans, however, the situation could even be reversed: interannual variations may sometimes lead their annual counterparts for many geophysical variables. As far as sea surface temperature (SST) is concerned, the most typical example is the El Niño/La Niña phenomena, which occur quasi periodically every 2–7 years and may last for more than half of the overall time (e.g., Trenberth 1997). Numerous definitions and indices have been proposed to characterize these scientifically important and publicly concerned events (Hanley et al. 2003). Qualitatively, El Niño has been defined as the name given to the occasional return of unusually warm water in the normally cold water (upwelling) region along the Peruvian coast. It is also a Pacific basinwide increase in sea surface temperatures in the central and/or eastern equatorial Pacific Ocean and thus has three phases: warm tropical Pacific SSTs (El Niño), cold tropical Pacific SSTs (La Niña), and near-neutral conditions (e.g., Glantz 2001). Quantitatively, El Niño has been defined in terms of simple indices, among others, as corresponding to times when sea surface temperature anomalies in the Niño-3 region (see Fig. 4d) exceed 0.5°C or when SST anomalies in the Niño-3.4 region (see Fig. 4g) exceed 0.4° (e.g., Trenberth and Stepaniak 2001) or 0.5°C (e.g., Larkin and Harrison 2005). It is also found that there are considerable shifts in the origin of El Niños during the past decades. El Niño events tend to originate along the coast of South America and then migrate westward during 1951–72 (Rasmusson and Carpenter 1982). In the following two decades or so, El Niño events develop first in the central Pacific and then spread eastward (e.g., Wang 1995); thus, the Niño-3.4-based Niño indices (Bamston et al. 1997) become most popular for the last decade (Wolter and Timlin 2011). More recently, both cold tongue El Niños and warm pool El Niños have been identified, which further extend the origin to the western Pacific (e.g., Kug et al. 2009). Despite the diversity in El Niño/La Niña definition and their geographical pattern, it is helpful to keep in mind that these events occur about 55% of the time so that more often than not the tropical oceans are regarded as being in one phase or the other and average conditions are less common (Trenberth 1997). The El Niño/La Niña associated with the Southern Oscillation are known to be the dominant signals in SST variability, ranging from intraseasonal to multidecadal in the tropical Pacific (e.g., Chen and Li 2008; Chen et al. 2010a). The nature of complex year-to-year SST changes clearly requires a comprehensive characterization of the interannual variability at the air–sea interface.

It should be pointed out that almost all El Niño–Southern Oscillation (ENSO) indices presently used implicitly assume that El Niño/La Niña are surface-dominant phenomena, since sea surface temperature is universally used as the single or key variable in ENSO index definitions. This assumption, however, has been challenged by existing evidence from both field observations and model studies. As an early example, Guilderson and Schrag (1998) found that an abrupt shift in subsurface temperatures of the equatorial undercurrent (EUC) in the tropical Pacific is associated with changes in the 1976 El Niño. Chao et al. (2002) identified a curved surface similar to the depth of 20°C isotherm with maximum sea temperature anomaly in the tropical Pacific, along which the evolution of El Niño/La Niña trajectories are better revealed. The importance of subsurface signatures to ENSO was also noted by Zhang et al. (2009), who first looked at the ENSO asymmetry from the subsurface. They suggested that the subsurface temperature anomalies may be a good proxy as a measure of the relationship between ENSO variability and asymmetry, because there seems to be no correlation between SST variance and SST skewness. More recently, Meehl et al. (2011) found that, in decades with a slightly negative global mean surface temperature trend, the ocean above 300 m takes up significantly less heat whereas the ocean below 300 m takes up significantly more compared to non-hiatus decades. They further emphasized that a hiatus period is a relatively common climate phenomenon and may be linked to La Niña–like conditions. These studies imply that ENSO-related interannual activities may initially take place in the ocean interior with a delayed surface manifestation. Based on existing observations over the past two decades, we believe that variations of tropical air–sea interaction, underwater energy transmission, planetary wave propagation, and ocean current evolution all have a prominent subsurface component on interannual time scales and thus contribute substantially to the overall thermal variability of the upper ocean.

The recent availability of decade-long Array for Real-Time Geostrophic Oceanography (Argo) data, which provide temperature and salinity measurements of the upper ocean globally and operationally since 2004, allows us for the first time to perform a systematic analysis of the four-dimensional (4D; x, y, z, and t) structure of sea temperature variability on interannual time scales, leading to a number of inspiring results which are potentially important for the understanding of the mechanisms of El Niño/La Niña episodes along with warm pool/cold tongue activities, as well as the dynamics of tropical Rossby/Kelvin waves and equatorial current systems. The rest of the paper is organized as follows: A brief description of the Argo and SST data used in this investigation along with an extended modal extraction scheme is provided in section 2. The individual behaviors of the identified interannual modes and their combined effects are presented in parallel in sections 3 and 4, respectively, followed by a discussion on their possible mechanisms in section 5. Last, a summary with conclusions is given in section 6.

2. Data and method

Ten years of Argo data spanning from 2004 to 2013 are used in this study. Argo floats are designed to observe large-scale (seasonal and longer; 1000 km and larger) subsurface ocean variability globally (Roemmich et al. 2009). As the first observation system of global subsurface ocean in history and one of the main sources of in situ temperature–salinity (TS) measurements, Argo project has an unprecedented spatiotemporal sampling and coverage, the aim of which is to provide continuous TS observations of the ~0–2000-m upper ocean in near–real time. By January 2014, there were as many as 3613 active floats disseminated around the global ocean spaced at every 3° of longitude and latitude (http://www.argo.ucsd.edu/). In this analysis, gridded Argo temperature data are obtained from the China Argo Real-Time Data Center (http://www.argo.org.cn/). The original observations used to generate the product are the D-mode Argo data with pressure offsets corrected (Barker et al. 2011). There are 48 vertical layers in our dataset ranging from 5 to 1950 m with a spatial resolution of 1° × 1° and a temporal resolution of 1 month.

The SST data used in this analysis are extracted from an Extended Reconstructed SST (ERSST) product (version 3b) released by the NOAA/National Climatic Data Center (Smith et al. 2008). The ERSST product is generated on the basis of either the Comprehensive Ocean Atmosphere Dataset (COADS) or the International COADS (ICOADS) SST anomalies for its version 1 and versions 2–3, respectively. The SST dataset extracted from ERSST for our study is a monthly record with a global coverage at 2° × 2° grids for the period of January 2004 through December 2013 (http://www.ncdc.noaa.gov/ersst/).

The method used in this study is an extension of the three-dimensional frequency-varying harmonic analysis (Chen 2006). The accumulation of continuous time series from available Argo floats has exceeded one decade for the first time, which is approximately twice the normal ENSO cycle. According to the Nyquist criterion, such a dataset will allow major interannual signals to be reasonably resolved. The harmonic component of sea temperature with a periodicity of p for a given layer can be derived through Fourier analysis of the temperature time series T(x, y, z, t) at each grid point (x, y) of depth z, T(x, y, z, t) = A(x, y, z, p) × cos[2πt/p + φ(x, y, z, p)], where A is the amplitude of the p component, φ is the phase angle that determines the time when the maximum of the p harmonic occurs, and t varies from 0 to p months. Given a complete spatiotemporal dataset T(x, y, z, t, p), A(x, y, z, p), and φ(x, y, z, p) can be simultaneously retrieved and are supposed to carry full information of interannual sea temperature variability in terms of amplitude and phase for the global upper ocean.

3. Interannual modes: Individual behaviors

As a measure of the strength of oceanic variability, an “integrated spectrum” of sea temperature time series is derived for the interannual band (Fig. 1): that is, the harmonic amplitude with respect to depth and period,

 
formula

where N(zk) is the total number of valid grid points at depth zk. In the upper layers between 1 and 22 (corresponding to 0–220 m), a striking subsurface belt of maximum interannual variability is well defined. Three prominent ENSO-related modes centered around 90–100 m (as labeled with A, B, and C in Fig. 1) can be clearly identified. It is also obvious that the general trend of sea temperature variability exhibits an upward decrease with moderate surface manifestation, and a downward decrease with very little interannual change below layer 37 (corresponding to 500 m).

Fig. 1.

Period–depth diagram of layer-integrated spectrum of sea temperature variability for the interannual band derived from Argo data of 2004–13. The color scale depicts the globally averaged harmonic amplitude for each depth layer.

Fig. 1.

Period–depth diagram of layer-integrated spectrum of sea temperature variability for the interannual band derived from Argo data of 2004–13. The color scale depicts the globally averaged harmonic amplitude for each depth layer.

As a regional verification, the power spectrum analysis is performed at depths of 0, 10, and 100 m over the Niño-3.4 region for 2004–13 (Fig. 2a). It turns out that the three-mode structure remains unchanged at these depths. The exact periodicity for a given mode, however, is slightly shifted with depth: 17–20 months for mode A, 31–34 months for mode B, and 49–50 months for mode C (as can be observed in Fig. 2a). Moreover, it is apparent that the strength of the three principal modes increases systematically from the sea surface to ~100-m depth, consistent with what is revealed in Fig. 3a. This verification indicates that the three-mode vertical structure is robust and stable for the ENSO-sensitive region of the upper Pacific in terms of interannual variability.

Fig. 2.

(a) The modal structure of interannual sea temperature variability at depths 0 (black), 10 (blue), and 100 m (red) over the Niño-3.4 region for the period 2004–13. (b) As in (a), but for sea level (black) and outgoing longwave radiation (red) variabilities over the equatorial Pacific region (5°S–5°N) derived from merged satellite altimeter data and NOAA polar satellites data, respectively.

Fig. 2.

(a) The modal structure of interannual sea temperature variability at depths 0 (black), 10 (blue), and 100 m (red) over the Niño-3.4 region for the period 2004–13. (b) As in (a), but for sea level (black) and outgoing longwave radiation (red) variabilities over the equatorial Pacific region (5°S–5°N) derived from merged satellite altimeter data and NOAA polar satellites data, respectively.

Fig. 3.

(a) Recovered peak amplitude of sea temperature variability for identified interannual modes: A ( months), B ( months), and C ( months). (b) Dispersion diagram of central periodicity as a function depth corresponding to (a).

Fig. 3.

(a) Recovered peak amplitude of sea temperature variability for identified interannual modes: A ( months), B ( months), and C ( months). (b) Dispersion diagram of central periodicity as a function depth corresponding to (a).

At least two salient characteristics can be further derived for modes A–C: namely, peaking depth and period drift (as shown in Figs. 3a,b, respectively). As far as mode A (B or C) is concerned, the interannual variability is found to reach its maximum at the depth of 100 m (90 m) (Fig. 3a) and becomes very weak at about 200 m (see also Fig. 1: note that the peak corresponding to mode C becomes insignificant and cannot be effectively resolved below 220 m). Moreover, within the range of significant interannual variability, a period drift is evident for modes B and C: the former increases from 31 months at sea surface to 35 months at 90 m and remains unchanged until 300 m, with an average of 33.8 months, while the latter increases from 48 months at sea surface to 51–52 months below 70 m with an average of 50.3 months. In contrast, mode A is rather stable in periodicity with only a slight drop from 20 to 19 months at the depth of 20 m. It can therefore be argued that, as far as sea temperature is concerned, interannual variabilities mainly occur within the top 220-m subsurface layers and their intensities are depth dependent with a generally small frequency shift.

The spatial distributions of the amplitude and phase of sea temperature variability for the “top,” “peak,” and “bottom” layers of modes A–C are plotted in Figs. 4 and 5, respectively. As far as amplitude is concerned, ENSO signatures are apparent at all depths with maximum strength in the peak layers (Figs. 4b,e,h); highly consistent with Fig. 3a. It is worth noting that the ENSO signals at the sea surface derived from ERSST data (Figs. 4a,d,g being rather similar to the 5-m top layer of Argo measurement: not shown) are relatively weak compared to the internal layers of 90–100 m revealed by Argo data, implying that sea surface may not be the optimal layer for the diagnosis of El Niño episodes. Accordingly, the ENSO index might be improved by taking into account sea temperature anomalies in the upper ocean at ~100 m, which appears to be more sensitive than SST alone. Going through all subpanels in Fig. 4, one also finds an east–west contrast that is related to the cold tongue (Figs. 4a,b,d,e,g,h) and warm pool (Figs. 4b,c,e,f,h,i) El Niños, respectively, and the latter is found to be very active in recent decades after 1990 (Kug et al. 2009). For the conventional cold tongue El Niños in the eastern Pacific, unlike the surface situation, no obvious southern preference of associated high variability is apparent in the ocean interior, even the opposite is true for the peak layer (e.g., Fig. 4e). In fact, during the post-2000 hiatus of global warming, the Pacific trade winds intensified substantially (England et al. 2014). As a result, an increase in the equatorial upwelling and a spinup of the subtropical gyre are underway, leading to a cooling over the tropical central and eastern Pacific. This cooling extends poleward along the west coast of the Americas, giving rise to consistent enhancement of ENSO-related activities on both sides of the equator. All these features confirm that there are considerable differences in the pattern of ENSO-induced variabilities between surface and internal layers.

Fig. 4.

Global distribution of recovered harmonic amplitude of sea temperature variability for selected periodicities (p; months) and depths (z; m). Mode A: (a) p = 20, z = 0; (b) p = 19, z = 100; and (c) p = 19, z = 200. Mode B: (d) p = 31, z = 0; (e) p = 35, z = 90; and (f) p = 35, z = 200. Mode C: (g) p = 48, z = 0; (h) p = 51, z = 90; and (i) p = 51, z = 220. Here, (a),(d),(g) are derived from ERSST data, whereas the rest of the panels are derived from Argo data. The white boxes in (a),(d),(g) indicate the Niño-1/2 (0°–10°S, 270°–280°E), Niño-3 (5°S–5°N, 210°–270°E), Niño-4 (5°S–5°N, 160°–210°E), and Niño-3.4 (5°S–5°N, 190°–240°E) regions, respectively.

Fig. 4.

Global distribution of recovered harmonic amplitude of sea temperature variability for selected periodicities (p; months) and depths (z; m). Mode A: (a) p = 20, z = 0; (b) p = 19, z = 100; and (c) p = 19, z = 200. Mode B: (d) p = 31, z = 0; (e) p = 35, z = 90; and (f) p = 35, z = 200. Mode C: (g) p = 48, z = 0; (h) p = 51, z = 90; and (i) p = 51, z = 220. Here, (a),(d),(g) are derived from ERSST data, whereas the rest of the panels are derived from Argo data. The white boxes in (a),(d),(g) indicate the Niño-1/2 (0°–10°S, 270°–280°E), Niño-3 (5°S–5°N, 210°–270°E), Niño-4 (5°S–5°N, 160°–210°E), and Niño-3.4 (5°S–5°N, 190°–240°E) regions, respectively.

Fig. 5.

Global distribution of recovered harmonic phase of sea temperature variability for selected depths and periodicities corresponding to Fig. 3. The phase values in all subpanels are normalized to 12 months.

Fig. 5.

Global distribution of recovered harmonic phase of sea temperature variability for selected depths and periodicities corresponding to Fig. 3. The phase values in all subpanels are normalized to 12 months.

We further examine the corresponding phase evolution for modes A–C at different depths (Fig. 5). A solar-induced hemispheric opposition dominates the surface layer (Figs. 5a,d,g) for all three modes and is considerably weakened at depths of ~90–100 m (Figs. 5b,e,h) before vanishing below ~200 m (Figs. 5c,f,i). The large-scale westward propagation of the cold tongue signature originating from the upwelling along the Peruvian coast remains clearly evident at all selected depths for mode B (Figs. 5d–f). Following the fine structures of the tongue edges corresponding to the north and south equatorial currents, one finds that it takes at least ~30 months for the ENSO signals to spread from the eastern to western Pacific, forming a typical cycle of an El Niño event. A similar trans-Pacific migration is also visible for modes A and C, though it is less obvious. In the tropical Indian Ocean, an interesting characteristic is the reversal of propagation direction with respect to depth for mode A: shifting from largely northeastward within 0–50 m to southwestward at ~90 m (Figs. 5a,b). This might be a reflection of the upwelling/monsoon interaction: coastal upwelling generated in the eastern Indian Ocean is evident at depths around 100 m but is strongly modified by the Asian–Australian monsoons in the surface layers, thus preventing the Indian Ocean El Niño from being effectively diagnosed using sea surface variables alone. A joint view of Figs. 4 and 5 suggests that a sort of amplitude/phase coherence exists between corresponding subpanels: areas of high interannual variability usually associate with well-defined systematic propagations in the phase domain, such as the central and eastern tropical Pacific (Figs. 4b and 5b and Figs. 4h and 5h, respectively) and Indian Oceans (Figs. 4b and 5b). These are probably the real “hearts” of different El Niños/La Niñas where an efficient ENSO index should be targeted.

4. Interannual modes: Combined effects

Viewing from a zonal dimension, the latitude–depth-averaged amplitudes of recovered tropical sea temperature (30°S–30°N) as a function of longitude for combined interannual modes are shown in Fig. 6. As far as the upper tropical oceans are concerned, at least six ENSO-active sectors can be identified around P1 (141°E) and P2 (232°E) in the Pacific Ocean, I1 (48°E) and I2 (78°E) in the Indian Ocean, and A1 (317°E) and A2 (13°E) in the Atlantic Ocean. The existence of these peaks suggest that ENSO-like variability is geographically dependent in all major basins, with high, median and low strengths in the Pacific, Indian, and Atlantic Oceans, respectively. In the Pacific Ocean, P2 coincides with Niño-3 (see Fig. 4d) and represents the core region of the cold tongue El Niño, while P1 is to the west of Niño-4 and corresponds to the mean position of the recently intensified warm pool El Niño (Kug et al. 2009). A sensitivity test has been made in which the amplitude for the ENSO region (5°S–5°N) is overlaid on Fig. 6 (not shown). It turns out that the locations of the four major peaks (P1, P2, A1, and I1) remain largely unchanged, while the two minor ones (A2 and I2) are shifted to some extent, implying that the temperature variabilities in the ENSO region and tropical region are highly consistent to a large extent in the Pacific but to a less extent in the two other basins. Also note that the zonal pattern of depth-integrated amplitude of interannual sea temperature variability clearly shows that the definition of the traditional Niño regions based on surface variables only may not be fully effective for capturing the most sensitive zones of El Niños/La Niñas in the ocean interior. In addition, it is speculated that the locations of the three pairs of peak in Fig. 6 are linked to the ascending/descending branches of the Walker cells within each ocean basin, and the atmospheric Walker circulation is at least partially responsible for the double-peak structure of ENSO-related variabilities in each of the ocean basins.

Fig. 6.

Latitude/depth-averaged amplitude of recovered tropical sea temperature (30°S–30°N) as a function longitude for combined modes A–C. The range of depth used in the calculation is 5–220 m.

Fig. 6.

Latitude/depth-averaged amplitude of recovered tropical sea temperature (30°S–30°N) as a function longitude for combined modes A–C. The range of depth used in the calculation is 5–220 m.

Turning to the meridional dimension and confined to the Pacific Ocean, one finds a dominant primary ENSO zone centered at L0 (1°S), along with two secondary ones at L1 (33°S) and L2 (37°N) in the Southern and Northern Hemispheres, respectively (Fig. 7). This basic pattern exhibits a considerable consistency for the three individual modes (A, B, and C; not shown) in terms of longitude–depth-averaged amplitude of recovered sea temperature. Since the peaks corresponding to L0 and L1/L2 are located along ascending and descending branches of the atmospheric Hadley cell, they are likely to be associated with the effect of an “atmosphere bridge” (Alexander et al. 2002; Liu and Alexander 2007). It can also be readily confirmed that, under normal conditions, the zonal distribution of oceanic precipitation (see Fig. 2 of Chen et al. 2003) is very similar to Fig. 7. The fluctuations of sea temperature with large amplitude in the equatorial Pacific during El Niños provide anomalously great heat supply, which intensifies the Hadley circulation and makes it maintain more than the normal flux of angular momentum to the midlatitude belt of westerly winds. As a result, the sea temperature around ±35° may vary accordingly on an interannual basis to such an extent that a visible response to the Hadley circulation should occur. A stronger response appears in the North Pacific (see Fig. 7), where its baroclinicity is known to be greater.

Fig. 7.

Longitude/depth-averaged amplitude of recovered Pacific sea temperature (120°–285°E) as a function latitude for combined modes A–C. The range of depth used in the calculation is 5–220 m.

Fig. 7.

Longitude/depth-averaged amplitude of recovered Pacific sea temperature (120°–285°E) as a function latitude for combined modes A–C. The range of depth used in the calculation is 5–220 m.

Next, it is desirable to have a vertical view of the El Niño/La Niña–related structures along the two horizontal dimensions. As far as the Pacific Ocean is concerned, keeping in mind the integrated zonal and meridional patterns of interannual sea temperature variation (Figs. 6 and 7), a longitude–depth cross section along 1°S and two latitude–depth cross sections along 141° and 232°E are plotted for averaged modes A–C in Figs. 8a, 9a, and 9b, respectively. Combining the three panels, a striking “tick”-shaped variability maximum can be recognized across the upper equatorial Pacific. We call this well-defined dominant feature a “Niño pipe,” which extends from ~120° to 270°E in longitude and from ~0 to 250 m in depth (Fig. 8a) with a diameter of ~150–200 m (Fig. 9). For comparison, the current velocity along the same cross section of 1°S in the Pacific Ocean is derived from Simple Ocean Data Assimilation (SODA) data for the same period as shown in Fig. 8b. It appears that the core of the Niño pipe nicely follows the pathway of the EUC for the east part of the tick after the date line, but it is found to be shallower than the latter on the west side in the latitude–depth diagram (not shown). According to Izumo (2005), the EUC has a meridional width of about 200–400 km centered on the equator in the thermocline between about 50- and 200-m depths across almost the whole length of the basin. A close inspection of Figs. 8 and 9, however, reveals several deviations of the Niño pipe with regard to existing descriptions of the EUC, such as the asymmetry of its core region about the equator (Figs. 9a,b), the meandering of its main path outside the traditionally described 2°S–2°N band (McPhaden 1986), and the “ticking” nature of the feature near the date line (Fig. 8) instead of a pure tilt from west to east. These discrepancies may expand our understandings to the spatiotemporal structures of the EUC in greater details, especially in the context of global warming and its current hiatus.

Fig. 8.

(a) Averaged longitude–depth cross section of recovered harmonic amplitude of sea temperature variability along 1°S of the Pacific sector (120°–285°E) for combined modes A–C. (b) As in (a), but for current velocity derived from SODA data.

Fig. 8.

(a) Averaged longitude–depth cross section of recovered harmonic amplitude of sea temperature variability along 1°S of the Pacific sector (120°–285°E) for combined modes A–C. (b) As in (a), but for current velocity derived from SODA data.

Fig. 9.

Averaged latitude–depth cross section of recovered harmonic amplitude of sea temperature variability for combined modes A–C along (a) 141° and (b) 232°E of the Pacific Ocean. The vertical dashed lines indicate the equator.

Fig. 9.

Averaged latitude–depth cross section of recovered harmonic amplitude of sea temperature variability for combined modes A–C along (a) 141° and (b) 232°E of the Pacific Ocean. The vertical dashed lines indicate the equator.

5. Discussion

The eastern Pacific upwelling and western Pacific warm pool are two most sensitive zones as far as El Niño/La Niña are concerned. Li and Mu (1999) report that strong and continued positive subsurface sea temperature anomalies (100–200-m, 80–150-m, and 40–80-m depths in the western, central, and eastern equatorial Pacific, respectively) in the warm pool region occur prior to the El Niño event, and the collapse of an El Niño is directly associated with the eastward propagation of these anomalies. It is also observed that the thermocline in the western Pacific shoals by 20–40 m in 1997 because of Rossby waves excited by the initial weakening of the trade winds, and equatorial Kelvin waves with increasing amplitude propagating eastward across the Pacific depress the thermocline in the eastern Pacific by more than 90 m (McPhaden 1999). Very recently, England et al. (2014) argued that there is a building up consensus that the subsurface ocean, with its vast capacity for heat storage, is playing a significant role through enhanced heat uptake. In the central and western Pacific, there is evidence for increased equatorial pycnocline convergence of mass and heat, which drives a net heat gain in the ocean interior. Sea surface temperature has thus also changed as a result of the internal changes, with notable warming in the western Pacific warm pool and the western boundary and cooling over the tropical central and eastern Pacific.

Many previous studies suggest that EUC and El Niño/La Niña are closely coupled with each other. During the 1982/83 El Niño, when one of the most dramatic EUC observations was made, direct velocity measurements indicate the virtual disappearance of the undercurrent from September 1982 until January 1983 at 159°W (Firing et al. 1983) and during January and February 1983 at 110°W (Halpern 1987). Chavez et al. (1999) also pointed out a weak-to-strong shift of EUC during the 1997–99 exceptional El Niño/La Niña cycle of the past century. Using data from the Topical Atmosphere Ocean (TAO)/Triangle Trans-Ocean Buoy Network (TRITON) moorings and ocean general circulation models (OGCMs) for the period 1980–2002, Izumo (2005) investigated the role of EUC on El Niño formation and concluded that ENSO-related interannual variations of heat transport are mainly due to the EUC and the meridional overturning cells feeding it. Empirical orthogonal function (EOF) analysis was performed by Wang et al. (2009) to study the variability of the Pacific EUC and its relationship with ENSO using Topical Ocean and Global Atmosphere (TOGA), TAO, and SODA data. They show that the first two EOF modes correspond to the variations of the eastern and central Pacific EUC, respectively, with their time coefficients having positive and negative lags against the Niño index in terms of maximum correlation. Remember that the post-2000 warming hiatus is also linked with the fluctuation of the EUC (England et al. 2014). Estimates derived from reanalysis products suggest an acceleration of the equatorial jets and an increase in wind-driven Ekman divergence away from the equator over the past two decades. In the central and western Pacific, there is evidence for increased equatorial pycnocline convergence of mass and heat and an associated acceleration of the EUC.

Given the close spatial and dynamic relationships between EUC and El Niño/La Niña, it is natural to discuss them in the context of Niño pipe: The cold water upwelled in the eastern equatorial Pacific and diverged in the surface layer toward the subtropics during an El Niño event was originated from the EUC cold waters as a result of the subduction in the subtropics brought by meridional convergence in the pycnocline through the interior ocean or the western boundary currents (Goodman et al. 2005). Furthermore, the strength contrast between the Niño pipe and the conventional Niño regions at sea surface (see Figs. 4a,d,g) in terms of interannual sea temperature variability during a typical El Niño event (1.0°–1.5°C versus 0.5°–1.0°C; see also Figs. 8 and 9) suggest that the EUC could be reasonably viewed as a sort of “pumping pipe” for ENSO both geographically and geophysically.

To further understand the formation mechanism of the Niño pipe, we refocus on equatorial Rossby/Kelvin wave dynamics in relation with the EUC. At least three issues need to be clarified: 1) Why do the subsurface maxima of combined interannual sea temperature amplitude as represented by the Niño pipe coincide with the pathway of EUC in space (Fig. 8)? 2) Why is the core of the Niño pipe shifted slightly but significantly from the equator (Fig. 9)? 3) Why does the Niño pipe as well as the tropical Pacific zone have a three-peak structure in their interannual sea temperature modality (Figs. 1 and 3)? To examine the first issue, it is necessary to understand that Rossby waves play a critical role in the transient adjustment of ocean circulation to changes in large-scale atmospheric forcing (Chelton and Schlax 1996). Such forcing may include wind fetch and buoyancy heating/cooling at the eastern boundaries and over the ocean interior. Typical examples are intraseasonal Kelvin waves induced by westerly winds associated with Madden–Julian oscillation, off-equator annual Rossby waves generated by sea surface temperature anomalies near the Peru coast, interannual Rossby waves excited by the weakening of trade winds toward Indonesia and New Guinea (McPhaden 1999; Tozuka and Yamagata 2003), and so on. Such westward-propagating Rossby waves and eastward-propagating Kelvin waves with time scales from weeks to years are a common feature in the ocean and may introduce considerable upwelling/downwelling turbulences of similar frequencies in the equatorial Pacific, which may elevate or depress the thermocline by tens of meters (McPhaden 1999). More importantly, it is understood that the primary baroclinic mode of an internal Rossby wave has a velocity profile that changes sign at the depth of the thermocline, and variations of the sea surface property are mirrored as thermocline depth variations of the opposite sign with about three orders of magnitude greater amplitude (Gill 1982). Once the Rossby waves are generated, their systematic powerful penetrations naturally lead to the observed Niño pipe with a band of maximum interannual amplitude of sea temperature variability corresponding to the remarkable displacement of the thermocline (Fig. 8a). Coincidently, the EUC is known as a narrow ribbon of eastward flow centered on the equator in the upper thermocline (McPhaden 1986). The Niño pipe and the “EUC ribbon” are therefore overlapping with each other for a major part of their pathways in line with the tick-shaped thermocline (Figs. 8a,b).

To discuss the second issue, we need to go back to Fig. 9. It is observed that the core of the Niño pipe is centered at 3.75°N along the 141°E cross section (Fig. 9a) and at 2°S along the 232°E cross section (Fig. 9b). Moreover, instead of being confined to 2°S–2°N as the EUC (McPhaden 1986), the axis of the Niño pipe is found to meander gradually from about 4°N, 130°E to about 3°S, 270°E (not shown). This can be related to the finding of Chelton and Schlax (1996) that the latitudinal structure of sea level associated with the dominant tropical Rossby waves has a local minimum at the equator with symmetric maxima at about 4°N and 4°S. They point out that the observed westward propagation is much more apparent along 4°N and 4°S than along the equator. Consequently, the mirrored Niño pipe resulting from Rossby wave perturbation also exhibits a meridional shift within ±4°. Following the same mechanism and taking into account the annual and interannual ENSO interactions, Fig. 9 may also serve as a possible evidence of the off-equatorial warm and cold Rossby/Kelvin waves that propagate along two bands of 3°–5°N and 3°–5°S (Tozuka and Yamagata 2003).

Last, to examine the third issue raised above the modal structures of interannual sea level and outgoing longwave radiation (OLR) variabilities over the equatorial Pacific region (5°S–5°N) are derived from merged satellite altimeter data and NOAA polar satellites data using a methodology developed by Chen et al. (2010b) (Fig. 2b). Surprisingly, the three-peak pattern is also distinctly evident for both sea level and OLR with similar periodicities to modes A, B, and C: 18.2/19.0 versus 19.2 months, 28.5/33.0 versus 33.8 months, and 50.6/53.0 versus 50.3 months. In this cospectrum analysis of the tropical Pacific, OLR variability serves as a good proxy for the deep atmospheric convection conditions that generate atmospheric heating anomalies and force local and remote atmospheric circulation anomalies. The high correlation between OLR and upper-ocean sea temperature in their spectra suggests a direct oceanic thermal response to the atmospheric solar forcing. Specifically, the bandpassed time series of normalized amplitudes of OLR and sea temperature at 100-m depth for modes A–C are shown in Fig. 10. The well-defined phase reversal for all three modes suggests that atmospheric forcing is likely to be significant for each of the identified harmonic. Furthermore, as estimated by Chelton and Schlax (1996), a sea level variation of 5 cm may correspond to a thermocline displacement of about 50 m. Figure 2b thus also confirms that the primary sea level modality is transformed to sea temperature modality as a result of Rossby wave penetration into the thermocline. Combining existing findings with our own results, it is likely that ocean–atmosphere feedbacks and Rossby/Kelvin wave processes might be the major mechanism behind the identified Niño pipe.

Fig. 10.

Bandpassed time series of normalized amplitudes of OLR and sea temperature at 100-m depth for (a) mode A, (b) mode B, and (c) mode C.

Fig. 10.

Bandpassed time series of normalized amplitudes of OLR and sea temperature at 100-m depth for (a) mode A, (b) mode B, and (c) mode C.

6. Summary and conclusions

The past half century experienced an explosive expansion in the volume of oceanic and atmospheric data, thanks to the substantial advances in earth observation technologies such as satellite remote sensing and Argo floats. These new platforms and systems have greatly increased the domains of space, time and spectrum for earth observation in terms of resolution, duration, homogeneity, and continuity, leading to an unexpected “flooding” of big marine data in huge dimensions. An extended four-dimensional (x, y, z, and T) harmonic extraction scheme aimed at revealing spatially/temporally independent variability modes is adopted in this analysis, which has the advantage of being space–time decoupled and fully data adaptive. Ten years of Argo data with unprecedented quality for the period 2004–13 allows the 4D structure of interannual modality of upper-ocean temperature to be systematically derived, yielding a number of useful results and conclusions.

Three principal modes with central periodicities at 19.2, 33.8, and 50.3 months have been identified for the first time in the upper tropical ocean, and their relationship with ENSO is investigated in the context of EUC. A striking tick-shaped pipe-like feature of interannual variability maxima, named a “Niño pipe” in this paper, has been revealed within the 10°S–10°N upper Pacific. The pipe core extends downward from ~50 m at 130°E to ~250 m near the date line before tilting upward to the sea surface at about 275°E, coinciding nicely with the pathway of the Pacific EUC. The double-peak zonal modality pattern of the Niño pipe in the upper Pacific is echoed in the subsurface Atlantic and Indian Oceans through the Walker circulation, while the single-peak meridional modality pattern is mirrored in the subsurface North and South Pacific through Hadley circulation.

The 4D structure of interannual sea temperature modality in the Pacific Ocean is basically determined by heat budget in forms of warm pool and cold tongue evolution at the sea surface, while it is largely controlled, in the vast ocean interior and in the context of El Niño/La Niña cycle, by mass and energy redistribution associated with planetary waves and equatorial currents across the upper Pacific basin. A “resonant” three-peak modal structure implies a strong coupling between sea level variability at the surface and sea temperature variability around the thermocline. Accumulating evidence suggests that Rossby/Kelvin wave dynamics in tandem with EUC-based thermocline dynamics are the main mechanisms of the three-mode Niño pipe in ENSO cycles.

Acknowledgments

This research was jointly supported by the Global Change Research Program of China under project 2012CB955603, and the Natural Science Foundation of China under projects 41331172, U1406404, and 61361136001. Special thanks go to the China Argo Real-Time Data Center for providing us with the gridded Argo data product used in this study. The authors also thank the two anonymous reviewers for their thorough reviews and constructive suggestions on a previous version of this manuscript.

REFERENCES

REFERENCES
Alexander
,
M. A.
,
I.
Bladé
,
M.
Newman
,
J. R.
Lanzante
,
N.-C.
Lau
, and
J. D.
Scott
,
2002
:
The atmospheric bridge: The influence of ENSO teleconnections on air–sea interaction over the global oceans
.
J. Climate
,
15
,
2205
2231
, doi:.
Barker
,
P. M.
,
J. R.
Dunn
,
C. M.
Domingues
, and
S. E.
Wijffels
,
2011
:
Pressure sensor drifts in Argo and their impacts
.
J. Atmos. Oceanic Technol.
,
28
,
1036
1049
, doi:.
Bamston
,
A. G.
,
M.
Chelliah
, and
S. B.
Goldenberg
,
1997
:
Documentation of a highly ENSO-related SST region in the equatorial Pacific
.
Atmos.–Ocean
,
35
,
367
383
, doi:.
Chao
,
J.
,
S.
Yuan
,
Q.
Chao
, and
J.
Tian
,
2002
:
A data analysis study on the evolution of the El Niño/La Niña cycle
.
Adv. Atmos. Sci.
,
19
,
837
844
, doi:.
Chavez
,
F. P.
,
P. G.
Strutton
,
G. E.
Friederich
,
R. A.
Feely
,
G. C.
Feldman
,
D. G.
Foley
, and
M. J.
McPhaden
,
1999
:
Biological and chemical response of the equatorial Pacific Ocean to the 1997-98 El Niño
.
Science
,
286
,
2126
2131
, doi:.
Chelton
,
D. B.
, and
M. G.
Schlax
,
1996
:
Global observations of oceanic Rossby waves
.
Science
,
272
,
234
238
, doi:.
Chen
,
G.
,
2006
:
A novel scheme for identifying principal modes in geophysical variability with application to global precipitation
.
J. Geophys. Res.
,
111
,
D11103
, doi:.
Chen
,
G.
, and
H.
Li
,
2008
:
Fine pattern of natural modes in sea surface temperature variability: 1985–2003
.
J. Phys. Oceanogr.
,
38
,
314
336
, doi:.
Chen
,
G.
,
J.
Ma
,
C.
Fang
, and
Y.
Han
,
2003
:
Global oceanic precipitation derived from TOPEX and TMR: Climatology and variability
.
J. Climate
,
16
,
3888
3904
, doi:.
Chen
,
G.
,
B.
Shao
,
Y.
Han
,
J.
Ma
, and
B.
Chapron
,
2010a
:
Modality of semiannual-to-multidecadal oscillations in global sea surface temperature variability
.
J. Geophys. Res.
,
115
, C03005, doi:.
Chen
,
G.
,
Z.
Wang
,
C.
Qian
,
C.
Lv
, and
Y.
Han
,
2010b
:
Seasonal-to-decadal modes of global sea level variability derived from merged altimeter data
.
Remote Sens. Environ.
,
114
,
2524
2535
, doi:.
England
,
M. H.
, and Coauthors,
2014
:
Recent intensification of wind-driven circulation in the Pacific and the ongoing warming hiatus
.
Nat. Climate Change
,
4
,
222
227
, doi:.
Firing
,
E.
,
R.
Lukas
,
J.
Sadler
, and
K.
Wyrtki
,
1983
:
Equatorial undercurrent disappears during 1982-1983 El Niño
.
Science
,
222
,
1121
, doi:.
Gill
,
A. E.
,
1982
:
Atmosphere-Ocean Dynamics.
Academic Press, 662 pp.
Glantz
,
M. H.
,
2001
: Currents of Change: Impacts of El Niño and La Niña on Climate and Society. Cambridge University Press, 252 pp.
Goodman
,
P.
,
W.
Hazeleger
,
P.
de Vries
, and
M.
Cane
,
2005
:
Pathways into the Pacific equatorial undercurrent: A trajectory analysis
.
J. Phys. Oceanogr.
,
35
,
2134
2151
, doi:.
Guilderson
,
T. P.
, and
D. P.
Schrag
,
1998
:
Abrupt shift in subsurface temperatures in the tropical Pacific associated with changes in El Niño
.
Science
,
281
,
240
243
, doi:.
Halpern
,
D.
,
1987
:
Observations of annual and El Niño thermal and flow variations at 0°, 110°W, and 0°, 95°W during 1980–1985
.
J. Geophys. Res.
,
92
,
8197
8212
, doi:.
Hanley
,
D. E.
,
M. A.
Bourassa
,
J. J.
O’Brien
,
S. R.
Smith
, and
E. R.
Spade
,
2003
:
A quantitative evaluation of ENSO indices
.
J. Climate
,
16
,
1249
1258
, doi:.
Izumo
,
T.
,
2005
:
The equatorial undercurrent, meridional overturning circulation, and their roles in mass and heat exchanges during El Niño events in the tropical Pacific Ocean
.
Ocean Dyn.
,
55
,
110
123
, doi:.
Kug
,
J.-S.
,
F.-F.
Jin
, and
S.-I.
An
,
2009
:
Two types of El Niño events: Cold tongue El Niño and warm pool El Niño
.
J. Climate
,
22
,
1499
1515
, doi:.
Larkin
,
N. K.
, and
D. E.
Harrison
,
2005
:
On the definition of El Niño and associated seasonal average U.S. weather anomalies
.
Geophys. Res. Lett.
,
32
,
L13705
, doi:.
Li
,
C. Y.
, and
M. Q.
Mu
,
1999
:
El Niño occurrence and subsurface ocean temperature anomalies in the Pacific warm pool
.
Chin. J. Atmos. Sci.
,
23
,
513
521
.
Liu
,
Z.
, and
M.
Alexander
,
2007
:
Atmospheric bridge, oceanic tunnel, and global climatic teleconnections
.
Rev. Geophys.
,
45
,
RG2005
, doi:.
McPhaden
,
M. J.
,
1986
:
The equatorial undercurrent: 100 years of discovery
.
EOS, Trans. Amer. Geophys. Union
,
67
,
762
765
, doi:.
McPhaden
,
M. J.
,
1999
:
Genesis and evolution of the 1997-98 El Niño
.
Science
,
283
,
950
954
, doi:.
Meehl
,
G. A.
,
J. M.
Arblaster
,
J. T.
Fasullo
,
A.
Hu
, and
K. E.
Trenberth
,
2011
:
Model-based evidence of deep-ocean heat uptake during surface-temperature hiatus periods
.
Nat. Climate Change
,
1
,
360
364
, doi:.
Rasmusson
,
E. M.
, and
T. H.
Carpenter
,
1982
:
Variations in tropical sea surface temperatures and surface wind fields associated with the Southern Oscillation/El Niño
.
Mon. Wea. Rev.
,
110
,
354
384
, doi:.
Roemmich
,
D.
, and Coauthors
,
2009
:
The Argo program observing the global ocean with profiling floats
.
Oceanography
,
22
,
34
43
, doi:.
Smith
,
T. M.
,
R. W.
Reynolds
,
T. C.
Peterson
, and
J.
Lawrimore
,
2008
:
Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006)
.
J. Climate
,
21
,
2283
2296
, doi:.
Tozuka
,
T.
, and
T.
Yamagata
,
2003
:
Annual ENSO
.
J. Phys. Oceanogr.
,
33
,
1564
1578
, doi:.
Trenberth
,
K. E.
,
1997
:
The definition of El Niño
.
Bull. Amer. Meteor. Soc.
,
78
,
2771
2777
, doi:.
Trenberth
,
K. E.
, and
D. P.
Stepaniak
,
2001
:
Indices of El Niño evolution
.
J. Climate
,
14
,
1697
1701
, doi:.
Wang
,
B.
,
1995
:
Interdecadal changes in El Niño onset in the last four decades
.
J. Climate
,
8
,
267
285
, doi:.
Wang
,
H.
,
J.
Chen
, and
Y.
He
,
2009
:
Variation of equatorial undercurrent and its relationship with ENSO cycle (in Chinese with English abstract)
.
Acta Oceanol. Sin.
,
31
,
1
11
, doi:.
Wolter
,
K.
, and
M. S.
Timlin
,
2011
:
El Niño/Southern Oscillation behaviour since 1871 as diagnosed in an extended multivariate ENSO index (MEI.ext)
.
Int. J. Climatol.
,
31
,
1074
1087
, doi:.
Zhang
,
T.
,
D.-Z.
Sun
,
R.
Neale
, and
P. J.
Rasch
,
2009
:
An evaluation of ENSO asymmetry in the Community Climate System Models: A view from the subsurface
.
J. Climate
,
22
,
5933
5961
, doi:.