Indian Ocean Dipole – Relation of Deep Meridional Overturning Circulation in the Indian Ocean

For this paper the 42-year German Estimating and circulation and Climate of the Ocean (GECCO) synthesis data to analyse and examine the relation of the deep meridional overturning circulation (DMOC) with the mode of the Indian Ocean Dipole (IOD). Decomposing the DMOC into the Ekman and geostrophic transport, the external mode and a residual term are used to assess the contrition of various dynamical processes. The DMOC with a marginal residual term is successfully described by the first 3 terms. The following conclusions are obtained:

1.     The seasonal cycle of the DMOC is determined mainly by the Ekman component. During the transitional seasons (March-April and September-October) in the northern Indian Ocean Basin, where the geostrophic component dominates.

2.     The Ekman component dominates the structure of the DMOC at the beginning of the IOD phase (May-June); at and following the peak phase of the IOD (September-December, the structure of the DMOC is determined primarily by the geostrophic component in correspondence with the sea surface temperature anomalies which are well developed, while a secondary role is played by the wind, and therefore the Ekman component, south of 10^oS and contributes negatively within the zonal band of 10^o on both sides of the equator. Therefore, there is surface to a deep ocean connection through which wind and ocean temperature anomalies that are related to the IOD are transferred down to the deep ocean. Even in the deep ocean signals that propagate westwards are observed, which suggests possible roles for the Rossby waves in the transferring of the surface signal to the deep ocean.

Introduction

A deep meridional inflow is one of the characteristic features at the southern Indian Ocean where North Atlantic Deep Water, Antarctic Bottom Water, and Circumpolar Deep Water enter the Indian Ocean in the western basin off Madagascar, as well as off the coast of East Africa and in the east, along the ninety East Ridge (Mantyla & Reed, 1995; Schott & McCreary, 2001). The Indian Ocean deep meridional overturning circulation (DMOC) and its temporal and spatial variability remain open to be investigated in a manner that is more systematic.

Based on individual transoceanic hydrographic sections, either estimating the geostrophic transports (Toole & Warren, 1993; Robinson and Toole, 1997) or by an inverse box modelling approach (Macdonald, 1998; Ganachaud et al., 2000; Bryden & Beal, 2001; Sloyan & Rintoul, 2001; Ganachaud, 2003; Lumpkin & Speer, 2007; McDonagh et al., 2008), the time-mean structure of the DMOC was obtained earlier. The results of these earlier studies are often contaminated by strong seasonal variability in the Indian Ocean, as both earlier approaches depend heavily of hydrographic data sections, which are often sparse in the spatial and temporal resolution. In comparison, Numerical model efforts as an alternative way show weak, though noticeable, time-mean bottom inflow [2-5 Sv (1 Sv ≡ 10⁶ m³ s^-1); Garternicht & Schott, 1997; Lee & Marotzke, 1998; Stammer et al., 2012; Drijfhout & Garabato, 2008: Wang et al., 2012). The estimation of 17 Sv from Ferron & Marotzke (2003) is the only exception, which has been shown to not be maintainable when an adjoint method modifies the initial conditions and forcing fields that are applied for several hundred years (Palmer et al., 2007) instead of for a few years as was the case of Ferron & Marotzke (2003). It has also been acknowledged, on the other hand, that although the results of numerical modelling are dynamically consistent they are conditioned to prescribe weak vertical mixing, different surface heat and momentum fluxes products, model resolution, boundary conditions that are open, and deficits with respect to the formation of bottom water (Schott & McCreary, 2001; Drijfhout & Garabato, 2008; Rintoul et al., 2010; Wang et al., 2010).

Large seasonal variations over most latitudes are displayed by the DMOC and the DMOC is characterised by 2 counterrotating meridional overturning cells (Lee & Marotzke, 1998; Wang et al., 2012): In the boreal winter one cell is anticlockwise (facing the paper (or monitor) unless specified otherwise), located in the region couth of 20^oS, while the other is clockwise, and is located in the interior of the Indian Ocean centred on 10^oS. Responding to the seasonal reversal of the monsoon both cells reverse their rotation in the boreal summer. Almost the entire water column is covered year-round by the overturning streamfunction, which indicates the deep-reaching effect of the Indian monsoon (Schott et al., 2002).

Pronounced interannual variability is exhibited by the Indian Ocean DMOC. It has been found that there is a significant 2-year variation in the DMOC (Wang et al., 2012) as well as other oceanic variables that are essential, such as meridional ocean transport of heat in the Equatorial Indian Ocean (Chirokova & Webster, 2006) which is reminiscent of the tropospheric biennial oscillation (TBO; Meehl & Arblaster, 2002; Loschnigg et al., 2003). There is high correlation between the DMOC and the most prominent interannual mode, the Indian Ocean Dipole (IOD) (Wang et al., 2012), the positive phase of which is characterised by negative SST anomalies in the tropical southeastern Indian Ocean and positive anomalies in the tropical western Indian Ocean.

The characteristic dipole SST pattern of IOD events first occurs in May-June, peaking in September-October; its demise is likely to be the reversal of the monsoon that weakens winds along the equator and along the coast of Indonesia, which diminishes the importance of oceanic dynamics in the regulation of SST (Saji et al., 1999). It is suggested the acknowledgement that the IOD is parented by an anomalous monsoon through the generation of zonal temperature gradients between upwelling regions, it was suggested by Webster et al. (1999, 2002) that the east-west SST gradient that were initiated by the anomalous monsoon intensity can lead to the enhancements of the zonal gradients of the SST by modulating local upwelling and meridional transport of oceanic heat through involving equatorial wave dynamics. In this respect, they suggest that the IOD is an integral part of the TBO, as it could regulate the intensity of the monsoon system, introduces slow dynamics into the SST anomalies that are initiated by the monsoon, and help to penetrate these anomalies to the next year.

 In this study Wang et al., show that the Indian Ocean DMOC is linked to conditions of the ocean surface mainly through wind (thus the Ekman component) on seasonal scales and through anomalies of wind and sea temperature on interannual time scales that are related to the IOD, by decomposing the DMOC into 3 dynamical components and assessing their relative contributions. They extended the traditional view of the air-sea coupled system in the Indian Ocean, which is confined to only the upper Indian Ocean and the atmosphere overlying it so far (Webster et al., 2002; Feng & Meyers, 2003; Loschnigg et al., 2003; Annamalai &Murtugudde, 2004; Pillai, 2008). The organisation of this paper is:

·        section2 described data and methodology;

·        section 3 presents the roles of various dynamical components of the DMOC during IOD events;

·        section 4 is a discussion;

·        section 5 provides a summary and conclusions.

DMOC variations during IOD events

A high-pass Butterworth filter with a cutoff frequency of 7 years was applied to the meridional streamfunction, SST and the wind stress field at each model grid for the period 1960-2001, in order to function on interannual variations. Of these variables, components that were longer than 7 years which have been removed through the high-pass filtering, are presented in fig.3, and the remainder of the analysis (except for fig. 8) revers to the variables that are high-pass filtered. Resulting from this, there 2 dominant interannual periodicities and well as addition to seasonal cycles, are retained in their analysis: 2-3 years which correspond to TBO and 3-7 years which correspond to ENSO (Pillai & Mohankumar, 2009). Note, that prior to the composite analysis, the high-pass filtered variables are further subject to deseasonalisation; i.e., the monthly climatological states are removed from the respective variable.

Climatological states: DMOC and surface wind

Wang et al. first presented the annual cycle of the Indian Ocean deep meridional streamfunctions that have been high-pass filtered.

In the boreal winter (from November to February) and summer (from May-August), meridional overturning cells are most obvious double counterrotating, while in transitional seasons (March-April and September-October) the northern cell splits into 2 parts. A clockwise closed deep cell dominates the interior of the ocean (north of 20^oS) in the boreal winter and an anticlockwise deep cell locates in the southern Indian Ocean (south of 20^oS), both of which reverse their rotations in the boreal summer, responding to the reversal of the summer monsoon. The entire water column is covered by this pair of deep cells with maximum strength of 10-12 Su at a depth of 1,000 metres (located between 10^oS and 5^oN) and is separated noticeably at around 20^oS throughout the year.

In the boreal winter (November-February) positive wind stress curl occurs north of 10^oS and to its south negative wind stress curl occurs with the maximum centred along 20^oS. In the Boreal summer the wind stress pattern resembles that in winter months, though with opposite signs. According to Ekman dynamics, positive (negative) maximum curl corresponds to maximum downwelling (upwelling in the Southern Hemisphere, ha has been manifested by confluences of upwards (downwards) branches of both cells near 20^oS.

The low-frequency wind component (≥7 years ), which is primarily characterised by southeasterlies in the southern Indian Ocean and over the circumpolar ocean, strong westerlies, persists throughout the year and Wang et al suggest it may relate to the southern annular mode (Thompson & Wallace, 2000). The Ekman drift that results leads to convergence and downwelling between 60^oS and 20^oS where the maximum easterly component is located, and between the equator and 20^oS divergence and upwelling, which accounts for an upwelling (positive) cell within the southern Indian Ocean domain with maximum amplitude of about 2 Sv in the deep ocean.

Climatological states: DMOC and its dynamical components

The main structure of the DMOC is determined by the Ekman components and account for a large part of the amplitude throughout all months. In transitional seasons (March-April and September-October) there are exceptions in the northern Indian Ocean Basin, where the Ekman contribution is offset by the geostrophic component which dominates the local structure of the DMOC. There are 2 main features that are shown by the external component:

1)    South of 20^oS there is a coherent pastern throughout the year, which extends to 5,000 m, and is in phase with the Ekman contributions for most of the year.

2)    In summer (May-October) and winter months (November-April) its spatial features are rather similar though with opposite signs.

Because of the zonal topographic gradients, changes in horizontal ocean gyre, or in cross-basin zonal gradients of the meridional velocity, which is depth independent, in relation to the summer/winter monsoon shift, both features reflect only.

In summary, the structure of the seasonal DMOC can be reconstructed satisfactorily from its 3 dynamical components: Ekman flow and its barotropic compensation, the thermal wind component, and it external mode which is depth independent. In the following, relative contribution s of these 3 dynamical components to DMOC are assessed each month and are subject to a composite analysis in order to obtain a quantitative estimate of their roles in relation to the IOD events.

Positive IOD composites

Composite analysis for strong positive IOD events is used to study possible imprints on the Indian Ocean DMOC of IOD events. First, these selected events are subject to deseasonalisation: i.e., climatologically monthly DMOC is removed prior to the composite analysis.

The strength of the IOD is measured by differences in zonal sea surface temperature (SST) between the tropical western Indian Ocean (50^oE-70^oE), (10^oS-10^oN) and the tropical southeastern Indian Ocean (90^oE-110^oE, 10^oS-0^o), following the same definition as is given in Saji et al., 1999). As is shown in fig. 5 in this paper, the time series of the IOD index along with that of the Indian Ocean DMOC index is represented by the volume of transport of the bottom flow (ƴ¹ ≥ 28.11 kg/m³) across 34^oS. A 7-month running average is used to smooth both time series. There were 90 positive IOD that were obtained by Wang et al. (1965, 1967, 1972, 1976, 1977, 1982, 1987, 1994, and 1997), with the threshold of 1.5 (multiplied by the standard deviation). The DMOC index is out of phase with the IOD index, in most cases, whereas the years 1065 and 1987 are exceptional. The years 1965 and 1987 were excluded in order to highlight the commonality of the majority cases, thereby keeping 7 positive IOD cases in the following composite calculations.

Off the coast of Java cold SST anomalies first appear in May, accompanied by stronger than normal alongshore southeasterlies and easterly anomalies along the equator between 80^oE and 100^oE. 2-fold effects of these wind anomalies have been revealed by previous studies: stronger coastal upwelling off Java and Sumatra is led to by alongshore southeasterly wind, which is stronger than normal, while along the equator the easterly anomalies hamper the eastwards intrusion of the equatorial current, thereby reducing the supply of heat to the east. Cooling off Indonesia is enhanced by both effects, where another prerequisite for this cooling is provided the shoaling thermocline (Xie et al., 2002). This process may relate to coastal advection as well as equatorial and Kelvin wave dynamics (e.g., Feng & Meyers, 2003). E.g., (Xie et al., 2002) that a westwards downwelling Rossby wave was forced by the wind stress curl that was associated with the IOD leads to deeper thermocline and therefore and the warmer SST in the west, which is intensified further on the surface as a result of reduced wind speed and evaporation that is induced by anomalously extended northwestwards–extended southeasterly trade winds (Saji et al., 1999). Therefore, changes of the southeast trade winds on the equatorial region coupled with SST anomalies through ocean dynamics give rise to SST dipole events. The westwards propagation of cold SST anomalies along 10^oS from September to December is a partial manifestation of the major role played by equatorial wave dynamics in the regulation of the dipole SST dynamics. The IOD-related atmospheric circulation, on the other hand, is essentially a reflection of the zonal pressure difference in the equatorial Indian Ocean; it is thereby it is related closely to changes in equatorial wind systems, such as the Walker circulation (Saji & Yamagata, 2003).

Northwesterly wind anomalies appear in the central southern Indian Ocean in September-October, which indicates a weakening of the southeasterly trade winds. They form, together with the easterly wind anomalies along the equator, an anticlockwise wind stress curl that generates anomalous Ekman pumping within the zonal band from the equator to 20⁰S. In November-December this positive wind stress curl extends further westwards along the equator, and occupies almost the entire southern tropical Indian Ocean.

The DMOC composite is characterised in May-June by a dipole demarcated at about 10^oS; In July-August the associated anomalies almost vanish, exhibiting no distinguishable spatial patterns. An anomalous negative cell occurs between 20^oS and the equator in September-October, which expands into November-December into the entire southern Indian Ocean Basin accompanied by a positive cell in the upper 2,000 m in the northern Indian Ocean.

The DMOC features that are IOD-related shown above are, to a large extent, described by the reconstruction following Eq. 1 (in Figs. 7a,e,f) in this paper by Wang et al., 2014). At different phases of the IOD each dynamical component of the DMOC plays varying roles. In May-June, e.g., the DMOC anomalies are mainly contributed by the Ekman components which are characterised by a pair of cells that are reversely rotating separated at 10^oS, though in later months it is geostrophic current that determines the spatial feature of the DMOC anomalies; in particular, stratified DMOC structures in November-December in the Southern Ocean basin is attributed mainly to the geostrophic component.

Ekman-related DMOC anomalies are offset by wind contribution on both sides of the equator within 10^o of the equator by thermal wind contribution, though at south of 10⁰S, the Ekman thermal wind components share the same sign; therefore both contribute to the strong DMOC anomalies in the southern Indian Ocean. The DMOC anomaly that is related to the external mode is confined primarily to the deep southern Indian Ocean below 1,000 m and its highest loading in September-December is near 10^oS.

The IOD signals that are the most pronounced are manifested as wind and SST anomalies. These 2 variables link directly to the Ekman and thermal wind components of the DMOC, respectively, through which the IOD signals penetrate to the deep ocean. The SST anomalies are still weak at the starting phase of the IOD events, May-June, and so are the anomalous wind flows. The associated anomalous wind stress curl on the other side exhibits cross-basin coherent patterns: anomalous bands of positive wind stress curl anomalies along approximately 10^oN and 20^oS, which results in the local DMOC extrema. The wind stress curl anomaly loses its cross-basin coherent feature in July-August; correspondingly, the DMOC contribution that is Ekman-related reduces considerably. Meanwhile, near the Java coast the cold SST anomaly spreads and enhances further until December, which is consistent with the strong contribution from the thermal wind component of the DMOC.

Accompanying strengthening and the propagation to the west of the cold SST anomalies in the southern tropical Indian Ocean from September to December, the anomalous equatorial easterly and northwesterly winds in the southern Indian Ocean strengthen and, moreover, resume a cross-basin coherent feature with the zero (anomalous) wind stress curl line taking up a northwest-southeast orientation between 10^oS and 20^oS. This corresponds with the negative Ekman component of the DMOC that is centred along around 15^oS.

Discussion

In this paper Wang et al. adopted the method of Lee & Marotzke (1998) and Hirschi & Marotzke (2007) and applied it to the DMOC in the Indian Ocean on seasonal and IOD-related timescales. Wang et al. also analysed transitional months when the Asian monsoon reverses its direction (March-April and (September-October), and the DMOC features a 2-cell structure to the north of 20^oS in contrast to a coherent structure during monsoon-prevailing months as a supplement to Lee & Marotzke (1998) that discussed the DMOC feature in the monsoon-prevailing months (January and July). The previous prevailing monsoon weakens in the previous months and the subsequent monsoon regime is not yet fully established; a result from this is that the Ekman component weakens, which is seen more clearly to the north of the equator, whereas a coherent structure is obtained by the geostrophic component across the entire Indian Ocean Basin. The latter partly offsets the former and in March-April even supersedes it to the north of 10^oS and north of the equator in September-October, which contributes to the 2-cell feature north of 20^oS in the transitional months.

At around 20^oS there is a clear demarcation line which separates 2 circulation cells that are rotating reversely throughout the year. This reflects the effects of the monsoonal winds and intertropical convergence zone which is associated with it Southern Ocean. In January this demarcation line is less pronounced and in July, is almost indistinguishable (Lee & Marotzke, 1998). According top Wang et al. the difference may mainly from the difference of wind product: GECCO datasets used NCEP wind stresses and Lee & Marotzke (1998) used surface fluxes from Hellerman & Rosenstein (1983). It may be partially due to the strong constraints in the sponge layer in connection to the Antarctic Circumpolar Current (ACC) (Lee & Marotzke, 1998). Interannual variability of the ACC, as well as its interaction with the internal Indian Ocean Basin which, according to Wang et al. may be nontrivial in shaping the circulation of the feature in the southern Indian Ocean, are inhibited by these strong constraints. It is demonstrated by the composite analysis for the positive IOD events that the link between the Indian Ocean DMOC and surface conditions of the ocean: namely, the characteristically IOD-related and SST anomalies are linked directly to the structure of the DMOC on Interannual timescales.

Wang et al. extended by this means the view of the air-sea coupled system in the Indian Ocean which is merely confined to the upper Indian Ocean and the overlying atmosphere so far (Webster et al., 2002; Feng & Myers, 2003; Loschnigg et al., 2003; Annamalai & Murtugudde, 2004; Pillai, 2008), to the deep ocean. As a result of such a connection between the surface and the deep ocean events on the surface of the ocean in the form of anomalies of wind and SST anomalies are linked dynamically to the variability of the DMOC; it is therefore not surprising that DMOC is observed exhibits concentrated energy on timescales that are characteristic of the surface air-sea coupled system, such as on the typical time scales of TBO an d ENSO (Wang et al., 2012). The strong contribution of the Ekman component to the variability of the Indian Ocean DMOC that is revealed through the year supports the suggestion that wind forcing on a large scale has a significant role in sustaining overturning in the deep ocean by way of mixing processes (Guan & Huang, 2008; Huang, 1999; Nycander et al., 2007; Huussen et al., 2012).

Through which mechanism do temperature anomalies related to the IOD influence the deep ocean down to 5,000 m, is a question that naturally arises, particularly from September to December. Temperature and Meridional velocity fields are clearly observed to propagate to the west, even at a depth of 3,000 m, as is shown in a snapshot of temperature anomalies at 2,000 m at 20^oS for the period 1971-1973. When approaching the equator this propagation to the west is more rapid. Wang et al., conjectured, therefore, that the surface signal that is related to the IOD may be carried down in the form of the annual Rossby waves that are driven by the wind which force the deep ocean by vertical by movements of the thermocline because of convergence and divergence in the upper ocean which is wind-driven (Kessler & McCreary, 1993; Johnson, 2011). A strong presence of annual wind-driven Rossby waves has, in fact, been observed in hydrographic profiles of the Arabian Sea (Brandt et al., 2002) and in the equatorial Pacific (Kessler & McCreary, 1993) and is reported to be coupled with the overlying atmosphere on annual and Interannual scales (White, 2000a, 2001). It is most likely that these wind-driven Rossby waves are the messenger responsible for communicating the surface variations down to the deep ocean. According to Wang et al. a detailed analysis related to this aspect is planned for future studies.

Wang et al. say they would like to stress that the choice of positive events might vary, depending of the dataset and the definition used (e.g. Saji et al., 1999; Saji & Yamagata, 2003; Du et al., 2013). The analysis that was presented in their paper was based on the GECCO ocean reanalysis dataset. Wang et al. missed the IOD events in 1961 and 1963, compared to previous studies (e.g. Saji et al., 1999; Saji & Yamagata, 2003; Du et al., 2013). They conjecture that the mistargeting in 1961 and 1963 may relate to the slow adjustment to the initial conditions in the assimilation process in the GECCO data. They attempted to minimise this effect by disregarding the first 8 simulation years (1952-1958), which may not be enough, which can be seen by the abrupt increase in the DMOC index from 1960 to 1063.

Also, a similar analysis has been performed for negative IOD cases. 6 negative events are selected based on the threshold of -1.5: 1964, 1970, 1973, 1975, 1996 and 1998. It is shown by a closer look at the annual course of the IOD index that positive and negative IOD events are not symmetric around zero. Though all of the positive IOD events that were selected mature in autumn, 3 out of the 6 IOD events (1964, 1970 and 1975) reach the minimum between December and March of the following year and in 1973 and 1998 the minimum IOD index appears near September. Many studies have already noticed this symmetry between the opposite phases of the IOD (Ummenhofer et al., 2013). The wide spread of the development course of negative IOD events makes the interpretation of the time-based component less meaningful, though the DMOX index shows an out-of-phase relation with the IOD index during most of the negative IOD events. The results presented here may, therefore, pertain only to positive phases.

Summary and conclusions

Wang et al. have examined the variations of the Indian Ocean deep meridional overturning circulation (DMOC) in relation to IOD events by use of GECCO synthesis. In order to assess the relative importance of various dynamical processes in causing the seasonal cycle of the DMOC and in imprinting influences of the IOD down into the deep ocean, the DMOC is decomposed into 4 components (following Hirschi & Marotzke, 2007):

1)    The Ekman flow, a barotropic compensation,

2)    Vertical sheared flows comprising the thermal wind,

3)    the external mode that accounts for zonally varying topography, and

4)    a residual term that includes a geostrophic flow that is associated with friction and nonlinear effects.

On seasonal to Interannual time scales the residual term remains relatively small, and was therefore not discussed in detail in this study.

The overall seasonal structure of the DMOC is determined by the Ekman component and also contributes to a substantial part of the amplitude throughout all months. In the transitional seasons (March to April and September to October), there are exceptions, when over the entire Indian Ocean Basin the thermal wind component has the same sign, with an amplitude contribution that is comparable to the Ekman component. The thermal wind component offsets the Ekman contribution and determines the local DMOC structure, in the northern Indian Ocean Basin. Moderate seasonal variations at latitudes where the Somali Current and the corresponding gyre reverse their directions on a seasonal basis are caused by the external mode (Lee & Marotzke, 1998). South of 20^oS it shares the same sign with the corresponding Ekman component, which suggests its close link to the seasonally reversed wind pattern on both sides of 20^oS, and therefore also to the related Ekman drift.

It is revealed by the composite analysis of the 3 dynamical components of the DMOC that each component plays a different role during different phases of the IOD. The DMOC structure related to the IOD is characterised in May-June by a dipole pattern that is separated at about 10^oS, which is contributed mainly by the Ekman component, whereas the contribution of the geostrophic component increases considerably and determines the main feature of the DMOC. These changes in the relative contribution of the Ekman and thermal wind components of the DMOC relate closely to the surface features at different phases of the IOD: namely the SST anomalies related to the IOD and the associated wind anomalies.

Sources & Further reading

Wang, W., et al. (2014). "Deep Meridional Overturning Circulation in the Indian Ocean and Its Relation to Indian Ocean Dipole." Journal of Climate 27(12): 4508-4520.