1. Introduction
Salinity, along with temperature, is known as a key parameter in physical oceanography. Since salinity controls the density of the seawater, understanding its spatiotemporal distribution is of great importance for an accurate description of dynamics and thermodynamics of the ocean. However, because of technical difficulty in salinity observation, its variability has not been understood compared to that of temperature. While temperature directly affects the atmosphere as a heat source, salinity does not have such explicit impacts. For this reason, salinity in the tropical ocean had long been considered to be a passive tracer and its role had not been investigated (Cooper 1988), although the importance of salinity in driving thermohaline circulation had been recognized (Bryan 1986).
In the early 1990s, intensive observations carried out in the tropical Pacific led to the identification of strong salinity stratification in the upper ocean. For example, Lukas and Lindstrom (1991) pointed out that the excess of precipitation over evaporation and the subduction of salty water from subtropics generate strong salinity stratification in the western tropical Pacific. Such salinity stratification in the upper ocean plays a key role in determining mixed layer depth (MLD), which is a key parameter in determining sea surface temperature (SST). They named the layer between the mixed layer and the isothermal layer the “barrier layer” since it insulates the surface layer from cold water below the thermocline and prevents entrainment of the cold water like a barrier. Sprintall and Tomczak (1992) revealed that the barrier layer is also found in other tropical and subtropical oceans. Since the discovery of the barrier layer, numerous observational (Picaut et al. 1996; Ando and McPhaden 1997) and modeling studies (Vialard and Delecluse 1998a) have focused on upper ocean salinity variability in the tropical Pacific and its relation with the El Niño–Southern Oscillation (ENSO). Impacts of salinity on the mean state of the tropical Pacific and the development of ENSO events are quantitatively diagnosed by several sensitivity experiments using ocean general circulation models (OGCMs) (Vialard and Delecluse 1998b; Vialard et al. 2002) and coupled general circulation models (CGCMs) (Maes 2002; Maes et al. 2005). More recently, Zhu et al. (2014) showed that predictability of the ENSO is significantly influenced by the initialization of salinity anomalies.
Because of fewer salinity observations and the difficulties in accurate simulation of salinity, little is known about salinity variability in the tropical Indian Ocean compared to the tropical Pacific. Similar to the ENSO in the tropical Pacific, there is a dominant climate mode called the Indian Ocean dipole (IOD) in the tropical Indian Ocean (Saji et al. 1999; Webster et al. 1999). When a positive IOD (pIOD) event occurs, SST in the southeastern (western) tropical Indian Ocean becomes anomalously low (high). This phenomenon is phase-locked to the seasonal cycle with its peak in boreal fall. Since SST anomalies associated with the IOD exert strong impacts on precipitation in the Indian Ocean rim countries as well as the global climate (e.g., Saji and Yamagata 2003), its accurate understanding and prediction is of great importance. Since there is a remarkable zonal contrast of surface salinity (Rao and Sivakumar 2003) and strong salinity stratification (Qu and Meyers 2005) in the tropical Indian Ocean as in the tropical Pacific, it is natural to ask whether salinity plays important roles in the development of the IOD as is the case for the ENSO.
Recently, variations in sea surface salinity (SSS) associated with the IOD have been revealed by a number of studies (Thompson et al. 2006; Grunseich et al. 2011; Zhang et al. 2013). According to their results, negative (positive) SSS anomalies emerge in the central-eastern equatorial (southeastern tropical) Indian Ocean in concert with the development of the pIOD. These SSS anomalies are caused not only by anomalous rainfall, but also by changes in the ocean circulation associated with the pIOD (Zhang et al. 2013; Li et al. 2016). Anomalies with the opposite sign appear during the negative IOD (nIOD) (Grunseich et al. 2011). These SSS signals are also captured by recent satellite salinity observation (Durand et al. 2013; Du and Zhang 2015). In addition, Qiu et al. (2012) found that variability of barrier layer thickness in the tropical Indian Ocean is closely linked to the occurrence of the IOD by analyzing Argo data. They pointed out that variability of barrier layer thickness is primarily controlled by that of isothermal layer depth, and argued that it may affect the strength of the thermocline feedback, which plays an important role in the generation of SST anomalies.
Although general features of the IOD-induced SSS and barrier layer thickness variation have been discussed, little attention has been paid to subsurface salinity variations. Since stratification is determined by the vertical gradient of temperature and salinity, understanding subsurface salinity variations is as important for proper description of the upper ocean structure. Regarding subsurface variability, Rao et al. (2002) and Shinoda et al. (2004) demonstrated that temperature near the thermocline undergoes large variations associated with the IOD. They pointed out that such subsurface temperature variation is triggered by the vertical displacement of thermocline associated with the ocean adjustment in response to wind anomalies (Sayantani and Gnanaseelan 2015). Considering the above mechanism, it is natural to infer that salinity near the halocline (the depth with maximum vertical salinity gradient) also exhibits such variation.
Many previous studies regarding IOD-induced SSS anomalies speculate that they may play some role in the evolution of SST and affect the strength of air–sea coupled processes (Durand et al. 2013; Zhang et al. 2016; Li et al. 2016). In their pioneering work on the role of salinity in the IOD, Masson et al. (2004) performed numerical experiments to investigate the role of salinity stratification in vertical mixing, the ocean circulation, and SST evolution, targeting a strong pIOD event in boreal fall of 1997. They concluded that the existence of salinity stratification strengthens the equatorial circulations and enhances the eastern cooling during the pIOD. However, since they only focused on the importance of “total” salinity stratification and did not decompose it into mean and anomaly components, the roles played by salinity anomalies are not clear from their results. Therefore, quantitative understanding of influences of salinity anomalies on the upper ocean stratification and temperature evolution has not been established yet. This motivates us to explore this issue in more detail.
In this paper, we investigate salinity anomalies associated with the pIOD and their roles in the upper ocean stratification and temperature evolution. The rest of this paper is organized as follows. In section 2, we describe the observational dataset and the ocean reanalysis product used in this study. Observed salinity variability associated with the pIOD and their generation mechanisms are discussed in section 3. Section 4 focuses on the relative importance of temperature and salinity anomalies in the upper ocean stratification by decomposing density anomalies. Through sensitivity experiments using a one-dimensional mixed layer (1D ML) model, we quantify the impacts of salinity on vertical mixing and temperature in section 5. A summary of main results and discussions are presented in section 6. Unless otherwise specified, we will only show results from the pIOD throughout this paper, but we briefly comment on the nIOD in section 6.
2. Data
We use 1° × 1° gridded temperature and salinity data (Roemmich and Gilson 2009) to examine the thermohaline variation during IOD events. These gridded data are constructed based only on individual Argo profiles using an optimum interpolation method. The vertical resolution is about 10 m in the upper 150 m, and we analyze monthly data from January 2004 to December 2014. From 2004 to 2007, the spatial coverage of Argo profiles was relatively sparse in the tropical Indian Ocean, although it exceeded 80% of the planned level (1 float per 3° longitude × 3° latitude box) (Cai and Qiu 2013). For this reason, we need to be careful when we compare results from this dataset with those from other datasets during the earlier period.
To complement the short data period of the Argo observation, we also use data from the Ocean Reanalysis System version 4 (ORAS4) provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) (Balmaseda et al. 2013) . This product has been widely used for studies in the tropical Indian Ocean and validated against various in situ observations (Nyadjro and McPhaden 2014; Chen et al. 2015, 2016). The dynamical core model of the ORAS4 is the Nucleus for European Modeling of the Ocean (NEMO) and it is forced by daily atmospheric fluxes (solar radiation, total heat flux, evaporation-minus-precipitation, and wind stress) obtained from the ERA-40 reanalysis (from September 1957 to December 1988) (Uppala et al. 2005), the ERA-Interim reanalysis (from January 1989 to December 2009) (Dee et al. 2011), and the ECMWF operational archive (after January 2010). The ORAS4 assimilates temperature, salinity, and satellite-derived sea surface height (SSH) anomalies using a variational assimilation system called NEMOVAR (Mogensen et al. 2012). This product has a 1° × 1° horizontal resolution and 42 vertical levels with 18 of them in the upper 200 m. Monthly temperature, salinity, and velocity data from 1958 to 2014 are used for the present analysis. Since vertical velocity is not available, we estimate it by integrating horizontal divergence downward from the surface, with the surface value set to the time derivative of SSH.
We note that qualitatively the same results are obtained even if we use the Grid Point Value of the Monthly Objective Analysis (MOAA-GPV) (Hosoda et al. 2008) and the Simple Ocean Data Assimilation (SODA) version 2.2.4 datasets (Carton and Giese 2008). Since we focus on phenomena with interannual time scale in this study, we apply 3-month running mean after removing the mean seasonal cycle when calculating anomalies to remove subseasonal signals.
3. Salinity variation associated with the pIOD
a. Features of observed salinity variation
First, we take a look at pIOD-related salinity anomalies captured by the Argo observation. As an example of a typical pIOD event, we select the 2006 event, which is the strongest pIOD year during the Argo observation period. As documented by previous studies (Grunseich et al. 2011; Zhang et al. 2013), there are prominent negative SSS anomalies in the central-eastern equatorial Indian Ocean (Fig. 1, left). These anomalies develop during boreal fall and reach their peak toward the end of the year, followed by gradual decay in January–March of the next year. Through salinity budget analysis, previous studies have shown that a decrease in the transport of high salinity water from the western Indian Ocean associated with weakening of the climatological eastward current in boreal fall (the so-called Wyrtki jet; Wyrtki 1973) is the main factor for the generation of these negative SSS anomalies (Zhang et al. 2013; Li et al. 2016). In addition, weak positive SSS anomalies extend over the southeastern tropical Indian Ocean. However, the amplitude of these positive anomalies seems to be underestimated because of the insufficient number of Agro floats in this region (figure not shown). These positive SSS anomalies are caused by reduced precipitation associated with the pIOD (Zhang et al. 2013; Li et al. 2016).
Time evolution of Argo-observed salinity anomalies (psu) at (left) the surface and (right) 60-m depth in the tropical Indian Ocean from June 2006 to February 2007. Contour intervals (CI) are 0.2 psu.
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
On the other hand, the spatial pattern of salinity anomalies near the pycnocline (we select 60 m as an example) is somewhat different from that of SSS (Fig. 1, right). Despite the salient surface freshening, there are remarkable positive salinity anomalies in the eastern equatorial Indian Ocean. They appear and develop during boreal summer to fall and decay suddenly after the demise of the IOD. Their persistence seems to be slightly shorter than that of the surface signals. Other notable features are negative salinity anomalies in the eastern off-equatorial Indian Ocean (Fig. 1, right). To some extent, they seem to be related to the surface signals, but their spatial structure (a pair of negative anomalies straddling the equator) is markedly different. These results imply that processes contributing to the development of these subsurface salinity anomalies are greatly different from those of SSS.
To see whether these anomalies are common features of all pIOD years, we construct composites of salinity anomalies based on an ocean reanalysis product (ORAS4). For this purpose, we calculate the dipole mode index (DMI), which is defined as the difference between area-averaged SST anomalies over the western box (10°S–10°N, 50°–70°E) and the eastern box (10°S–0°, 90°–110°E) (Saji et al. 1999). Then, we define the pIOD year as the year when the DMI in September–November (SON) all exceeds 1 standard deviation. This definition excludes so-called unseasonable IOD (Du et al. 2013) and aborted IOD (Rao and Yamagata 2004) events so that only canonical pIOD events are selected. According to this criterion, six years (1961, 1963, 1972, 1994, 1997, and 2006) are identified as the pIOD years.
Figure 2 displays composites of salinity anomalies constructed from all pIOD years. We can see that SSS anomalies are similar to the observed pattern, although their amplitude is slightly different from the Argo observation. In addition, subsurface salinity anomalies are in good agreement with the Argo observation. These results suggest that surface and subsurface salinity anomalies captured by the Argo observation in 2006 are robust features of the pIOD.
Composite of salinity anomalies (psu; CI = 0.2 psu) at (left) the surface and (right) 60-m depth in the tropical Indian Ocean during the positive Indian Ocean dipole (pIOD) events (constructed from ORAS4). Year 0 denotes the year when the pIOD develops and year 1 means the following year. Anomalies significant at the 90% confidence level by a two-tailed t test are shaded. The orange and green boxes in the middle panels indicate the eastern equatorial Indian Ocean (EEIO: 2°S–2°N, 92°–98°E) and the southeastern tropical Indian Ocean (STIO: 9°–5°S, 83°–90°E) regions, respectively (see the main text).
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
To further illustrate the vertical structure of salinity anomalies presented above, we have prepared zonal sections of composited salinity anomalies along the equator (meridionally averaged over 2°S–2°N) and the off-equatorial band (9°–5°S). For the equatorial region (Fig. 3a), negative salinity anomalies near the surface extend in the central-eastern part with a peak around 75°E, whereas strong positive salinity anomalies are found near the halocline around 60-m depth in the east. These anomalies are generally located over regions where climatological salinity gradient is large. On the other hand, salinity anomalies along the off-equatorial band are characterized by surface saltening in the east and subsurface freshening in the central part (Fig. 4a). Negative salinity anomalies near the halocline have zonally slanted shape with their peak around 80°–90°E. These results suggest that both surface and subsurface salinity anomalies have their unique features and play key roles in the determination of the thermohaline structure.
(a) Equatorial section (averaged over 2°S–2°N) of composited salinity anomalies (psu; CI = 0.1 psu) during September–November (SON) of the pIOD years. (b)–(d) As in (a), but for (b) zonal, (c) meridional, and (d) vertical velocity anomalies (m s−1; CI = 0.05, 0.05, and 2.0 × 10−6 m s−1, respectively) during July–September (JAS) of the pIOD years. Anomalies significant at the 90% confidence level by a two-tailed t test are shaded. Also shown (black contours) is the climatological salinity in JAS and contour intervals are 0.2 psu.
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
As in Fig. 3, but for the off-equatorial band (averaged over 9°–5°S).
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
The zonal section of partial regression coefficients of salinity anomalies in the equatorial and off-equatorial band against the DMI and the Niño-3.4 index is shown in Fig. 5. For both bands, IOD-related signals have large amplitudes and bear a close resemblance to composited salinity anomalies (Figs. 5b,d), whereas ENSO-related signals are very weak (Figs. 5a,c). Therefore, we may conclude that the IOD is the main driver of salinity anomalies in boreal fall. Although previous studies showed the dominant influence of the IOD over the ENSO on thermocline (Rao and Behera 2005; Yu et al. 2005) and SSS (Zhang et al. 2013) variability in the tropical Indian Ocean north of 10°S, this study is the first to show such dominance in subsurface salinity variability.
(a) SON salinity anomalies (psu °C−1) along the equator (averaged over 2°S–2°N) against the December–February (DJF) Niño-3.4 index, having removed the influence of the SON dipole mode index (DMI). (b) Partial regression coefficients of SON salinity anomalies (psu °C−1; CI = 0.1 psu °C−1) along the equator against the SON DMI, having removed the influence of the DJF Niño-3.4 index. Regression coefficients significant at the 95% confidence level by a two-tailed t test are shaded. (c),(d) As in (a),(b), but for the off-equatorial section (averaged over 9°–5°S).
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
b. Mechanisms of salinity variation
In section 3a, we have seen that variations in subsurface salinity are quite different from those of SSS. Then, what make such differences? Previous studies showed that SSS anomalies associated with the IOD are primarily caused by anomalous freshwater flux and horizontal advection (Zhang et al. 2013; Li et al. 2016). However, the generation mechanisms of subsurface salinity anomalies are not clarified.
We compute each term on the right-hand side of Eq. (1) at each grid point and take an average over the eastern equatorial Indian Ocean (EEIO: 2°S–2°N, 92°–98°E; the orange box in Fig. 2) and southeastern tropical Indian Ocean (STIO: 9°–5°S, 83°–90°E; the green box in Fig. 2), where significant salinity anomalies are found.
Figure 6 shows the time evolution of these anomalies during pIOD events. In the EEIO box, the surface freshening is mainly caused by zonal advection anomalies (Fig. 6d), as pointed out by previous studies (Zhang et al. 2013; Li et al. 2016). On the other hand, saltening near the halocline is primarily caused by vertical advection anomalies (Fig. 6l). By decomposing them into each component, we find that all vertical advection terms (terms 7–9) play an important role in the generation of positive salinity anomalies, with terms 8 and 9 having their peak shallower than term 7. On the other hand, zonal advection anomalies contribute to the surface freshening mainly due to term 1, while term 2 partially cancels out subsurface saltening associated with vertical advection anomalies. The magnitude of the meridional components is relatively small. Although there are some discrepancies between the sum of advection terms (Fig. 6m) and the tendency anomalies (Fig. 6n), the positive anomalies near 60-m depth are clearly seen in both.
Composite of salinity advection anomalies [each term on the right-hand side, sum of terms 1–9, the left-hand side, and residual of Eq. (1)] averaged over the EEIO box (2°S–2°N, 92°–98°E) during the pIOD years (psu month−1; CI = 0.1 psu month−1). Anomalies significant at the 90% confidence level by a two-tailed t test are shaded.
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
The above results can be understood by inspecting depth–longitude sections of current anomalies. Figures 3b–d show composites of velocity anomalies during July–September (JAS) of the pIOD year, which is 2 months prior to the peak. In the eastern equatorial Indian Ocean, positive vertical velocity anomalies (w′ > 0) are formed near the halocline (Fig. 3d) in response to easterly wind anomalies (Chen et al. 2016). This anomalous intensification of upwelling increases the upward transport of salty water below the pycnocline [i.e., 
For the STIO box, subsurface freshening is due to zonal and vertical advection anomalies (Fig. 7). Both of them are dominated by effects of current anomalies (terms 1 and 7), suggesting the importance of the changes in the upper ocean circulation. In the pIOD composites (Figs. 4b–d), westward current anomalies (u′ < 0) are found in the east (Fig. 4b). In addition, negative vertical velocity anomalies (w′ < 0) extend near the pycnocline in the middle of the basin (Fig. 4d). Previous studies demonstrated that these downwelling anomalies are excited by anticyclonic wind stress curl anomalies associated with the equatorial easterly wind anomalies (Rao and Behera 2005; Yu et al. 2005). Since the zonal and vertical gradients of climatological salinity in this region are also negative (i.e., 
As in Fig. 6, but for the STIO box (9°–5°S, 83°–90°E) with contour intervals of 0.04 psu month−1.
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
As mentioned above, offline calculations of salinity balance terms using an ocean reanalysis product are usually hampered by large residuals, especially in region with energetic subseasonal variations and/or mesoscale eddies. Given that there are strong intraseasonal current variability (Masumoto et al. 2005) and mesoscale eddy activities (Feng and Wijffels 2002; Yu and Potemra 2006) in the tropical Indian Ocean, we cannot totally rule out the possibility that other effects not captured in this analysis may play an important role in the formation of salinity anomalies. In addition, changes in the diffusion terms, which are not considered here, may also contribute to the evolution of subsurface salinity anomalies. For comprehensive understanding of their generation mechanisms, an online calculation of salinity budget using an ocean model that can realistically simulate salinity field is desirable. A study in this direction using a regional ocean model is underway.
4. Impacts of salinity anomalies on density structure
In the previous section, we have seen that the upper ocean salinity in the tropical Indian Ocean undergoes significant variations associated with the pIOD, in response to variations in precipitation pattern and ocean circulation. An intriguing question is whether these salinity anomalies can affect the upper ocean stratification. To address this issue, we decompose observed density anomalies into contribution from temperature and salinity and compare their relative importance.
a. Decomposition of density anomalies
Although the density of the seawater ρ is a nonlinear function of temperature T, salinity S, and pressure, nonlinearity in the equation of state and the thermobaric effect (density dependence on pressure) is negligible within a typical parameter range in the tropical upper ocean. Therefore, density anomalies can be decomposed into contributions from temperature and salinity anomalies. More specifically, we first compute the total density [ρ(T, S)] based on the interannual temperature and salinity profile using the UNESCO version of the equation of state. Then, we calculate the density by replacing interannual salinity with climatological salinity [ρ(T, Sclim)]. The difference between two densities, [ΔρS = ρ(T, S) − ρ(T, Sclim)], is the impact of interannual salinity anomalies on density anomaly. The contribution from temperature anomalies to density anomalies ΔρT can be estimated in a similar manner [ΔρT = ρ(T, S) − ρ(Tclim, S)]. Because of linearity, Δρ can be approximated by the sum of ΔρS and ΔρT. In this way, we can quantify the relative importance of temperature and salinity anomalies in generating density anomalies by comparing ΔρT and ΔρS with Δρ. This method was proposed by Zheng and Zhang (2012), who applied it to the tropical Pacific Ocean, but it has never been applied to the tropical Indian Ocean. Here, we focus on the squared buoyancy frequency, which is a key parameter for describing upper ocean dynamics and thermodynamics (Zheng and Zhang 2012, 2015; Zheng et al. 2014).
b. Impacts of salinity anomalies on density stratification
Composites of temperature and salinity anomalies associated with the pIOD along the equatorial section are presented in Figs. 8a and 8b. There is strong cooling (weak warming) near the thermocline in the east (west). These temperature anomalies are manifestation of the vertical heaving of thermocline associated with the pIOD (Rao et al. 2002; Rao and Behera 2005). For salinity, there is prominent surface freshening in the central-eastern part and subsurface saltening in the east, as discussed in section 3. By converting these anomalies into density anomalies, their contribution to density anomalies is estimated (Figs. 8c–e). We can identify a dipole structure in density anomalies, with a positive (negative) peak near the pycnocline (surface) in the eastern (central) part (Fig. 8e). If we neglect salinity anomalies when calculating density anomalies, the positive peak in the east is underestimated and the negative peak in the central part diminishes (Fig. 8c). On the other hand, the density anomalies purely originating from salinity variations (Fig. 8d) resemble total density anomalies, but their amplitude is weaker than total density anomalies, especially in the east. These results suggest that both temperature and salinity anomalies play important roles in modulating the density structure over this region. The salinity contribution is especially large in the central-eastern equatorial Indian Ocean, where significant salinity variations are observed.
Depth–longitude section of composited (a) temperature (°C; CI = 0.5°C) and (b) salinity anomalies (psu; CI = 0.1 psu) during SON of the pIOD along the equator (averaged over 2°S–2°N). (c),(d) As in (a),(b), but for density anomalies (kg m−3; CI = 0.2 kg m−3) and (e) total contribution from (c) temperature and (d) salinity anomalies. (f)–(h) As in (c)–(e), but for the squared buoyancy frequency anomalies (s−2; CI = 5 × 10−5 s−2). Anomalies significant at the 90% confidence level by a two-tailed t test are shaded.
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
To assess the impacts of these density anomalies on the upper ocean stratification, we take their vertical derivative to convert to squared buoyancy frequency anomalies (Figs. 8f–h). In the eastern part, the combination of negative density anomalies near the surface and positive density anomalies near the pycnocline leads to significant increase in density gradient and vertical stability (Fig. 8h). Comparing the amplitude of temperature and salinity contribution, we find that salinity-related anomalies have comparable or even larger magnitude.
When similar analyses are conducted for the off-equatorial region (Fig. 9), a zonal dipole structure in temperature anomalies is found near the thermocline (Fig. 9a). Positive anomalies in the west are related to downwelling signals induced by anomalous wind stress curl, whereas negative anomalies in the east may be associated with coastal upwelling and/or upwelling Rossby waves reflected from the eastern boundary (Rao et al. 2002; Shinoda et al. 2004). Positive (negative) anomalies in the east (west) are evident from the salinity composite (Fig. 9b), as discussed in section 3.
As in Fig. 8, but for the off-equatorial region (averaged over 9°–5°S).
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
Reflecting temperature and salinity anomalies, total density anomalies in the off-equatorial region (Fig. 9e) have a dipole structure with the opposite signs compared to those of the equatorial section. They are primarily dominated by contribution from temperature anomalies (Fig. 9c), but the salinity contribution has nonnegligible amplitude and amplifies the signals (Fig. 9d).
A composite of squared buoyancy frequency anomalies (Fig. 9h) reveals zonally elongated negative anomalies in the central-eastern part, indicating destabilization of the upper water column. These anomalies are explained by positive density anomalies near the surface and negative anomalies below the pycnocline. Again, the contribution from temperature anomalies prevails over that of salinity anomalies (Figs. 9f,g). However, especially in the central-eastern region, weakening of salinity stratification plays an important role in determining the structure of total anomalies.
In summary, the salinity stratification in the central-eastern equatorial Indian Ocean is strengthened due to the negative (positive) salinity anomalies near the surface (subsurface). On the other hand, subsurface freshening and surface saltening in the southeastern tropical Indian Ocean lead to a decrease in the vertical gradient of density and destabilization of the upper ocean. In both regions, the changes in salinity stratification are comparable in magnitude to those in temperature stratification in terms of contribution to density stratification. These results suggest that the IOD-related salinity anomalies have significant influences on the stability of the upper ocean and they have potential to modulate the strength of the vertical mixing and evolution of temperature.
5. Sensitivity experiments
In the previous section, we have quantified the influence of the pIOD-related salinity anomalies on the upper ocean stratification and shown that they have large impacts comparable to those of temperature. Then, to what extent do they affect the strength of the vertical mixing in the upper ocean and the evolution of SST during the pIOD? To bridge the gap between the changes in salinity stratification and the evolution of temperature, we perform sensitivity experiments using a 1D ML model.
a. Experimental design of sensitivity experiments
The 1D ML model we adopt here is a level-2.5 turbulence closure model (Mellor and Yamada 1982; Nakanishi and Niino 2009) with modified closure constants (Furuichi et al. 2012). Through comparisons with large eddy simulations and observations, this model has been shown to have a good skill in simulating vertical mixing in the upper ocean (Furuichi et al. 2012; Furuichi and Hibiya 2015). A detailed formulation and description of the 1D ML model are given in the appendix.
To quantify the impacts of salinity on the strength of the upper-ocean mixing and temperature evolution, we perform two kinds of sensitivity experiments. One is the control (CTL) experiment, which uses temperature, salinity, and velocity profiles derived from the ORAS4 as the initial condition. This experiment is designed to reproduce the observed thermohaline structures and serves as a benchmark for comparison. For this purpose, Newtonian damping terms to the observed values are added to temperature, salinity, and velocity equations as the heat, salt, and momentum flux correction terms, respectively. These terms serve as surrogates of three-dimensional processes, which play crucial roles in determining their evolution, but are not included in the 1D ML model. This method is essential for preventing the model from drifting to an unrealistic state. For a more detailed description of this method, readers may refer to section b of the appendix.
In the other experiment, called the climatological salinity (Sclim) experiment, the model is initialized with the same conditions as the CTL experiment, except that those of salinity are replaced by the climatological values. To focus on the impacts on the strength of the vertical mixing, we assume that other processes contributing to the evolution of temperature and momentum (such as horizontal and vertical advection) are not influenced by salinity anomalies. Based on this assumption, we force the model with the same heat and momentum boundary conditions as the CTL experiments, while salinity is restored to the climatological values. Since the only difference between CTL and Sclim experiments is initial and restoring salinity profiles, their disparity is purely caused by salinity anomalies. By comparing them, we can quantitatively evaluate the influences of salinity anomalies on the vertical mixing and the evolution of temperature. The concept of these experiments is schematically summarized in Fig. 10.
Schematic diagram illustrating the configuration of CTL and Sclim experiments.
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
We carry out these experiments at each grid point in 1° longitude × 1° latitude of the central-eastern tropical Indian Ocean (10°S–10°N, 80°–110°E) to identify regions where salinity anomalies have large impacts on SST. Integrations start from 15 August of each pIOD year and integrated for 3 months with the time steps of 900 s. The maximum depth is 300 m and the vertical resolution is uniform at 2 m. Daily shortwave/longwave radiation, 2-m air temperature/specific humidity, 10-m winds, and precipitation from the ECMWF atmospheric reanalysis are used to force the model. Turbulent heat fluxes, evaporation, and wind stress are calculated using bulk formulas of Kara et al. (2005).
b. Results of sensitivity experiments
Figure 11 shows a composite of the difference in SST (temperature at 1-m depth) between CTL and Sclim experiments after 3-month integration. Note that regions with positive (negative) values indicate that inclusion of salinity anomalies results in warmer (cooler) SST. We can see significant warming along the equator (~0.3°–0.5°C), where both surface and subsurface salinity anomalies strengthen the upper ocean stratification. On the other hand, there are regions with significant cooling (~0.2°–0.5°C) in the southeastern tropical Indian Ocean. These regions roughly coincide with the region of reduced salinity stratification. Although the amplitude of these differences varies among the events, the maximum reaches 0.8°C in strong pIOD events of 1961 and 1997 (figure not shown). Since the differences between the two experiments are statistically significant, we conclude that they are typical features of the pIOD events. These results evidence the idea that the modulations of stratification induced by salinity anomalies have significant impacts on the evolution of SST associated with the pIOD in the tropical Indian Ocean.
Composite of difference between CTL and Sclim experiments in SST (°C; CI = 0.1°C) after 3-month integration. The differences significant at the 90% confidence level are shaded. The red and blue boxes denote box 1 and box 2, respectively (see the main text).
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
To understand the detailed mechanisms of the above differences, we check the area-averaged time evolution of temperature, salinity, squared buoyancy frequency, horizontal currents, turbulent kinetic energy, turbulent length scale, and vertical diffusion coefficient in regions where significant SST differences are found. For this purpose, we introduce box 1 (1°S–1°N, 86°–93°E) and box 2 (9°–6°S, 87°–96°E), which serve as a representative areas with significant warming and cooling, respectively.
In box 1, a combination of surface freshening and subsurface saltening (Fig. 12b) leads to stronger stratification (Fig. 12e). Enhanced stratification hampers the generation of the turbulent kinetic energy [Fig. 12f: see also the second term on the right-hand side of Eq. (A7)] and limits the spatial scale of turbulent eddies [Fig. 12g: see also lB in Eq. (A19)]. Therefore, the strength of the vertical mixing within the mixed layer is greatly reduced (Fig. 12h). The reduction of vertical diffusion coefficients is around 2 × 10−3 m2 s−1, which amounts to 30%–40% of their total values. Consequently, the exchange of heat between the upper and lower layers is reduced, resulting in surface warming and subsurface cooling of approximately 0.4°C (Fig. 12a). In addition, stronger stratification near the surface traps input of northwestward momentum by southwesterly winds and enhances northwestward (northeastward) currents in the mixed layer (below the mixed layer) (Figs. 12c,d) by about 0.05 m s−1.
Time evolution of composited difference in (a) temperature (°C; CI = 0.05°C), (b) salinity (psu; CI = 0.1 psu), (c) zonal and (d) meridional velocity (m s−1; CI = 0.01 m s−1), (e) squared buoyancy frequency (s−2; CI = 4 × 10−5 s−2), (f) turbulent kinetic energy (m2 s−2; CI = 2.5 × 10−6 m2 s−2), (g) turbulent length scale (m; CI = 5 × 10−2 m), and (h) vertical diffusion coefficient (m2 s−1; CI = 1 × 10−4 m2 s−1), averaged over box 1 (1°S–1°N, 86°–93°E). Differences significant at the 90% confidence level by a two-tailed t test are shaded.
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
On the other hand, positive salinity anomalies near the surface and relatively weak negative salinity anomalies near the halocline destabilize the upper ocean stratification in box 2 (Figs. 13b,e), as mentioned in the previous section. Accordingly, the turbulent kinetic energy and length scale of turbulence become larger (Figs. 13f,g) and the vertical mixing within the mixed layer is enhanced (the diffusion coefficients increase by 40%–50%; Fig. 13h). As a result, significant cooling (warming) near the surface (thermocline) of about 0.4°C is induced (Fig. 13a). The differences in the velocity between the two experiments are less than 0.01 m s−1 and much smaller than box 1 (Figs. 13c,d). This may be due to the fact that MLD in box 2 is deeper than that in box 1 by 10–20 m, and the momentum trapping is not as efficient. All of these results derived from sensitivity experiments support the conjecture presented in the previous section that the changes in salinity stratification lead to modulation of vertical mixing and strongly affect the thermal structure in the upper ocean during the pIOD.
As in Fig. 12, but for box 2 (9°–6°S, 87°–96°E). Note that color bars and contour intervals are different from Fig. 12 for (f)–(h), and their CI = 1 × 10−5 m2 s−2, 1.5 × 10−1 m, and 8 × 10−4 m2 s−1, respectively.
Citation: Journal of Climate 30, 19; 10.1175/JCLI-D-17-0133.1
6. Summary and discussion
a. Summary
In this study, salinity variations associated with the pIOD and their impacts on the upper ocean are first investigated using an observational dataset and an ocean reanalysis product. As already shown by previous studies (Thompson et al. 2006; Grunseich et al. 2011; Zhang et al. 2013), negative (positive) SSS anomalies appear in the central-eastern equatorial Indian Ocean (southeastern tropical Indian Ocean) during pIOD events. In addition to these well-documented SSS anomalies, we have found positive salinity anomalies near the halocline in the eastern equatorial Indian Ocean, and negative subsurface salinity anomalies in the central off-equatorial region. These salinity anomalies are not due to El Niño, which sometimes co-occurs with the pIOD.
The generation mechanisms of these salinity anomalies can be explained by the ocean adjustment associated with the pIOD. During the pIOD, easterly wind anomalies along the equator cause weakening of the climatological eastward current in boreal fall (Wyrtki jet), which transports salty water from the western Indian Ocean. Consequently, SSS in the central-eastern equatorial Indian Ocean becomes anomalously low. The easterly wind anomalies also strengthen the upwelling in the eastern equatorial Indian Ocean, and reinforce the vertical advection of high salinity water from the subsurface that contributes to the saltening near the halocline. At the same time, anticyclonic wind stress curl anomalies generate downwelling anomalies in the off-equatorial region. These downwelling anomalies push the halocline downward and generate negative salinity anomalies in tandem with westward current anomalies. These mechanisms are quantitatively verified by salinity balance analysis.
The influences of these salinity anomalies on the upper ocean stratification are then diagnosed using the density decomposition method. It is found that the vertical density gradient and the stability of water column are significantly increased in the upper layer of the eastern equatorial Indian Ocean because of negative SSS anomalies along with positive salinity anomalies near the halocline. Conversely, near-surface saltening and subsurface freshening destabilize the upper ocean in the southeastern tropical Indian Ocean. Thus, amplitude of density and squared buoyancy frequency anomalies are greatly underestimated if we do not take account of the above salinity anomalies into the computation. These results highlight the importance of salinity anomalies in determining thermohaline structure in the upper ocean of the tropical Indian Ocean and their potential of controlling the vertical mixing process and temperature evolution.
For quantitative assessment of the salinity impact, we have developed a 1D ML model based on Furuichi et al. (2012) and conducted two sensitivity experiments with (CTL) and without (Sclim) salinity anomalies in initial conditions and flux corrections for each pIOD events. We have found that SST over the eastern equatorial Indian Ocean (southeastern tropical Indian Ocean) in the CTL experiment is significantly higher (lower) than that of the Sclim experiment. Time evolutions of the vertical diffusion coefficients suggest that these differences are attributed to reduced (enhanced) vertical mixing and entrainment of cold water. The magnitude of SST difference between two experiments is about 0.4°C, but it reached up to 1.0°C in an extreme event. Therefore, pIOD-related salinity anomalies can exert large impacts on SST through the modification of vertical mixing processes and potentially feed back onto the evolution of the IOD itself. This is the first study that quantitatively demonstrates the impacts of pIOD-related salinity anomalies on the SST, although many previous studies suggested their potential importance.
b. Discussion
In this study, we have shown that both surface and subsurface salinity undergo significant variation during pIOD events. Throughout this paper, we have only discussed the pIOD-related salinity variation and their importance on the stratification and have not dealt with the negative IOD (nIOD). During the nIOD, surface and subsurface salinity in the tropical Indian Ocean exhibit fluctuations similar to those during the pIOD, but with the opposite sign. However, because of the nonlinear nature of the IOD (Hong et al. 2008; Cai and Qiu 2013), salinity anomalies associated with nIOD are not merely a mirror image of those of the pIOD (figure not shown). Generally, the amplitude of nIOD-related salinity anomalies is weaker than that of pIOD-related anomalies.
In addition, we have focused on typical pIOD events in this study. Recent studies revealed that the IOD can be classified into several categories based on its seasonality (Du et al. 2013; Weller et al. 2014) and spatial pattern of SST anomalies (Endo and Tozuka 2016; Tozuka et al. 2016). It will be interesting to see how salinity anomalies and the roles played by them are different among these different types of the IOD.
Another interesting topic is the relationship between decadal variations of the IOD (Tozuka et al. 2007) and those of salinity. Recently, Du et al. (2015) reported the significant freshening of the upper ocean salinity in the southeastern tropical Indian Ocean since the mid-1990s. Although the regions with the decreasing trend are located on the southeastern side of the eastern pole of the IOD, the freshening may affect the salinity anomalies associated with the IOD via changes in the climatological salinity gradient. The influence of these low-frequency variations in mean thermohaline structure on the property of the IOD should be pursued further in future studies. In addition to long-term natural variability, the acceleration of the hydrological cycle associated with global warming (Held and Soden 2006) is projected to amplify the contrasts in the mean SSS pattern (Durack and Wijffels 2010). Since advection of climatological salinity by anomalous current is the main generation mechanism of the IOD-related salinity anomalies, stronger mean salinity gradient in a warmer climate may augment variations in salinity stratification and alter the spatial pattern and amplitude of the IOD itself. Therefore, unraveling the relationship between the low-frequency modulation/long-term trend of the mean salinity structure and interannual salinity variability is also important for understanding future changes of the IOD (Cai et al. 2013, 2014).
Through sensitivity experiments, we have shown that the impacts of salinity anomalies on SST are about 0.3°–0.4°C. Given that the amplitude of SST anomalies associated with the IOD is about 0.5°–1°C, SST differences between the two experiments with and without salinity anomalies amount to about 30%–80% of IOD-related SST anomalies. Since the SST in the eastern tropical Indian Ocean is close to 28°C, which is the critical threshold of deep convection (Gadgil et al. 1984; Graham and Barnett 1987), small fluctuations of SST associated with salinity anomalies may exert profound impacts on the air–sea interaction processes and hence the IOD evolution itself. The present results may also have important implications for coupled modeling. Many CGCMs have biases in simulating the main characteristics of the IOD such as its amplitude and spatial pattern of SST anomalies (Liu et al. 2011; Cai and Cowan 2013; Liu et al. 2014). Since SST anomalies and associated changes in the strength of convective activity in the tropics affect the global climate through atmospheric teleconnection, their accurate simulation is of great importance. In addition, the predictability of the IOD is lower than that of the ENSO (Luo et al. 2007; Shi et al. 2012; Liu et al. 2017). The reasons for simulation biases and lower predictability of the IOD may be attributed to various factors such as improper representation of mean thermocline structure, stochastic intraseasonal variability, and the presence of monsoonal circulation. In addition to these factors, our results suggest that an accurate representation of salinity anomalies may also be a key factor for successful simulation and prediction of the IOD since they play crucial roles in the SST evolution.
Furthermore, SSS in the tropical Indian Ocean has strong intraseasonal variability (Drushka et al. 2014; Li et al. 2015; Horii et al. 2016) besides the interannual variability discussed in this study. The main sources of intraseasonal variations are the energetic atmospheric variability associated with the Madden–Julian oscillation (MJO; Madden and Julian 1971) and resulting modulation of ocean currents and freshwater fluxes (Li et al. 2015). Horii et al. (2016) suggested a possibility that these SSS variations affect the SST evolution via changes in the MLD. In addition, Drushka et al. (2014) and Horii et al. (2016) pointed out that the amplitude of intraseasonal SST variations may be closely related to the strength of background thermohaline conditions. These intriguing hypotheses can be quantified by performing similar 1D ML model experiments. The 1D ML modeling framework used in this study is applicable to other regions and phenomenon with various time scales and may help us to reveal the essential processes that govern the vertical mixing of the upper ocean and evolution of physical parameters.
We note that there is a caveat for interpreting results from 1D ML model experiments. It is assumed that salinity anomalies only modulate the strength of vertical diffusion and other processes such as horizontal and vertical advection are not influenced by salinity anomalies. However, since salinity affects pressure gradient in addition to vertical diffusion, oceanic velocities and associated advective processes are also expected to be altered by salinity anomalies. To assess the relative importance of changes in such processes and vertical mixing, well-designed sensitivity experiments using an OGCM are required. Our results derived from simplified experiments will be a cornerstone for understanding the results of elaborated experiments that entail multiple interactions of various processes.
In addition, we forced the model by prescribed atmosphere and neglected air–sea coupling processes. In reality, SST perturbations induced by salinity anomalies affect the overlying atmosphere. The resulting change in the atmospheric circulation may influence the evolution of the original salinity anomalies through changes in freshwater flux and currents. Clarifying the roles played by such feedback processes is an important and interesting issue that requires further investigation. More observations of such processes and modeling experiments using atmosphere–ocean coupled models are essential to disentangle complex nature of coupling processes.
Acknowledgments
We thank two anonymous reviewers for providing constructive comments that helped us to improve our original manuscript. We would like to express our gratitude to Profs. Yukio Masumoto and Ichiro Yasuda for providing insightful suggestions. The present research was supported by the Japan Society for Promotion of Science through Grant-in-Aid for Scientific Research (B) JP16H04047 and partly carried out for the iDEWS project supported by SATREPS Program of JICA/AMED in Japan and ACCESS (NRF/DST) in South Africa. The first author is financially supported by Research Fellowship of Japan Society for the Promotion of Science (JSPS) and Leading Graduate Course for Frontiers of Mathematical Sciences and Physics, MEXT, Japan.
APPENDIX
Formulation of One-Dimensional Mixed Layer Model
a. General formulation of model
List of closure constants.
Equations (A1), (A2), (A3), (A4), and (A7) are discretized on a staggered grid [T, S, U, and V are placed on the vertical ith grid, while κT, κS, κV, and q2 are on (i + 1/2)th grid] with implicit time stepping. These equations are reduced to a tridiagonal system.
b. Flux correction method
Because of the absence of 3D processes such as horizontal/vertical advection and horizontal diffusion, integration of 1D models results in a drift to an unrealistic state. To prevent such spurious behavior, we use a flux correction approach described below.
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.
Balmaseda, M. A., K. Mogensen, and A. T. Weaver, 2013: Evaluation of the ECMWF ocean reanalysis system ORAS4. Quart. J. Roy. Meteor. Soc., 139, 1132–1161, doi:10.1002/qj.2063.
Bryan, F., 1986: High-latitude salinity effects and interhemispheric thermohaline circulations. Nature, 323, 301–304, doi:10.1038/323301a0.
Cai, W., and T. Cowan, 2013: Why is the amplitude of the Indian Ocean dipole overly large in CMIP3 and CMIP5 climate models? Geophys. Res. Lett., 40, 1200–1205, doi:10.1002/grl.50208.
Cai, W., and Y. Qiu, 2013: An observation-based assessment of nonlinear feedback processes associated with the Indian Ocean dipole. J. Climate, 26, 2880–2890, doi:10.1175/JCLI-D-12-00483.1.
Cai, W., X.-T. Zheng, E. Weller, M. Collins, T. Cowan, M. Lengaigne, W. Yu, and T. Yamagata, 2013: Projected response of the Indian Ocean dipole to greenhouse warming. Nat. Geosci., 6, 999–1007, doi:10.1038/ngeo2009.
Cai, W., A. Santoso, G. Wang, E. Weller, L. Wu, K. Ashok, Y. Masumoto, and T. Yamagata, 2014: Increased frequency of extreme Indian Ocean dipole events due to greenhouse warming. Nature, 510, 254–258, doi:10.1038/nature13327.
Carton, J. A., and B. S. Giese, 2008: A reanalysis of ocean climate using Simple Ocean Data Assimilation (SODA). Mon. Wea. Rev., 136, 2999–3017, doi:10.1175/2007MWR1978.1.
Chen, G., W. Han, Y. Li, D. Wang, and M. J. McPhaden, 2015: Seasonal-to-interannual time-scale dynamics of the Equatorial Undercurrent in the Indian Ocean. J. Phys. Oceanogr., 45, 1532–1553, doi:10.1175/JPO-D-14-0225.1.
Chen, G., W. Han, Y. Shu, Y. Li, D. Wang, and Q. Xie, 2016: The role of Equatorial Undercurrent in sustaining the Eastern Indian Ocean upwelling. Geophys. Res. Lett., 43, 6444–6451, doi:10.1002/2016GL069433.
Cooper, N. S., 1988: The effect of salinity on tropical ocean models. J. Phys. Oceanogr., 18, 697–707, doi:10.1175/1520-0485(1988)018<0697:TEOSOT>2.0.CO;2.
Currie, J. C., M. Lengaigne, J. Vialard, D. M. Kaplan, O. Aumont, S. W. A. Naqvi, and O. Maury, 2013: Indian Ocean dipole and El Niño/Southern Oscillation impacts on regional chlorophyll anomalies in the Indian Ocean. Biogeosciences, 10, 6677–6698, doi:10.5194/bg-10-6677-2013.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828.
Drushka, K., J. Sprintall, and S. T. Gille, 2014: Subseasonal variations in salinity and barrier-layer thickness in the eastern equatorial Indian Ocean. J. Geophys. Res. Oceans, 119, 805–823, doi:10.1002/2013JC009422.
Du, Y., and Y. Zhang, 2015: Satellite and Argo observed surface salinity variations in the tropical Indian Ocean and their association with the Indian Ocean dipole mode. J. Climate, 28, 695–713, doi:10.1175/JCLI-D-14-00435.1.
Du, Y., W. Cai, and Y. Wu, 2013: A new type of the Indian Ocean dipole since the mid-1970s. J. Climate, 26, 959–972, doi:10.1175/JCLI-D-12-00047.1.
Du, Y., Y. Zhang, M. Feng, T. Wang, N. Zhang, and S. Wijffels, 2015: Decadal trends of the upper ocean salinity in the tropical Indo-Pacific since mid-1990s. Sci. Rep., 5, 16050, doi:10.1038/srep16050.
Durack, P. J., and S. E. Wijffels, 2010: Fifty-year trends in global ocean salinities and their relationship to broad-scale warming. J. Climate, 23, 4342–4362, doi:10.1175/2010JCLI3377.1.
Durand, F., G. Alory, R. Dussin, and N. Reul, 2013: SMOS reveals the signature of Indian Ocean dipole events. Ocean Dyn., 63, 1203–1212, doi:10.1007/s10236-013-0660-y.
Endo, S., and T. Tozuka, 2016: Two flavors of the Indian Ocean dipole. Climate Dyn., 46, 3371–3385, doi:10.1007/s00382-015-2773-0.
Feng, M., and S. Wijffels, 2002: Intraseasonal variability in the South Equatorial Current of the east Indian Ocean. J. Phys. Oceanogr., 32, 265–277, doi:10.1175/1520-0485(2002)032<0265:IVITSE>2.0.CO;2.
Furuichi, N., and T. Hibiya, 2015: Assessment of the upper-ocean mixed layer parameterizations using a large eddy simulation model. J. Geophys. Res. Oceans, 120, 2350–2369, doi:10.1002/2014JC010665.
Furuichi, N., T. Hibiya, and Y. Niwa, 2012: Assessment of turbulence closure models for resonant inertial response in the oceanic mixed layer using a large eddy simulation model. J. Oceanogr., 68, 285–294, doi:10.1007/s10872-011-0095-3.
Gadgil, S., P. V. Joseph, and N. V. Joshi, 1984: Ocean–atmosphere coupling over monsoon regions. Nature, 312, 141–143, doi:10.1038/312141a0.
Graham, N. E., and T. P. Barnett, 1987: Sea surface temperature, surface wind divergence, and convection over tropical oceans. Science, 238, 657–659, doi:10.1126/science.238.4827.657.
Grunseich, G., B. Subrahmanyam, V. S. N. Murty, and B. S. Giese, 2011: Sea surface salinity variability during the Indian Ocean dipole and ENSO events in the tropical Indian Ocean. J. Geophys. Res. Oceans, 116, C11013, doi:10.1029/2011JC007456.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, doi:10.1175/JCLI3990.1.
Hong, C.-C., T. Li, LinHo, and J.-S. Kug, 2008: Asymmetry of the Indian Ocean dipole. Part I: Observational analysis. J. Climate, 21, 4834–4848, doi:10.1175/2008JCLI2222.1.
Horii, T., I. Ueki, K. Ando, T. Hasegawa, K. Mizuno, and A. Seiki, 2016: Impact of intraseasonal salinity variations on sea surface temperature in the eastern equatorial Indian Ocean. J. Oceanogr., 72, 313–326, doi:10.1007/s10872-015-0337-x.
Hosoda, S., T. Ohira, and T. Nakamura, 2008: A monthly mean dataset of global oceanic temperature and salinity derived from Argo float observations. JAMSTEC Rep. Res. Dev., 8, 47–59, doi:10.5918/jamstecr.8.47.
Kara, A. B., H. E. Hurlburt, and A. J. Wallcraft, 2005: Stability-dependent exchange coefficients for air–sea fluxes. J. Atmos. Oceanic Technol., 22, 1080–1094, doi:10.1175/JTECH1747.1.
Keerthi, M. G., M. Lengaigne, J. Vialard, C. de Boyer Montégut, and P. M. Muraleedharan, 2013: Interannual variability of the tropical Indian Ocean mixed layer depth. Climate Dyn., 40, 743–759, doi:10.1007/s00382-012-1295-2.
Li, J., C. Liang, Y. Tang, C. Dong, D. Chen, X. Liu, and W. Jin, 2016: A new dipole index of the salinity anomalies of the tropical Indian Ocean. Sci. Rep., 6, 24260, doi:10.1038/srep24260.
Li, Y., W. Han, and T. Lee, 2015: Intraseasonal sea surface salinity variability in the equatorial Indo-Pacific Ocean induced by Madden–Julian oscillations. J. Geophys. Res. Oceans, 120, 2233–2258, doi:10.1002/2014JC010647.
Liu, H., Y. Tang, D. Chen, and T. Lian, 2017: Predictability of the Indian Ocean dipole in the coupled models. Climate Dyn., 48, 2005–2024, doi:10.1007/s00382-016-3187-3.
Liu, L., W. Yu, and T. Li, 2011: Dynamic and thermodynamic air–sea coupling associated with the Indian Ocean dipole diagnosed from 23 WCRP CMIP3 models. J. Climate, 24, 4941–4958, doi:10.1175/2011JCLI4041.1.
Liu, L., S.-P. Xie, X.-T. Zheng, T. Li, Y. Du, G. Huang, and W. Yu, 2014: Indian Ocean variability in the CMIP5 multi-model ensemble: The zonal dipole mode. Climate Dyn., 43, 1715–1730, doi:10.1007/s00382-013-2000-9.
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.
Luo, J., S. Masson, S. K. Behera, and T. Yamagata, 2007: Experimental forecasts of the Indian Ocean dipole using a coupled OAGCM. J. Climate, 20, 2178–2190, doi:10.1175/JCLI4132.1.
Madden, R. A., and P. R. Julian, 1971: Detection of a 40–50 day oscillation in the zonal wind in the tropical Pacific. J. Atmos. Sci., 28, 702–708, doi:10.1175/1520-0469(1971)028<0702:DOADOI>2.0.CO;2.
Maes, C., 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, and S. Belamari, 2005: Importance of the salinity barrier layer for the buildup of El Niño. J. Climate, 18, 104–118, doi:10.1175/JCLI-3214.1.
Masson, S., J.-P. Boulanger, C. Menkes, P. Delecluse, and T. Yamagata, 2004: Impact of salinity on the 1997 Indian Ocean dipole event in a numerical experiment. J. Geophys. Res., 109, C02002, doi:10.1029/2003JC001807.
Masumoto, Y., H. Hase, Y. Kuroda, H. Matsuura, and K. Takeuchi, 2005: Intraseasonal variability in the upper layer currents observed in the eastern equatorial Indian Ocean. Geophys. Res. Lett., 32, L02607, doi:10.1029/2004GL021896.
Mellor, G. L., and T. Yamada, 1982: Development of a turbulence closure model for geophysical fluid problems. Rev. Geophys., 20, 851–875, doi:10.1029/RG020i004p00851.
Mogensen, K., M. A. Balmaseda, and A. Weaver, 2012: The NEMOVAR ocean data assimilation system as implemented in the ECMWF ocean analysis for System 4. ECMWF Tech. Memo. 668, 59 pp. [Available online at https://www.ecmwf.int/sites/default/files/elibrary/2012/11174-nemovar-ocean-data-assimilation-system-implemented-ecmwf-ocean-analysis-system-4.pdf.]
Nakanishi, M., and H. Niino, 2009: Development of an improved turbulence closure model for the atmospheric boundary layer. J. Meteor. Soc. Japan, 87, 895–912, doi:10.2151/jmsj.87.895.
Nyadjro, E. S., and M. J. McPhaden, 2014: Variability of zonal currents in the eastern equatorial Indian Ocean on seasonal to interannual time scales. J. Geophys. Res. Oceans, 119, 7969–7986, doi:10.1002/2014JC010380.
Paulson, C. A., and J. J. Simpson, 1977: Irradiance measurements in the upper ocean. J. Phys. Oceanogr., 7, 952–956, doi:10.1175/1520-0485(1977)007<0952:IMITUO>2.0.CO;2.
Picaut, J., M. Ioualalen, C. Menkes, T. Delcroix, and M. J. McPhaden, 1996: Mechanism of the zonal displacements of the Pacific warm pool: Implications for ENSO. Science, 274, 1486–1489, doi:10.1126/science.274.5292.1486.
Qiu, Y., W. Cai, L. Li, and X. Guo, 2012: Argo profiles variability of barrier layer in the tropical Indian Ocean and its relationship with the Indian Ocean dipole. Geophys. Res. Lett., 39, L08605, doi:10.1029/2012GL051441.
Qu, T., and G. Meyers, 2005: Seasonal variation of barrier layer in the southeastern tropical Indian Ocean. J. Geophys. Res., 110, C11003, doi:10.1029/2004JC002816.
Rao, R. R., and R. Sivakumar, 2003: Seasonal variability of sea surface salinity and salt budget of the mixed layer of the north Indian Ocean. J. Geophys. Res., 108, 3009, doi:10.1029/2001JC000907.
Rao, S. A., and T. Yamagata, 2004: Abrupt termination of Indian Ocean dipole events in response to intraseasonal disturbances. Geophys. Res. Lett., 31, L19306, doi:10.1029/2004GL020842.
Rao, S. A., and S. K. Behera, 2005: Subsurface influence on SST in the tropical Indian Ocean: Structure and interannual variability. Dyn. Atmos. Oceans, 39, 103–135, doi:10.1016/j.dynatmoce.2004.10.014.
Rao, S. A., S. K. Behera, Y. Masumoto, and T. Yamagata, 2002: Interannual subsurface variability in the tropical Indian Ocean with a special emphasis on the Indian Ocean dipole. Deep-Sea Res. II, 49, 1549–1572, doi:10.1016/S0967-0645(01)00158-8.
Roemmich, D., and J. Gilson, 2009: The 2004–2008 mean and annual cycle of temperature, salinity, and steric height in the global ocean from the Argo Program. Prog. Oceanogr., 82, 81–100, doi:10.1016/j.pocean.2009.03.004.
Saji, N. H., and T. Yamagata, 2003: Possible impacts of Indian Ocean dipole mode events on global climate. Climate Res., 25, 151–169, doi:10.3354/cr025151.
Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401, 360–363.
Sayantani, O., and C. Gnanaseelan, 2015: Tropical Indian Ocean subsurface temperature variability and the forcing mechanisms. Climate Dyn., 44, 2447–2462, doi:10.1007/s00382-014-2379-y.
Shi, L., H. H. Hendon, O. Alves, J.-J. Luo, M. Balmaseda, and D. Anderson, 2012: How predictable is the Indian Ocean dipole? Mon. Wea. Rev., 140, 3867–3884, doi:10.1175/MWR-D-12-00001.1.
Shinoda, T., H. H. Hendon, and M. A. Alexander, 2004: Surface and subsurface dipole variability in the Indian Ocean and its relation with ENSO. Deep-Sea Res. I, 51, 619–635, doi:10.1016/j.dsr.2004.01.005.
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.
Thompson, B., C. Gnanaseelan, and P. S. Salvekar, 2006: Variability in the Indian Ocean circulation and salinity and its impact on SST anomalies during dipole events. J. Mar. Res., 64, 853–880, doi:10.1357/002224006779698350.
Tozuka, T., J.-J. Luo, S. Masson, and T. Yamagata, 2007: Decadal modulations of the Indian Ocean dipole in the SINTEX-F1 coupled GCM. J. Climate