1. Introduction
Traditional descriptions of the evolution of ENSO focus on the role of upper-ocean temperature variations but neglect the role of salinity variations (e.g., Jin 1997). However, there is a growing appreciation that salinity variations may play an active role in the evolution of ENSO. Salinity stratification of the Pacific warm pool mixed layer can lead to development of a barrier layer, which can act to cut off subsurface entrainment cooling into the mixed layer (e.g., Lukas and Lindstrom 1991; Maes et al. 2002, 2005) and enhance warm pool displacements by concentrating the response to atmospheric wind forcing to a shallower layer (e.g., Lukas and Lindstrom 1991; Ando and McPhaden 1997; Vialard et al. 2002; Maes et al. 2006; Bosc et al. 2009). Development of the barrier layer can enhance zonal advective feedbacks at the edge of the warm pool during onset of El Niño (e.g., Picaut and du Penhoat 1997; Delcroix and Picaut 1998; Delcroix and McPhaden 2002; Maes et al. 2002, 2005). Subsurface salinity anomalies could also be a source of long lead predictability of El Niño because they can act as a long-lived source for density anomalies that appear in the equatorial waveguide that may trigger a coupled response (e.g., Yang et al. 2010; Hackert et al. 2011; Zhao et al. 2013).
Although a consistent picture is emerging of the evolution and possible active role of salinity variations during El Niño based on a range of observational analyses, datasets, and analysis periods, many of the above-mentioned studies have been based on limited record lengths because of the paucity of in situ salinity observations. Only since the early 2000s has there been good global sampling of subsurface salinity as a result of the Argo ocean profiling network (http://www.argodatamgt.org/) that allows for direct analysis of salinity variations at the edge of the warm pool (e.g., Maes et al. 2006; Bosc et al. 2009) and during ENSO (e.g., Qu et al. 2014). Remotely sensed observations of surface salinity are even more restricted, becoming available only since around 2010. Although consistent depictions of an active role for salinity variations during ENSO are emerging from these datasets (e.g., Qu and Yu 2014), a major drawback of this restricted data availability is that the full spectrum of ENSO behavior—including development of extreme events such as those that occurred in 1982 and 1997, the apparent increased frequency of central Pacific El Niño since 1999, and the asymmetry of El Niño and La Niña—has not been well diagnosed from the perspective of salinity variations.
As an alternative to the use of direct observations of salinity, ocean reanalyses that have become widespread both for initialization of coupled model seasonal prediction (e.g., Balmaseda et al. 2013) and for climate analysis (e.g., Carton et al. 2000) can provide a dynamically constrained description of salinity variations by making optimum use of sparse observations in a data assimilation cycle. However, the first generation of ocean assimilation systems focused primarily on assimilation of temperature (e.g., Smith et al. 1991; Balmaseda et al. 2008; Xue et al. 2011) often results in spurious or erroneous analyses of salinity (e.g., Balmaseda et al. 2008; Xue et al. 2011; Zhao et al. 2013, 2014). New ocean assimilation systems have been developed that better assimilate available in situ data while providing dynamically consistent analyses of temperature and salinity (e.g., Balmaseda et al. 2013; Zhao et al. 2013, 2014). We make use of one such new-generation ocean reanalysis: the ensemble-based ocean reanalysis system called the POAMA Ensemble Ocean Data Assimilation System (PEODAS; Yin et al. 2011), which has been developed to support seasonal climate prediction at the Bureau of Meteorology (Australia). PEODAS assimilates both temperature and salinity observations using time-varying error cross covariances (Yin et al. 2011) so that dynamically consistent salinity increments can be generated even in the absence of salinity observations.
Although we acknowledge that there is no completely satisfying substitute for the lack of global salinity observations prior to the Argo and satellite era, the PEODAS reanalyses have been shown to match available independent in situ observations well (Yin et al. 2011; Xue et al. 2012; Shi et al. 2016). Furthermore, we will show that the interannual variations of temperature and salinity associated with ENSO, which are the key focus of this study, are much larger than any trends in the reanalyses (either spurious because of changing availability of in situ observations or real because of climate variability). Thus, we have some confidence that the interannual variations that we detect for the period from 1980 to 2012, including the differences in El Niño behavior for canonical and extreme events and the contrast between El Niño and La Niña, are reflective of observed El Niño behavior.
Based on 33 years (1980–2012) of PEODAS reanalyses, this paper aims to document the spatial (three dimensional) and temporal structure of tropical Pacific salinity variations associated with ENSO and to identify the role of salinity variations, including the formation and evolution of the barrier layer, during different types of El Niño as well as the contrast between El Niño and La Niña. We will also assess the relative contribution to the near-surface salinity variations from atmosphere forcing [evaporation minus precipitation (E − P)] and ocean dynamical terms (advection) as well as the relative contribution by salinity to upper-ocean density variations, which is the ultimate link between salinity variability and ENSO dynamics. As an outcome of our focus on salinity variations, we will also support the notion that central Pacific El Niño, which is sometimes referred to as Modoki (e.g., Ashok et al. 2007), is best viewed as being more typical or canonical, whereas eastern Pacific El Niño, which is often referred to as canonical El Niño (e.g., Capotondi et al. 2015), is the exception (e.g., Takahashi et al. 2011), at least based on the currently available observational record. A brief description of PEODAS reanalyses, the method of empirical orthogonal function (EOF), and the definition of the barrier layer thickness (BLT) are presented in section 2. The spatial and temporal variations of salinity and the barrier layer associated with ENSO are shown in section 3. A composite of three main types of ENSO (strong El Niño, moderate El Niño, and La Niña) and the role of salinity during each will be presented in sections 4 and 5, respectively. Forcing of the salinity variation will be discussed in section 6, followed by the discussion and conclusions in section 7.
2. PEODAS reanalyses and the EOF method
a. PEODAS reanalyses
PEODAS uses the Australia Community Ocean Model 2 (ACOM2; Schiller et al. 2002), which is a global configuration of the Modular Ocean Model, version 2 (MOM2; Pacanowski 1995). PEODAS assimilates observations of in situ temperature T and salinity S from conductivity–temperature–depth (CTD); expendable bathythermograph (XBT; temperature only); and Argo profiles, sourced from the Enhanced Ocean Data Assimilation and Climate Prediction (ENACT), version 3 (hereafter EN3), analysis based on a quality-controlled database (Ingleby and Huddleston 2007). In the assimilation, temperature, salinity, and velocity fields are all updated at all model levels using flow-dependent, three-dimensional error cross covariances (Yin et al. 2011). During the assimilation cycle, the ocean model is driven by surface fluxes (stress, radiation, and freshwater) from ERA-40 (Uppala et al. 2005) prior to 2002 and NCEP–NCAR reanalyses (Kalnay et al. 1996) thereafter. In addition, sea surface salinity (SSS) is slowly relaxed back to climatology (1-yr relaxation time). Surface temperature is also strongly relaxed to daily observational analyses (Reynolds et al. 2002) with a 1-day relaxation time scale. The PEODAS reanalyses are available daily on a 2° × 0.5° grid. Here we use monthly mean analyses of temperature, salinity, and velocity in the upper 15 vertical levels (from the surface to 260 m) for the period 1980–2012.
Xue et al. (2012) showed that the temperature analyses from PEODAS have a comparable or better fit to the EN3 observational record as compared to other ocean reanalysis products. Zhao et al. (2014) further showed good agreement between monthly variations of analyzed salinity averaged over the top 300 m and the objective analysis of monthly input quality-controlled EN3. Shi et al. (2016) have recently shown that the PEODAS salinity analyses are comparable with the other major centers’ ocean reanalysis products that have contributed to the Ocean Reanalyses Intercomparison Project (ORA-IP).
b. EOF method
Our primary analysis tool is to conduct EOF analysis of temperature and salinity in the depth–longitude plane along the equator so as to highlight the coherent vertical and zonal variations associated with El Niño and La Niña. We perform the EOF analysis on monthly temperature and salinity anomalies for the period 1980–2012 (climatological seasonal cycle removed). The EOF analysis is done separately for temperature and salinity and is based on the correlation matrix so as to better capture coherent signals in the vertical. However, results are similar if we use the covariance matrix, with the primary difference being the order of the second and third EOFs of salinity are swapped. Instead of using combined EOF analysis, separate EOF analysis clearly shows that leading EOFs of salinity are distinct, ordered the same, and well correlated with temperature EOFs.
We display the spatial structure of the EOFs, which are based on standardized input data, by multiplying the eigenvectors by the standard deviation of the input data at each grid point and by the standard deviation of the associated principal component (PC), thereby yielding spatial patterns that have real units with magnitude typical of a one standard deviation anomaly of each EOF. We display coherent variability in other fields by regression onto the normalized principal components, thereby yielding anomalies with magnitude associated with a one standard deviation anomaly of each EOF.
c. Definition of the BLT
We diagnose the barrier layer thickness following the approach of Sprintall and Tomczak (1992): the barrier layer thickness is the difference in the depth of the isothermal layer and the depth of the mixed layer. The depth of the isothermal layer is given by the depth at which temperature decreases by 0.2°C compared to the temperature at the reference depth at 10-m depth. The closest level in PEODAS to this reference level is at the first model level (7.5 m). The depth of the mixed layer is given by the depth at which the density increases from that at 7.5 m by an equivalent change due to a 0.2°C temperature change.
3. EOF results
We begin by first displaying the mean and standard deviation of temperature and salinity with depth along the equator (Figs. 1b,d) and at the surface (Figs. 1a,c) in order to highlight the contrast of the distribution of variability between temperature and salinity. Variability of temperature in the western Pacific warm pool is seen to be relatively low, while the maximum standard deviation of surface temperature occurs in the eastern Pacific cold tongue. Surface salinity exhibits a mean minimum in the western Pacific (i.e., in the fresh pool), with a local minimum in salinity extending eastward along the northern branch of the ITCZ that merges with another fresh pool off the west coast of Central America. Maximum salinity occurs in the central Pacific and extends equatorward from the subtropical South Pacific, presumably reflecting waters that experience strong mean evaporation in the trade regimes. In contrast to temperature, surface salinity variability peaks in the western Pacific at the eastern edge of the fresh pool. In the vertical, mean temperature monotonically decreases with depth (Fig. 1b), but mean salinity along the equator (Fig. 1d) exhibits minima at the surface and maxima near the thermocline, which will have important ramifications for salinity variations induced by vertical displacements of the thermocline. However, salinity variability in the vertical is seen to peak near the surface (Fig. 1d). Together, Fig. 1 emphasizes distinctive differences in spatial distribution of salinity and temperature variability in the tropical Pacific.
Time mean (contours) and std dev (shaded) of (a) SST (°C), (b) temperature along equator (°C), (c) SSS (psu), and (d) salinity along equator (psu). The contour interval in (a) and (b) is 0.5°C above 28°C and 2°C below 28°C. The contour interval in (c) and (d) is 0.25 psu.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
The spatial structures (depth–longitude) of the leading three EOFs of temperature and salinity (calculated separately) along the equatorial Pacific are displayed in Fig. 2. Their explained variances are indicated and are repeated in Table 1, which also includes estimates of sampling uncertainty (North et al. 1982). Table 1 indicates that the first three EOFs for both temperature and salinity are well separated from each other and well separated from the respective EOF4. EOF4 (not shown) captures trends in both temperature and salinity, which are not the major focus of the present study, but the relatively small explained variance associated with EOF4 (7%) in comparison to the explained variance by the first three EOFs (>45%) indicates the dominance of the interannual variability captured in the first three EOFs.
Spatial patterns of EOF1–EOF3 for (a)–(c) temperature (°C) and (d)–(f) salinity (psu). The anomalies are shown by multiplying respective eigenvectors by a one std dev anomaly of each PC after renormalizing the eigenvector by the std dev of the input fields at each grid point. The black curves in (a)–(f) signify the 29° (edge of warm pool) and 20°C isotherms and the blue curves in (d)–(f) signify the 34.8-psu isohaline (edge of fresh pool). The variance explained by each EOF (%) is indicated.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
Eigenvalues (%; explained variance) and sampling uncertainty for the leading four EOFs of temperature and salinity.
Although the EOFs were computed separately for temperature and salinity, there is a one-to-one correspondence between the EOFs from each. That is, the zero-lag correlation of the principal component time series of EOF1 of temperature (PC1T) with EOF1 from salinity (PC1S) is 0.92. Similarly, the correlation of the second pair is 0.87 and for the third pair is 0.81 (these zero-lag correlations are also indicated in Fig. 3). Furthermore, these cross correlations peak at zero lag (not shown). However, we note that the eigenvalues (i.e., explained variance) of the salinity EOFs are generally smaller than those for temperature EOFs, indicating that the salinity variability has larger complexity than the temperature variability. Furthermore, as discussed below, the patterns of salinity EOFs are more complex and of smaller spatial scale compared to those from temperature. Nonetheless, we make the case that the leading three EOFs of salinity are distinctive and match one to one with their counterparts in temperature, and together the leading three EOFs capture the key features of El Niño evolution, including the contrasts between El Niño and La Niña and eastern Pacific/central Pacific El Niño.
Lag correlation of PC1T with PC2T, PC1S with PC2S, and PC1T with PC2S. The zero-lag correlations of PC1T with PC1S, PC2T with PC2S, and PC3T with PC3S are also indicated in the bottom right.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
EOF1 for temperature (EOF1T; Fig. 2a) has a dipole variation in temperature along the thermocline (indicated by the mean 20°C isotherm) and was identified with a similar EOF approach as Kumar and Hu (2014) as the tilt mode that depicts mature El Niño conditions (or La Niña with opposite sign). EOF2T (Fig. 2b), which has a more uniform sign along the thermocline across the basin with modest surface signature, depicts the discharge/recharge phase of El Niño (Jin 1997; Meinen and McPhaden 2000; Kumar and Hu 2014). It lags EOF1T by ~8 months (Fig. 3).
EOF1 for salinity (EOF1S; Fig. 2d) depicts the equivalent mature phase of El Niño in salinity. In contrast to EOF1T, which is characterized as a dipole across the basin, EOF1S is localized in the western Pacific mixed layer. Maximum loading straddles the mean edge of the fresh pool as indicated by the mean 34.8-psu isohaline, which typically delineates where the maximum zonal gradient in SSS occurs (e.g., Delcroix and McPhaden 2002; Maes et al. 2004; Maes 2008). Thus, mature El Niño conditions are associated with an eastward-expanded fresh pool (e.g., Qu and Yu 2014). The small salinity anomalies at depth along the thermocline associated with EOF1S can be understood as resulting from the vertical displacement of the thermocline as inferred from EOF1T acting on the mean vertical salinity gradient (Fig. 1).
EOF2S (Fig. 2e) depicts the salinity anomaly at the discharge (recharge) phase of El Niño, and it too lags EOF1S and EOF1T by ~8 months (Fig. 3). In contrast to EOF2T, which has limited surface signature (e.g., Jin and An 1999), EOF2S exhibits a pronounced near-surface west–east dipole structure, which can be viewed as a farther eastward displacement of the fresh pool at the end of the El Niño cycle (or conversely, with opposite sign, a buildup of the mixed layer fresh anomaly in the western Pacific prior to onset of El Niño). There is also a zonally symmetric salinity anomaly along the thermocline, which again can be understood by the implied vertical displacement of the thermocline captured by EOF2T acting on the mean salinity gradient.
The surface expressions of these EOFs are shown in Fig. 4 by regressing SST and SSS onto the PCs of salinity. We also include the regression of surface wind stress and upper-ocean currents (mean over the top 50 m). Because of the high correlation of the EOFs of salinity and temperature, these plots are nearly identical to the regressions onto the corresponding PCs of temperature.
(a)–(c) SST (°C; shaded) with surface wind stress (N m−2; vectors with scale of 0.12 × 10−1 N m−2; only vectors with magnitude >0.25 × 10−1 N m−2 are displayed) and (d)–(f) SSS (psu; shaded) with upper 50-m currents (cm s−1; vectors with scale of 10 cm s−1; only vectors with magnitude >2.5 cm s−1 are displayed) associated with PC1–PC3 of salinity. The anomalies are computed as a regression onto the normalized PCs.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
For EOF1, positive SST anomalies in the central and eastern Pacific at the mature phase of El Niño are associated with convergent westerly winds in the western and central Pacific (Fig. 4a) and eastward current anomalies (Fig. 4d), which are typical of mature El Niño conditions in the central Pacific (e.g., Jin and An 1999). We note that peak surface warming associated with EOF1T is in the central-eastern Pacific as opposed to closer to the South American coast as sometimes results from EOF analyses of surface temperature. The westerly upper-ocean current anomalies associated with EOF1T, which are interpreted to be the direct response to the anomalous westerly wind forcing (e.g., Jin and An 1999), appear to play a role in advecting the fresh pool eastward, thus giving rise to the strong negative SSS anomaly along the equator in the western Pacific (Fig. 4d). The specific role for advection and freshwater forcing of the salinity EOFs is addressed further below.
For the discharge phase (EOF2), a modest cold SST anomaly spans the entire basin (Fig. 4b), the wind anomalies are weak, and the surface current anomalies are now westward (Fig. 4e), which reflects the geostrophic response to the reversal in the meridional gradient of heat content (minimum now on equator) at the peak of El Niño and which acts to discharge heat off of the equator after the peak of El Niño (e.g., Jin and An 1999). The fresh SSS anomaly is shifted farther east as compared to mature El Niño conditions, and positive SSS anomalies are now evident in the far west. Again, a role for advection of the mean zonal salinity gradient by the anomalous currents is suggested and will be quantified below.
EOF3 (Figs. 2c,f and 4c,f) depicts a farther shift of the warm/fresh westerly anomalies into the eastern Pacific as well as eastward-shifted westerly currents along the equator. The appearance of the positive salinity anomaly in the far-western Pacific due to strong loading on EOF3 even suggests an eastward detachment of the fresh pool in years of strong loading onto EOF3. As will be discussed below, EOF3 is interpreted to capture behavior of extreme El Niño.
The temporal behavior of EOF1 (i.e., its principal component) along with its association with eastward expansion of the warm/fresh pool during El Niño (and contraction during La Niña) is displayed in Fig. 5. Here we show the time series of PC1T superposed on the equatorial Hovmöller plot of total SST (but only shaded for SST > 28.5°C; Fig. 5a) and PC1S superposed on the total SSS field (but only shaded for SSS > 34.3 psu; Fig. 5b). We will discuss the barrier layer variation shown in Fig. 5c below. This display of PC1 on top of the total SST and SSS fields emphasizes that El Niño is well described as an eastward shift of the warm/fresh pool such as that occurred during 1982, 1987, 1992, 1994, 1997, 2002, and 2009 (these years refer to the year in which the event began). Similarly, La Niña is well described as a westward contraction such as that occurred in 1988, 1998, 1999, 2007, 2010, and 2011. The two super El Niño years (1982 and 1997) stand out as the only times that the warm pool (SST > 28.5°C) and fresh pool (SSS < 34.7 psu) extended across the entire Pacific.
Hovmöller (time–longitude) plots of equatorial 5°N–5°S (a) SST (°C; shaded for SST > 28.5°C), (b) SSS (psu; shaded for SSS > 34.3 psu), and (c) BLT (m; shaded for BLT > 10 m) for 1980–2012. The black curve in (a) is a time series of PC1T (normalized units on right-hand y axis) and in (b) is a time series of PC1S. These curves using the same y axis are repeated in (c) with PC1T in black and PC1S in green.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
We do not display the temporal behavior of PC2, but its behavior can be inferred by the strong cross correlation with PC1 (e.g., Fig. 3). That is, PC2 has large positive loading 6–9 months after PC1 peaks and large negative loading 6–9 months before PC1 peaks.
The temporal behavior of EOF3 is displayed in Figs. 6a and 6b superposed on the principal component of EOF1. For both PC3T and PC3S, large positive excursions occurred only twice, during the super El Niños in 1982 and 1997. On the other hand, moderate negative excursions occur during moderate El Niño events such as 1987, 1992, 1994, 2002, and 2009. Also, moderate positive excursions occur during La Niña such as those in 1988, 1998, 1999, 2007, 2010, and 2011.
Time series of (a) PC1T and PC3T and (b) PC1S and PC3S (standardized units). The scatter of monthly standardized values of PC3 vs PC1 is shown for (c) PCT and (d) PCS.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
The scatter of PC3 versus PC1 (Figs. 6c,d) demonstrates the distinctive difference in behavior between La Niña/moderate El Niño and super El Niño. For moderate El Niño and La Niña (i.e., PC1 ≤ 1), there is a negative linear relationship between PC1 and PC3. Taking EOF1 and EOF3 together (e.g., Figs. 4a,c,d,f) indicates that La Niña and moderate El Niño are more focused in the central Pacific, thus representing modest westward and eastward displacements of the warm/fresh pool in the western Pacific. However, for strong or super El Niño (i.e., PC1 > 2), the relationship between EOF1 and EOF3 flips to being strongly positive such that super El Niño is associated with an emphatic eastward expansion of the warm/fresh pool across the Pacific.
This interpretation of the differences between strong El Niño and moderate El Niño/La Niña is similar to Dommenget et al. (2013). However, following Takahashi et al. (2011), we group together the behavior of moderate El Niño and La Niña to represent a typical or canonical El Niño that is more focused in the central Pacific. In contrast, El Niño that is most concentrated in the eastern Pacific (captured by EOF3) reflects the nonlinear extreme of El Niño and is considered to be the exception rather than a reflection of canonical behavior. This distinction between the linear behavior of moderate El Niño/La Niña that is concentrated in the central Pacific (and is sometimes referred to as Modoki) and the nonlinear behavior of super El Niño that is shifted into the eastern Pacific is brought out even more strongly with the salinity EOFs than for the SST EOFs as used by Takahashi et al. (2011).
4. Canonical evolution of moderate El Niño/La Niña and super El Niño
This contrasting evolution of the EOFs for moderate El Niño/La Niña and super El Niño is further highlighted by forming composites of the leading three PCs for each case. We define events based on the PC1T anomaly averaged in December–February and identify moderate El Niño (1 < PC1T < 2) in 1987, 1992, 1994, 2002, and 2009; La Niña (PC1T < −1) in 1988, 1998, 1999, 2007, 2010, and 2011; and super El Niño (PC1T > 2) in 1982 and 1997 (the years here refer to December, which we call year 0). Although only two super El Niño events occurred in the record used here (1982/83 and 1997/98), forming the composite based on two events is justified because the evolution of the leading principal components in each event is similar (e.g., Fig. 6) and markedly distinguishable from the other moderate El Niño events [see also Fig. 4 in Takahashi et al. (2011)]. We form composites of the PCs of temperature and salinity, but only show the composites based on the temperature PCs because they are nearly identical to those based on salinity. The PCs are expressed in normalized units.
These composites of the PCs are displayed in Fig. 7. For super El Niño, PC1 and PC3 develop in unison and peak late in year 0, indicating that maximum warmth/freshness is shifted into the eastern Pacific. The recharge phase (negative PC2) strongly leads this development by 8–10 months, and the peak discharge (positive PC2) strongly occurs 6–8 months after the peak of the event at the end of year 0. Accompanying the strong discharge, PC1 swings negative in the following year, indicating a tendency for La Niña events to follow strong El Niño (e.g., Cai et al. 2014).
Composite evolution of PC1T, PC2T, and PC3T for (a) strong El Niño, (b) moderate El Niño, and (c) La Niña. Composites are displayed from January of year 0 to December of year 1. Units are standardized.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
In contrast, moderate El Niño (Fig. 7b) has near-zero anomaly in PC3 prior to the peak of the event, which occurs slightly later than for super El Niño (i.e., PC1 peaks around February of year 1). But PC3 then peaks negative some 3–4 months later (i.e., in June of year 1), thus capturing a westward shift of the warm/fresh anomaly in the central Pacific. As for super El Niño, the development of PC1 for moderate El Niño is led by the recharge phase (i.e., negative PC2); however, there is a less obvious preceding peak in PC2, and the peak lag with PC1 is much shorter (i.e., 2–3 months in contrast to 8–10 months for strong El Niño), indicating less predictive capability of these moderate events that are more concentrated in the central Pacific (e.g., McPhaden 2012; Capotondi et al. 2015). Furthermore, PC1 decays back to zero by the end of year 1, thus indicating little tendency for a swing to La Niña after a moderate El Niño. La Niña exhibits nearly opposite behavior to moderate El Niño, with near-zero anomaly in PC3 and weak positive PC2 (i.e., discharge) prior to the peak of the event, followed by a positive excursion of PC3 some 3–4 months after the peak negative phase of PC1 and weak recharge (i.e., weak swing to negative PC2).
The spatial–temporal evolution of super El Niño, moderate El Niño, and La Niña is depicted by compositing the reconstructed surface temperature and salinity fields based on the respective leading three EOFs. We also form the vertical mean temperature anomaly from the surface to 250 m (indicative of the upper-ocean heat content) by vertically integrating the reconstructed temperature anomaly using the three EOFs of temperature. The composites of these reconstructed fields are formed as above for the strong El Niño, moderate El Niño, and La Niña events and displayed as Hovmöller plots along the equator from January of year 0 to December of year 1 (Fig. 8).
Composite evolution of (top) SST (°C; shaded) and vertically integrated temperature anomalies for 0–250 m (°C; contours with an interval of 0.5°C) and (bottom) SSS anomalies (psu; shaded) and total BLT (m; contours with an interval of 10 m). Composites are formed for (a),(b) strong El Niño; (c),(d) moderate El Niño; and (e),(f) La Niña. Anomalies are averaged over 5°N–5°S and shown monthly from January of year 0 to December of year 1 (see text).
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
For both surface salinity and temperature, moderate El Niño and La Niña are seen to be best described as central and western Pacific phenomena. Extension and concentration of El Niño into the eastern Pacific occurs only during rare super El Niños. Moderate El Niño and La Niña display nearly equal and opposite modest eastward expansion and westward contractions of the western Pacific warm/fresh pool, while super El Niño shows a distinctive detachment of the fresh pool from the western Pacific and extension into eastern Pacific together with peak temperature anomaly occurring in the eastern Pacific. The contrast in the evolution of the subsurface temperature as discussed in Fig. 7 is also well depicted in Fig. 8. Super El Niño is more strongly associated with a preceding recharge of heat 8–10 months prior and a subsequent strong discharge of heat 8–10 months after, while moderate El Niño exhibits a much shorter time scale for the preceding recharge and a much weaker and shorter period of discharge (e.g., Capotondi et al. 2015).
5. Role of salinity for ENSO evolution
One way to assess the possible role of salinity variations for evolution of ENSO is to compute the salinity contribution to the density perturbations associated with each EOF. We form the density anomaly implied by each pair of temperature and salinity EOFs using a linearized equation of state (e.g., Gill 1982) and then highlight where the relative salinity contribution to the density anomaly is large (Fig. 9). We see that for EOF1 (tilt mode), the east–west dipole density anomaly along the thermocline (reduced density in the east where the thermocline is anomalously deep and increased density in the west where the thermocline is anomalously shallow) has little contribution from salinity. However, the localized region of enhanced stability (i.e., reduced density) around the date line, which is associated with the eastward expansion of the warm/fresh pool during El Niño, has a primary contribution from salinity. For EOF2 (discharge or recharge), the zonally symmetric density anomaly along the thermocline is primarily due to the temperature perturbation, but in the far-western Pacific, the positive density anomaly in the mixed layer has a primary contribution from salinity. EOF3, which captures the extreme eastward shift of super El Niño and the westward displacement of moderate El Niño and La Niña, also has a density perturbation in the mixed layer in the far-western Pacific that is primarily the result of the salinity anomaly.
Equatorial (5°N–5°S) density anomalies (kg m−3; contours with an interval of 0.1 kg m−3, negative anomalies dashed) and the relative contribution from salinity (%; shaded) associated with EOF1–EOF3.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
These density anomalies in the mixed layer of the western and central Pacific have been hypothesized to affect El Niño in two distinctive fashions. First, the enhanced stability provided by a freshened mixed layer at the edge of the warm pool prior to and during El Niño implies that impulsive wind forcing will be trapped in the mixed layer and thus result in a stronger dynamical response (e.g., westerly forcing during El Niño will result in stronger near-surface westerly current anomalies, thus producing a stronger eastward shift of the warm pool; e.g., Vialard et al. 2002). Second, the enhanced stratification provided by the salinity anomaly in the mixed layer can result in an enhanced barrier layer, whereby the mixed layer that is controlled by salinity stratification will become more isolated from the bottom of the deeper isothermal layer (Lukas and Lindstrom 1991). A greater barrier layer thickness will act to maintain the surface warm anomaly at the edge of the warm pool during development of El Niño against cooling effects of vertical entrainment and so possibly lead to stronger El Niño development.
The temporal evolution of the barrier layer thickness along the equator is displayed in Fig. 5c, where we have shown positive barrier layer values superposed on the time series of PC1T and PC1S in order to emphasize concomitant shifts of the barrier layer with the east–west shifts of the warm/fresh pool. Synchronized eastward movement of enhanced barrier layer thickness at the edge of the warm pool during El Niño is evident (e.g., Ando and McPhaden 1997; Bosc et al. 2009), as is the contraction westward to the far-western Pacific during La Niña. The barrier layer is also seen to disappear at the demise after the peak of the two super El Niño events.
The composite evolution of the barrier layer thickness for super El Niño, moderate El Niño, and La Niña are displayed in Figs. 8b, 8d, and 8f. The close correspondence of the enhanced barrier layer thickness with the fresh salinity anomaly during El Niño is evident, as is the tendency for enhanced barrier layer thickness to develop in the far-western Pacific 6–12 months prior to onset of El Niño (e.g., Maes et al. 2005). The barrier layer tends to disappear after super El Niño but to contract westward after the peak moderate El Niño.
Another clear distinction between super El Niño and moderate El Niño is the tendency for enhanced barrier layer thickness to shift continuously into the eastern Pacific during evolution of super El Niño but be confined to the edge of the western Pacific fresh pool during moderate El Niño. There is also an asymmetry between El Niño and La Niña, with the barrier layer contracting to the western Pacific during La Niña but showing very little temporal evolution during the event, whereas for El Niño there is a tendency for development of enhanced barrier layer in the far west prior to El Niño that then shifts eastward and strengthens as the warm pool expands eastward during El Niño (for both moderate and super). Finally, there is an apparent tendency for a stronger, more concentrated region of enhanced barrier layer in the far-western Pacific to precede super El Niño as compared to moderate El Niño, which perhaps could be a distinguishing feature for development of super El Niño.
The amplitude of the barrier layer variations generally agree with those diagnosed by Qu and Yu (2014), who made use of a gridded analysis based on Argo and satellite observations of salinity. However, they show a much stronger and concentrated contraction of the barrier layer into the western Pacific during La Niña. Such difference may be due to different samples of La Niña (we drew from 1980–2012 while their study was limited to 2005–13). However, we also see in Fig. 5c that the barrier layer is clearly smaller before 2002 and larger after 2002 during La Niña, which might reflect decadal changes due to the swing of the Pacific decadal oscillation to a stronger Walker circulation and more contracted western Pacific fresh pool since the end of the twentieth century (e.g., England et al. 2014).
6. Causes of salinity variations
We assess the cause of the near-surface salinity variations associated with the leading three EOFs of salinity by regressing the salt flux (i.e., evaporation minus precipitation) and total advective salinity tendency (horizontal and vertical) in the top 50 m onto the standardized PCs. The mixed layer advective tendency is computed using the PEODAS reanalyses following the method of Johnson et al. (2002). We form the composite tendencies by regression of the advective tendency and salt flux onto each of the PCS values, with the PCS values lagging by 3 months so as to better capture the tendency relationship between the forcing and response (Fig. 10). The results are not qualitatively sensitive to lags of 2–6 months. For salinity EOF1 the fresh anomaly on the eastern edge of fresh pool as seen in Fig. 4d appears to be driven by a combination of negative advective tendency in the western portion of the fresh pool (primarily due to eastward advection of low-salinity water; not shown) and an excess of precipitation over evaporation farther to the east near the date line as a result of the eastward shift of the convective anomaly into the central Pacific during El Niño. Also apparent is an excess of evaporation over precipitation in the far-western portion of the fresh pool that partially compensates the eastward advection of low-salinity water. Together, these tendency terms act to create a low-salinity anomaly centered near 165°E that forms at the peak of El Niño (i.e., Fig. 4d).
Total advective tendency of (a)–(c) salinity and (d)–(f) salt flux (E − P) into the top 50 m formed by a regression onto the leading three standardized PCs of salinity, with the PCs lagging by 3 months. The units are expressed as an equivalent freshwater flux (mm day−1) in the mixed layer (50 m deep).
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
The zonal dipole in surface salinity associated with EOF2 (Fig. 4e) appears to be primarily driven by a similar west–east dipole in E − P: excess evaporation over precipitation in the far-western Pacific (and the opposite in the central Pacific) during the discharge phase of El Niño results in a positive salinity anomaly in the far west and a negative anomaly in the central Pacific. Similarly, during the recharge phase prior to El Niño, excess precipitation over evaporation in the far-western Pacific contributes to a negative salinity anomaly, which would help explain the appearance of enhanced barrier layer thickness in the lead-up to El Niño (i.e., Figs. 8b,d). The farther-eastward detachment of the fresh pool during super El Niño as captured by EOF3 (Fig. 4f) is also seen to develop as a combination of advective effects primarily in the west and a dipole anomaly of E − P farther to the east.
The eastward detachment of the fresh pool into the eastern Pacific sets super El Niño apart from moderate (or normal) El Niño (e.g., Figs. 5b and 8b). So what causes the shift of the fresh pool into the eastern Pacific during super El Niño? Examination of the temporal variation of the PCs suggests that a positive contribution from all three EOFs leads to the development of the fresh anomaly in the eastern Pacific at or just after the peak of an event. Examination of the tendency forcing terms (Fig. 10) indicates for all three EOFs that development of fresh anomalies in the equatorial eastern Pacific stems primarily from an excess of precipitation over evaporation due to contraction of the ITCZ to the equator during El Niño (e.g., Vecchi and Harrison 2006; Zhang and McPhaden 2006), which is especially prominent for super El Niño (e.g., Cai et al. 2014). A role for anomalous meridional advection of climatologically fresher waters onto the equator from the north in the far-eastern Pacific is also indicated (Figs. 10a,c).
Concentrating on the behavior of E − P in the eastern Pacific cold tongue, the scatter of total E − P versus total SST in the equatorial eastern Pacific (Fig. 11a) reveals an expected negative feedback for SST < ~25°C. That is, when SST is cool in this region of trade easterlies, evaporation increases monotonically with SST (i.e., acts to damp SST), with a slope equivalent to ~9.4 W m−2 (°C)−1 as expected from the Clausius–Clapeyron effect on the saturation deficit in the boundary layer (e.g., Zhang and McPhaden 1995). However, after a threshold of ~26°C, excess of precipitation over evaporation begins to occur, and surface wind speed drops (Fig. 11b), which is associated with increasing westerly anomalies (or decreasing easterly winds; Fig. 11c). This presumably reflects that the threshold of SST for deep atmospheric convection is exceeded and that deep precipitating convection in the cold tongue region is supported by converging westerly anomalies in those rare instances when the cold tongue temperature exceeds 27°C during super El Niño (e.g., Takahashi and Dewitte 2016). As a result, beyond ~27°C the damping effect of evaporation on SST is much reduced and there is a net freshwater flux into the mixed layer. Reduced easterly wind in the eastern Pacific also acts to reduce equatorial upwelling and therefore can further enhance a positive SST anomaly in the cold tongue. This nonlinear behavior of E − P in the cold tongue appears to be one contributing factor that sets apart super El Niño (Takahashi and Dewitte 2016).
Scatterplot of monthly (a) E − P (mm day−1), (b) surface wind speed (m s−1), and (c) surface zonal stress (dyn cm−2; 1 dyn = 10−5 N) vs SST in a box 5°N–5°S, 120°–140°W. The sloping red curve in (a) depicts the expected increase of evaporation with SST as a result of increasing saturation deficit. The slope is equivalent to 9.4 W m−2 (°C)−1.
Citation: Journal of Climate 29, 6; 10.1175/JCLI-D-15-0650.1
7. Discussion and conclusions
We have examined interannual variation of salinity in the equatorial Pacific as depicted in the PEODAS reanalyses based on 33 years of data (1980–2012). We have focused on interannual variations associated with ENSO and have highlighted a one-to-one correspondence of the leading three EOFs of temperature, which capture the canonical mature state, discharge phase, and extreme El Niño, with the leading three EOFs of salinity. Although temperature variations are strongest in the eastern Pacific thermocline, the salinity variations are strongest in the western Pacific mixed layer and largely represent the concomitant east–west shifts of the western Pacific fresh pool during El Niño and La Niña. Even though the variations in temperature and salinity occur in vastly different portions of the Pacific, they both result from the same atmospheric forcing and subsequent dynamical response in the ocean. That is, mature El Niño occurs in conjunction with westerly anomalies in the western Pacific that are driven by increased rainfall in the central Pacific as a result of the warm SST anomaly. These westerly anomalies support the tilt mode in temperature but also act to advect the western Pacific fresh pool eastward, with a further tendency to shift the fresh pool eastward owing to the anomalous freshwater flux in the central Pacific as a result of the rainfall anomaly there. Similarly, the discharge phase of the El Niño is associated with little surface wind anomaly, but westward surface currents develop in response to the reversal of the meridional temperature gradient (Jin and An 1999). These westward surface currents act to advect saline water westward into the far-western Pacific and so feed an increase in mixed layer salinity there at the demise of El Niño (or, in the opposite phase, a freshened mixed layer prior to onset of El Niño).
The possible role of salinity variations for evolution of El Niño, consistent with previous studies, was indicated to occur via variations of mixed layer stratification in the western Pacific. At the onset of El Niño, freshening of the far-western Pacific results in a more stably stratified mixed layer and enhanced barrier layer that would favor advective expansion of the warm pool and insulation of warm pool anomalies from entrainment of colder subsurface waters, thus promoting enhanced El Niño development. However, we emphasize here, consistent with Qu and Yu (2014), that an enhanced barrier layer synchronously shifts eastward with the warm pool during El Niño, thus indicating that anomalous barrier layer development contributes to El Niño beyond just in the buildup phase (e.g., Maes et al. 2005). Also, the western limit of the contraction of the enhanced barrier layer during La Niña suggests a possible contribution of salinity to the asymmetry in the strength of El Niño and La Niña. That is, the barrier layer will act to intensify the response to westerly winds during El Niño, but the westward contraction of a barrier layer means that there is not then a concomitant intensification of the response to the easterly anomalies during La Niña.
Our results, consistent from both a temperature and salinity perspective, suggest that there are two sorts of ENSO events: moderate El Niño/La Niña, which tend to be focused in the western-central Pacific and are associated with modest east–west shifts of the warm/fresh pool in the western-central Pacific, and super El Niño, whereby the warm pool extends all the way across the Pacific and the fresh pool detaches from the western Pacific. Because super El Niño is rare, our view is that central Pacific El Niño is viewed as being typical, consistent with Takahashi et al. (2011). This grouping of moderate El Niño and La Niña as being central Pacific phenomena distinct from super El Niño that extends into the eastern Pacific is more naturally derived using EOFs of temperature and especially salinity in the equatorial depth plane rather than the more traditional depiction of ENSO using EOFs of surface temperature (e.g., Ashok et al. 2007).
This identification of moderate El Niño and La Niña as central Pacific phenomena that more commonly occur versus super El Niño that extends into the eastern Pacific but rarely occurs is also largely consistent with Dommenget et al. (2013). However, they further distinguished differences in structure between moderate La Niña and strong La Niña (shifted farther west in the Pacific). They also argued that there is a negative skewness of central Pacific events. That is, cold events are stronger than warm events in the central Pacific as can be discerned in the scatterplot of PC3 versus PC1 (Figs. 6c,d), which shows that negative values of PC1 (with positive PC3) occur up to about −2 but positive PC1 (with negative PC3) only occurs up to about 1, especially for salinity field. PC1 above about 1 then is associated with positive PC3, indicating that El Niño then shifts into the eastern Pacific above a certain threshold. This negative skewness of central Pacific El Niño can be understood simply as the sudden shift of positive E − P into the central and eastern Pacific, as the threshold for deep convection is achieved in the cold tongue (e.g., Takahashi and Dewitte 2016) and so El Niño cannot be constrained in the central Pacific.
However, we have not assessed what might cause this threshold to be surpassed so that super El Niño occurs. The timing and intensity of stochastic forcing by westerly wind events has been indicated to be one of the key determinants of whether an event evolves into a central Pacific moderate El Niño event or an eastern Pacific super El Niño event (e.g., Shi et al. 2011; Fedorov et al. 2015). However, changes in the background climate that would favor eastern Pacific El Niño as a result of stronger thermocline feedback in the eastern or central Pacific El Niño owing to stronger zonal advective feedback in the west has also been proposed (e.g., Kug et al. 2009). The swing from the warm phase to the cold phase of the interdecadal Pacific oscillation at the end of the twentieth century (e.g., England et al. 2014) has been postulated to have recently favored central Pacific El Niño and reduced the likelihood of eastern Pacific El Niño (e.g., Chung and Li 2013; Hu et al. 2013; Xiang et al. 2013; Zhao et al. 2016, manuscript submitted to J. Climate).
We pointed out some of the limitations of our ocean reanalyses, especially the lack of global salinity observations prior to around 2000. We justified our results by noting that the interannual variations depicted in these reanalyses are stronger than the trends in the data, whether real or spurious, and that our results are largely consistent with many previous studies of the variation of salinity associated with El Niño. Nonetheless, there are apparent multiyear variations of temperature, salinity, and barrier layer (e.g., Fig. 5) that require further investigation. For instance, a warming trend in the western Pacific warm pool is evident after around 1998 and establishment of a nearly constant barrier layer is evident after around 2000, which coincides with the availability of Argo data. However, there are known variations of Pacific climate that could be contributing to these trends. In particular, since 1998, the Pacific Walker circulation has been accelerated, in response to increased SST and concomitant increased rainfall in the western Pacific (e.g., England et al. 2014), which could be contributing to the ongoing enhanced barrier layer as seen in Fig. 5c. The realism and possible role of salinity-induced changes in ENSO dynamics is the subject of an ongoing study.
Acknowledgments
Support for this study was provided by the Managing Climate Variability (MCV) project and the Victorian Climate Initiative (VicCI).
REFERENCES
Ando, K., and M. J. McPhaden, 1997: Variability of surface layer hydrography in the tropical Pacific Ocean. J. Geophys. Res., 102, 23 063–23 078, doi:10.1029/97JC01443.
Ashok, K., S. K. Behera, S. A. Rao, H. Weng, and T. Yamagata, 2007: El Niño Modoki and its possible teleconnections. J. Geophys. Res., 112, C11007, doi:10.1029/2006JC003798.
Balmaseda, M. A., A. Vidard, and D. L. T. Anderson, 2008: The ECMWF ocean analysis system: ORA-S3. Mon. Wea. Rev., 136, 3018–3034, doi:10.1175/2008MWR2433.1.
Balmaseda, M. A., K. Mogensen, and A. Weaver, 2013: Evaluation of the ECMWF ocean reanalysis system ORAS4. Quart. J. Roy. Meteor. Soc., 139, 1132–1161, doi:10.1002/qj.2063.
Bosc, C., T. Delcroix, and C. Maes, 2009: Barrier layer variability in the western Pacific warm pool from 2000 to 2007. J. Geophys. Res., 114, C06023, doi:10.1029/2008JC005187.
Cai, W., and Coauthors, 2014: Increasing frequency of extreme El Niño events due to greenhouse warming. Nat. Climate Change, 4, 111–116, doi:10.1038/nclimate2100.
Capotondi, A., and Coauthors, 2015: Understanding ENSO diversity. Bull. Amer. Meteor. Soc., 96, 921–938, doi:10.1175/BAMS-D-13-00117.1.
Carton, J. A., G. Chepurin, and X. Cao, 2000: A Simple Ocean Data Assimilation analysis of the global upper ocean 1950–95. Part II: Results. J. Phys. Oceanogr., 30, 311–326, doi:10.1175/1520-0485(2000)030<0311:ASODAA>2.0.CO;2.
Chung, P. H., and T. Li, 2013: Interdecadal relationship between the mean state and El Niño types. J. Climate, 26, 361–379, doi:10.1175/JCLI-D-12-00106.1.
Delcroix, T., and J. Picaut, 1998: Zonal displacement of the western equatorial Pacific “fresh pool.” J. Geophys. Res., 103, 1087–1098, doi:10.1029/97JC01912.
Delcroix, T., and M. McPhaden, 2002: Interannual sea surface salinity and temperature changes in the western Pacific warm pool during 1992–2000. J. Geophys. Res., 107, 8002, doi:10.1029/2001JC000862.
Dommenget, D., T. Bayr, and C. Frauen, 2013: Analysis of the non-linearity in the pattern and time evolution of El Niño Southern Oscillation. Climate Dyn., 40, 2825–2847, doi:10.1007/s00382-012-1475-0.
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:10.1038/nclimate2106.
Fedorov, A. V., S. Hu, M. Lengaigne, and E. Guilyardi, 2015: The impact of westerly wind bursts and ocean initial state on the development, and diversity of El Niño events. Climate Dyn., 44, 1381–1401, doi:10.1007/s00382-014-2126-4.
Gill, A. E., 1982: Atmosphere–Ocean Dynamics. International Geophysics Series, Vol. 30, Academic Press, 662 pp.
Hackert, E., J. Ballabrera-Poy, A. J. Busalacchi, R.-H. Zhang, and R. G. Murtugudde, 2011: Impact of sea surface salinity assimilation on coupled forecasts in the tropical Pacific. J. Geophys. Res., 116, C05009, doi:10.1029/2010JC006708.
Hu, Z.-Z., A. Kumar, H.-L. Ren, H. Wang, M. L’Heureux, and F.-F. Jin, 2013: Weakened interannual variability in the tropical Pacific Ocean since 2000. J. Climate, 26, 2601–2613, doi:10.1175/JCLI-D-12-00265.1.
Ingleby, B., and M. Huddleston, 2007: Quality control of ocean temperature and salinity profiles—Historical and real-time data. J. Mar. Syst., 65, 158–175, doi:10.1016/j.jmarsys.2005.11.019.
Jin, F. F., 1997: An equatorial ocean recharge paradigm for ENSO. Part I: Conceptual model. J. Atmos. Sci., 54, 811–829, doi:10.1175/1520-0469(1997)054<0811:AEORPF>2.0.CO;2.
Jin, F. F., and S. I. An, 1999: Thermocline and zonal advective feedbacks within the equatorial ocean recharge oscillator model for ENSO. Geophys. Res. Lett., 26, 2989–2992, doi:10.1029/1999GL002297.
Johnson, E. S., G. S. E. Lagerloef, J. T. Gunn, and F. Bonjean, 2002: Surface salinity advection in the tropical oceans compared with atmosphere freshwater forcing: A trial balance. J. Geophys. Res., 107, 8014, doi:10.1029/2001JC001122.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, doi:10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
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:10.1175/2008JCLI2624.1.
Kumar, A., and Z. Z. Hu, 2014: Interannual and interdecadal variability of ocean temperature along the equatorial Pacific in conjunction with ENSO. Climate Dyn., 42, 1243–1258, doi:10.1007/s00382-013-1721-0.
Lukas, R., and E. Lindstrom, 1991: The mixed layer of the western equatorial Pacific Ocean. J. Geophys. Res., 96, 3343–3357, doi:10.1029/90JC01951.
Maes, C., 2008: On the ocean salinity stratification observed at the eastern edge of the equatorial Pacific warm pool. J. Geophys. Res., 113, C03027, doi:10.1029/2007JC004297.
Maes, C., J. Picaut, and S. Belamari, 2002: Salinity barrier layer and onset of El Niño in a Pacific coupled model. Geophys. Res. Lett., 29, 2206, doi:10.1029/2002GL016029.
Maes, C., J. Picaut, Y. Kuroda, and K. Ando, 2004: Characteristics of the convergence zone at the eastern edge of the Pacific warm pool. Geophys. Res. Lett., 31, L11304, doi:10.1029/2004GL019867.
Maes, C., J. Picaut, and S. Belamari, 2005: Importance of salinity barrier layer for the buildup of El Niño. J. Climate, 18, 104–118, doi:10.1175/JCLI-3214.1.
Maes, C., K. Ando, T. Delcroix, W. S. Kessler, M. J. McPhaden, and D. Roemmich, 2006: Observed correlation of surface salinity, temperature and barrier layer at the eastern edge of the western Pacific warm pool. Geophys. Res. Lett., 33, L06601, doi:10.1029/2005GL024772.
McPhaden, M. J., 2012: A 21st century shift in the relationship between ENSO SST and warm water volume anomalies. Geophys. Res. Lett., 39, L09706, doi:10.1029/2012GL051826.
Meinen, C. S., and M. J. McPhaden, 2000: Observations of warm water volume changes in the equatorial Pacific and their relationship to El Niño and La Niña. J. Climate, 13, 3551–3559, doi:10.1175/1520-0442(2000)013<3551:OOWWVC>2.0.CO;2.
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, doi:10.1175/1520-0493(1982)110<0699:SEITEO>2.0.CO;2.
Pacanowski, R. C., 1995: MOM2 documentation user’s guide and reference manual, version 1.0. GFDL Tech. Rep. 3, 232 pp.
Picaut, J. F. M., and Y. du Penhoat, 1997: An advective–reflective conceptual model for the oscillatory nature of the ENSO. Science, 277, 663–666, doi:10.1126/science.277.5326.663.
Qu, T., and J. Yu, 2014: ENSO indices from sea surface salinity observed by Aquarius and Argo. J. Oceanogr., 70, 367–375, doi:10.1007/s10872-014-0238-4.
Qu, T., T. Song, and C. Maes, 2014: Sea surface salinity and barrier layer variability in the equatorial. J. Geophys. Res., 119, 15–29, doi:10.1002/2013JC009375.
Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 1609–1625, doi:10.1175/1520-0442(2002)015<1609:AIISAS>2.0.CO;2.
Schiller, A., J. S. Godfrey, P. C. McIntosh, N. R. Smith, O. Alves, G. Wang, and R. Fiedler, 2002: A new version of the Australian Community Ocean Model for seasonal climate prediction. CSIRO Marine Research Rep. 240, 79 pp. [Available online at www.cmar.csiro.au/publications/cmrreports/240/text240.pdf.]
Shi, L., H. H. Hendon, O. Alves, M. C. Wheeler, D. Anderson, and G. Wang, 2011: On the importance of initializing the stochastic part of the atmosphere for forecasting the 1997/1998 El Niño. Climate Dyn., 37, 313–324, doi:10.1007/s00382-010-0933-9.
Shi, L., and Coauthors, 2016: An assessment of upper ocean salinity content from the Ocean Reanalyses Intercomparison Project (ORA-IP). Climate Dyn., doi:10.1007/s00382-015-2868-7, in press.
Smith, N. R., J. E. Blomley, and G. Meyers, 1991: A univariate statistical interpolation scheme for subsurface thermal analysis in tropical oceans. Prog. Oceanogr., 28, 219–256, doi:10.1016/0079-6611(91)90009-B.
Sprintall, J., and M. Tomczak, 1992: Evidence of the barrier layer in the surface layer of the tropics. J. Geophys. Res., 97, 7305–7316, doi:10.1029/92JC00407.
Takahashi, K., and B. Dewitte, 2016: Strong and moderate nonlinear El Niño regimes. Climate Dyn., doi:10.1007/s00382-015-2665-3, in press.
Takahashi, K., A. Montecinos, K. Goubanova, and B. Dewitte, 2011: ENSO regimes: Reinterpreting the canonical and Modoki El Niño. Geophys. Res. Lett., 38, L10704, doi:10.1029/2011GL047364.
Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131, 2961–3012, doi:10.1256/qj.04.176.
Vecchi, G. A., and D. E. Harrison, 2006: The termination of the 1997–98 El Niño. Part I: Mechanisms of oceanic change. J. Climate, 19, 2633–2646, doi:10.1175/JCLI3776.1.
Vialard, J., P. Delecluse, and C. Menkes, 2002: A modeling study of salinity variability and its effects the tropical Pacific Ocean during the 1993–1999 period. J. Geophys. Res., 107, 8005, doi:10.1029/2000JC000758.
Xiang, B., B. Wang, and T. Li, 2013: A new paradigm for the predominance of standing central Pacific warming after late 1990s. Climate Dyn., 41, 327–340, doi:10.1007/s00382-012-1427-8.
Xue, Y., B. Huang, Z. Z. Hu, A. Kumar, C. Wen, D. Behringer, and S. Nadiga, 2011: An assessment of oceanic variability in the NCEP Climate Forecast System reanalysis. Climate Dyn., 37, 2511–2539, doi:10.1007/s00382-010-0954-4.
Xue, Y., and Coauthors, 2012: A comparative analysis of upper-ocean heat content variability from an ensemble of operational ocean reanalyses. J. Climate, 25, 6905–6929, doi:10.1175/JCLI-D-11-00542.1.
Yang, S. C., M. Rienecker, and C. Keppenne, 2010: The impact of ocean data assimilation on seasonal-to-interannual forecasts: A case study of the 2006 El Niño event. J. Climate, 23, 4080–4095, doi:10.1175/2010JCLI3319.1.
Yin, Y. H., O. Alves, and P. R. Oke, 2011: An ensemble ocean data assimilation system for seasonal prediction. Mon. Wea. Rev., 139, 786–808, doi:10.1175/2010MWR3419.1.
Zhang, G. J., and M. J. McPhaden, 1995: The relationship between sea surface temperature and latent heat flux in the equatorial Pacific. J. Climate, 8, 589–605, doi:10.1175/1520-0442(1995)008<0589:TRBSST>2.0.CO;2.
Zhang, X., and M. J. McPhaden, 2006: Wind stress variations and interannual sea surface temperature anomalies in the eastern equatorial Pacific. J. Climate, 19, 226–241, doi:10.1175/JCLI3618.1.
Zhao, M., H. Hendon, O. Alves, Y. H. Yin, and D. Anderson, 2013: Impact of salinity constraints on the simulated mean state and variability in a coupled seasonal forecast model. Mon. Wea. Rev., 141, 388–402, doi:10.1175/MWR-D-11-00341.1.
Zhao, M., H. Hendon, O. Alves, and Y. H. Yin, 2014: Impact of improved assimilation of temperature and salinity for coupled model seasonal forecasts. Climate Dyn., 42, 2565–2583, doi:10.1007/s00382-014-2081-0.
