Observations from TRITON buoys in the warm/fresh pool and a global ocean general circulation model are used to study the interannual variability of the equatorial western Pacific and the relationship between the zonal warm water transport, meridional convergence, and the warm water volume (WWV). The simulated temperature, salinity, and zonal warm water transport are validated with the mooring observations for the period 2000–14. The model results are then used to examine the WWV balance in ENSO cycles in an extended period from 1980 to 2014. It is shown that the zonal transport is highly correlated with meridional convergence and leads by about 4–5 months, and their phase offset determines the WWV changes. This result differs from the recharge paradigm in which the meridional convergence is supposed to be mainly responsible for the WWV changes. There is also no apparent change in relationship between zonal and meridional transports since 2000, unlike that between WWV and SST. The study suggests that the zonal warm water transport from the western boundary could have major implications for ENSO dynamics.
The El Niño–Southern Oscillation (ENSO) is the strongest interannual signal around the globe and has been the focus for many studies. A great deal of previous studies have demonstrated that the changes of warm water volume (WWV) associated with the changes of the depth of the main thermocline in the equatorial zone are closely linked with the ENSO cycle (Wyrtki 1985; Cane and Zebiak 1985; Zebiak 1989; Springer et al. 1990; Meinen and McPhaden 2000). Prior to El Niño, the main thermocline typically becomes deeper and the warm water builds up in the central-eastern equatorial Pacific. During El Niño, the main thermocline rebounds and the warm water is discharged to higher latitudes. As it leads SST evolution and plays an important dynamical role in ENSO cycle, WWV is believed to be a good predictor for ENSO. Jin (1997a,b) proposed an equatorial ocean recharge paradigm that describes how changes in WWV regulate El Niño and La Niña events. In this theory, the recharge oscillation relies on the nonequilibrium between the zonal-mean equatorial thermocline depth and wind stress. For example, the anomalous equatorial easterly (westerly) wind stress induces meridional convergence (divergence) of zonally integrated Sverdrup transport during cold (warm) events, which recharges (discharges) warm water from (to) the off-equatorial Pacific. The discharge and recharge of warm water as a result of anomalous equatorial winds transform the equatorial Pacific between the warm and cold phases.
The zonal flow in the equatorial western Pacific is fundamental to the zonal displacement of the warm/fresh pool (Picaut et al. 1996; Delcroix and Picaut1998; Bosc et al. 2009; Zhang and Clarke 2015). The existence of reversing jets in the equatorial western Pacific at the beginning of El Niño, for example, promotes the eastward migration of the warm/fresh pool, deepens the barrier layer, and suppresses the cool upwelling water (Picaut et al. 1996; Delcroix and Picaut 1998). The deep barrier layer provides a favorable condition for SST buildup and consequently gives rise to vigorous ocean–atmosphere interaction in the equatorial Pacific (Ando and McPhaden 1997; Maes et al. 2002, 2005). The role of salinity change in determining the zonal flow has also been noted (Zhang and Busalacchi 2009; Zhu et al. 2014).
Many previous studies have attempted to quantify major contributions to the WWV changes based on observations (Meinen and McPhaden 2000, 2001; Alory and Delcroix 2002; Bosc and Delcroix 2008) and numerical models (Brown and Fedorov 2010; Lengaigne et al. 2012; H.-C. Chen et al. 2015). These studies though have not reached a definite conclusion. Using gridded temperature data, Meinen and McPhaden (2001) found that the buildup of WWV prior to El Niño primarily resulted from the enhanced eastward transports across 156°E and weakened southward transports across 8°S. However, to derive geostrophic transports from temperature data alone, they had to use an empirical temperature–salinity (T–S) relation. Moreover, the gridded equatorial Pacific temperature analyses are based on rather sparse observations. Their geostrophic transport estimates therefore could be subject to a significant error. For example, they found that the diapycnal transport, the residual between the estimated net horizontal transport, and the rate of change of WWV often was as large as the horizontal transport. Using sea surface heights (SSHs) from satellite altimetry, Bosc and Delcroix (2008) emphasized the importance of meridional transport in controlling the WWV changes. They found that the WWV changes are a consequence of the opposite effect of the meridional Ekman and geostrophic transports. However, to derive volume transports from SSH-derived surface geostrophic currents, they too had to assume an empirical vertical velocity profile. Given their large potential error, it is difficult to characterize from observational analyses alone the WWV changes in relation to the zonal, meridional, and diapycnal transports.
The ocean general circulation models or global ocean reanalysis products have also been used to quantify the WWV changes (Brown and Fedorov 2010; Lengaigne et al. 2012; H.-C. Chen et al. 2015). H.-C. Chen et al. (2015) found that the western boundary transport generally is opposite the meridional transport, but the latter dominates the WWV changes. The western boundary transport in their study is defined as the integrated meridional transport from the west coasts to 140°E at 9°N and 160°E at 9°S, respectively. Brown and Fedorov (2010), on the other hand, found that the zonal and meridional transports are comparable and the vertical diapycnal transport is relatively small; the western boundary transport is defined as the zonal transport across 156°E. The volume transport from ocean models is conserved and therefore can be more easily quantified compared to the observational analyses. However, because of the lack of comprehensive observational network, validating the simulated zonal and meridional transports is very challenging.
El Niño events are commonly divided into two types (flavors). The eastern Pacific (EP) or canonical El Niño events mainly occur in the eastern equatorial Pacific where the cold tongue is located (Rasmusson and Carpenter 1982). In contrast, El Niño events in which anomalous SST are concentrated in the central equatorial Pacific around the date line are defined as central Pacific (CP) El Niño or El Niño Modoki (Kao and Yu 2009; Ashok et al. 2007). CP El Niños are more frequent in the 2000s. The relationship between WWV and SST also appears to have changed since 2000 (McPhaden 2012). Singh and Delcroix (2013) argued that the discharge–recharge paradigm only applies to EP ENSO, while a totally different signature shows in CP ENSO. Ren and Jin (2013), on the other hand, indicated that both EP and CP ENSO obey the same discharge–recharge oscillator mechanism, though CP ENSO has reduced feedback processes.
The objective of this study is to examine the relationship between the zonal transport from the western boundary and the WWV changes in the equatorial Pacific during ENSO. An extensive set of mooring observations in the warm/fresh pool in 2000–14 is used to validate the zonal warm water transport calculated from a global ocean model. Using model and observations, the role of the zonal and meridional transports in the context of the discharge–recharge paradigm is quantified. The rest of the paper is organized as follows: A brief description of the mooring data and numerical model is presented in section 2. In section 3, the temperature, salinity, and zonal transport from the model are compared with mooring observations. The characteristics of the zonal transport and contributions of temperature and salinity to its variability are analyzed in section 4. In section 5, the effects of zonal transport on meridional transports and WWV changes are examined for the 2000–14 period. To find the differences of transports and WWV changes between EP and CP ENSO, the relationships among zonal and meridional transports and WWV during 1980–99 are examined in section 6. Conclusions and discussions are given in section 7.
2. Data and model description
The 2000–14 daily temperature and salinity data at 137°, 147°, and 156°E meridians from Triangle Trans-Ocean Buoy Network (TRITON) moored buoys are provided by Tropical Atmosphere–Ocean (TAO) project office (http://www.pmel.noaa.gov/tao/). The locations of TRITON moored buoys are shown in Fig. 1. TRITON moorings measure temperature and salinity in the upper 750 m at 12 levels: 1, 25, 50, 75, 100, 125, 150, 200, 250, 300, 500, and 750 m (Ueki and Ando 2013). From temperature and salinity, the seasonal mean is calculated, and the geopotential anomaly ζ referenced to the depth of 500 m is obtained. The geostrophic flow Ug off the equator can be determined:
where β = df/dy, where f represents the Coriolis frequency.
The global ocean simulations are performed with the Parallel Ocean Program (POP; Smith and Gent 2002). A displaced pole grid is used in POP to avoid singularities at the North Pole. The latitudinal grid spacing varies smoothly in the Southern Hemisphere from 0.27° on the equator to over 0.5° at 80°S and in the Northern Hemisphere from 0.27° on the equator to about 0.5° at the midlatitudes and 0.1°–0.5° toward the North Pole. The longitudinal grid spacing is 1.125°. The vertical grid has 32 levels, with intervals varying gradually from 10 m at the surface to 500 m toward the bottom. The Gent and McWilliams parameterization (Gent and McWilliams 1990) is employed for horizontal tracer mixing, and the anisotropic viscosity scheme of Smith and McWilliams (2003) is used for the horizontal momentum mixing. Vertical mixing and convection coefficients are calculated using the Richardson number–dependent mixing scheme (Pacanowski and Philander 1981).
The World Ocean Atlas (WOA09; Locarnini et al. 2010; Antonov et al. 2010) climatological temperature and salinity fields are used for the model initial field. The surface heat flux and freshwater flux forcing are formulated by restoring sea surface temperature (SST) and sea surface salinity (SSS) to observations with a 30-day time scale rather than using the bulk formulae (Haney 1971; Kamenkovich and Sarachik 2004). A control run of 80 yr was performed with wind forcing from Hellerman climatological monthly mean wind stress (Hellerman and Rosenstein 1983) and buoyancy forcing from WOA09 climatological monthly mean SST and SSS. Following the control run, the model is forced from 1979 to 2014 by monthly wind stress (monthly average of daily means) derived from ERA-Interim (Dee et al. 2011) and monthly SST and SSS derived from the ECMWF Ocean Reanalysis System, version 4 (ORAS4; Balmaseda et al. 2013).
3. Characteristics of the equatorial western Pacific
The mean temperature structure in the equatorial Pacific is marked by a shoaling thermocline to the north (Wyrtki and Kilonsky 1984). The 20°C isotherm surface shoals from about 200 m at 5°S to about 150 m at 8°N (Ueki and Ando 2013). This characteristic of south–north tilting also exists in the mean salinity structure where the subsurface salinity maximum approximately coincides with the thermocline. The equatorial western Pacific is characterized by a large body of fresh surface water known as the fresh pool, and the water north of equator is fresher than in the south (Delcroix and Picaut 1998). Therefore, in the upper 100 m the salinity gradient is much larger than the temperature gradient. The mixed layer in this region is controlled by the halocline that inhibits mixing and causes a barrier layer between the bottom of the mixed layer and the top of the thermocline (Lukas and Lindstrom 1991). The surface salinity increases eastward, while the subsurface salinity decreases eastward. The simulated mean temperature and salinity structures are similar to the observations (not shown).
We focused first on the buoy observations. To identify the major interannual variations in the upper ocean, empirical orthogonal function (EOF) analyses were applied respectively to the seasonal-mean temperature, salinity, and geopotential anomaly in the upper 300 m with the three combined meridian sections. The zonal geostrophic velocity is then calculated from EOF of the geopotential anomaly using (1) and (2). There are large data gaps in several moorings at the 137° and 147°E sections during the first 2 yr (2000/01), which are filled with spatiotemporal interpolation. Figure 2 shows the first EOF modes for observed temperature, salinity, and zonal velocity in the upper 300 m, and Fig. 3 shows the corresponding principle components.
The first EOF mode of observed temperature explains about 66% of the total variance. The temperature fluctuations are concentrated in the thermocline (Fig. 2). The largest changes occur north of the equator at 137°E. The principal component indicates shoaling of thermocline during El Niño (2002/03, 2006/07, and 2009/10) and deepening of thermocline during La Niña (2008/09 and 2010/11; Fig. 3). The first EOF mode of observed salinity explains about 52% of the total variance and is mainly confined to the upper 100 m with the core centered at the equator. The salinity also exhibits distinct interannual variability and is highly coherent with the temperature variability (γ = 0.83). The water in the upper 100 m becomes fresher during El Niño and saltier during La Niña. The salinity changes are confined to the south of the equator, and the largest salinity changes are at 156°E.
The first EOF mode of observed geopotential anomaly explains about 74% of the total variance. The corresponding zonal geostrophic velocity anomalies show significant spatial variations (Fig. 2). The largest changes occur south of the equator (1°S) at 156°E and affect almost the whole upper 300-m layer. The changes north of the equator at 156°E show an obvious sheared structure that the variations above and below 50 m are of opposite sign. The amplitudes are relatively small at 147°E, which is likely due to missing buoy measurements south of the equator. At 137°E the amplitudes are large and of opposite sign north and south of 5°N. The zonal geostrophic flow exhibits strong interannual variability (Fig. 3) and is highly correlated with temperature (γ = −0.97) and salinity (γ = −0.81). The zonal flow strengthens (eastward) during El Niño and weakens (westward) during La Niña.
The EOF analyses were also applied to the simulated temperature, salinity, and zonal velocity. The model results are averaged to create seasonal means and are interpolated into a uniform grid with 1° spacing at the three sections from 8.5°N to the southern land boundary. The first EOF modes of simulated temperature and salinity explain about 63% and 37% of the total variance, respectively. The simulation reproduces well the observed temperature and salinity structures, showing strong temperature fluctuations in the thermocline and distinct salinity meridional gradients in the upper 100 m (not shown). The corresponding principle components are highly correlated with the observed (γT = 0.93, γS = 0.91; Fig. 3). The first EOF mode of simulated zonal velocity explains about 37% of the total variance. The largest amplitudes are confined to the upper 100 m. The principal component shows interannual variability comparable to the observed (Fig. 3). The model though shows stronger seasonal variability associated with the monsoon forcing off the coasts of Papua New Guinea, located outside the TRITON buoy network (Butt and Lindstrom 1994; Ueki et al. 2003).
4. Zonal transport across 156°E
The 156°E section has the most complete spatial and temporal coverage and shows the most remarkable interannual changes of salinity and zonal velocity. We therefore use the 156°E section to analyze the interannual variability of zonal geostrophic transport of the western boundary. The observed mean currents at 156°E are characterized by a strong Equatorial Undercurrent (EUC) with the core centered at 200 m and velocity maxima approaching 0.5 m s−1. The simulated mean currents show a similar EUC as well as a South Equatorial Current (SEC) south of the equator and a North Equatorial Countercurrent (NECC) around 6°N (not shown). The maxima velocities of SEC and NECC are around 0.15 m s−1.
The EOF analyses were applied, respectively, to the observed and simulated geopotential anomalies at the 156°E section. The first modes of the observed and simulated geopotential anomalies both explain about 81% of the total variance. The corresponding flow patterns are characterized by large amplitudes in the surface layer (0–100 m). The amplitudes are largest south of the equator in the observed, while the amplitudes are more symmetrical about the equator in the simulation (not shown). In the subsurface, both observation and simulation show variations of opposite sign north and south of the equator. The principal components indicate a dominant interannual variability, and the simulated and observed are highly correlated (γ = 0.94) (Fig. 4). Compared to the combined three-section analysis (Fig. 3), the observed currents remain basically the same, but the simulated currents from the single section are void of strong seasonal fluctuations.
The observed interannual variabilities of temperature, salinity, and zonal geostrophic velocity are well captured in the model simulation. The observed transport, however, is restricted to the geostrophic component. Since the zonal flows at or close to the equator are highly nonlinear, the transport estimate based on the geostrophy could be biased. The assumption of a level of no motion (500 m) could also affect the geostrophic transport estimate. To check the consistency, the warm water transport anomaly calculated from the observed geostrophic zonal velocity is compared with that calculated from the simulated total zonal velocity (Fig. 5). The warm water transport across 156°E is integrated from the surface to the 20°C isotherm and between 5°S and 8°N. A 15-month, low-pass filter is applied to remove the strong seasonal signals apparent in the first EOF mode of geostrophic currents (Fig. 4; Lengaigne et al. 2012; Singh and Delcroix 2013). The two transport estimates have comparable amplitudes and are highly correlated (γ = 0.90). This indicates that the geostrophic component dominates the total warm water transport. The transport anomalies are positive during El Niño and negative during La Niña.
The changes in zonal geostrophic velocity are accompanied by large changes of the upper-ocean salinity, suggesting that the salinity could have a significant control of the upper-geostrophic transport. As the interannual variabilities of zonal geostrophic velocity and salinity are concentrated in the upper 100 m, the upper-layer (0–100 m) geostrophic transports associated with the meridional temperature and salinity gradients are separately calculated. The geostrophic transport associated with the meridional temperature (salinity) gradient is estimated by replacing the time-varying salinity (temperature) with a climatological mean. The contributions of salinity and temperature are generally in phase, but their roles change over time (Fig. 6). The salinity plays an important role in determining the upper-geostrophic transport during the 2002/03 and 2006/07 El Niño events. The temperature, on the other hand, is primarily responsible for the upper-geostrophic transport during the 2009/10 El Niño. Zhang and Clarke (2015) emphasized the dominant role of meridional salinity gradient on the upper-geostrophic flows. In their study, the geostrophy is referenced to the isothermal layer depth (ILD), whereas the temperature contribution mostly arises from the thermocline.
5. Relationship between zonal transport and WWV in 2000–14
To learn the relationship between zonal transport and ENSO, the model results are used to calculate the transport balance of warm water volume (defined as the volume of warm water with temperatures greater than 20°C) within a box of 8°N–8°S and 156°E–80°W (e.g., Meinen and McPhaden 2001). Many previous studies have used a smaller box of 5°N–5°S in the transport calculation (e.g., Meinen and McPhaden 2000). To assure that our results are robust, we have repeated the transport calculation with the smaller box and found only minor difference between the two. A 15-month filter is applied to all model results to remove seasonal variations. The accumulation of warm water is a long-term (interannual) process that the seasonal variations contribute <10% of the total variance. The WWV anomaly is compared to the Niño-4 (5°N–5°S, 160°E–150°W) SST index (Fig. 7). The warm water volume gradually increases preceding El Niño and decreases sharply at the mature phase. The maximum lagged correlation coefficient between WWV and Niño-4 is 0.65 with WWV leading by 1 month. Using Niño-3.4 index yields a similar correlation coefficient (not shown).
The WWV is also compared with the observed and simulated sea surface height (SSH; Fig. 7). The observed SSH is derived from the Ssalto/Duacs gridded multimission altimeter products. The two SSHs are averaged over the same horizontal box as WWV. The SSH, which essentially reflects the 20°C isotherm, has been routinely used to estimate the WWV (Wyrtki 1985; Bosc and Delcroix 2008). The simulated SSH agrees well with the observed SSH (γ = 0.97) and is also highly correlated with WWV (γ = 0.89). This indicates that the interannual sea surface height variability in the equatorial Pacific mainly arises from dynamical variations. The outstanding agreement between observed and simulated SSH independently validates the simulated WWV.
Figure 8 shows the warm water transports across the three lateral sides of the box (the volume transport scale in Fig. 8a is twice that of Figs. 8b and 8c). For the study region, the inflow is defined as positive, and the outflow is defined as negative. The warm water transport across 156°E is generally positive (eastward), and it increases during El Niño and decreases, sometimes even reverses, during La Niña. The warm water transport across 8°N and 8°S are generally away from the equator. The meridional divergence typically increases during El Niño and decreases during La Niña; however, there is no strong relationship between the 8°N and 8°S sections.
The mean eastward water transport across 156°E is 11.7 Sv (1 Sv ≡ 106 m3 s−1), whereas the mean poleward transports are −19.1 and −12.2 Sv at 8°N and 8°S, respectively (Table 1). Thereby, there is a strong mean vertical transport across the base of the 20°C isotherm (D20). To determine the source of the upwelled water, the horizontal transports are also calculated above the fixed depth of 500 m below the core of EUC, where the mean vertical motion of interannual time scale is expected to be small. The new transport time series are included in Fig. 8 and the means are included in Table 1. The mean transports in the lower layer between D20 and 500 m are small across 8°N and 8°S. On the other hand, there is a large mean eastward flow across 156°E in the lower layer associated with shoaling of EUC. The mean zonal convergence below D20 supports the upwelled water that balances the mean meridional divergence above D20. The lower-layer transport is relatively steady, which contributes little to the interannual variability.
Figure 9 shows the zonal warm water transport anomaly across 156°E, the meridional warm water convergence anomaly, the corresponding net horizontal warm water transport, and the rate of change of WWV (the tendency, dWWV/dt). The zonal transport and meridional convergence have comparable amplitudes, and the two are closely linked with the former leading the latter by about 4 months (maximum lagged correlation γ = 0.89). The zonal warm water transport increases rapidly starting at the prior winter before the peak of El Niño, transitioning from a westward anomaly (discharge) to an eastward anomaly (recharge). The meridional convergence follows suit a few months later, starting as an equatorward anomaly (recharge) and then a poleward anomaly (discharge). Their phase offset leads to a prolonged period of recharging and WWV buildup prior to the peak of El Niño (Fig. 7). Immediately afterward, the zonal transport reverses to westward, while the meridional convergence remains poleward. The transition coincides with the peak zonal (westward) and meridional (poleward) transports, causing a sharp decrease of WWV and the emergence of La Niña. The net horizontal warm water transport is highly correlated with the warm water volume change with 1-month lag (maximum lagged correlation γ = 0.88; Fig. 9). The residual, the difference between net horizontal transport and WWV tendency, might be attributed to diapycnal transport. The interpolation error though could be a factor: in calculating warm water transport the velocity profile must be projected to the interpolated 20°C isotherm. The standard deviations for zonal transport, meridional transport, and residual are 8.4, 7.3, and 4.3 Sv, respectively. Obviously, the zonal warm water transport plays a major role in modulating the meridional warm water convergence and altering the warm water volume.
At the equator, the primary force balance is between the zonal wind stress and zonal pressure gradient (Fig. 10). While it is beyond the scope of this study to examine the momentum balance, it is useful to find out the relative contribution of the zonal pressure gradient caused by the temperature and salinity. The contribution of the zonal temperature (salinity) gradient at 156°E is obtained by replacing the time-varying salinity (temperature) with the corresponding climatological mean. The temperature and salinity gradients have opposite effects on zonal pressure gradient. The temperature dominates most of the time, whereas the salinity only offsets the temperature effect. This is consistent with the findings in the warm/fresh pool during the TOGA Coupled Ocean–Atmosphere Response Experiment (COARE; Cronin et al. 2000).
6. Relationship between zonal transport and WWV in 1980–99
To explore the differences between the two ENSO types, we also calculated the warm water transports and WWV for the period of 1980–99. Since no continuous TRITON mooring data are available during this time period, it is not possible to independently verify the simulated zonal transport. Figure 11 shows WWV, SSH, and Niño-3.4 SST index. The Niño-3.4 (5°N–5°S, 120°–170°W) SST index is used, as the 1980–99 period is marked by several major EP El Niño events, for example, 1982/83 and 1997/98. The maximum lagged correlation coefficient between WWV and Niño-3.4 is 0.76 with WWV leading by 4 months (for Niño-4, γ = 0.78, lag = 3 months). The lead relationship between WWV and SST is longer compared to the 2000–14 period, and the lag correlation coefficient is also higher. This result is in line with McPhaden (2012). The WWV and SSH are highly correlated with WWV leading by 1 month (maximum correlation = 0.98).
The zonal transport and meridional convergence anomalies (referenced to 1980–99) are shown in Fig. 12. The mean warm water transport across 156°E is 14.9 Sv, and the mean poleward warm water transports across 8°N and 8°S are −20.7 and −13.3 Sv, respectively. Both zonal and meridional transports are larger than during 2000–14 (Table 1). The baseline shift is perhaps associated with the change of sign of the Pacific decadal oscillation in the late 1990s (England et al. 2014). The meridional convergence and the zonal transport are closely linked, and the former lags the latter by about 5 months (maximum lagged correlation γ = 0.81). This result is similar to the 2000–14 period. The net horizontal warm water transport anomaly and the warm water volume change are also shown in Fig. 12. The net horizontal warm water transport anomaly and the warm water volume change have comparable magnitudes, and they are highly correlated with a 2-month lag (maximum lagged correlation γ = 0.84). The standard deviations for the zonal and meridional transports are about the same, 8.5 Sv, and are comparable to the 2000–14 period. This suggests that there is no apparent change in the relationship between the zonal transport, meridional convergence, and WWV between the two periods. The residual of about 7.5 Sv, however, is much more significant during 1980–2000. The larger residual is mainly caused by the net horizontal transport underestimating the WWV tendency whose standard deviation is about 11.3 Sv. The two time series though are highly correlated, suggesting that the residuals could be artificial (interpolation error; Fig. 12). Unfortunately, the horizontal transport cannot be independently verified because of lack of buoy observations during this period.
7. Discussion and conclusions
In previous observational studies, for example, Meinen and McPhaden (2001) and Bosc and Delcroix (2008), ad hoc assumptions are necessary in order to obtain the geostrophic transports from gridded temperature and altimeter sea surface height data. These observational analyses consequently could be subject to large potential error. In this study, the zonal geostrophic transport from the western boundary at 156°E is directly calculated from continuous temperature and salinity measurements in the closely spaced TRITON buoys for 2000–14. The observed zonal geostrophic transport shows large interannual variability, about 10 Sv, associated with ENSO. This indicates that the western boundary transport must be an important factor in modulating the warm water volume. Using a global ocean model, validated carefully with the observed warm water transport and sea surface height (a proxy for WWV), the warm water mass balance is examined. The model result, covering an extended period from 1980 to 2014, shows that the WWV changes are regulated by the zonal and meridional transports; contribution from the diapycnal transport appears to be secondary. Our result is consistent with Brown and Fedorov (2010) who showed from their model study that the zonal and meridional transports are comparable during the 1997/98 El Niño event.
Though the phase relationship between WWV and SST has changed since 2000, the role of zonal transport from the western boundary does not appear to differ during the entire model study period of 1980–2014. Previously, Singh and Delcroix (2013) found large differences in the WWV tendency between EP and CP events. Since their analysis is based on regression of model results to various Niño indices, it is difficult to tell if the difference simply reflects the fact that the WWV tendency is generally much larger during EP events, for example, 1982/83 and 1997/98. Moreover, Ren and Jin (2013) demonstrated that the discharge–recharge paradigm operates in both EP and CP ENSO, upon regression of model results with detrended CP indices. They did not compute the warm water mass balance; rather, the meridional and zonal gradients of SSHA were used as proxies for the eastward and poleward heat transports. Nevertheless, in their synthesis the meridional SSH gradient did lead zonal SSH gradient in both ENSO types (their Fig. 1).
In the discharge–recharge theory (Jin 1997a,b), the warm water transformation in the equatorial Pacific is mainly dependent on the meridional transport driven by wind stress curl. The zonal transport, on the other hand, is assumed to play only a secondary role in relaxing thermocline to its equilibrium position. Our result, in contrast, shows that the western boundary transport is of fundamental importance. Indeed, the WWV change appears to be controlled by the 4-month phase lead between zonal transport and meridional convergence. The strong western boundary transport suggests an off-equatorial forcing. Previously, Weisberg and Wang (1997) emphasized the impact of wind stress curl on off-equator thermocline in the western Pacific; though, they did not explore the oceanic link between the off-equator thermocline and WWV.
In this study, we apply a 15-month filter to remove short-term (<1 yr) variability. Our approach is in the spirit of classical theories in which ENSO is treated as a self-sustaining interannual oscillation. On the other hand, it is well recognized that the westerly wind bursts (WWBs) could have major impact on ENSO. D. Chen et al. (2015) recently proposed that the diversity of El Niño events mainly results from the westerly wind bursts. When WWBs are located over the equator, they could excite large downwelling equatorial Kelvin waves that deepen the thermocline in the central-eastern Pacific. WWBs could also generate strong surface currents that displace the edge of the warm/fresh pool eastward. Either effect would favor extreme El Niño events and give rise to different El Niño types. The off-equator WWBs have received less attention, but they too could have sustaining effects on charging the WWV (McGregor et al. 2016). Although WWBs are included in the model’s atmospheric forcing, their impact cannot be easily isolated in the presence of other short-term fluctuations. It is beyond the scope of this study to argue whether a particular westerly or easterly burst would initialize or inhibit El Niño events (Lengaigne et al. 2002; Levine and McPhaden 2016).
In summary, we use TRITON data combined with a global ocean model to investigate the characteristic of the western boundary transport and its relationship with ENSO. We show that the zonal and meridional warm water transports are an integral process in regulating the WWV changes. We find no obvious differences between EP and CP ENSO in the relationship of zonal and meridional warm water transports. We also demonstrate that the upper-layer salinity plays an important but not controlling role in the meridional and zonal pressure gradients. How the warm/fresh pool impacts or even regulates ENSO is an intriguing question worthy of careful further observational and model studies.
This work is supported by grants from the National Basic Research Program of China (2013CB430302), the National Natural Science Foundation of China (41321004), and the scientific research fund of the Second Institute of Oceanography of China (JG1510). The Hellerman wind product was downloaded from the IRI/LDEO Climate Data Library, and the ERA-Interim wind stress was obtained from European Centre for Medium-Range Weather Forecasts (ECMWF). The World Ocean Atlas and ECMWF ORA-S4 datasets were provided by the Asia–Pacific Data Research Centre (APDRC). The altimeter products were produced by Ssalto/Duacs and distributed by AVISO, with support from the Centre National d’Etudes Spatiales (CNES). All the Niño indexes were obtained from Earth System Research Laboratory (ESRL), supported from National Oceanic and Atmospheric Administration (NOAA).