1 Introduction

Ocean salinity is an important property of the water mass (Delcroix and Hénin 1991). Ocean salinity is a key indicator in the global hydrological cycle (Durack and Wijffels 2010; Durack et al. 2012; Schmitt 2008; Schmitt and Blair 2015; Skliris et al. 2014) and is closely related to both the dynamics and thermodynamics of the oceans (Lagerloef 2002; Rao and Sivakumar 2003; Ren and Riser 2009). Salinity affects ocean stability (Nyadjro et al. 2012; Schiller et al. 1997), the dynamic variability of the ocean (Stammer 1997; Sun et al. 2015; Thompson et al. 2006) and air–sea interactions (Grunseich et al. 2013; Guan et al. 2014; Horii et al. 2016; Simon et al. 2006; Williams et al. 2010). The Array for Real-Time Geostrophic Oceanography (Argo) provides observations from the global ocean that allow detailed studies of the water mass and related ocean dynamics and has led to increasing numbers of salinity and temperature profiles for the upper 2000 m (Gould et al. 2004; Riser et al. 2008; Roemmich and Owens 2000; Roemmich et al. 2009). These data offer an opportunity to examine the variability in salinity and related ocean dynamics of the Indian Ocean (Jensen 2003; Masson et al. 2004).

The tropical Indian Ocean (TIO) has a unique role in the global climate system (Sharma et al. 2010; Subrahmanyam et al. 2011). The Indian Ocean has a high rate of precipitation and the most typical monsoon climate of the global ocean. Strong westerlies dominate the equatorial Indian Ocean (EIO) in the boreal spring and autumn during the transition seasons of the monsoon around May and October–November and these westerlies drive eastward surface currents known as Wyrtki jets (Wyrtki 1973; Reppin et al. 1999). Under normal conditions, variations in salinity affect the stratification and air–sea heat flux of the Indian Ocean, triggering a response in the atmospheric circulation that significantly influences the thermodynamics of the upper layer (Liang et al. 2018; Zeng and Wang 2017). These interactions lead to further variations in the sea temperature and convection, which indirectly influence the intensity and development of the southwest monsoon and affect the Indian Ocean and neighboring countries (Yadav et al. 2019; Sharma et al. 2020). It is therefore particularly important to study the variability of salinity in the TIO.

As a result of the significant difference in salinity between the Arabian Sea and the Bay of Bengal, the salinity in the upper layer of the TIO is generally higher in the west and lower in the east. The Indonesian throughflow and the south equatorial current transport low salinity water to the western Indian Ocean at 20–10° S. Different mechanisms are responsible for the variability of salinity at different depths and on different timescales. The surface salinity is largely controlled by the surface freshwater flux (FWF), whereas the salinity in the upper layer is also affected by horizontal advection, vertical entrainment, the turbulence of mixing, diffusion and nonlinear effects (Hasson et al. 2013, 2014; Martins and Stammer 2015).

The interannual variability of salinity in the TIO is closely related to the Indian Ocean dipole (IOD; Grunseich et al. 2011; Zhang et al. 2013; Nyadjro and Subrahmanyam 2014; Du and Zhang 2015) and the El Niño–Southern Oscillation (ENSO; Clarke and Liu 1994; Meyers 1996; Wijffels and Meyers 2004; Feng et al. 2003, 2010), which are the important climate events affecting coupled air–sea processes in the TIO (Schott et al. 2009; Phillips et al. 2021). Some studies have shown that these two interannual modes have significant impacts on the dynamics of the Indian Ocean via modulation of the Walker circulation. IOD and ENSO events sometimes appear simultaneously (Saji and Yamagata 2003; Stuecker et al. 2017)—for example, a positive IOD (pIOD) can co-occur with an El Niño event, whereas a negative IOD (nIOD) can co-occur with a La Niña event (Cai et al. 2009; Gnanaseelan et al. 2012; Hong et al. 2008).

Previous studies have focused on the variability of salinity and its relationship with precipitation and ocean processes in the TIO. Kido et al. (2019) concluded that positive sea surface salinity (SSS) anomalies in the southeastern TIO are primarily caused by a reduction in precipitation and partly by enhanced evaporation due to increased wind speeds, whereas negative SSS anomalies in the central-eastern EIO are generated by zonal advection anomalies induced by anomalous wind stress. Their results showed that large-scale ocean changes in response to pIOD-related atmospheric anomalies are the key drivers of the observed anomalies in salinity. Zhang et al. (2013) found different variations in salinity in the EIO under the influence of the pIOD and nIOD. Anomalous westward equatorial currents, attributed to anomalous easterlies, led to low salinity anomalies along the equator during the developing and mature phases of the pIOD (Du and Zhang 2015; Nyadjro and Subrahmanyam 2014; Vinayachandran and Nanjundiah 2009). The anomalies during an nIOD were the opposite of those during a pIOD.

The variations in salinity are also significant during the decay phases and the following year of IOD events in the TIO in regions where the circulation system has an important role in maintaining the heat and salt balance of the entire area, including the Wyrtki jets and the south equatorial current (Sun et al. 2019; Zhang et al. 2013; Li et al. 2018). The ENSO teleconnection also induces anticyclonic (cyclonic) atmospheric circulation anomalies in the TIO when an El Niño (La Niña) event appears (Xie et al. 2002; Wang et al. 2003) and the evolution of the ENSO is often associated with the development of the IOD. The variations in salinity observed throughout the EIO are largely explained by the dynamics responsible for dipole events. They also feed back to the climate events, mainly through ocean stratification, the barrier layer, temperature inversion, the sea temperature and large-scale circulation patterns (Kido and Tozuka 2017; Bhavani et al. 2017).

Most studies have considered variations in salinity under the influence of multiple climate events or the processes of salinity variation during the development of certain climate events. There have been few studies of the interannual anomalies of salinity in the TIO and comparisons of the mechanism between two abnormal events of the same type. We therefore studied the mechanisms responsible for the differences in significant anomalous salinity events in different time periods. Our results will help to predict future variations in salinity in the Indian Ocean under different climate events.

We found significant positive salinity anomalies in 2017 based on 6 years of observational data from the conductivity, temperature, depth (CTD) sonde used during the Indian Ocean spring voyage. We then used Argo gridded data from 2007 to 2018 to confirm this phenomenon (Fig. 1). We found extremely high salinity anomalies in the central TIO from the boreal autumn of 2010 to the boreal spring of 2011 (hereafter referred to as 10A11S; i.e., September 2010–May 2011; Fig. 1b) and from the boreal autumn of 2016 to the boreal spring of 2017 (hereafter referred to as 16A17S; i.e., September 2016–May 2017; Fig. 1c). We found high salinity anomalies (> 0.2 psu) relative to the climatology derived from the Argo data for the time period 2007–2018. We used the main area of high salinity anomalies at (70–90° E, 10° S–0°; Fig. 1, black boxes) as the focus of our research.

Fig. 1
figure 1

Annual climatological mean spatial distribution of the upper level a salinity (units: psu) and d temperature (units: °C) in the tropical and subtropical Indian Ocean based on eArgo data from 2007 to 2018. Spatial distribution of the salinity anomaly (units: psu) in the tropical and subtropical Indian Ocean during b 10A11S and c 16A17S. Spatial distribution of the temperature anomaly (units: °C) in the tropical and subtropical Indian Ocean during e 10A11S and f 16A17S. Only the anomalies > 95% confidence level based on Student’s t test are shown. The black boxes in parts b and c show the main area of the positive salinity anomaly and those in parts e and f indicate the position corresponding to the positive salinity anomaly

Full size image

Negative temperature anomalies were also present, although their location deviated from that of the significant positive salinity anomalies (Fig. 1e and f) and they presented a reverse relationship. There were strong positive salinity anomalies and negative temperature anomalies during 10A11S and weaker anomalies during 16A17S. Hovmöller diagrams of the salinity/temperature anomalies with longitude–latitude indicate a clear change from positive to negative from 2007 to 2018 (Fig. 2). The positive salinity anomalies lasted for the time periods 10A11S and 16A17S (Fig. 2, red boxes). The salinity anomalies in 10A11S were stronger than those in 16A17S, the range of the anomalies was larger and the salinity anomaly events took different times to reach their maximum.

Fig. 2
figure 2

Hovmöller diagrams of the salinity (shading; units: psu) and temperature (contourse; units: °C) anomalies along a the zonal section, averaged meridionally over 10° S–0° and b the meridional section, averaged zonally over 70–90° E based on the Argo data for 2007–2018. The red boxes indicate extremely high salinity anomalies during 10A11S and 16A17S

Full size image

We combined salinity data from various datasets with satellite and reanalysis data to analyze the influence of surface buoyancy fluxes and ocean dynamics on salinity anomalies in the Indian Ocean. We explored the underlying dynamics of the extremely high salinity anomalies in the central TIO during 10A11S and 16A17S and quantitatively analyzed their mechanisms using the salinity budget equation.

The remainder of this paper is organized as follows. Section 2 describes the data and methods used and our results are presented in Sect. 3. Section 3.1 discusses the spatial characteristics of the extremely high salinity anomalies. Section 3.2 explores the variations in the salinity anomalies caused by surface buoyancy fluxes, whereas Sect. 3.3 considers the anomalies caused by ocean dynamics, which are modulated by large-scale climate events. Section 3.4 analyzes the contribution of each item in the salinity budget equation causing the anomalous variations in salinity during 10A11S and 16A17S. Section 4 discusses our results and summarizes our study.

2 Data and methods

2.1 Data

We used the Scripps Institution of Oceanography gridded monthly Argo products to analyze the variability in salinity during the time period 2007–2018. The Argo datasets have a horizontal resolution of (1° × 1°) at each standard pressure level. The products provide temperature/salinity profiles after quality control. We also downloaded other salinity data for comparison to check the quality of the Argo data.

The datasets used for comparison purposes included the L3 V5.0 Soil Moisture Active Passive (SMAP) SSS produced by NASA’s Jet Propulsion Laboratory (Fore et al. 2020). These data have a spatial resolution of 0.25° and are available from April 2015 to April 2018. We also used the European Centre for Medium-Range Weather Forecasts (ECMWF) Ocean Reanalysis System 5 (ORAS5) salinity data with a horizontal resolution of (1° × 1°) and a near-surface resolution of 1 m in the vertical direction for 2007–2018. We downloaded the Simple Ocean Data Assimilation (SODA) salinity data with a resolution of (0.5° × 0.5°) for 2007–2015. We used another composite SSS level-3 ocean salinity debiased product generated operationally by the Centre Aval de Traitement des Données Soil Moisture and Ocean Salinity (SMOS) satellite. The nominal requirements for SMOS retrieval are to achieve a 0.1 accuracy (Mecklenburg et al. 2008; Font et al. 2010).

We also used in situ CTD measurements for the Indian Ocean voyage in spring 2014–2019. The data were analyzed in two sections: across the 0° (equator) section between 80 and 92° E and across the 80° E section between 6° S and 2° N. The spatial correlation coefficients when we compared the salinity data from the CTD sonde and the Argo and ORAS5 datasets in these two sections were mostly > 0.5 (figure not shown). Figure 3 shows the average salinity change trend for the two sections. The CTD data are for dozens of days from March to May in any one year, not the average of the whole month, but the Argo and ORAS5 downloaded data are monthly averages. The data used are all from the same area, but there will still be some deviation when calculating the salinity anomaly (e.g., in 2015), although the difference is relatively small. The datasets have large correlation coefficients and relatively small root-mean-square errors (RMSEs), so there are clear consistencies between these different types of data.

Fig. 3
figure 3

eTime series of the salinity anomaly (units: psu) in the CTD (blue), Argo (red) and ECMWF ORAS5 (yellow) datasets area-averaged at 80° E (left-hand panel) and the equator (right-hand panel) from 2014 to 2019. The correlation coefficients and RMSEs between the time series are estimated for the periods when data are available in both datasets

Full size image

To calculate the surface FWF, we used the Global Precipitation Climatology Project (GPCP) Version 2.3 monthly data with a resolution of (2.5° × 2.5°), supported by the NOAA Climate Data Record Program, both available since 1979 (Adler et al. 2003; Huffman et al. 2009), and the evaporation at a resolution of (1° × 1°) from the objectively analyzed air–sea heat fluxes (the OA Flux; Yu and Weller 2007). The flux was mapped onto a (1° × 1°) grid using linear interpolation by re-griding the GPCP data onto the OA Flux data.

We used the Oceanic General Circulation Model (Masumoto et al. 2004; Sasaki et al. 2004) for the Earth Simulator model to obtain the ocean current data. This model is forced by National Centers for Environmental Prediction (NCEP) winds and has a horizontal resolution of (0.1° × 0.1°) with 54 vertical levels. The NCEP monthly pressure 10–1000 hPa wind fields with a resolution of (2.5° × 2.5°) were used to analyze the wind anomalies.

The IOD is characterized by the dipole mode index (DMI), defined as the difference in the area-averaged sea surface temperature anomaly (SSTA) between the western (50–70° E, 10° S–10° N) and eastern (90–110° E, 10° S–0°) TIO (Saji et al. 1999). The ENSO is characterized by the Oceanic Niño Index (ONI), defined as the area-averaged SSTA in the Niño 3.4 region (120–170° W, 5° S–5° N). We used SSTA data from NOAA Extended Reconstructed Sea Surface Temperature Version 5 (ERSSTv5) dataset with a resolution of (2° × 2°) to calculate the ONI and DMI and selected ± 0.5 times the standard deviation as their threshold.

All the anomalous data, apart from the general features of salinity, were obtained by subtracting the climatological annual cycles from the monthly time series during the time period 2007–2018. Most of the data were monthly means downloaded from the Asia–Pacific Data-Research Center of the International Pacific Research Center at the University of Hawaii (http://apdrc.soest.hawaii.edu).

2.2 Data validation

We compared the Argo, ECMWF ORAS5, SMAP and SODA salinity datasets from 2007 to 2018 to validate the accuracy of the salinity data. The spatial distributions of these salinity data were similar (figure not shown). To quantify the comparison, we derived the spatial average salinity anomalies for time series analysis (Fig. 4) and the spatial distributions of the salinity RMSEs between the Argo data and the other salinity data (Fig. 5). The area-averaged Argo salinity anomalies compared well with the anomalies in the ECMWF ORAS5, SODA and SMAP datasets (Fig. 4). The area-averaged correlation coefficients between Argo data and the other datasets were > 0.2 and their RMSEs were small. The RMSEs for salinity were relatively small in the Indian Ocean between the Argo data and other datasets (Fig. 5), especially in the main area of high salinity anomalies (Fig. 1, black boxes), indicating the reliability of the Argo data. The Argo and ECMWF ORAS5 datasets had the highest similarity and the smallest RMSE and we therefore used the ECMWF ORAS5 salinity data to calculate the salinity budget equation.

Fig. 4
figure 4

Time series of the salinity anomaly (unit: psu) in the Argo (blue), SMAP (red), ECMWF ORAS5 (yellow) and SODA (purple) datasets area-averaged over (70–90° E, 10° S–0°) from 2007 to 2018. The correlation coefficients and RMSEs between the time series are estimated for the periods when data are available in both datasets

Full size image
Fig. 5
figure 5

Spatial distributions of the salinity RMSEs a between the Argo and ECMWF ORAS5 data, b between the Argo and SMOS data and c between the Argo and SMAP data

Full size image

2.3 Methods

Salinity budget analysis has been widely used to identify the key processes governing the evolution of the salinity anomalies. The salinity budget equation (Gao et al. 2014; Qu et al. 2013) can be written as:

$$\frac{\partial \left[ S \right]}{{\partial t}} = - \frac{{\left( {P - E} \right)}}{h} \left[ S \right] - \nabla_{H} \left( {uS, vS} \right) - \frac{1}{h} \left( {\left[ S \right] - S_{ - h} } \right) w_{e} - \partial_{Z} \left( {wS} \right) + \varepsilon$$

where the square brackets mean the depth average within the selected depth. \(\frac{\partial [S]}{\partial t}\) is the salinity tendency, P and E are the precipitation and evaporation, respectively, h is 50 m, \(u\), \(v\) and \(w\) are the zonal, meridional and vertical velocities, respectively, and \({S}_{-h}\) is the salinity 15 m below the selected depth base (Ren et al. 2011). The subscripts H and Z denote the horizontal and vertical components of the variables, respectively. The first and second terms on the right-hand side of the equation represent the surface FWF and horizontal advection, respectively. The horizontal advection is divided into zonal advection and meridional advection. The vertical entrainment terms consist of the third and fourth terms on the right-hand side of the equation. The change in the depth of the mixed layer is not involved because we mainly consider the salinity anomaly within the upper 50 m and the analyses are mainly within this depth. We mainly analyzed the effects of the horizontal advection terms and FWF. The terms other than these two are classified as the residual term \(\varepsilon\). We nominally consider the salinity and temperature at 10 m (first level) as the salinity of the sea surface.

The residual term ε was calculated from the salinity tendency minus the sum of the FWF and horizontal advection. To separate the interannual variability from the seasonal cycle, each variable was divided into two parts: the climatological mean seasonal cycle and the anomaly variability separated from the seasonal cycle (e.g., \(u = \overline{u} + u^{\prime }\)). By neglecting the higher order nonlinear terms, the salinity horizontal advection term (Zhang et al. 2013) in the salinity equation can be rewritten as:

$$- \nabla_{H} \left( {uS, vS} \right) = - \left( {\overline{u}\frac{{\partial \left[ S \right]^{\prime } }}{\partial x} + u^{\prime } \frac{{\partial \overline{\left[ S \right]} }}{\partial x} + \overline{v}\frac{{\partial \left[ S \right]^{\prime } }}{\partial y} + v^{\prime } \frac{{\partial \overline{\left[ S \right]} }}{\partial y}} \right)$$

where the horizontal advection term contains two parts, one caused by the variability of the ocean currents (e.g., \(- u^{\prime } \frac{{\partial \overline{\left[ S \right]} }}{\partial x}\)) and the other caused by the variability of the salinity gradient (e.g., \(- \overline{u}\frac{{\partial \left[ S \right]^{\prime } }}{\partial x}\)).

3 Results

3.1 Spatial characteristics of extremely high salinity anomalies

Figure 6 shows the monthly Argo area-averaged salinity/temperature variations and their anomalies with depth from 2007 to 2018. In particular, extremely high salinity anomalies and a depth of > 50 m were clearly identified during 10A11S and 16A17S (Fig. 6b). The intensities and development trends of the salinity anomalies were different; in addition, when positive salinity anomalies occurred, significant negative temperature anomalies also appeared (Fig. 6b, contours). Such extremely high salinity anomalies during these two periods were less common than other low salinity events. The strong temporal correspondence between the life cycle of salinization events in these two periods clearly indicated the dominance of these events in the interannual variability of salinity in the central TIO. Only these two periods showed abnormally strong salinization events during the 12 years of this study.

Fig. 6
figure 6

Time–depth distribution of the area-averaged a salinity (shading; units: psu) and temperature (contours; units: °C) and b salinity anomaly (shading; units: psu) and temperature anomaly (contours; units: °C) from 2007 to 2018 based on the Argo data

Full size image

The spatial distribution of the high-value centers of the three-month moving mean extremely high salinity anomalies was different in all time periods in the main area of high salinity anomalies (Fig. 1, black boxes) during 10A11S and 16A17S (Fig. 7, shading). In 10A11S (Fig. 7a–c), the positive high salinity anomalies first appeared in the eastern Indian Ocean and gradually became stronger before decreasing again, with the highest anomalies in January. By contrast, in 16A17S (Fig. 7d–f), the positive salinity anomalies were relatively small at first and then gradually became larger, with the highest anomalies in March. The salinity anomalies during 10A11S had a more obvious tendency to move westward and southward, whereas during 16A17S the anomalies extended eastward in the later period. The intensity of the former event was greater than that of the latter from September to February, but the latter was slightly stronger in March–May.

Fig. 7
figure 7

Spatial distribution of the three-month moving mean high salinity anomaly (shading; units: psu) and SSTA (contours; units: °C) during a–c 10A11S and d–f 16A17S based on the Argo data in the main high salinity anomaly area. The figure shows the development trend of the salinity anomaly throughout the events and compares the two events. Only the anomalies exceeding the 95% confidence level based on Student’s t-test are shown

Full size image

Previous studies have shown that variations in salinity anomalies are mainly related to two factors: the surface buoyancy fluxes and the ocean dynamics. The ocean dynamic processes are mainly caused by climate modes, such as the IOD and ENSO. Figure 8 shows the time series of the ONI and DMI and the correlation coefficients between the ONI/DMI and salinity anomalies during the time period 2007–2018. The salinity anomalies in the research area were negatively correlated with the ENSO and IOD and the correlation coefficients reached a maximum when the salinity anomalies lagged the ENSO by three months and the IOD by four months.

Fig. 8
figure 8

Time series of the ONI (blue) and DMI (red) (units: °C) in 2007–2018. The blue and red dashed lines denote the threshold of ± 0.5 times the standard deviation for the ONI and DMI, respectively. The yellow curve denotes the time series of the salinity anomaly (units: psu) area-averaged over (70–90° E, 10° S–0°)

Full size image

The development of the climate events in 2010–2011 and 2016–2017 were different. 2010 was a strong La Niña year and 2016 was a weaker La Niña year, almost close to normal. Both 2010 and 2016 were strong nIOD years, but the nIOD in 2016 lasted for longer. In summary, the La Niña event in 2016 was much weaker than that in 2010, but the IOD events in 2016–2017 were a little stronger than that in 2010–2011. Because both years were nIOD and La Niña events, extremely positive salinity anomalies occurred during 10A11S and 16A17S. We will focus later on analyzing how climate models affect the variations in the salinity anomalies.

3.2 Variations in the salinity anomalies caused by surface buoyancy fluxes

There were significantly high salinity anomalies during 10A11S and 16A17S. We first analyzed the variations in the positive salinity anomalies caused by the surface buoyancy fluxes in the main area of high salinity anomalies during 10A11S and 16A17S.

Figure 9 shows the spatial distributions of the cumulative precipitation anomaly and the surface FWF for three months during 10A11S and 16A17S. There were two centers of anomalously high precipitation in autumn 2010 (Fig. 9a, contours) and spring 2017 (Fig. 9f, contours), indicating that precipitation was high in these two periods and low at other times. High precipitation mainly occurred east of 82° E. There was more precipitation in autumn 2010 than in autumn 2016 and more precipitation in spring 2017 than in spring 2011.

Fig. 9
figure 9

Spatial distribution of the cumulative precipitation anomaly (contours; units: mm day−1) and the surface FWF (i.e., evaporation minus precipitation; shading; units: mm day−1) in three months during a–c 10A11S and d–f 16A17S. The distributions of the precipitation anomaly and the FWF correspond to the distribution of salinity anomalies shown in Fig. 7. The most significant anomalies are in autumn 2010 and spring 2017 (red boxes in parts a and f)

Full size image

The abnormal amounts of evaporation during 10A11S and 16A17S were small and the difference in evaporation between the two periods was also very small (figure not shown). The net surface FWF (Fig. 9, shading) obtained by subtracting the precipitation from the evaporation had roughly the opposite distribution to the precipitation anomaly. The surface FWF in autumn 2010 was smaller than that in autumn 2016 and the FWF in spring 2017 was smaller than that in spring 2011; these two periods have more significantly negative values.

Based on our previous analyses (Fig. 7), the region was also an area of high salinity anomalies. The salinity anomalies in autumn 2010 were higher than those in autumn 2016 and those in spring 2017 were higher than those in spring 2011. The variations in the FWF during these two time periods did not correspond to the salinity anomalies. The FWF therefore does not make a significant contribution to the variations in salinity during the same time period. We also found that the precipitation in other periods had a negative anomaly, the FWF was either positive or slightly negative, and the positive salinity anomalies were significant. This suggests that the precipitation/FWF may influence the salinity anomaly at this time, although the exact degree of impact requires a quantitative analysis of the salinity budget. These analyses show that the FWF can only explain a small part of the positive salinity anomalies and that another factor, perhaps ocean dynamic processes, affects the generation of highly positive salinity anomalies.

3.3 Variations in the salinity anomalies caused by ocean dynamics

We analyzed the surface buoyancy fluxes and initially concluded that they are not the main cause of the extremely high salinity anomalies. The ocean dynamic processes that cause variations in salinity may be related to large-scale dynamic processes, such as the IOD and ENSO (Fig. 8). We therefore analyzed how these climate modes modulate the ocean dynamic processes, which, in turn, affect the salinity anomalies.

Figures 10 and 11 show the evolution of the monthly mean salinity anomalies and related processes during 10A11S and 16A17S, respectively. The zonal wind anomalies in the EIO are positive during September–November (Figs. 10a–c and 11a–c), indicating that there were abnormal westerlies at this time. The anomalies in 2010 were stronger than those in 2016. Based on our previous analysis (Fig. 8), the La Niña in 2010 was much stronger than that in 2016, whereas the intensities of the nIOD were similar in both years. A La Niña in the tropical Pacific leads to an increase in the Walker circulation and atmospheric bridges also increase the Walker circulation in the TIO. Therefore the counterclockwise circulation in the TIO strengthened in the strong La Niña year of 2010 and the westerly anomalies were stronger. During the same period in 2016, the process of changing from southwesterlies the westerlies led to weaker eastward currents in 2016 than in 2010 and to the intrusion of high salinity water from west to east in the TIO. The salinity anomalies in September–November 2010 were therefore higher under the influence of strong westerlies and eastward currents than those during the corresponding months in 2016. The contours in Figs. 10b and c show that the high salinity anomalies in 2010 reached the coast of Sumatra in the eastern Indian Ocean and were > 0.4 psu.

Fig. 10
figure 10

Spatial distribution of the monthly mean Argo salinity anomaly (blue contours; units: psu), ocean currents (vectors; units: m s−1) and zonal wind anomaly (shading; unit: m s−1) during 10A11S. The figure represents the change in the ocean dynamic processes over time during the whole event. Positive zonal wind anomalies indicate westerlies. The solid blue contour lines represent the positive salinity anomalies. Only areas with salinity anomalies > 0.2 psu are shown

Full size image
Fig. 11
figure 11

Spatial distribution of the monthly mean Argo salinity anomaly (blue contours; units: psu), ocean currents (vectors; units: m s−1) and zonal wind anomaly (shading; unit: m s−1) during 16A17S. The figure represents the change in the ocean dynamic processes over time during the whole event. Positive zonal wind anomalies indicate westerlies. The solid blue contour lines represent the positive salinity anomalies. Only areas with salinity anomalies > 0.2 psu are shown

Full size image

The central-eastern EIO zonal wind anomalies weakened during December–February of the following year (Figs. 10d–f and 11d–f). As the westerly anomalies weakened and changed to enorthwesterlies, the eastward currents near the equator also weakened and the eastward currents became westward currents. These currents were more obvious in 2011. The extremely high salinity anomalies in the eastern Indian Ocean caused by the eastward currents under the combined action of the early stage of a strong La Niña and nIOD were transported westward, leading to high salinity anomalies in the central TIO. The center of the high salinity anomalies (Fig. 10d–f, contours) shows that the trends of these anomalies gradually moved westward. The high salinity anomalies did not move significantly westward in 2017 because they were affected by the weak positive salinity anomalies in the eastern Indian Ocean caused by the weaker eastward currents under the combined action of an early-stage weak La Niña and nIOD and weaker westward currents. These phenomena were mainly caused by weakening of La Niña and nIOD events of different intensities. The salinity anomalies began to gradually decrease as a result of the decrease in the intensity of these climate events and the salinity anomalies in these 3 months of 2011 were higher than in the equivalent months in 2017.

The La Niña event was still in recession in March–May 2011. The easterly anomalies were mainly on the equator in 2011, especially from April to May, whereas there were clear westerly anomalies in 2017. There were still westward currents in 2011 (Fig. 10g–i) and the upwelling Rossby waves caused by the previous period continued to move westward, which caused the positive salinity anomalies to gradually move both westward and southward. The positive salinity anomalies decreased in the central TIO. The ENSO gradually become normal in 2017 (Fig. 11g–i) as a result of the weak La Niña in 2016. The easterlies became westerlies and were clearly stronger in April. The westward currents turned into eastward currents, which resulted in positive salinity anomalies in the central TIO. The positive salinity anomalies were therefore stronger in 2017 than in 2011.

There were significant negative salinity anomalies between these two extremely high salinity anomaly events (2012 and 2016; see Fig. 2), especially from autumn 2011 to spring 2013 (11A13S) and from autumn 2015 to summer 2016 (15A16S). Figure 8 shows that 2011 was a weak pIOD year with a weak La Niña, whereas 2012 was a strong pIOD year without either an El Niño or La Niña; 2015 was a strong pIOD year with a strong El Niño. An analysis of the salinity anomalies and related processes during 15A16S (Fig. 12) showed that they were the opposite of those in 10A11S and 16A17S (Figs. 10 and 11). Under the influence of the strong pIOD at the end of 2015, the westward currents caused by the abnormal easterlies resulted in significant negative salinity anomalies in the central-eastern Indian Ocean (Fig. 12, contours).

Fig. 12
figure 12

Spatial distribution of the monthly mean Argo salinity anomaly (blue contours; units: psu), ocean currents (vectors; units: m s−1) and zonal wind anomaly (shading; unit: m s−1) during 15A16S. The solid blue contour lines represent the negative salinity anomalies. Only areas with salinity anomalies less than − 0.2 psu are shown

Full size image

The Walker circulation weakens when an El Niño event occurs, leading to weakening of the clockwise circulation formed under the influence of the pIOD. This causes abnormal weakening of the salinity, but its impact is relatively small compared with that of the pIOD. The same phenomenon occurred during 11A13S (figure not shown). The IOD intensities for 11A13S and 15A16S were similar, but no El Niño event occurred during 11A13S and therefore the salinity anomalies caused were stronger than in 15A16S. The salinity in the central-eastern TIO therefore showed significant negative anomalies in some years, indicating a major influence of climate events on salinity anomalies. This is highly consistent with the positive salinity anomalies in 2010–2011 and 2016–2017 observed in this study.

The La Niña event that occurred during 2007 was as strong as that in 2010, but, without an nIOD event, it did not produce a significant positive salinity anomaly in 2007–2008 (Fig. 8). We also analyzed the relationships between the salinity anomalies in the central TIO from 1979 to 2018 based on the ECMWF ORAS5 data and the IOD/ENSO (figure not shown). The years 1998 and 2010 had similar La Niña intensities, but the nIOD was stronger in 2010 and the resulting salinity was abnormally strong. The intensity of the 2016 La Niña event was significantly weaker than that in 1998, but the nIOD intensity was stronger, resulting in an abnormally strong salinity in 2016–2017. A strong El Niño occurred in 1997 and 2015 and a weak El Niño occurred in 1994 and 2006. However, the negative salinity anomalies were weaker in 1997 and 2015 than in 1994 and 2006. The years 1982 with 2015 had similar pIOD intensities. A strong El Niño occurred in 1982 and a much stronger El Niño in 2015; the salinity anomalies were the opposite of the El Niño strength. The strong El Niño in 1982 caused strong negative salinity anomalies, but led to weak negative salinity anomalies in 2015.

These analyses indicate that the interannual variability of the oceanic and atmospheric conditions associated with the IOD may have a crucial role in the modulation of the variability of salinity in the central TIO. The ENSO influences the anomalous intensity of salinity. When IOD events of a similar intensity occur, ENSO events of different intensities will cause an abnormal difference in salinity. Our results show some similarities with the findings of Zhang et al. (2013), such as the primary control of the interannual variability of salinity in the TIO by IOD events rather than by ENSO events, and the view of Burns and Subrahmanyam (2016) that IOD events are the dominant influence on salinity during the peak IOD months of co-occurring event years.

3.4 Analyses of the salinity budget of two extremely high salinity events

The salinity anomalies and related processes during 10A11S and 16A17S were very different in the central TIO under the influence of different intensities of nIOD and La Niña events. To further understand the mechanisms associated with the variations in the salinity anomaly, we carried out salinity budget analyses to determine the main processes contributing to the variability of salinity. The salinity budget obtained from datasets with uncertainties from different sources contains a large residual term. We therefore only provide qualitative analyses to help us to understand the relative contributions of horizontal advection and the FWF during 10A11S and 16A17S.

Figure 13 shows the salinity variation tendency and the main processes during 10A11S and 16A17S based on the salinity budget equation. The value of each term in the salinity budget equation is different in different time periods. During 10A11S (Fig. 13a–c), the salinity first increases and then gradually decreases from December, before increasing again. At the same time, the horizontal advection (the S-adv term in Fig. 13a), which represents ocean dynamic processes, is first positive, then negative and then positive again. The S-adv term is the largest term even if the residual item is large and is therefore the main factor in the variation of salinity for the two time periods. There is a large positive advection term in September–October 2010, which leads to a more rapid increase in salinity. This advection term became negative in early 2011, indicating that it had an inhibitory effect on the increase in salinity, which gradually decreased, before increasing again as a result of a smaller positive advection. These variations were mainly caused by zonal advection, especially the abnormal changes in the zonal currents. During 16A17S (Fig. 13d and e), the changes in salinity were very small in the early stages and then became larger. The changes in the S-adv term were the same. Advection therefore had a major role in the variation in salinity, with larger values in March–May. In general, advection promoted an increase in salinity, especially zonal advection.

Fig. 13
figure 13

Time series of the averaged anomaly of different budget components (units: psu per month) during a–c 10A11S and d–f 16A17S based on the ECMWF ORAS5 salinity data in the main high salinity area. S/t S-adv and FWF represent the salinity anomaly tendency, horizontal advection term and surface FWF term, respectively. The residual term ε is the S/t term minus the S-adv and FWF terms. S-advx and S-advy represent the zonal advection and meridional advection terms, respectively. Us′, u′S, Vs′ and v′S \(- \overline{u}\frac{{\partial \left[ S \right]^{\prime } }}{\partial x}\), \(- u^{\prime}\frac{{\partial \overline{\left[ S \right]} }}{\partial x}\), \(- \overline{v}\frac{{\partial \left[ S \right]^{\prime } }}{\partial y}\) and \(- v^{\prime}\frac{{\partial \overline{{\left[ {\text{S}} \right]}} }}{\partial y}\), respectively

Full size image

The FWF was always negative, indicating a negative effect on the variation in salinity. The influence was more significant in autumn 2010 and spring 2017, with little effect on salinity in other periods. In the early stage of the event, the S-adv item was greater in 2010 than in 2016, whereas the S-adv term in spring 2017 was much stronger than that in spring 2011. The S-adv term was strongly negative in early 2011 and then became weakly positive. The S-adv term changed from small to large in early 2017. The high salinity anomalies were therefore caused by the mutual effects of the advection and FWF terms, but the contribution of the advection term was greater.

Ocean horizontal advection was therefore the main factor in the variation of the high salinity anomaly during 10A11S and 16A17S. The FWF had a small, negative effect on the positive salinity anomalies in two events. The mutual effects of horizontal advection and FWF led to extremely high salinity anomalies in the central TIO in two time periods.

4 Summary and discussion

We identified significant high salinity anomalies in the central TIO during 10A11S and 16A17S from the 2007–2018 Argo products and various satellite and reanalysis dataset. The paper gives a detailed description of the spatial characteristics of these high salinity anomalies and their underlying mechanisms of formation.

Our results show that the salinity anomalies in autumn 2010 and spring 2017 were stronger than those in autumn 2016 and spring 201 and that precipitation in the first two periods was larger than that in the latter two periods. The changes in the FWF were inconsistent with the variations in the salinity anomaly and therefore the impact of the FWF made a small, negative contribution to the high salinity anomalies in these two periods. The high salinity anomalies cannot be fully explained by the changes in the regional exchanges of air–sea freshwater.

The co-occurrence of the nIOD and La Niña events in 2010 and 2016 was the main reason for the high salinity anomalies. The strong eastward currents caused by the nIOD and La Niña events during September–December 2010 and 2016 transported high salinity water eastward, triggering positive high salinity anomalies in the eastern TIO. During the decay phases of the nIOD and La Niña in January–May 2011 and 2017, the westward currents transported the high salinity water in the eastern Indian Ocean westward, resulting in high salinity anomalies in the central Indian Ocean. As the La Niña event in 2010 was much stronger than the event in 2016, the strong Walker circulation caused by the strong La Niña event enhanced the dynamic processes in the Indian Ocean and the high salinity anomalies at the end of 2010 appeared earlier and were stronger than those in 2016. The decrease of the high salinity anomalies at the end of 2010 lasted longer than those in 2016. The salinity anomalies therefore gradually weakened in early 2011, whereas they decreased and then increased in early 2017 and the maximum salinity anomalies occurred in different months (January 2011 and March 2017, respectively). The salinity anomalies from March to May were more significant in 2017 than in 2011. The salinity anomalies at the end of 2010 were therefore stronger than those in 2016, whereas the salinity anomalies in spring 2017 were stronger than those in 2011. These results are summarized in Fig. 14.

Fig. 14
figure 14

Schematic diagrams of the key processes influencing the extremely high salinity anomalies during a 10A11S and b 16A17S. The red (blue) arrows indicate the wind (current) fields. The purple arrows indicate the atmospheric circulation and the red ‘P’ means precipitation. ‘Recession’ indicates the recession in the nIOD and La Niña events

Full size image

Our analyses verify the view that the interannual variation in salinity in the TIO was mainly affected by IOD events. When a nIOD event occurred, the positive salinity anomalies caused by the westerlies were seen at the end of the year and at the beginning of the following year. Positive salinity anomalies of different intensities occurred depending on the intensity of the La Niña event. La Niña events promoted the enhancement of positive salinity anomalies during nIOD events. By contrast, when a pIOD event occurred, there were negative salinity anomalies of different intensities depending on the intensities of the El Niño event. An El Niño event suppressed the enhancement of negative salinity anomalies in pIOD years.

Quantitative analyses of the salinity budget equation showed the advection term was positive and at a maximum at the end of 2010, then became negative and gradually weakened. The advection term was relatively small during 16A17S and strong positive advection did not appear until spring 2017. The FWF had a small, negative influence on the salinity anomalies. Ocean horizonal advection was therefore the main factor in the positive salinity anomalies in the central TIO. The positive salinity anomalies at the end of 2010 and 2016 were mainly due to positive zonal advection anomalies. By contrast, the decrease in salinity anomalies in early 2011 was mainly caused by negative zonal current anomalies. The salinity anomalies in early 2017 first decreased and then increased, similar to the results of previous ocean dynamic analyses. As a result, the mutual effects of horizontal advection and the FWF led to extremely high salinity anomalies in the central TIO during 10A11S and 16A17S.

Based on the salinity budget equation, there was one residual term ε in addition to the horizontal advection term and FWF terms. The main reason for the residual term is that the data are from different sources. There is no dynamic balance among the different sources to maintain a closed salinity budget. Various centers reconstruct the observations from different sources using different methods, leading to different errors. The item generally includes some nonlinear terms, such as the horizontal and vertical mixing disturbance, turbulent diffusion and the accumulation of errors from the other terms, but these items are relatively complex. We only present qualitative analyses here to help understand the salinity anomalies. We need to obtain higher resolution data and improve the methods of computing all the terms contributing to the salinity anomalies.

Our results highlight the important role of oceanic dynamics in the variability of salinity in the TIO. This study provides new perspectives on the interannual variability of salinity in the TIO, which has implications for the stratification of the upper ocean in the central-eastern TIO and potential impacts on air–sea interactions and climate variability in this region.