Deep-Sea Research IIndian Ocean Circulation abstract Spiciness variability in the thermocline is believed to be an important subsurface ocean process affecting sea surface temperature

Thermocline spiciness variations in the tropical Indian Ocean observedduring 2003–2014

Yuanlong Li a,b, Fan Wang b,n

a Department of Atmospheric and Oceanic Sciences, University of Colorado, Boulder, CO, USAb Key Laboratory of Ocean Circulation and Waves, Institute of Oceanology, Chinese Academy of Sciences, Qingdao, China

a r t i c l e i n f o

Article history:Received 8 August 2014Received in revised form9 December 2014Accepted 19 December 2014Available online 29 December 2014

Keywords:SpicinessIndian OceanCirculation

a b s t r a c t

Spiciness variability in the thermocline is believed to be an important subsurface ocean process affectingsea surface temperature (SST) and climate variability over the tropical oceans. Analysis of in-situobservations during 2003–2014 reveals pronounced spiciness variations at interannual-to-decadaltimescale within the thermocline of the tropical Indian Ocean (TIO). Isopycnal potential temperatureand salinity anomalies between σθ¼24.0–25.0 kg m�3 have typical magnitudes of 0.2 1C and 0.08 psu inthe southeastern Arabian Sea and the southern TIO, comparable with those observed in the Pacific basin.In the southeastern Arabian Sea, spiciness variations are dominated by a decadal trend, showing positive(warm, salty) anomalies in 2003–2006 and negative (cold, fresh) anomalies in 2009–2013. The majorcause is the mixed-layer property change in the northern Arabian Sea, which induces variation in bothspiciness and amount of water detrained down to the thermocline. In the southern TIO, largest variationsoccur at two zonal spiciness fronts where different thermocline water masses converge. These signalsare primarily produced by wind-driven geostrophic advection. Anomalies at the northern front (11–61S)exhibit westward spreading tendency and quick diffusion, which reflect mainly the signatures of the 1stbaroclinic mode Rossby waves. Anomalies at the southern front (18–131S) move westward to thewestern TIO via the advection of the South Equatorial Current (SEC). These low-frequency subsurfacespiciness variations can alter the background vertical thermal gradient in the thermocline ridge region(55–851E, 12–41S), although such impact on the SST variability is generally small.& 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND

license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Potential temperature θ and salinity S of sea water at a givenisopycnal surface are addressed as “spiciness” by oceanographers (e.g.,Veronis, 1972; Munk, 1981; Flament, 2002; Huang, 2011). Correspond-ingly, density-compensated isopycnal θ and S variability is referred toas spiciness variability (e.g., Schneider, 2000). Positive (warm, salty)and negative (cold, fresh) subsurface spiciness anomalies are usuallyproduced in the formation areas of thermocline water masses by air–sea interaction (Bindoff and McDougall, 1994), anomalous subduction(Johnson, 2006), and convective mixing (Yeager and Large, 2004).Otherwise, away from the outcropping region, anomalies can also begenerated by subsurface isopycnal advection across spiciness fronts(Schneider, 2000; Kilpatrick et al., 2011; Li et al., 2012b). Aftergeneration, spiciness variations spread passively in the ocean via thewind-driven upper-ocean circulation. The important role of large-scale, low-frequency spiciness variations in the tropical climate

modulations is drawing increasing attention. Especially, anomaliesfrom the extra-tropics are brought equatorward by the lower branchof the shallow overturning cell (e.g., McCreary and Lu, 1994; Lee andMarotzke, 1998) and affect sea surface temperature (SST) variabilitythrough the equatorial upwelling (e.g., Schneider, 2004; O’Kane et al.,2014), which is believed to be one of the mechanisms responsible forthe decadal climate variability (Latif and Barnett, 1996; Gu andPhilander, 1997).

Descriptions of subsurface thermal variations in the Pacific basinbased on historical hydrographic data have been available since the1990s (Deser et al., 1996; Zhang et al., 1998; Schneider et al., 1999; Luoand Yamagata, 2001). Decadal timescale anomalies emerge at mid-latitudes of both hemispheres and propagate to the tropical Pacificalong themean geostrophic streamlines in themain thermocline. It wasalso pointed out that these anomalies undergo intensive diffusion alongthe pathways and have been greatly weakened as reaching the tropics(Schneider et al., 1999). Due to the shortage of subsurface conductivitymeasurements, it was difficult to separate spiciness signals fromplanetary wave signals in historical data. On the other hand, theconnection between tropical and subtropical oceans through subsurfacespiciness signal transmission were confirmed by many numericalmodeling studies (e.g., Pierce et al., 2000; Nonaka and Xie, 2000;

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/dsri

Deep-Sea Research I

http://dx.doi.org/10.1016/j.dsr.2014.12.0040967-0637/& 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

n Correspondence to: Key Laboratory of Ocean Circulation and Waves, Institute ofOceanology, Chinese Academy of Sciences, 7 Nanhai Road, Qingdao 266071, China.

E-mail address: [email protected] (F. Wang).

Deep-Sea Research I 97 (2015) 52–66

Schneider, 2000, 2004; Yeager and Large, 2004; Luo et al., 2005;Nonaka and Sasaki, 2007). Modeling researches also suggested thatspiciness variations, especially those from the South Pacific, have aconsiderable impact on the decadal climate transitions (e.g., Giese et al.,2002; Yeager and Large, 2004; Luo et al., 2005; O’Kane et al., 2014).

The launch of Argo program (Roemmich et al., 2009) in the2000s provides a powerful tool to routinely monitor the tempera-ture and salinity variations below the sea surface. Analyses of Argodata have revealed robust spatial-temporal characteristics ofregional and basin-scale spiciness variability (e.g., Johnson, 2006;Yeager and Large, 2007; Sasaki et al., 2010; Kolodziejczyk andGaillard, 2012; Li et al., 2012a, 2012b; Katsura et al., 2013;Kolodziejczyk et al., 2014). Anomalies with large magnitudes areformed in the eastern Pacific basin in both hemispheres andcommunicated to the central equatorial Pacific by the interiorequatorward flow (Johnson and McPhaden, 1999). Albeit inten-sively diffused along the pathways, arrivals of these signals are stillsufficient to modify the background thermal structure of theequatorial region upon which SST evolves (e.g., Kolodziejczykand Gaillard, 2012; Li et al., 2012a). These results provide observa-tional support for the climatic importance of thermocline spicinessvariations proposed by earlier modeling studies.

While most of the existing researches were focused on thespiciness variations in the Pacific Ocean (e.g., Schneider et al.,1999; Schneider, 2000, 2004; Kessler, 1999; Yeager and Large,2004; Luo et al., 2005; Sasaki et al., 2010; Kolodziejczyk and

Gaillard, 2012, 2013) or the Atlantic Ocean (e.g., Lazar et al., 2001;Laurian et al., 2006, 2009; Kolodziejczyk et al., 2014), theircounterpart in the Indian Ocean has never been examined before.However, the important role of the tropical Indian Ocean (TIO) inglobal climate has been increasingly recognized (e.g., Annamalaiet al., 2007; Xie et al., 2009; Luo et al., 2010, 2012). Especially, thestrong SST variance in the Seychelles–Chagos thermocline ridge(SCTR) (McCreary et al., 1993; Hermes and Reason, 2008) of thesouthwest TIO has profound local and remote impacts on theweather and climate (e.g., Xie et al., 2002; Annamalai et al., 2005;Duvel and Vialard, 2007; Izumo et al., 2010a, 2010b; Vialard et al.,2009; Li et al., 2013). Spiciness anomalies can travel in thecomplicated Indian Ocean overturning circulation (Lee andMarotzke, 1998; Schott et al., 2002; Lee, 2004) and modulate,directly or indirectly, the TIO SST variability associated with theimportant climate modes, such as the Madden–Julian oscillations(Madden and Julian, 1971) and the Indian Ocean dipole (IOD) (Sajiet al., 1999). In this regard, knowledge of the thermocline spicinessvariations in the TIO could contribute to our understanding of thetropical air–sea interaction.

In this study we attempt to investigate the low-frequency (inter-annual-to-decadal timescale) spiciness variability in the thermoclineof the TIO. In-situ and satellite observational data (Section 2) areanalyzed to reveal the spatial structure (Section 3.1) and temporalvariability (Section 3.2) of the thermocline spiciness, explore thegeneration mechanism of low-frequency anomalies in important

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 20140

0.5

1

1.5

2

2.5x 10

4

Pro

file

num

ber

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 20140

0.5

1

1.5

2

2.5

T er

ror [

°C]

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 20140

0.05

0.1

0.15

0.2

S e

rror

[psu

]

Year

Fig. 1. (a) Number of monthly profile of the MOAA GPV dataset in the TIO (30–1201E, 201S–201N). (b) Mean interpolation errors of temperature between σθ¼24.0–25.0 kg m�3 of the MOAA GPV in the TIO. (c) is the same as (b) but for salinity.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–66 53

regions (Sections 3.3 and 3.4), and discuss their possible impact on theTIO SST (Section 4). Primary findings of the paper are summarized inSection 5.

2. Data and methods

The Grid Point Value of the Monthly Objective Analysis (MOAAGPV) (Hosoda et al., 2008) is a monthly ocean-state estimate productbased on data profiles from Argo floats, buoymeasurements, and castsof research cruises. For each month since January 2001, temperatureand salinity data from various sources are interpolated onto standardpressure levels between 10 and 2000 dbar using the Akima spline(Akima, 1970) and then mapped onto 11�11 fields using the two-dimensional optimal interpolation method. The final MOAA GPVproduct provides gridded monthly estimates of temperature andsalinity together with their interpolation errors. The vertical pressurelevels in the upper 500 dbar are at 10, 20, 30, 50, 75,100,125,150, 200,250, 300, 400, and 500 dbar. A detailed description of data processing,gridding, and error estimation of MOAAGPV is provided in Hosodaet al. (2008). Since available, MOAA GPV has been widely utilized byrecent researches to investigate surface and subsurface oceanicvariability on both regional and global scales (e.g., Li et al., 2012a,2012b; Katsura et al., 2013; Prakash et al., 2013; Qiu and Chen, 2013).

In the TIO (30–1201E, 201S–201N) there is a steady increase ofdata profiles for MOAA GPV during the past 15 years (Fig. 1a),primarily benefiting from the proceeding of the Argo program. Themonthly profile number has exceeded 5000, 10,000, and 15,000since 2003, 2005, and 2011, which indicates that there are onaverage more than 2.4, 4.8, and 7.1 profiles in each 11�11 grid box,respectively. The interpolation errors of temperature and salinityin the σθ¼24.0–25.0 kg m�3 layer (which will be shown to be agood representation of the TIO main thermocline) decrease rapidlywith time (Fig. 1b and c). Here potential density σθ is calculatedwith reference to the sea surface. Since 2003 the mean interpola-tion errors have been below 0.8 1C and 0.07 psu (unit based onpractical salinity scale PSS-78) and further reduced to around0.3–0.5 1C and 0.02–0.04 psu since 2006. Because the typicalmagnitudes of low-frequency θ and S anomalies are respectively0.2 1C and 0.08 psu (see Section 3.2), the signal-to-noise ratio ismuch higher for S than for θ. Here we mainly use the monthlymean S between 24.0 and 25.0 kg m�3 as the representative of thespiciness level in the TIO thermocline. Data spatial distributionand equipment precision are also factors we should consider ininterpreting the results, especially before 2005.

In this study, we mainly use the MOAA GPV data during January2003–November 2013. Data records are resampled into 0.1 kg m�3

σθ bins through linear interpolation. As an indicator of isopycnalgeostrophic stream function, acceleration potential (AP) is calcu-lated by vertically integrating specific volume anomaly δ from the2000-dbar reference level. Zonal and meridional geostrophiccurrents are further estimated with the isopycnal gradients of APthrough

ðU;VÞ ¼ 1f

�∂AP∂y

;∂AP∂x

� �; ð1Þ

where f is the Coriolis parameter. To obtain low-frequency varia-tions, climatological seasonal cycle is first removed from theoriginal data, and a 13-month low-pass Hanning filter is used toremove high-frequency signals (addressed as the “low-pass fil-tered” data hereafter). To keep the data length, the MOAA GPVdata during July–December 2002 and December 2013–May 2014are also involved in the filtering but discarded from the low-passfiltered data.

Besides MOAA GPV, other datasets used in our analysis includethe 0.251�0.251 sea surface height (SSH) product of the Archiving,

Validation, and Interpretation of Satellite Oceanographic (AVISO)(Ducet et al., 2000) and the 0.71�0.71 10 m wind data from theEuropean Centre for Medium-Range Weather Forecasts (ECMWF)Reanalysis Interim (ERA-Interim) product (Dee et al., 2011). Zonaland meridional surface wind stress, τx and τy, are calculated fromthe ERA-Interim 10 m wind speed |W10| using the standard bulkformula,

τx ¼ ρacd W10j ju10 and τy ¼ ρacd W10j jv10; ð2Þwhere ρa¼1.175 kg m�3 is the air density, cd¼0.0015 is the dragcoefficient, and u10 and v10 are the zonal and meridional 10 mwind components. The two datasets are available at respectivelyweekly and daily resolutions, but here they are averaged intomonthly resolution to facilitate our analysis.

At interannual timescale, El Niño-Southern Oscillation (ENSO) andthe IOD are the two primary climate modes influencing the TIO. Herewe adopt respectively the Niño-3.4 index from the Climate PredictionCenter (CPC) of the National Oceanic and Atmospheric Administra-tion (NOAA) and dipole mode index (DMI) from the Frontier ResearchCenter for Global Change of the Japan Agency for Marine-EarthScience and Technology (JAMSTEC) as indicators of them.

3. Results

3.1. Mean structure

Before investigating temporal variability, it is instructive to firstcast a glance at the mean spatial structure of the thermocline spicinesswhich mainly reflects the distribution of several major thermoclinewater masses. The meridional salinity section at 65.51E provides arough image of water mass distribution for the Arabian Sea and thesouthwestern Indian Ocean (Fig. 2a). The σθ¼24.0–25.0 kg m�3 layeris at a mean depth of �100 m at low latitudes and reaches the mixedlayer base (about 30–40 m) in the northern Arabian Sea (north of181N) and the subtropical South Indian Ocean (south of 251S). In thefollowing content, we will show that this layer corresponds roughly tothe depth of the main thermocline in the TIO. Two high-S waters areformed in the two outcropping regions and spread equatorward assubsurface high-S tongues. The two water masses are named respec-tively as the North Indian Water (NIW) and the South IndianSubtropical Water (STW) (Rochford, 1964; Talley and Baringer, 1997;Wijffels et al., 2002). They are separated by a body of fresh waterbetween 14 and 81S. Checking the lateral distribution between 24.0and 25.0 kg m�3 reveals that this fresh water locates at the westerntip of a low-S tongue in the southern TIO (Fig. 2b). This subsurfacefresh water tongue originates from the Maritime Continent andwedges into the prevailing high-S Indian Ocean water by the advec-tion of the South Equatorial Current (SEC) between 14 and 81S. It isnamed as the Australasian Mediterranean Water (AAMW), reflectingits origin (Mamayev, 1975; You and Tomczak, 1993). All the threewaters exist as subsurface salinity (or spiciness) maxima, except thatthe NIW and STW have relatively much higher spiciness levels (warm,salty) than the AAMW. Convergence of the three water masses formstwo zonal spiciness fronts in the central-to-eastern TIO basin (e.g.,Gordon et al., 1997; Wijffels et al., 2002). The northern front at 11–61Sseparates the NIW and AAMW, while the southern front at 18–131Smarks the boundary between the AAMW and STW.

The σθ¼24.0–25.0 kg m�3 layer is at a mean depth of 80–140mover most areas of the TIO (Fig. 2c). In the SCTR region (55–851E,12–41S), it is elevated to 70–80m due to the strong wind-drivenupwelling. The mean θ distribution of this layer (shown in Fig. 3b)closely resembles that of S. The typical θ value in the TIO is between20 and 23 1C, consistent with the widely adopted definitions of themain thermocline in the TIO, e.g., Z20 or Z23, depth of the 20 1C or23 1C isotherm (e.g., Xie et al., 2002; Hermes and Reason, 2008). We

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–6654

also checked the θ and depth maps for each season (not shown).Though with prominent seasonal fluctuations, the typical values of θand depth are always within the reasonable ranges for the mainthermocline of the tropical ocean.

3.2. General features of spiciness variability

Standard deviation (STD) of the low-pass filtered isopycnal salinitymeasures the strength of low-frequency spiciness variability (Fig. 3a).Large STDs with magnitudes up to 0.08 psu can be seen in thesoutheastern Arabian Sea and the southern TIO. Variations in theArabian Sea are adjacent to the outcropping area in the north. In thisstudy, the outcropping area refers to the area where the mixed layerpotential density falls between 24.0 and 25.0 kg m�3. These signalsare basically confined north of 51N, having little influence on theequatorial region. In contrast, the large variations in the southern TIOoccur in areas away from the outcropping area. Two separated high-STD patches stand out, with large variance spreading westward. Onelarge patch with STD¼0.07–0.08 psu locates between 11 and 61S,while the other patch with maximal STD of 0.06–0.07 psu appearsbetween 18 and 131S. These salinity signals are well above the 0.02–0.04 psu mean interpolation error and O (0.01 psu) measurementprecision of Argo floats (e.g., Riser et al., 2008). We also checked theregional interpolation errors in these high-STD areas, which are closeto the basin mean value. If described in term of isopycnal θ, the typicalSTD value of these variations is around 0.2 1C (Fig. 3b), which will bebelow the mean 0.3–0.5 1C interpolation error. This makes isopycnalsalinity a better measure of spiciness variability than isopycnal θ.Spiciness variability in the TIO has comparable magnitudes to itscounterpart in the Pacific basin (see also Kolodziejczyk and Gaillard,2012; Li et al., 2012a), albeit with relatively smaller spatial scale. Suchpronounced spiciness changes in the thermocline are quite worthy ofin-depth investigation.

It is interesting to note that the STD maxima appear at the twozonal fronts where meridional spiciness gradient is large, which

can be crucial for the local production of spiciness anomalies. Onthe other hand, the STD of AP shows large values at the centralsouthern TIO basin (Fig. 3c), suggesting strong geostrophic currentvariations in the surrounding areas. Large lateral spiciness gradi-ents together with strong current variability suggest the possiblerole of anomalous advection across the mean spiciness fronts inproducing spiciness variations (e.g., Li et al., 2012b). The mean APcontours are nearly zonal in the southern TIO, reflecting thewestward flows of the SEC, which favors the westward spreadingof the spiciness signals from the eastern to the western basin.Spiciness signals generated at the southern front are carried to thewestern basin by the SEC, and further northward along theMadagascar coast by the North Madagascar Current (e.g.,Lutjeharms et al., 1981) to the SCTR region (55–851E, 12–41S),although the signals have been rather weak (STDo0.05 psu) whenreaching the SCTR due to the along-path mixing. The northernfront, on the other hand, partly overlaps the SCTR region. Spicinessvariations produced at the northern front directly influence thesubsurface temperature of the SCTR.

To capture the spatio-temporal characteristics of these variations,yearly mean maps of salinity anomaly are displayed in Fig. 4. Spicinessin the southeastern Arabian Sea exhibits an evident decadal trend,underscored by large positive anomalies during 2003–2006 andnegative anomalies during 2008–2013. These anomalies emerge nearthe southeastern rim of the outcropping area of the northern ArabianSea, i.e., the formation region of the NIW. Mixed layer property changeand anomalous subduction in the outcropping may be important forthe generation of the thermocline spiciness anomalies.

In the southern TIO, signals generated in the central-to-easternbasin are translated westward, which is more evident at thesouthern front. For example, the negative anomaly at 60–701E in2003 moves to 40–601E in 2004 and gets diminished in 2005; apositive anomaly first emerges in 2005 in the eastern basin at thesouthern front, moves slowly to the central basin, gets strength-ened during 2006–2007, and is finally diffused after 2008 in the

34.8

34

35

35

35

35.2

35.2

35.2

35.2

35.4

35.4

35.4

35.4

35.6

35.6

35.6

35.8

35.8

36

36 36.236.4

24 24

24

2525

25

25

Dep

th [m

]

30S 25N 20S 15S 10S 5S EQ 5N 10N 15N 20N500

400

300

200

100

0

34.6 35.1 35.6 36.1 36.6

34.6

34.8

34.8

35

35

35

35.2

35.2

35.2

35.4

35.635.8

36

NIW

AAMW

STW

40E 60E 80E 100E 120E

20S

10S

EQ

10N

20N

34.5

35

35.5

36

36.5

80 80

100

100120

140

40E 60E 80E 100E 120E

20S

10S

EQ

10N

20N

40

70

100

130

160

Fig. 2. Climatologic annual-mean fields based on the MOAA GPV data during 2003–2014: (a) meridional salinity (psu) section at 65.51E, (b) mean salinity (psu) and (c) meandepth (m) of the σθ¼24.0–25.0 kg m�3 layer. The black thick curves in (a) denote the 24.0 kg m�3 and 25.0 kg m�3 isopycnal surfaces. In (b), the two straight lines mark thetwo zonal fronts, and major thermocline water masses are also marked: the Australasian Mediterranean Water (AAMW), the North Indian Water (NIW), and the South IndianSubtropical Water (STW).

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–66 55

western basin. These anomalies have basin-wide spatial scales andmagnitudes larger in the central-to-eastern basin than in thewestern basin. Interestingly, variations at the two fronts are out-of-phase in some years (e.g., 2007, 2010, and 2011) and in-phase insome other years (e.g., 2005, 2006, and 2012). In the following twosubsections, we will further explore detailed characteristics andgeneration mechanisms of the spiciness variations separately forthe Arabian Sea (Section 3.3) and the southern TIO (Section 3.4).

3.3. The Arabian Sea

Observed spiciness variability in the southeastern Arabian Sea ischaracterized by a decadal trend. Fig. 5 shows the evolutions ofisopycnal salinity anomaly together with the mixed-layer salinity(MLS) anomaly at 65.51E, 17.51N and 70.51E, 14.51N. Here the mixedlayer depth (MLD) is defined as the depth at which the σθ differencefrom the 10m value is equal to the σθ change by 0.2 1C temperaturedecrease (de Boyer Montégut et al., 2004). At 65.51E, 17.51N, the upperportion of the σθ¼24.0–25.0 kg m�3 layer is within the mixed layer insome months (Fig. 5a). Under the monsoonal wind forcing, the MLD(unfiltered) exhibits evident semiannual variability, shoaling and

diving twice a year. From late-winter (boreal) to spring, the MLD islifted from 70–80 m to 20–30 m, due to the increasing solar radiationand weakening of the winter monsoon. During late-summer and fall,the MLD elevates again from 60m to 40m, as the summer monsoonretreats. The MLD shoaling detrains mixed layer water to the subsur-face layer and also feeds the mixed layer property anomalies to thethermocline. The consistency between MLS and thermocline S is quiteclear. There is also a tendency that thermocline S lags MLS by severalmonths. It is interesting to note that S in the deeper layer (below25.0 kg m�3) shows some unique variations, such as the negativeanomalies in 2004 and the positive ones in 2009. The two bulks ofspiciness anomalies are not likely related with the changes in themixed layer. They might be brought from their outcropping area infurther north, e.g., the Red Sea, or produced locally by anomalousisopycnal advection.

Strongest spiciness variations are observed at the southeastern rimof the outcropping area. Anomalies 40.15 psu can be seen betweenσθ¼24.0–24.5 kg m�3 at 70.51E, 14.51N (Fig. 5b), where this layer hasbeen well separated from the MLD. At this station, significantdiscrepancies are clearly discernible between MLS and thermoclineS. MLS features mainly biennial (2-year period) fluctuations with a

34.434.6

34.8

34.8

34.8

34.8

35

35

35 35

35

35.2

35.2

35.235.4

35.435.4

35.6

35.6

35.635.836

30E 60E 90E 120E 150E 180 150W 120W 90W

20S

10S

EQ

10N

20N

0.04 0.05 0.06 0.07 0.08

19

20

20

2021

21

21 21

21

22

2222

23

2324

30E 60E 90E 120E 150E 180 150W 120W 90W

20S

10S

EQ

10N

20N

0.05 0.08 0.11 0.14 0.17 0.2

21 21

21

21

22

22

22

22

22

23

23

23

24

24252627

30E 60E 90E 120E 150E 180 150W 120W 90W

20S

10S

EQ

10N

20N

0.1 0.14 0.18 0.22 0.26 0.3

Fig. 3. Interannual standard deviation (STD) maps of (a) salinity (psu), (b) potential temperature (1C), and (c) acceleration potential (AP) relative to 2000 dbar (m2 s�2)averaged in σθ¼24.0–25.0 kg m�3 layer. STDs are calculated with 13-month low-pass filtered MOAA GPV data during 2003–2013. The corresponding climatologic annual-mean fields of these variables are superimposed as black contours.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–6656

positive year followed by a negative year, while thermocline S isdominated by quasi-decadal variation as shown in Fig. 4. Alternatively,thermocline S bears more resemblance with the MLS in the upstreamarea where the thermocline seasonally ventilates (Fig. 5a). It shouldalso be noted that, the 24.0 kg m�3 isopycnal is quite close to theMLD, especially in fall, when the MLD is deep, and the thermocline isshallow. Thermocline spiciness can be modified locally by the mixedlayer through convective mixing (Yeager and Large, 2004, 2007;Kolodziejczyk and Gaillard, 2013).

A more visual comparison between subsurface S and outcrop-ping MLS can be achieved by plotting their yearly time seriestogether (Fig. 6a). Here subsurface S is averaged over the south-eastern Arabian Sea (65–751E, 10–151N) where the interannual STDachieves maximum, while the outcropping MLS is averaged overgrid points with the mixed layer σθ between 24.0 and 25.0 kg m�3

in the central-to-eastern Arabian Sea (60–701E, 12–261N). We alsocalculated the MLS for only MLD “shoaling months” (such asFebruary–April and August–October), but it is found that theshoaling months differ spatially and vary from year to year, andthe result has little difference from the all-year-round MLS. Theoutcropping MLS (blue in Fig. 6a) shows similar quasi-decadalchanges as the subsurface S (black in Fig. 6a), with positiveanomalies during 2003–2004, negative anomalies during 2011–2012, and a rebound in 2013. The changes of the outcropping MLSare however determined by two processes: the mixed layer salinitychange of the outcropping area which is induced by atmosphericfreshwater forcing and ocean mixed layer dynamics; and thegeographical change of outcropping region, i.e., the sea surfacedensity variation. To roughly separate the two effects, we compute

the “area-fixed” MLS (red in Fig. 6a), which is the MLS averagedover the climatologic mean outcropping area of 24.0–25.0 kg m�3

layer. Although this area-fixed MLS evolves quite differently fromthe subsurface S and time-varying MLS, it does have positiveanomaly in 2003–2004 and negative anomaly in 2011–2012 andthus partly contributes to the decadal trend of subsurface S.However, this contribution is smaller compared with the geogra-phical change of the outcropping region.

Discrepancies between MLS (blue) and thermocline S (black) arealso discernable. The subsurface S has stronger decadal changes (seealso Fig. 5) and weaker year-to-year fluctuations than MLS. Anotherprocess affecting the subsurface S is the yearly subduction rate, whichis the amount of the mixed layer water subducted to the thermoclineeach year. If anomalously more salty mixed layer is subducted into thethermocline, the southeastern Arabian sea will show positive subsur-face S anomaly. However, the computation of the subduction raterequires the tracing of each water particle to determine whether it isthe “effectively” detrained to the thermocline (e.g., Qiu and Huang,1995), which is quite difficult to apply with the 11�11 MOAA GPVdata in a coastal region. Alternatively, we computed the integrateddetrainment W (e.g., de Szoeke, 1980),

W ¼ �ZZ

wmþu� ∇hmþ∂hm

∂t

� �dxdy; ð3Þ

where wm is the vertical velocity at the base of the mixed layer whichis estimated as the Ekman pumping velocity with ERA-Interim winddata, hm is MLD, u is the geostrophic current at hm. The three terms inthe integration represent respectively the detrainment induced byvertical advection across the mixed layer base, horizontal advection,

35

35

35

Yr = 2003

20S

10S

EQ

10N

20N

35

35

Yr = 2004

35

35

Yr = 2005

35

35

Yr = 2006

35

Yr = 2007

20S

10S

EQ

10N

20N

35

Yr = 2008

35

Yr = 2009

35

35.5

Yr = 2010

35

35

35.5

Yr = 2011

40E 60E 80E 100E

20S

10S

EQ

10N

20N

35

35

Yr = 2012

40E 60E 80E 100E

35

35

Yr = 2013

40E 60E 80E 100E−0.15

−0.1

−0.05

0

0.05

0.1

0.15

S anomaly

S

Fig. 4. Yearly maps of isopycnal salinity anomaly (color shading; in psu) and mean salinity (black contours with 0.1-psu intervals) between σθ¼24.0–25.0 kg m�3 during2003–2013. The two green dots mark the locations of the stations in Fig. 5.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–66 57

and MLD variation. W is computed by integrating the three terms inthe time-varying outcropping area. We find that W does showpronounced interannual changes (green in Fig. 6b), and it alsocontributes to the 2004–2011 decreasing trend of subsurface S,because of the positive W anomalies in 2003–2005 and the negativeones in 2009–2011. Further analysis suggests that the three terms onthe rhs of Eq. (3) have very limited year-to-year variability (notshown). In fact, the key factor determining yearly-mean W value isthe outcropping area A¼∬ dxdy (pink in Fig. 6b). The outcroppingarea is relatively large during 2003–2005 covering more than half ofthe Arabian Sea, and shrinks to the northwestern Arabian Sea during2009–2011 (Fig. 6c). Therefore, W variations are also associated withthe geographical change of the outcropping region. When it covers alarger high-MLS area, a positive subsurface S anomaly will beproduced through subduction.

We also examined the possible impact from other processes. Forexample, lateral geostrophic advection can be estimated through

ADV¼ � USxþVSy� �

; ð4Þwhere Sx and Sy are zonal and meridional S gradients. ADV can causesome of the subsurface S variations, but its contribution is mainly atbiennial period and not likely responsible for the decadal changes (notshown). Therefore, based on the clues gathered here, variation of themixed-layer properties in the northern Arabian Sea, such as spicinessand density, is likely the primary cause for the quasi-decadal varia-tions of the thermocline spiciness. The mixed-layer property varia-bility in the outcropping area is in turn largely influenced byatmospheric fluxes. Heat and freshwater fluxes determine not onlythe spiciness level of the mixed layer but also mixed-layer densitythrough buoyancy forcing. Change of mixed-layer density furtherleads to geographical variation of the outcropping region and thus

the properties of the water subducted to the thermocline. Besides,ocean mixed-layer dynamical processes also act to modulate themixed-layer property variability. A more in-depth understanding ofthe underlying mechanisms requires a careful heat and salt budgetanalysis as has been done in Kolodziejczyk and Gaillard (2013), butsuch investigation has been beyond the scope of the present study.This should be examined in the future with more observational dataor numerical models.

3.4. The southern TIO

Behaviors of spiciness anomalies in the southern TIO can be betterobserved in a time-longitude plot (Fig. 7a). At the northern front(11–61S), strong S variations (40.1 psu) with a typical timescale of2–3 years are generated between 70 and 901E and diffused quicklyas spreading westward. S signals in the western basin are somewhatsmaller in magnitude and lower in frequency. Strong spiciness gra-dients and large AP variability suggest isopycnal geostrophic advectionas a possible mechanism producing the spiciness anomalies. Between11 and 61S, large subsurface AP anomalies propagate westward rapidlyfrom the central to western basin (Fig. 7b). AP has similar timescaleand a close correlation with S. Negative AP anomalies in 2005–2006,2008, and 2010–2011 are accompanied with positive S anomalies totheir east, while positive AP anomalies in 2006–2007, 2009–2010, and2012 leave behind negative S anomalies. This can be understood byconsidering the meridional advection. Negative (positive) APs drivesouthward (northward) anomalous geostrophic flows to their east(Fig. 7c) which bring the salty NIW from the north (the fresh AAMWfrom the south) and produce positive (negative) S anomalies. These APanomalies also drive large V anomalies to their west, with oppositesigns to those at 70–901E. This gives rise to out-of-phase S variationsbetween the western and central basins. However, since the meanmeridional Sy is weaker in the western basin, produced S anomaliesare much smaller in magnitude. Spatial-temporal variations of ther-mocline AP are strikingly consistent with SSH variations (Fig. 7d).Their propagation speeds are generally between 16 and 24 cm s�1.These characteristics tally with 1st-mode baroclinic Rossby waves (e.g.,Chelton and Schlax, 1996; Masumoto and Meyers, 1998). Straight linesin Fig. 7b indicate the theoretically predicted 1st-mode baroclinicRossby wave phase speed CR¼�βc2/f2, where β¼∂f/∂y is the mer-idional gradient of f, and c is the phase speed of baroclinic gravitywave. Some AP anomalies travel slower than CR, which may be due tothe influence of high modes or modification induced by dissipationand the background flow (e.g., Qiu et al., 1997; de Szoeke and Chelton,1999). Interestingly, S anomalies produced at 70–901E also showwestward moving behaviors at a similar speed to AP and SSH. Themean current is in fact not westward at these latitudes (Fig. 7g) andhence cannot be responsible for such swift transition. One reasonableexplanation is that the westward moving tendency in S anomalyreflects primarily the Rossby waves’ signatures. As Rossby wavespropagating westward at a zonal S front, it continuously producesnew S anomalies through cross-front advection around them, whileolder anomalies are diffused quickly, making up the visual impressionthat S signals are moving westward.

One may notice that there are differences in propagation featurebetween AP and V, although V is calculated from AP. The propagationof V looks generally slower than AP. In fact, our estimation of the APpropagation speed is based on the several well-structured westwardsignal tracks. A close inspection of Fig. 7b suggests that there are alsofragmental “stationary” AP anomalies, e.g., the negative one in 2004 inthe central basin, the positive one in 2004 in the eastern basin, andthe positive one in 2008 in the central-to-eastern basin. These APanomalies are also wind-driven Rossby waves, but when they movewestward, they are quickly cancelled by the out-of-phase wind forcingto their west. This speculation is confirmed by the complicated zonalstructure of wind stress curl (WSC) in Fig. 7e. V anomaly, as the zonal

24 24

24

24

24 24 24

2424

24 24

25

Dep

th [m

]

65.5°E,17.5°N

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013200180160140120100806040200

−0.2 −0.1 0 0.1 0.2

24 24

24

24

25

25Dep

th [m

]

Year

70.5°E,14.5°N

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013200180160140120100806040200

−0.2 −0.1 0 0.1 0.2

Fig. 5. Depth-time plots of the low-pass filtered isopycnal salinity anomaly (shownbelow the MLD; in psu) and mixed layer salinity anomaly (shown above the MLD;in psu) at (a) 65.51E, 17.51N and (b) 70.51E, 14.51N in the Arabian Sea. The black thickcurve denotes the unfiltered monthly MLD, while the two thin curves denote theunfiltered 24.0 kg m�3 and 25.0 kg m�3 isopycnal surfaces. Locations of the twostations are marked as green dots in Fig. 4. Blank area below the MLD is due to theoccasional outcropping of the isopycnals, where spiciness anomaly cannot beestimated.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–6658

gradient of AP anomaly, is influenced by both westward AP tracks andstationary AP signals, and therefore shows generally smallerpropagation speed.

At the southern front (18–131S), spiciness variations are domi-nated by decadal changes (Fig. 8a). In the eastern basin (90–1101E),large positive (negative) S anomalies are produced during 2003–2007 (after 2012). There are also weaker anomalies generated inthe western basin, such as the positive ones after 2011. Westwardmovement speed of S anomalies is visibly much slower than thoseat the northern front. Tracking the positive anomaly envelopeacross the basin suggests a mean westward speed of 4–5 cm s�1,roughly consistent with the zonal-mean subsurface strength of theSEC (e.g., Hastenrath and Greischar, 1991). It is interesting that itsmovement is accelerated between 2007 and 2009 in the western

basin. This reflects the zonal distribution of the SEC (Fig. 8g). Itaccelerates from 0–5 cm s�1 in the eastern basin to 7–12 cm s�1 inthe western basin. AP signals at this latitude range (Fig. 8b) havelarger propagation speed (10–20 cm s�1) than S signals whichmove with the SEC (4–5 cm s�1). All these features suggest thatthe advection of the SEC is the primary process translating Ssignals westward at the southern front.

During 2003–2008 positive APs dominate the western-to-centralbasin. They drive northward anomalous flows in the far eastern basin(Fig. 8c), which bring the saline STW from the south and producepositive S anomalies. During 2011–2013, negative APs emerge in thecentral basin. They drive northward V anomalies to its west andsouthward V anomalies to its east, which in turn lead to positive andnegative S anomalies in the western and eastern basins, respectively.

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

−0.15−0.1

−0.050

0.050.1

0.15

S [p

su]

subsurface salinityMLSfixed−area MLS

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013−1.5

−1

−0.5

0

0.5

1

1.5

W [S

v] &

A [1

05 km

2 ]

Year

detrainment (W)outcropping area (A)

50E 55E 60E 65E 70E 75E 80E

10N

15N

20N

25N 20032004200520062007200820092010201120122013mean

35.5

35.7

35.9

36.1

36.3

36.5

36.7

Fig. 6. (a) Yearly time series of subsurface (24.0–25.0 σθ) salinity (black) averaged over the southeastern Arabian Sea (65–751E, 10–151N; green rectangle in (c)) the mixed-layer salinity MLS (blue) averaged over the time-varying outcropping area of σθ¼24.0–25.0 kg m�3 in the central-to-eastern Arabian Sea (60–701E, 12–261N; pink rectanglein (d)), and the MLS (red) averaged over the fixed (climatologic mean) outcropping area. (b) Yearly time series of the integrated detrainment W (Sv) in the time-varyingoutcropping area of σθ¼24.0–25.0 kg m�3 and the outcropping area A (105 km2). In (a) and (b), mean values of these variables have been removed before plotting. (c) Coloredcurves denote the yearly FMAS (February, March, August, and September)-mean outcropping lines of 24.0 kg m�3 (represented by the contours of 24.0 kg m�3 mixed-layerσθ) during 2003–2013 (colored curves); the black curve is the climatologic FMAS-mean outcropping line; and the gray shading represents climatologic annual-mean MLS(psu).

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–66 59

AP signals here showmore discrepancies from the corresponding SSHsignals (Fig. 8d), implying more influence of higher baroclinic modes.Also, AP and V fields are more noisy here than at the northern front,which may be due to the active ocean internal instabilities at theselatitudes (e.g., Jochum and Murtugudde, 2005). This also acts to fuzzthe relationship between AP, V, and S anomalies.

AP and SSH variations at the two fronts can be explained by WSCchanges (Figs. 7e and 8e). Positive (negative) WSCs induce Ekmandownwelling (upwelling) in the southern hemisphere, causing posi-tive (negative) SSH and subsurface AP anomalies. Besides local windforcing in the TIO, the SSH and AP variations are also influenced by theeastern boundary forcing. Especially at the southern front, the SSH andAP anomalies along the western Australia coast cannot be explainedby the overlyingWSC anomalies. A large portion of them are remotelyforced by the ENSO-related western Pacific wind variations. ENSO-induced wave signals penetrate across the Indonesian Archipelago tosouthern TIO, existing as coastal Kelvin waves along the easternboundary and radiate free Rossby waves into the interior TIO (e.g.,Potemra, 2001; Wijffels and Meyers, 2004). For example, negative(positive) SSH anomalies during 2003–2006 (2010–2012) are related

with the long-lasting warm (cold) ENSO condition during these years,as indicated by positive (negative) Niño-3.4 values in Fig. 8f. SSH andAP variations in the interior TIO are therefore influenced by both localwind forcing in the TIO and remote wind forcing from the westernPacific Ocean. We display both Niño-3.4 index and DMI in Figs. 7f and8f, but it is difficult to distinguish their impacts on AP variability. Theyare significantly correlated with both indexes, consistent with pre-vious studies (e.g., Xie et al., 2002; Tozuka et al., 2010; Li et al., 2014).Therefore, the low-frequency spiciness variations in the southern TIOare, to a large extent, generated by anomalous geostrophic advectionacross the mean spiciness fronts induced by ENSO- and IOD-relatedwind forcing.

There are also some S variations cannot be satisfactorily explainedby meridional advection. Zonal advection may also play a role inproducing some of them. However, it is difficult to isolate the relativecontributions of zonal and meridional advections to the S variations inthe anomaly generation areas (Fig. 9a and d). It can be seen that Uanomalies (Fig. 9c and f) are larger in magnitude than V anomalies(Fig. 9b and e). On the other hand, themeridional salinity gradient Sy isone order greater than the zonal salinity gradient Sx. It is noticeable

0.1

−0.1

Yea

r

40E 60E 80E 100E2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

−0.15−0.1−0.05 0 0.05 0.1 0.15

0.2

0.2

0.2

0.2

−0.4

−0.2−0.2

−0.2

−0.2

40E 60E 80E 100E

−0.4 −0.2 0 0.2 0.4

2

2

2

2

2 −2

−2

40E 60E 80E 100E

−2 −1 0 1 2

5

5

5

5

5

5

10

10

−10

−5

−5

−5

Yea

r

40E 60E 80E 100E2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

−10 −5 0 5 10

2

2

2

2

2

22

2

2

2

2

2

46

−2

−2

−2

40E 60E 80E 100E2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

−4 −2 0 2 4

−3 −2 −1 0 1 2 3

DMI & Nino−3.4

DMI

Nino−3.4

10−10

−10

−10

−10

−10

−10−10

−10−10

−10

−10−10

−10

−10

−10

Yea

r

40E 60E 80E 100E2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

−15 −10 −5 0 5 10 15

Fig. 7. Time-longitude plots of (a) salinity anomaly (psu), (b) AP anomaly (m2 s�2), and (c) meridional geostrophic current V anomaly (cm s�1) between σθ¼24.0–25.0 kg m�3, (d) AVISO SSH anomaly (cm), and (e) WSC anomaly (10�8 N m�3) based on the ERA-Interim wind data averaged between 11 and 61S. (f) Normalized dipolemode index DMI (black) and Niño-3.4 index (gray). All the variables in (a)–(f) are 13-month low-pass filtered. (g) The unfiltered zonal geostrophic current U in the σθ¼24.0–25.0 kg m�3 layer averaged between 11 and 61S. The straight lines in (b) denote the mean characteristic of the 1st-mode baroclinic Rossby waves between 11 and 61S.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–6660

that U and V anomalies are amplified at low latitudes, and low-latitude signals seem to lead the higher-latitude signals. Those featuresare compatible with long Rossby waves (e.g., Masumoto and Meyers,1998). For the same wave amplitude, if measured by AP anomalymagnitude, geostrophic current anomalies are larger at low latitudesbecause of smaller f. Also wave signals propagate faster at lowlatitudes, leading to the meridional tilt of U and V anomalies in Fig. 9.

To assess the contribution of geostrophic advection to total Svariations, we calculate and compare ADVwith the temporal tendencyof salinity St¼∂S/∂t in the two most important anomaly generationareas. Between 70 and 901E at the northern front (Fig. 10a), the low-pass filtered ADV and St have similar amplitudes (0.41/0.37 in STD)and a linear correlation of r¼0.71 (499% confidence level). The zonaland meridional components,

ADVx ¼ �USx and ADVy ¼ �VSy: ð5Þhave similar degrees of contribution to total ADV (Fig. 10b). Theiramplitudes are roughly the same, and their correlations with the totalADV are both as high as r¼0.91. These results suggest that zonal andmeridional geostrophic advection is the primary generation mechan-ism for spiciness variations at the northern front. The situation is

somewhat different at the southern front. Between 90 and 1101E, ADVhas a large-enough amplitude for causing St changes (Fig. 10c), andtheir correlation is r¼0.49 (495% confidence level). Geostrophicadvection is at least one of the primary causes for the spicinessvariations at the southern front. In this area, ADVy is more importantthan ADVx, although contributions of both are significant (Fig. 10d).The degraded ADV/St correlation at the southern front may be due tothe effect of ocean internal processes, such as the active mesoscaleeddies and turbulent mixing in the southeastern subtropical IndianOcean (e.g., Jochum and Murtugudde, 2005; Cole and Rudnick, 2012).Also, the southern front has been close the outcropping area in thesouth, local generation though vertical processes, such as spiceinjection (Yeager and Large, 2004), may also play a role in producingspiciness variations, which demands further investigation in thefuture.

4. Discussion

To some extent, assessing the potential effect of the thermoclinespiciness variability on SST is of more importance than understanding

0.10.1

−0.1

Yea

r

40E 60E 80E 100E2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

−0.15−0.1−0.05 0 0.05 0.1 0.15

0.2

0.2

0.2

−0.2

−0.2

−0.2

−0.2

40E 60E 80E 100E

−0.4 −0.2 0 0.2 0.4

40E 60E 80E 100E

−2 −1 0 1 2

5

5

5

55

5

5

−5

−5−5

Yea

r

40E 60E 80E 100E2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

−10 −5 0 5 10

22

22

2

2

2

2

2

2

2

2

22

2

2

4

4

68

−6−4

−2

−2

−2

−2−2

40E 60E 80E 100E2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

−4 −2 0 2 4

−3 −2 −1 0 1 2 3

DMI & Nino−3.4

DMI

Nino−3.4

−1−1

0

−10

−10

−10

−10

−10

−10

−10

−10

Yea

r40E 60E 80E 100E

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

−15 −10 −5 0 5 10 15

Fig. 8. Time-longitude plots of (a) salinity anomaly (psu), (b) AP anomaly (m2 s�2), and (c) meridional geostrophic current V anomaly (cm s�1) between σθ¼24.0–25.0 kg m�3, (d) AVISO SSH anomaly (cm), and (e) WSC anomaly (10�8 N m�3) based on the ERA-Interim wind data averaged between 181 and 131S. (f) Normalized dipolemode index DMI (black) and Niño-3.4 index (gray). All the variables in (a)–(f) are 13-month low-pass filtered. (g) The unfiltered zonal geostrophic current U in the σθ¼24.0–25.0 kg m�3 layer averaged between 181 and 131S.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–66 61

its generation mechanism. In this section, we attempt to examine sucheffect in the SCTR region which is important in the TIO air–seainteraction (e.g., Xie et al., 2002; Annamalai et al., 2005; Vialard et al.,2009). Fig. 11a and b shows the temporal evolutions of the verticalthermal structure for the western part (SCTR-W; 55–701E) and easternpart (SCTR-E; 70–851E) of the SCTR. With pronounced seasonal andinterannual variations, MLD fluctuates between 20 and 50m, whilethe 24.0 kg m�3 isopycnal is between 50 and 90m. Because MLD andthe thermocline show generally in-phase variation due to the natureof wave adjustment (e.g., Li et al., 2014), there is always a verticaldistance of 25–40m between them.

Fig. 11c shows the time-longitude plots of the subsurface potentialtemperature θSUB (24.0–25.0 kg m�3). Similar to the salinity variationsin Fig. 8a, large θSUB variations of 0.2–0.4 1C are generated at 70–901E,while those in the western basin are rather weak. As a result, theSCTR-E region is more influenced by subsurface spiciness variationsthan the SCTR-W region. On the other hand, MLT changes (Fig. 11d), asa good representative of SST according to the comparison with theHadley Centre Sea Ice and Sea Surface Temperature data set (HadISST)(Rayner et al., 2003), are visibly stronger than θSUB. Close inspectionreveals that MLT changes in the SCTR are associated with thethermocline depth variations (represented by the mean depth of the24.0–25.0 kg m�3 layer), with warming (cooling) anomalies accom-panied with deeper (shallower) than normal thermocline. This is inline with the notion that in the SCTR region low-frequency MLT (orSST) variability is tightly coupled with thermocline dynamics (e.g., Xieet al., 2002). Spiciness variations affect MLT variability mainly byaltering the vertical temperature difference between the mixed layer

and thermocline, Δθ. Li et al. (2014) showed that the amplitude ofintraseasonal SST variability in the SCTR-E region is sensitive to theslow changes of the background temperature stratification. Here, wedefine Δθ simply as the difference between MLT and θSUB, Δθ¼MLT–θSUB. In the SCTR-W region (Fig. 11f), Δθ (STD¼0.17 1C) is predomi-nately determined by MLT (STD¼0.17 1C; r¼0.85), while the θSUBimpact is relatively smaller (STD¼0.08 1C; r¼�0.28).

In the SCTR-E region (Fig. 11g), the effect of θSUB (STD¼0.11 1C) onΔθ (STD¼0.27 1C) is much more significant (r¼�0.75). θSUB has anegative correlation with Δθ and MLT. It means that subsurfacespiciness provides a possible negative feedback for the low-frequencyMLT variability in some years. Warm (cold) MLT tends to encounternegative (positive) subsurface spiciness which acts to damp the MLTperturbation through vertical entrainment and mixing. This effect isdifferent from the well-known thermocline feedback (e.g., Xie et al.,1989, 2002; Neelin et al., 1998) which acts to amplify SST perturba-tions. The effect of θSUB on Δθ is evident only during the 2006–2007and 2010–2011 events. Interestingly, there are strong positive events ofboth ENSO and IOD during 2006–2007 and negative events during2010–2011 (Fig. 11e), which are the only two events with concurrentin-phase ENSO and IOD conditions during the observation period. Inother events, such as the 2012 one, even thoughMLTanomaly achieveslarge magnitudes, the thermocline spiciness has rather weak changes.This can be understood by considering the distinct generationmechanisms of MLT and θSUB anomalies. The MLT variability in theSCTR region is directly produced by the upwelling/downwelling Rossbywaves. The MLT anomalies in Fig. 11d closely follow the thermoclinedepth changes (e.g., Xie et al., 2002). On the other hand, large

Latit

ude

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

18S

14S

10S

6S

Latit

ude

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

18S

14S

10S

6S

Latit

ude

Year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

18S

14S

10S

6S

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

−0.15−0.1−0.0500.050.10.15

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

−1.5−1−0.500.511.5

Year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

−6−4−20246

Fig. 9. Latitude-time plots of (a) salinity (black contours; in psu) and salinity anomaly (shading; in psu), (b) V anomaly (cm s�1), and (c) U anomaly (cm s�1) between 24.0and 25.0σθ. All the maps are plotted with low-passed variables averaged between 70 and 901E. (d)–(f) are the same as (a)–(b) but for variables averaged between 90 and1101E.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–6662

thermocline spiciness anomalies are mainly produced by anomalousmeridional geostrophic current (recall Fig. 7) which is driven by thezonal gradient of thermocline depth (or AP). In Fig. 11c, it is clear thatthe large θSUB anomalies in 2006–2007 and 2010–2011 emerge to theeast of maximal thermocline depth changes. As a result, in Fig. 11g,there is a several months’ phase lag between positive (negative)MLT and negative (positive) θSUB anomalies. Rechecking Fig. 7 suggeststhat during 2006–2007 event, AP/SSH anomaly is positive in thecentral basin due to the positive IOD, while they are negative near theeastern boundary due to the concurrent El Niño (upwelling wavesignals penetrated to the southern TIO through the IndonesianArchipelago). The large zonal gradient between them leads to largenorthward V anomaly and thus negative spiciness anomaly. During2010–2011, the situation is absolutely reversed, and large positivespiciness anomaly is produced. In other years, similar condition is not

sufficiently satisfied. For example, in 2012, although the positive IODinduces positive AP/SSH anomaly in the central basin, the warm ENSOevent is too weak to induce negative AP/SSH anomaly near the easternboundary. Therefore, V anomaly and spiciness anomaly are relativelyweaker.

In spite of these interesting relationships, we have to state that theimpact of θSUB on MLT does not seem to be significant. First, becauseof the 25–40 m vertical distance, it is difficult to directly entrain waterfrom the 24.0–25.0 kg m�3 layer to the mixed layer. Second, MLTvariations are much stronger than θSUB and dominate Δθ changes(r¼0.93 in the SCTR-E). The relationship proposed here is based onrather preliminary discussion. It functions only in the SCTR region. Inother regions, such as the southern front, the thermocline is muchdeeper (Fig. 2c), leaving the mixed layer even more invulnerable tothermocline spiciness variability.

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013−1.5

−1

−0.5

0

0.5

1

1.5

[10−8

psu

s−1

] r = 0.71 St (STD = 0.37)

ADV (STD = 0.41)

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013−1.5

−1

−0.5

0

0.5

1

1.5

[10−8

o C s

−1]

70°−90°E, 11°−6°S

70°−90°E, 11°−6°S

r (ADV, ADVx ) = 0.91

r (ADV, ADVy ) = 0.91

ADVx (STD = 0.22)

ADVy (STD = 0.22)

ADV (STD = 0.41)

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013−1.5

−1

−0.5

0

0.5

1

1.5

[10−8

psu

s−1

] r = 0.48 St (STD = 0.15)ADV (STD = 0.24)

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013−1.5

−1

−0.5

0

0.5

1

1.5

[10−8

o C s

−1]

Year

90°−110°E, 18°−13°S

90°−110°E, 18°−13°S

r (ADV, ADVx ) = 0.60

r (ADV, ADVy ) = 0.84

ADVx (STD = 0.13)

ADVy (STD = 0.19)

ADV (STD = 0.24)

Fig. 10. (a) Low-pass filtered salinity tendency St and geostrophic advection term ADV between σθ¼24.0–25.0 kg m�3 averaged over 70–901E, 11–61S. (b) Low-pass filteredADV and its zonal and meridional components, ADVx and ADVy, between σθ¼24.0–25.0 kg m�3 averaged over 70–901E, 11–61S. (c) and (d) are analogs to (a) and (b) butaveraged over 90–1101E, 18–131S.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–66 63

5. Summary

Subsurface spiciness variations are believed to be potentiallyimportant in modulating the tropical SST and climate variability. Inthis study we examine the low-frequency (interannual-to-decadal)spiciness variability in the main thermocline (σθ¼24.0–25.0 kg m�3)of the TIO by analyzing data from the MOAA GPV product and satellite

measurements during 2003–2014. Spatial and temporal characteris-tics, generation mechanism, and possible impact on SST of thethermocline spiciness variations are described. Primary findings areas follows.

(1) According to our analysis of the MOAA GPV data, thermoclinespiciness variations in the TIO have been documented by Argo

Dep

th [m

]

Year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013140130120110100

908070605040302010

Dep

th [m

]

Year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

140130120110100908070605040302010

θ [°

C]

14

16

18

20

22

24

26

28

30

50E 60E 70E 80E 90E 100E2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

50E 60E 70E 80E 90E 100E

−0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5

−3 −2 −1 0 1 2 32003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

Yea

r

DMI & Nino−3.4

DMI

Nino−3.4

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013−1.5

−1

−0.5

0

0.5

1

1.5

Year

θ [°

C]

r (Δθ, MLT) = 0.85r (Δθ, θSUB) = −0.28

Δθ (STD = 0.17°C)

MLT (STD = 0.17°C)θSUB (STD = 0.09°C)

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013−1.5

−1

−0.5

0

0.5

1

1.5

Year

θ [°

C]

r (Δθ, MLT) = 0.93r (Δθ, θSUB) = −0.75

Δθ (STD = 0.27°C

MLT (STD = 0.20°CθSUB (STD = 0.11°C

Fig. 11. Depth-time plots of the unfiltered θ averaged in (a) the SCTR-W area (55–701E, 12–41S) and (b) the SCTR-E area (70–851E, 12–41S). The green curve denotes the MLD,while the black curves denote the 24.0 kg m�3 and 25.0 kg m�3 isopycnals. Time-longitude plots of (c) low-pass filtered subsurface potential temperature θSUB anomaly(σθ¼24.0–25.0 kg m�3) and (d) low-pass filtered mixed layer potential temperature MLT averaged over the latitude range of the SCTR, 12–41S. The low-pass filtered depth ofthis layer is superimposed as black contours with 5 m intervals. (e) Normalized low-pass filtered dipole mode index DMI (blue) and Niño-3.4 index (red). Low-pass filteredanomaly time series of MLT (blue), θSUB (red), and their difference Δθ (black) in (f) the SCTR-W and (g) the SCTR-E. The three thick dashed lines in (c) and (d) define thelongitude ranges of the SCTR-W and SCTR-E.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–6664

and other in-situ measurements. Low-frequency isopycnal θand S variability with STD values of 0.2 1C and 0.08 psu aredetected in the southeastern Arabian Sea and the southernTIO, which are comparable with those observed in the Pacificbasin. Salinity signals are well above the 0.02–0.04 psu inter-polation error and O(0.01 psu) measurement precision.

(2) In the southeastern Arabian Sea, spiciness variations aredominated by a decadal trend, showing positive anomalies in2003–2006 and negative anomalies in 2008–2013. Thesevariations are confined north of the equator. Large anomalies(40.15 psu) emerge in the southeastern Arabian Sea, close tothe outcropping area of the σθ¼24.0–25.0 kg m�3 layer.Further analysis suggests that the variation of the mixed layerproperty in the northern Arabian Sea is the primary cause forthe subsurface spiciness decadal trend. It leads to changes innot only the spiciness but also the amount of the mixed layerdetrained down to the thermocline.

(3) Spiciness in the southern TIO exhibits maximal variance at twozonal fronts formed by the convergence of three thermoclinewater masses: the salty NIW, the fresh AAMW, and the salty STW.At the northern front (6–111S), large anomalies (40.1 psu) at 2–3year periods emerge between 70 and 901E, showing a seeminglywestward moving tendency and quick diffusion, which reflectmainly the signatures of the 1st baroclinic mode Rossby waves. Atthe southern front (18–131S), variations are generally weaker anddominated by quasi-decadal signals. They move westward at4–5 cm s�1 across the TIO basin via the advection of the SEC.

(4) Further analysis reveals that ENSO- and IOD-related wind-driven geostrophic advection is a major cause for the thermo-cline spiciness anomalies in the southern TIO, especially at thenorthern front. Both zonal and meridional advections havesignificant contribution.

(5) Thermocline spiciness variations in the SCTR can alter thevertical temperature different Δθ between the mixed layer andthermocline, although its impact on SST variability does notseem to be significant.

Upon ending the description of this study, we need to note thatresults presented in this study are based on analysis of available in-situobservations. Limited by the quantity and quality of the observations,some estimation reported here are rather rough. This paper howeverprovides the first description for the pronounced thermocline spici-ness variations in the Indian Ocean basin, while those in the Pacificbasin have been reported by many studies. Conclusions and specula-tions made here should be examined with further accumulated dataor numerical models. Reasonably designed parallel model experi-ments isolating effects of different processes will be useful for under-standing the underlying physics, which is our underway work.

Acknowledgements

This research is supported by the National Basic Research(973) Program of China through Grant 2012CB417401, the StrategicPriority Research Program of the Chinese Academy of Sciences(XDA11010201), the NSFC Innovative Group Grant (41421005), andthe NSFC-Shandong Joint Fund for Marine Science ResearchCenters (U1406401). Y. Li is also supported by the NASA OceanSalinity Science Team Grant NNX14AI82G and NSF CAREER Award0847605.

MOAA GPV data are provided by Dr. Shigeki Hosoda throughthe FTP site ftp://ftp2.jamstec.go.jp/pub/argo/MOAA_GPV/; AVISOsea level product is downloaded from http://www.aviso.oceanobs.com/; ERA-Interim surface wind data are available at http://www.ecmwf.int/; Niño-3.4 index is adopted from CPC of NOAA throughhttp://www.cpc.ncep.noaa.gov; DMI is taken from the Frontier

Research Center for Global Change of JAMSTEC through http://www.jamstec.go.jp/frsgc/research/d1/iod/; data analysis andgraphing work in this study are finished with a licensed Matlabprogram.

References

Akima, H., 1970. A new method for interpolation and smooth curve fitting based onlocal procedures. J. Assoc. Comput. Mech. 17, 589–602.

Annamalai, H., Liu, P., Xie, S.-P., 2005. Southwest Indian Ocean SST variability: itslocal effect and remote influence on Asian Monsoons. J. Clim. 18 (20),4150–4167.

Annamalai, H., Okajima, H., Watanabe, M., 2007. Possible impact of the IndianOcean SST on the Northern Hemisphere circulation during El Niño. J. Clim. 20(13), 3164–3189.

Bindoff, N.L., McDougall, T.J., 1994. Diagnosing climate change and ocean ventila-tion using hydrographic data. J. Phys. Oceanogr. 24, 1137–1152.

Chelton, D.B., Schlax, M.G., 1996. Global observations of oceanic Rossby waves.Science 272 (5259), 234–238.

Cole, S.T., Rudnick, D.L., 2012. The spatial distribution and annual cycle of upperocean thermohaline structure. J. Geophys. Res. Oceans 117 (C2), C02027. http://dx.doi.org/10.1029/2011jc007033.

de Boyer Montégut, C., Madec, G., Fischer, A.S., Lazar, A., Iudicone, D., 2004. Mixedlayer depth over the global ocean: an examination of profile data and a profile-based climatology. J. Geophys. Res. 109 (C12), C12003. http://dx.doi.org/10.1029/2004JC002378.

de Szoeke, R.A., 1980. On the effects of horizontal variability of wind stress on thedynamics of the ocean mixed layer. J. Phys. Oceanogr. 10, 1439–1454.

de Szoeke, R.A., Chelton, D.B., 1999. The modification of long planetary waves byhomogeneous potential vorticity layers. J. Phys. Oceanogr. 29 (3), 500–511.

Dee, D., Uppala, S., Simmons, A., Berrisford, P., Poli, P., Kobayashi, S., Andrae, U.,Balmaseda, M., Balsamo, G., Bauer, P., 2011. The ERA‐Interim reanalysis:configuration and performance of the data assimilation system. Q. J. R.Meteorolog. Soc. 137 (656), 553–597.

Deser, C., Alexander, M.A., Timlin, M.S., 1996. Upper-ocean thermal variations in theNorth Pacific during 1970–1991. J. Clim. 9, 1840–1855.

Ducet, N., Le Traon, P., Reverdin, G., 2000. Global high-resolution mapping of oceancirculation from TOPEX/Poseidon and ERS-1 and -2. J. Geophys. Res. 105 (C8),19477–19498.

Duvel, J.P., Vialard, J., 2007. Indo-Pacific sea surface temperature perturbationsassociated with intraseasonal oscillations of tropical convection. J. Clim. 20 (13),3056–3082.

Flament, P., 2002. A state variable for characterizing water masses and theirdiffusive stability: spiciness. Prog. Oceanogr. 54 (1-4), 493–501.

Giese, B.S., Urizar, S.C., Fuckar, N.S., 2002. Southern Hemisphere origins of the 1976climate shift. Geophys. Res. Lett. 29 (2), 1014. http://dx.doi.org/1010.1029/2001GL013268.

Gordon, A.L., Ma, S., Olson, D.B., Hacker, P., Ffield, A., Talley, L.D., Wilson, D., Baringer, M.,1997. Advection and diffusion of Indonesian throughflow water within the IndianOcean south equatorial current. Geophys. Res. Lett. 24 (21), 2573–2576.

Gu, D., Philander, S.G.H., 1997. Interdecadal climate fluctuations that depend onexchange between the tropics and extratropics. Science 275, 805–807.

Hastenrath, S., Greischar, L., 1991. The monsoonal current regimes of the tropicalIndian Ocean: observed surface flow fields and their geostrophic and wind-driven components. J. Geophys. Res. Oceans 96 (C7), 12619–12633.

Hermes, J., Reason, C., 2008. Annual cycle of the South Indian Ocean (Seychelles–Chagos) thermocline ridge in a regional ocean model. J. Geophys. Res. 113,C04035. http://dx.doi.org/10.1029/2007JC004363.

Hosoda, S., Ohira, T., Nakamura, T., 2008. A monthly mean dataset of global oceanictemperature and salinity derived from Argo float observations. JAMSTEC Rep.Res. Dev. 8, 47–59.

Huang, R.X., 2011. Defining the spicity. J. Mar. Res. 69, 545–559.Izumo, T., Masson, S., Vialard, J., de Boyer Montegut, C., Behera, S.K., Madec, G.,

Takahashi, K., Yamagata, T., 2010a. Low and high frequency Madden–Julianoscillations in austral summer: interannual variations. Clim. Dyn. 35 (4),669–683.

Izumo, T., Vialard, J., Lengaigne, M., de Boyer Montegut, C., Behera, S.K., Luo, J.-J.,Cravatte, S., Masson, S., Yamagata, T., 2010b. Influence of the state of the IndianOcean Dipole on the following year’s El Nino. Nat. Geosci. 3 (3), 168–172.

Jochum, M., Murtugudde, R., 2005. Internal variability of Indian Ocean SST. J. Clim.18 (18), 3726–3738.

Johnson, G.C., 2006. Generation and initial evolution of a mode water θ–S anomaly.J. Phys. Oceanogr. 36, 739–751.

Johnson, G.C., McPhaden, M.J., 1999. Interior pycnocline flow from the subtropicalto the equatorial Pacific Ocean. J. Phys. Oceanogr. 29 (12), 3073–3089.

Katsura, S., Oka, E., Qiu, B., Schneider, N., 2013. Formation and subduction of NorthPacific tropical water and their interannual variability. J. Phys. Oceanogr. 43,2400–2415.

Kilpatrick, T., Schneider, N., Di Lorenzo, E., 2011. Generation of low-frequencyspiciness variability in the thermocline. J. Phys. Oceanogr. 41, 365–377.

Kessler, W.S., 1999. Interannual variability of the subsurface high salinity tonguesouth of the equator at 1651E. J. Phys. Oceanogr. 29 (8), 2038–2049.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–66 65

Kolodziejczyk, N., Gaillard, F., 2012. Observation of spiciness interannual variabilityin the Pacific pycnocline. J. Geophys. Res. 117, C12018. http://dx.doi.org/10.1029/2012JC008365.

Kolodziejczyk, N., Gaillard, F., 2013. Variability of the heat and salt budget in thesubtropical Southeastern Pacific mixed layer between 2004 and 2010: spiceinjection mechanism. J. Phys. Oceanogr. 43 (9), 1880–1898.

Kolodziejczyk, N., Reverdin, G., Gaillard, F., Lazar, A., 2014. Low-frequency thermo-haline variability in the Subtropical South Atlantic pycnocline during 2002–2013. Geophys. Res. Lett. 41 (18), 6468–6475. http://dx.doi.org/10.1002/2014gl061160.

Latif, M., Barnett, T.P., 1996. Decadal climate variability over the North Pacific andNorth America: dynamics and predictability. J. Clim. 9, 2407–2423.

Laurian, A., Lazar, A., Reverdin, G., 2009. Generation mechanism of spicinessanomalies: an OGCM analysis in the North Atlantic subtropical gyre. J. Phys.Oceanogr. 39 (4), 1003–1018.

Laurian, A., Lazar, A., Reverdin, G., Rodgers, K., Terray, P., 2006. Poleward propaga-tion of spiciness anomalies in the North Atlantic Ocean. Geophys. Res. Lett. 33,L13603. http://dx.doi.org/10.1029/2006GL026155.

Lazar, A., Murtugudde, R., Busalacchi, A.J., 2001. A model study of temperatureanomaly propagation from the subtropics to tropics within the South Atlanticthermocline. Geophys. Res. Lett. 28 (7), 1271–1274.

Lee, T., 2004. Decadal weakening of the shallow overturning circulation in theSouth Indian Ocean. Geophys. Res. Lett. 31, L18305. http://dx.doi.org/10.1029/2004GL020884.

Lee, T., Marotzke, J., 1998. Seasonal cycles of meridional overturning and heattransport of the Indian Ocean. J. Phys. Oceanogr. 28, 923–943.

Li, Y., Han, W., Shinoda, T., Wang, C., Lien, R.-C., Moum, J.N., Wang, J.-W., 2013.Effects of the diurnal cycle in solar radiation on the tropical Indian Ocean mixedlayer variability during wintertime Madden-Julian Oscillations. J. Geophys. Res.Oceans 118, 4945-4964. http://dx.doi.org/4910.1002/jgrc.20395.

Li, Y., Han, W., Shinoda, T., Wang, C., Ravichandran, M., Wang, J.-W., 2014. Revisitingthe Wintertime intraseasonal SST variability in the tropical South Indian Ocean:impact of the ocean interannual variation. J. Phys. Oceanogr. 44 (7), 1886–1907.

Li, Y., Wang, F., Sun, Y., 2012a. Low-frequency spiciness variations in the tropicalPacific Ocean observed during 2003–2012. Geophys. Res. Lett. 39, L23601. http://dx.doi.org/10.1029/2012GL053971.

Li, Y., Wang, F., Zhai, F., 2012b. Interannual variations of subsurface spiciness in thePhilippine Sea: observations and mechanism. J. Phys. Oceanogr. 42, 1022–1038.

Luo, J.-J., Sasaki, W., Masumoto, Y., 2012. Indian Ocean warming modulates Pacificclimate change. PNAS 109 (46), 18701–18706.

Luo, J.-J., Yamagata, T., 2001. Long-term El Niño-Southern Oscillation (ENSO)-likevariation with special emphasis on the South Pacific. J. Geophys. Res. Oceans106 (C10), 22211–22227.

Luo, J.-J., Zhang, R., Behera, S.K., Masumoto, Y., Jin, F.-F., Lukas, R., Yamagata, T., 2010.Interaction between El Niño and extreme Indian Ocean dipole. J. Clim. 23 (3),726–742.

Luo, Y., Rothstein, L.M., Zhang, R.H., Busalacchi, A.J., 2005. On the connectionbetween south Pacific subtropical spiciness anomalies and decadal equatorialvariability in an ocean general circulation model. J. Geophys. Res. 110, C10002.http://dx.doi.org/10.1029/2004JC002655.

Lutjeharms, J.R.E., Bang, N.D., Duncan, C.P., 1981. Characteristics of the current eastand south of Madagascar. Deep Sea Res. 28, 879–901.

Madden, R.A., Julian, P.R., 1971. Detection of a 40–50 day oscillation in the zonalwind in the tropical Pacific. J. Atmos. Sci. 28 (5), 702–708.

Mamayev, O., 1975. Temperature–salinity Analysis of World Ocean Waters. ElsevierElsevier Oceanography Series, Amsterdam p. 374.

Masumoto, Y., Meyers, G., 1998. Forced Rossby waves in the southern tropicalIndian Ocean. J. Geophys. Res. 103 (C12), 27589–27602.

McCreary, J.P., Kundu, P.K., Molinari, R.L., 1993. A numerical investigation ofdynamics, thermodynamics and mixed-layer processes in the Indian Ocean.Prog. Oceanogr. 31 (3), 181–244.

McCreary, J.P., Lu, P., 1994. Interaction between the subtropical and equatorialocean circulations: the subtropical cell. J. Phys. Oceanogr. 24 (2), 466–497.

Munk, W., 1981. Internal waves and small-scale processes. In: Warren, B.A.,Wunsch, C. (Eds.), Evolution of Physical Oceanography. MIT Press, pp. 264–291.

Neelin, J.D., Battisti, D.S., Hirst, A.C., Jin, F.F., Wakata, Y., Yamagata, T., Zebiak, S.E.,1998. ENSO theory. J. Geophys. Res. 103 (C7), 14261–14290.

Nonaka, M., Xie, S.P., 2000. Propagation of North Pacific interdecadal subsurfacetemperature anomalies in an ocean GCM. Geophys. Res. Lett 27 (22),3747–3750.

Nonaka, M., Sasaki, H., 2007. Formation mechanism for isopycnal temperature–salinity anomalies propagating from the eastern south pacific to the equatorialregion. J. Clim. 20 (7), 1305–1315.

O’Kane, T.J., Matear, R.J., Chamberlain, M.A., Oke, P.R., 2014. ENSO regimes and thelate 1970’s climate shift: the role of synoptic weather and South Pacific Oceanspiciness. J. Comput. Phys. 271, 19–38.

Pierce, D.W., Barnett, T.P., Latif, M., 2000. Connections between the Pacific Oceantropics and midlatitudes on decadal timescales. J. Clim. 13 (6), 1173–1194.

Potemra, J.T., 2001. Contribution of equatorial Pacific winds to southern tropicalIndian Ocean Rossby waves. J. Geophys. Res. 106 (C2), 2407–2422.

Prakash, S., Gairola, R.M., Thapliyal, P.K., 2013. Sea surface salinity estimation in theBay of Bengal using multisatellite measurements. IEEE Geosci. Remote Sens.Lett. 10 (3), 525–527.

Qiu, B., Chen, S., 2013. Concurrent decadal mesoscale eddy modulations in thewestern North Pacific subtropical gyre. J. Phys. Oceanogr. 43, 344–358.

Qiu, B., Huang, R.X., 1995. Ventilation of the North Atlantic and North Pacific:subduction versus obduction. J. Phys. Oceanogr. 25, 2374–2390.

Qiu, B., Miao, W., Müller, P., 1997. Propagation and decay of forced and freebaroclinic Rossby waves in off-equatorial oceans. J. Phys. Oceanogr. 27 (11),2405–2417.

Rayner, N.A., Parker, D.E., Horton, E.B., Folland, C.K., Alexander, L.V., Rowell, D.P.,Kent, E.C., Kaplan, A., 2003. Global analyses of sea surface temperature, sea ice,and night marine air temperature since the late nineteenth century. J. Geophys.Res. 108 (D14), 4407. http://dx.doi.org/10.1029/2002jd002670.

Riser, S., Ren, L., Wong, A., 2008. Salinity in Argo: a modern view of a changingocean. Oceanography 21 (1), 56–67.

Rochford, D., 1964. Hydrology of the Indian Ocean. III. Water masses of the upper500 m of the South-east Indian Ocean. Aust. J. Mar. Freshwater Res 15 (1),25–55.

Roemmich, D.H., et al., 2009. The Argo program: observing the global ocean withprofiling floats. Oceanography 22, 34–43.

Saji, N., Goswami, B., Vinayachandran, P., Yamagata, T., 1999. A dipole mode in thetropical Indian Ocean. Nature 401 (6751), 360–363.

Sasaki, Y.N., Schneider, N., Maximenko, N., Lebedev, K., 2010. Observationalevidence for propagation of decadal spiciness anomalies in the North Pacific.Geophys. Res. Lett 37, L07708. http://dx.doi.org/10.1029/2010GL042716.

Schneider, N., 2000. Decadal spiciness mode in the tropics. Geophys. Res. Lett. 27,257–260. http://dx.doi.org/10.1029/1999GL002348.

Schneider, N., 2004. The response of tropical climate to the equatorial emergence ofspiciness anomalies. J. Clim. 17 (5), 1083–1095.

Schneider, N., Miller, A.J., Alexander, M.A., Deser, C., 1999. Subduction of decadalNorth Pacific temperature anomalies: observations and dynamics. J. Phys.Oceanogr. 29 (5), 1056–1070.

Schott, F.A., Dengler, M., Schoenefeldt, R., 2002. The shallow overturning circulationof the Indian Ocean. Prog. Oceanogr. 53 (1), 57–103.

Talley, L.D., Baringer, M.O., 1997. Preliminary results from WOCE hydrographicsections at 801E and 321S in the central Indian Ocean. Geophys. Res. Lett. 24(22), 2789–2792.

Tozuka, T., Yokoi, T., Yamagata, T., 2010. A modeling study of interannual variationsof the Seychelles Dome. J. Geophys. Res. 115, C04005. http://dx.doi.org/10.1029/2009JC005547.

Veronis, G., 1972. On properties of seawater defined by temperature, salinity andpressure. J. Mar. Res. 30, 227–255.

Vialard, J., Duvel, J., McPhaden, M., Bouruet-Aubertot, P., Ward, B., Key, E., Bourras,D., Weller, R., Minnett, P., Weill, A., 2009. Cirene: air–sea interactions in theSeychelles–Chagos thermocline ridge region. Bull. Am. Meteorol. Soc. 90,45–61.

Wijffels, S., Meyers, G., 2004. An intersection of oceanic waveguides: variability inthe Indonesian throughflow region. J. Phys. Oceanogr. 34 (5), 1232–1253.

Wijffels, S., Sprintall, J., Fieux, M., Bray, N., 2002. The JADE and WOCE I10/IR6throughflow sections in the southeast Indian Ocean. Part 1: Water massdistribution and variability. Deep Sea Res. Part II 49 (7), 1341–1362.

Xie, S.-P., Annamalai, H., Schott, F.A., McCreary, J.P., 2002. Structure and mechan-isms of south Indian Ocean climate variability. J. Clim. 15 (8), 864–878.

Xie, S.-P., Hu, K., Hafner, J., Tokinaga, H., Du, Y., Huang, G., Sampe, T., 2009. IndianOcean capacitor effect on Indo-western Pacific climate during the summerfollowing El Nino. J. Clim. 22 (3), 730–747.

Xie, S.-P., Kubokawa, A., Hanawa, K., 1989. Oscillations with two feedback processesin a coupled ocean-atmosphere model. J. Clim. 2 (9), 946–964.

Yeager, S.G., Large, W.G., 2004. Late-winter generation of spiciness on subductedisopycnals. J. Phys. Oceanogr. 34, 1528–1547.

Yeager, S.G., Large, W.G., 2007. Observational evidence of winter spice injection.J. Phys. Oceanogr. 37, 2895–2919.

You, Y., Tomczak, M., 1993. Thermocline circulation and ventilation in the IndianOcean derived from water mass analysis. Deep Sea Res. Part I 40, 13–56.

Zhang, R.-H., Rothstein, L.M., Busalacchi, A.J., 1998. Origin of upper-ocean warmingand El Niño change on decadal scales in the tropical Pacific Ocean. Nature 391(6670), 879–883.

Y. Li, F. Wang / Deep-Sea Research I 97 (2015) 52–6666