1 Introduction

The Indian Ocean Dipole (IOD) is one of the dominant coupled ocean–atmosphere modes in the tropical Indian Ocean (TIO) that is manifested in a zonal dipole pattern of sea surface temperature anomalies (SSTAs; Saji et al. 1999). The positive IOD (pIOD) is characterized by warm SSTAs in the western TIO (WIO) and surface cooling in the southeastern TIO (Saji et al. 1999; Webster et al. 1999). The occurrence of IOD exerts significant impacts on surrounding areas and leads to worldwide climate teleconnections via atmospheric bridges and oceanic pathways (Li and Mu 2001; Saji and Yamagata 2003). Therefore, it is necessary to comprehensively investigate the mechanisms responsible for IOD events.

IOD exhibits diverse temporal-spatial characteristics. Canonical IOD events are known to develop mostly in boreal summer and mature in the following autumn, presenting obvious seasonal phase-locking features (Saji et al. 1999; Webster et al. 1999). However, another type of IOD referred to as the unseasonable IOD, possesses an earlier onset and a shorter duration, mostly developing and maturing within the boreal summer (Du et al. 2013). Previous studies have suggested that both internal variability within the Indian Ocean and external forcing from other basins contribute to the diversity of IOD (Rao and Behera 2005; Terray et al. 2007; Hong et al. 2008c; Sun et al. 2015; Yang et al. 2015; Chen et al. 2016, 2021; Stuecker et al. 2017; Jiang et al. 2022; Jiang and Liu 2022; Ling et al. 2022; Zhang et al. 2022a). This implies that the evolution of IOD may be affected by multiple air–sea coupling processes, complicating our understanding and prediction of diverse IOD events.

ENSO has been identified as a major external factor influencing IOD (Annamalai et al. 2003; Fischer et al. 2005; Cai et al. 2011b; Roxy et al. 2011; Guo et al. 2015; Fan et al. 2017; Liu et al. 2017; Stuecker et al. 2017). About 40% of IOD events occur simultaneously with ENSO (Hong et al. 2008c). During an El Niño event, the weakened Walker circulation results in easterly wind anomalies along the equatorial Indian Ocean (Ashok et al. 2003; Behera et al. 2006). These easterly winds can initiate positive Bjerknes feedback, energizing pIOD events (Wang and Wang 2014; Zhang et al. 2015; Liu et al. 2017). The upper-ocean thermal structure in the Indian Ocean can also be influenced by ENSO through Rossby waves (Xie et al. 2009, 2016) and Indonesian Throughflow (ITF; Annamalai et al. 2003; Song et al. 2007; Tozuka et al. 2007), creating favorable conditions for IOD occurrence. Therefore, ENSO can significantly affect IOD by regulating both atmospheric circulation and oceanic pathways. However, not all IOD events are associated with ENSO, and vice versa (Ashok et al. 2003; Fischer et al. 2005; Behera et al. 2006; Wang et al. 2016; Lu and Ren 2020; Zhang et al. 2020, 2022b; Jiang and Liu 2022; Song and Ren 2022). Some IOD events may be triggered by local air-sea interactions within the Indian Ocean, independent of ENSO (Rao and Behera 2005; Behera et al. 2006; Wang et al. 2016). The crucial role of internal variability in triggering IOD has also been highlighted by Luo et al. (2008), as well as the importance of the subtropical Indian Ocean Dipole (SIOD; Behera and Yamagata 2001) acting as a precursor for IOD occurring several months later (Terray et al. 2007). The subtropical high linked to the SIOD can promote easterly wind anomalies along the eastern TIO, initiating IOD even in the absence of ENSO (Feng et al. 2014; Zhang et al. 2018, 2020; Huang et al. 2021).

Previous studies have suggested that cold SSTAs in the eastern TIO (EIO) are key factors in the development of pIOD events (Webster et al. 1999; Luo et al. 2010; Yang et al. 2015). The intensity, onset time, and spatial distribution of easterly wind anomalies, which are related to the formation of the eastern pole of IOD events, have been extensively investigated (Du et al. 2013; Sun et al. 2014; Guo et al. 2015; Endo and Tozuka 2016; Zhang et al. 2019; Cai et al. 2021; Jiang et al. 2022). However, the SST distribution in the WIO also differs during the evolution of IOD and plays an important role in the diversity of IOD events (Schott et al. 2009; Du et al. 2020, 2023; Zhang et al. 2020). Therefore, understanding the dynamic processes responsible for the evolution of the western pole is also crucial for improving predictions of different types of IOD events. Additionally, the WIO is considered an important moisture source for the Indian summer monsoon, which strongly influences the IOD (Ashok et al. 2001). Different SST distributions in the WIO can induce distinct monsoonal rainfall patterns, which locally modify air-sea interactions in the Indian Ocean (Drbohlav et al. 2007; Hong et al. 2008a, b; Effy et al. 2020). Thus, further investigation of the atmospheric-oceanic dynamics responsible for the western pole of IOD can deepen our understanding of the diversity of IOD events.

Given the distinct temporal-spatial characteristics of autumn IOD and summer IOD (respectively referred to as the canonical IOD and the unseasonable IOD in previous studies), this study aims to further investigate the different ocean–atmosphere coupling processes responsible for the evolution of the eastern and western poles of these two types of IOD events. The remaining part of this paper is organized as follows. Section 2 presents the data and methods used in this study. In Sect. 3.1, the temporal-spatial characteristics of summer IOD and autumn IOD are described. In Sect. 3.2, we analyze the mixed layer heat budget to reveal the dynamics and thermodynamics that are responsible for the evolution of the eastern and western poles of these two types of IOD events. In Sect. 3.3, we highlight the distinctive linkages between ENSO and these two types of IOD events. The results have been summarized and discussed in Sect. 4.

2 Data and methods

2.1 Data

Monthly SST data on a horizontal resolution of 1° × 1° is derived from the Met Office Hadley Centre SST dataset (HadISST1; Rayner 2003) to display the evolution of IOD events. Atmospheric data including surface wind and sea level pressure (SLP) are obtained from the National Centers for Environmental Prediction‐National Center for Atmospheric Research (NCEP/NCAR) Reanalysis with a 2.5° × 2.5° resolution (Kalnay et al. 1996). The datasets above cover the period from 1948 to 2022.

Ocean potential temperature, oceanic circulation data, net surface heat flux, and sea surface height (SSH) are all derived from the NCEP Global Ocean Data Assimilation System (GODAS) product (Behringer et al. 1998). The SSH anomaly (SSHA) is used as a proxy to represent the variation of the thermocline depth. The outgoing longwave radiation (OLR) is derived from National Oceanic and Atmospheric Administration (NOAA; Liebmann and Smith 1996). The OLR is regarded as a proxy for deep convection, clouds, and precipitation in the tropical region. The positive (negative) OLR anomaly (OLRA) represents negative (positive) rainfall and suppressed (enhanced) convection. The monthly outputs above cover the period from 1980 to 2022. The four components of surface heat flux (latent heat flux, sensible heat flux, longwave radiation, and shortwave radiation) available from 1984 to 2009 are supported by Objectively Analyzed air-sea Fluxes for the global oceans (OAFlux; Yu and Weller 2007). After removing the monthly mean climatology, all variable anomalies presented in this study are detrended to eliminate the long-term trends.

To demonstrate the robustness of the results, we also checked the SST data from the NOAA SST ERSST monthly v5 (Huang et al. 2017) and obtained similar results. We also perform the mixed layer heat budget analysis using the SODA 3.3.1 (Carton et al. 2018) and the Ocean Reanalysis System 5 (ORAS5) data (Zuo et al. 2019). The four components of surface heat flux from the ECMWF Reanalysis v5 (ERA5; Hersbach et al. 2020) dataset is utilized to investigate processes being responsible for changes in net heat flux.

2.2 Methods

2.2.1 Classification of IOD events

The Dipole Mode Index (DMI) is defined as the difference in the averaged SSTAs between the WIO (10° S–10° N, 50° E–70° E) and the southeastern TIO (10° S–0°, 90° E–110° E), following Saji et al. (1999). An IOD event is defined when the three-month moving average of DMI exceeds one standard deviation during the peak season. We mainly focus on the pIOD events in this study. Some IOD events peak in boreal summer (June–July–August, JJA), while some others mature in boreal autumn (September–October–November, SON). Thus, the IOD events are divided into two groups based on the peak season in this study, i.e., the summer IOD and the autumn IOD. From 1948 to 2022, eight summer IOD events and eight autumn IOD events are identified, consistent with the findings of previous studies (e.g. Sun et al. 2014; Zhang et al. 2019).

2.2.2 Mixed layer heat budget analysis

To understand the ocean–atmosphere coupling processes responsible for the evolution of the western and the eastern poles, we conduct the mixed layer heat budget analysis (Vijith et al. 2020) for the summer IOD and the autumn IOD. The mixed layer heat budget equation is derived as follows,

$$\frac{\partial T}{\partial t} = Q-\left(u\frac{\partial T}{\partial x}+v\frac{\partial T}{\partial y}\right)-\left(\frac{T-{T}_{-h}}{h}\right){w}_{e}+R$$
(1)

Variables \(T\), \(u\), and \(v\) represent potential temperature, zonal and meridional ocean current velocity averaged within the mixed layer, respectively. Variables \(w\) and \(h\) represent vertical velocity and the mixed layer depth (MLD). Suffix \(-h\) denotes the quantity at the base of the mixed layer. The terms in Eq. 1 from left to right, are the mixed layer temperature (MLT) tendency, net heat flux, horizontal advection, entrainment, and residual term.

The \({w}_{e}\) in the entrainment term can be decomposed into the ML tendency, vertical advection and lateral induction, which is given by,

$${w}_{e }= \frac{\partial h}{\partial t}+{w}_{-h}+{u}_{-h}\frac{\partial h}{\partial x}+{v}_{-h}\frac{\partial h}{\partial y}$$
(2)

\(Q\) represents the net air-sea heat flux retained within the mixed layer. Further, \(Q\) can be decomposed into the \({Q}_{0}\) and \({Q}_{pen}\), where \({Q}_{0}\) is the net surface heat flux and \({Q}_{pen}\) is the shortwave radiation that penetrates the base of the mixed layer. \({\rho }_{0}\) is the density of water (1026 kg m−3), \({C}_{p}\) is the specific heat capacity of water (3986 J kg−1 K−1).

$$Q = \frac{{Q}_{0}-{Q}_{pen}}{{\rho }_{0}{C}_{p}h}$$
(3)

\({Q}_{0}\) is defined as the sum of the latent heat flux (LH), sensible heat flux (SH), shortwave radiation (SW) and longwave radiation (LW) at surface.

$${Q}_{0} = LH+SH+SW+LW$$
(4)

\({Q}_{pen}\) is estimated as (Paulson and Simpson 1977; Santoso et al. 2010),

$${Q}_{pen} = SW\left[R{e}^{\left(-h/{\gamma }_{1}\right)}+\left(1-R\right){e}^{\left(-h/{\gamma }_{2}\right)}\right],$$
(5)

where \(R\) = 0.67, \({\gamma }_{1}\) = 1 and \({\gamma }_{2}\) = 17 (Dong et al. 2007).

After performing the Reynolds decomposition analysis for Eq. 1, we obtain the Mixed layer heat budget equation as follows,

$$\frac{\partial T^{\prime}}{\partial t} = {Q}^{\prime}-\left(u^{\prime}\frac{\partial \overline{T}}{\partial x }+\overline{u}\frac{\partial {T }^{\prime}}{\partial x}+u^{\prime}\frac{\partial T^{\prime}}{\partial x}+v^{\prime}\frac{\partial \overline{T}}{\partial y }+\overline{v}\frac{\partial {T }^{\prime}}{\partial y}+v^{\prime}\frac{\partial T^{\prime}}{\partial y}\right)-\left(\frac{{T}^{\prime}-{T}_{-h}^{\prime}}{h}\overline{{w }_{e}}+\frac{\overline{T }-\overline{{T }_{-h}}}{h}{w}_{e}^{\prime}+\frac{{T}^{\prime}-{T}_{-h}^{\prime}}{h}{w}_{e}^{\prime}\right)+R,$$
(6)

where overbars and primes present the monthly climatology and anomaly, respectively.

3 Results

3.1 Characteristics of the summer IOD and the autumn IOD

Two types of IOD events with distinctive phase-locking features are presented in this study. The summer IOD develops rapidly in early boreal spring and reaches its mature phase in July (Fig. 1a), while the autumn IOD develops from late spring to autumn and peaks around October (Fig. 1b). It can be noticed that the western and eastern poles of the summer IOD develop simultaneously, while the eastern pole of the autumn IOD precedes the western pole for nearly one season (Fig. 1). The growth rate of the western pole of the autumn IOD seems to be suppressed in summer, significant positive SSTAs in the western pole are therefore absent before August (Figs. 1b and 2e, 7 of 8 autumn IOD events display this feature). Although the eastern pole of both the summer IOD and the autumn IOD begins to develop in JJA, that of the autumn IOD is characterized by a longer lifetime owing to stronger easterly wind anomalies in SON (Fig. 2e, f). With the intensification of the easterly wind anomalies, warm SSTAs are formed in the WIO of the autumn IOD (Fig. 2f). That is to say, these strong easterly winds may be crucial to the formation of not only the cold SSTAs in the eastern pole but also the warm SSTAs in the western pole, leading to the occurrence of autumn IOD. The associated ocean–atmosphere coupling processes will be discussed in Sect. 3.2.

Fig. 1
figure 1

Temporal evolution of the DMI (solid line, units: °C), SSTAs averaged over the western pole (red bar, units: °C) and the eastern pole (blue bar, units: °C) of the summer IOD in (a) and the autumn IOD in (b). Solid black line represents the mean values of HadISST and ERSST whereas shading represents a 0.5 standard deviation of the DMI. The error bars represent 1 standard error

Full size image
Fig. 2
figure 2

Composite of SSTAs (shading, units: ℃) and surface wind anomalies (vectors, units: m s−1) for the summer IOD (a–c) and the autumn IOD (d–f), from the developing spring to the succeeding autumn. Datasets for SST: HadISST and ERSST. Black boxes on the left and right respectively indicate the location of the western pole (10° S–10° N, 50° E–70° E) and the eastern pole (10° S–0°, 90° E–110° E). Only values at the 90% confidence level according to a two-tailed Student’s t-test are shown

Full size image

The spatial evolution of the summer IOD and autumn IOD also displays large differences (Fig. 2). The broad-scaled warm SSTAs of the western pole are observed to be close to the central Indian Ocean during JJA in the summer IOD years (Fig. 2b). The warm sea surface during the summer monsoon season is favorable for the enhancement of convection and the induced local air-sea interactions in the WIO. However, warm SSTAs of the autumn IOD are absent until SON and locate more to the west intruding into Somalia’s coastal region, indicating distinctive underlying mechanisms for the development of the two types of IOD events.

The discussion above points out that the temporal-spatial characteristics of the summer IOD and autumn IOD are remarkably different. The summer IOD is featured by warm SSTAs in the WIO during JJA, while the autumn IOD is characterized by strong easterly wind anomalies during SON, implying distinctive ocean-atmospheric couplings may be responsible for their development. The autumn IOD is similar to the canonical IOD described in previous studies (Saji et al. 1999; Webster et al. 1999), the mature phase of which lags behind the summer IOD resembling the unseasonable IOD proposed by Du et al. (2013) by nearly 3 months. We will further investigate the dynamic and thermodynamic processes responsible for the two types of IOD events, in terms of both the western and eastern poles.

3.2 Mechanisms

3.2.1 Warming in the western pole

The warm western pole of the summer IOD mostly develops during JJA, while that of the autumn IOD primarily occurs in SON (Fig. 2), suggesting that the underlying mechanisms might differ between the two types of IOD events.

For the summer IOD, the warming tendency of WIO SSTAs during March-August is mainly contributed by changes in both the oceanic dynamics and the net surface heat flux (Fig. 3a). The oceanic dynamics especially the entrainment over the WIO is presented to be important at an early stage of the developing phase (Fig. 2a). These features may act as precursors to the WIO warming for the summer IOD. However, after June, the relatively weak oceanic processes in the WIO are insufficient to maintain the warm mixed layer temperature (MLT) tendency (Fig. 3a), which may be related to the absence of strong equatorial easterly wind anomalies (Fig. 2b, c).

Fig. 3
figure 3

Composite monthly evolution of the MLT tendency (\(\partial T^{\prime}/\partial t\), black bold solid line, units: °C month−1), the net surface heat flux term (\({Q}_{a}\), orange solid line, units: °C month−1), the total zonal advection (green solid line, units: °C month−1), the total meridional advection (blue solid line, units: °C month−1), the entrainment term (red solid line, units: °C month−1) for (a) the western pole of summer IOD, (b) the eastern pole of summer IOD, (c) the western pole of autumn IOD, and (d) the eastern pole of autumn IOD. The MLT tendency and oceanic processes are averaged across the GODAS, SODA, and ORAS5 datasets. The net surface heat flux is averaged across the OAFlux and ERA5 datasets. Shading represents 1 standard deviation

Full size image

During boreal summer, the surface heat flux takes the role of oceanic advection in contributing to the warm WIO SSTAs (as shown in Figs. 3a and 4a). The positive latent heat flux and longwave radiation downward into the upper ocean are responsible for the large amplitude of net surface heat flux, which in turn is related to the weakened wind speed and enhanced convection in the WIO (Figs. 5a–c and 6b). In the western pole, the decrease in wind speed (Fig. 5a, b) leads to a reduction in sea surface latent heat loss, facilitating the formation of warm SSTAs. Additionally, the strong convection (denoted by the negative OLRA; Fig. 6b) over the WIO also contributes to the positive effect of longwave radiation. Although the SST-cloud-radiation feedback causes a reduction in shortwave radiation into the sea surface (as shown in Fig. 4a), the warm SSTAs in the WIO lead to an increase in water vapor, further promoting atmospheric convection and longwave radiation reflected into the ocean. Thus, the dominant factors in the warm growth rate of the western pole of summer pIOD during JJA are the vigorous latent heat flux and longwave radiation into the upper ocean (as depicted in Fig. 4a).

Fig. 4
figure 4

Composite of the heat budget terms in JJA and SON (units: °C month−1) over (a) the western pole of summer IOD, (b) the eastern pole of summer IOD, (c) the western pole of autumn IOD, and (d) the eastern pole of autumn IOD. The subplots represent the four components of net surface heat flux, namely the latent heat flux (LH), sensible heat flux (SH), shortwave radiation (SW), and longwave radiation (LW). The MLT tendency and oceanic processes are averaged across the GODAS, SODA, and ORAS5 datasets. The net surface heat flux is averaged across the OAFlux and ERA5 datasets. The error bars represent 1 standard error

Full size image
Fig. 5
figure 5

Composite of wind speed anomalies (shading, units: m s−1) and SLPA (contour, units: hPa) for the summer IOD (a–c) and the autumn IOD (d–f), from the developing spring to the succeeding autumn. Wind speed anomalies at the 90% confidence level according to a two-tailed Student’s t-test are dotted

Full size image
Fig. 6
figure 6

Composite of OLR anomalies (shading, units: W m−2) for the summer IOD (a–c) and the autumn IOD (d–f), from the developing spring to the succeeding autumn. Values at the 90% confidence level according to a two-tailed Student’s t-test are in black dots

Full size image

In contrast to the summer IOD, the western pole of the autumn IOD experiences a delay in warming, with warm SSTAs not being evident until September (Fig. 3c). During JJA, the relatively weak atmospheric and oceanic processes are insufficient to warm the cooler sea surface in MAM (Fig. 2d), resulting in the absence of upper-ocean warming in the WIO during summertime. However, the positive MLT tendency of the western pole is rapidly established during SON, reaching its peak in October due to strong oceanic processes (Fig. 3c). This warming tendency is counteracted by the net surface heat flux, particularly shortwave radiation, which impedes the warming tendency. Despite this, the strong upper-ocean meridional advection and entrainment are sufficient to offset the negative net surface heat flux, indicating the dominance of oceanic processes for the western pole of the autumn IOD. These findings suggest that the ocean–atmosphere coupling processes responsible for the evolution of the western pole of the autumn IOD differ from those of the summer IOD, with oceanic processes playing a more dominant role.

Compared to the summer IOD, easterly wind anomalies are stronger during the autumn IOD in JJA, and they rapidly strengthen in SON (Fig. 2), leading to stronger responses in the upper ocean of the WIO. Westward propagating off-equatorial downwelling Rossby waves are observed during autumn IOD years (Fig. 7b), and the Ekman pumping feedback (\(- w^{\prime}\partial \overline{T}/\partial z\)) associated with anomalous descending motions plays a dominant role compared to other terms (Fig. 4c). The Rossby waves originate in the southeastern TIO (~ 80° E), propagate westward along 10° S–8° S (Fig. 7b), and reach the WIO after about three months with a phase speed of approximately 0.2 m s−1 (consistent with previous studies Xie et al. 2002; Huang et al. 2020; Zhang and Du 2022). Correspondingly, significant warm SSTAs appear in the western pole and reach the maximum in SON (Fig. 2f). It can be noticed that the warming spreads to the east coast of North Africa as easterly wind anomalies intrude further west (Fig. 2f). It is reasonable to assume that the formation of the western pole of the autumn IOD is closely related to the intensive easterly winds. Previous studies also proposed that the strong easterly wind anomalies along the equator and anticyclonic wind stress curl over the off-equatorial region can not only drive warm water accumulated westward (Guo et al. 2015; Jiang et al. 2022) but also excite the westward downwelling Rossby waves (Cai et al. 2021), facilitating the warm SSTAs in the western pole (Murtugudde and Busalacchi 1999; Xie et al. 2002; Du et al. 2020). Hence, these wind-induced downwelling Rossby waves are essential to the warming SSTAs in the western pole of the autumn IOD.

Fig. 7
figure 7

Time-longitude diagrams of anomalous SSH (shading, units: cm) and thermocline depth (contour, units: m) averaged in the off-equatorial region (10° S–8° S) and the equatorial Indian Ocean (2° S–2° N) for the summer IOD (a, c) and the autumn IOD (b, d). Values at the 90% confidence level are in black dots. The yellow asterisk represents the maximum value of the positive SSHA

Full size image

Overall, the dynamic processes accounting for the western pole of the two types of pIOD events are significantly different. During summer pIOD years, positive SSTAs in the WIO primarily result from local ocean–atmosphere feedback, particularly the positive downward latent heat flux and longwave radiation induced by enhanced convection. However, oceanic processes associated with downwelling Rossby waves dominate the formation of the western pole of the autumn pIOD, which is closely related to the strong easterly wind anomalies.

3.2.2 Cooling in the eastern pole

Negative SSTAs emerge along the Java/Sumatra coastal region during boreal summer, coinciding with the establishment of the Indian summer monsoon in both types of pIOD years. The formation of the cold eastern pole is closely related to the prevailing easterly wind anomalies over the EIO. During this monsoonal season, intensified easterly winds further shoal the thermocline and enhance coastal upwelling off Java/Sumatra, triggering the positive Bjerknes feedback (Bjerknes 1969), resulting in negative zonal advection and vertical advection in the eastern TIO, which are mostly in phase with the MLT cooling tendency (Fig. 3b, d). This indicates the importance of oceanic dynamics in the development of the eastern pole of IOD events. Easterly wind anomalies can also cool the eastern pole via wind-evaporation-SST feedback (WES; Xie and Philander 1994). The acceleration of the wind speed (Fig. 5b, e) and greater latent heat loss (Fig. 4b, d) from the EIO sea surface can be noticed in JJA. The amplitude of the anomalous latent heat flux is observed to be larger in autumn IOD, corresponding to stronger easterly anomalies in the EIO (Fig. 2c, f). Although changes in the latent heat flux are conducive to cooling the eastern pole, shortwave radiation dominates the net surface heat flux and acts as a damping term in both types of IOD (as shown in Hong et al. 2008c). This is because the negative SSTAs suppress atmospheric convection over the EIO in JJA (Fig. 6), which reduces the cloud amount (positive OLRA) and leads to positive downward shortwave radiation into the upper ocean (Fig. 4b, d, as shown in Francis et al. 2007). In the succedent autumn, with the intensification of negative SSTAs and weakened convection, the damping effect of net heat flux rapidly increases, especially in the autumn IOD years with a colder underlying sea surface (Fig. 4d). However, the cold SSTAs in the eastern Indian Ocean still possesses a longer lifetime in the autumn IOD years compared to the summer IOD years.

The differences in the evolution of the eastern pole may be related to changes in easterly winds (Fig. 2c, f) and associated oceanic dynamics. During SON, the intensity of horizontal advection and entrainment weakens in summer IOD years (as seen in Fig. 3b). These oceanic processes are insufficient to offset the effect of the positive net surface heat flux, resulting in a gradual decline in the cooling tendency in the eastern pole during the summer IOD. However, oceanic advection continues to develop in SON and is strong enough to overwhelm the damping effect of the positive net surface heat flux (Fig. 3d). The cooling tendency associated with the ocean dynamical processes in the autumn IOD is approximately five times stronger than that in the summer IOD during this season (Fig. 4b, d). The autumn IOD is therefore characterized by a longer lifetime. Previous studies have also pointed out that nonlinear oceanic advection is critical to the formation of IOD (Halkides and Lee 2009; Ng et al. 2015; Cai et al. 2021). As shown in Fig. 4d, nonlinear advection, which is related to the Bjerknes feedback, dominates over other processes during SON. The stronger ocean dynamic processes are closely related to the remarkable intensive easterly winds persisting in SON (Fig. 2f), which dominate the formation and long persistence of the dipole SSTA pattern of the autumn IOD. The source of the easterly winds observed in the autumn IOD could be attributed to the remote forcing of ENSO, which will be discussed in the next section.

3.3 Differences in the ENSO-IOD relationship

The composite patterns of the two types of IOD events indicate that autumn IOD events are intimately related to ENSO (Fig. 10). We perform a lead regression to remove the succeeding Niño 3.4 index (mean value of the Niño3.4 index during December(0)–January(1)). The occurrence of autumn IOD events during 1948–2022 is reduced from 8 to 3 after removing the succeeding Niño 3.4 index (Table 1), demonstrating that the autumn IOD is strongly modulated by ENSO variability. The monthly evolution of summer IOD and autumn IOD after removing the effect of ENSO has been shown in Fig. 8. Compared with Fig. 1, it can be observed that the amplitude of the DMI and SSTA averaged over both the western and eastern poles of the autumn IOD largely decreases when the impact of ENSO is removed (Figs. 1b, 8b, and 9f), indicating that ENSO contributes significantly not only to the eastern pole but also to the western pole of the autumn IOD. While those amplitude in the summer IOD have hardly changed. The result indicates that ENSO has little effect on the summer IOD but strongly affects the autumn IOD.

Table 1 The summer IOD events and the autumn IOD events during 1980–2020. The event in bold denotes the one still robust after removing ENSO
Full size table
Fig. 8
figure 8

Same as Fig. 1, but for using SSTAs with the influence of ENSO excluded. ENSO variability is represented by Niño 3.4 index, which is defined as the SSTAs averaged over the tropical Pacific region (5° S–5° N, 170° W–120° W). The solid black line represents mean values of HadISST and ERSST whereas shading represents a 0.5 standard deviation of the DMI. The error bars represent 1 standard error

Full size image
Fig. 9
figure 9

Composite of SSTAs (shading, units: ℃) and surface wind anomalies (vectors, units: m s−1) that exclude the influence of ENSO for the summer IOD (a–c) and the autumn IOD (d–f), from the developing spring to the succeeding autumn. Datasets for SST: HadISST and ERSST. Only values at the 90% confidence level according to a two-tailed Student’s t-test are shown

Full size image

Easterly wind anomalies are crucial for the development of the eastern pole of IOD, as mentioned before. However, the spatial distribution, intensity, and persistence of these anomalies differ between the summer IOD and autumn IOD (Fig. 2). During the development of the summer IOD, anomalous easterly winds are concentrated in a narrow belt and are more restricted to the EIO, while during the autumn IOD, they feature a broader meridional pattern and extend more to the west (Fig. 2). Further studies indicate that the strong easterly wind anomalies during the autumn IOD are related to the El Niño accompanied intensification of the Australian high (Figs. 5f and 10d–f). During the developing phase of the ENSO, noticeable positive SLPA in the Australian region can be observed. The enhanced Australian high is favorable for the strengthened easterly wind anomalies during the autumn IOD, which in turn are conducive to the development of oceanic processes via positive Bjerknes feedback, corresponding to the strong zonal advection and entrainment in SON (Fig. 4d). That’s to say, the persistence of the autumn IOD is mainly attributed to the El Niño accompanied Australian high anomaly. However, these atmospheric responses to ENSO are absent in summer IOD years (Fig. 10a–c), demonstrating that ENSO has little effect on the summer IOD. During the summer IOD, owing to the lack of strong positive SLPA over the Australian continent in SON (Fig. 5c), the easterly wind anomalies along the Java/Sumatra coastal region are relatively weak (Fig. 2c). Therefore, the summer IOD events gradually decline in SON. This result is consistent with our finding that local air-sea interactions in the Indian Ocean play a crucial role in the evolution of the summer IOD.

Fig. 10
figure 10

Regression of SSTAs (shading, units: ℃), surface wind anomalies (vectors, units: m s−1), and SLPA (contour, units: hPa) upon Niño 3.4 index for the summer IOD (a–c) and autumn IOD (d–f), from the developing spring to the succeeding autumn. Datasets for SST: HadISST and ERSST. Anomalies are standardized by the amplitude of ENSO before composite. Only wind vectors that exceed 0.1 m s−1 are shown

Full size image

The differences in the ENSO-IOD relationship have a significant impact on the evolution of the eastern and western poles of the two types of IOD events. The eastern pole of the autumn IOD has a longer lifetime than that of the summer IOD. The growth rate of the autumn IOD is much larger due to strong easterly wind anomalies along the equatorial Indian Ocean, intimately related to ENSO. These anomalies persist until boreal winter and have a larger amplitude, making autumn IOD events much stronger than summer ones. The zonal wind anomalies also contribute to the formation of the western pole via the oceanic downwelling Rossby waves. However, atmospheric processes dominate the evolution of the short-lived summer IOD. Thus, ENSO is the primary external forcing that promotes and prolongs the growth of IOD events. Large-scale atmospheric responses to ENSO variability effectively fuel the development of the autumn IOD, while they are absent in summer IOD years (Fig. 10).

4 Summary and discussion

In this study, we have shown that the summer IOD and the autumn IOD, respectively mature in boreal summer and autumn, exhibit distinctive temporal-spatial characteristics resulting from different climate dynamic processes. The summer IOD benefits from broad-scaled warm SSTAs in the WIO, which enhance local air-sea interactions and easterly wind anomalies in the EIO through the east–west temperature gradient. The warm sea surface promotes local atmospheric convection and reduces surface wind speed. Then it induces strong positive longwave radiation reflected back to the upper ocean and less latent heat loss from the sea surface, which in turn lead to a rapid growth of warming in the western pole of the summer IOD. Additionally, the eastern pole begins to develop through positive Bjerknes feedback when the Indian summer monsoon prevails in June and then decays as the monsoon system retreats. Thus, the summer IOD is primarily influenced by the air-sea interactions intrinsic to the Indian Ocean, particularly in the WIO.

Conversely, for the autumn IOD, strong easterly wind anomalies extending from the eastern TIO to the central TIO play a crucial role in the development of both the western and eastern poles. Once these wind anomalies prevail off Sumatra/Java, the positive Bjerknes feedback and WES feedback contribute to the establishment of cold SSTAs in the EIO. The intensified equatorial easterly winds can also induce off-equatorial anticyclonic wind stress curl that excites downwelling Rossby waves, further warming SSTAs in the WIO via oceanic processes. Thus, the evolution of the eastern pole precedes that of the western pole, similar to the formation of the canonical IOD shown in previous studies (Wang et al. 2019). Additionally, the eastern pole associated with strong easterly wind anomalies dominates the formation of the autumn IOD (Luo et al. 2010).

The easterly wind anomalies during autumn IOD years are fueled remotely by El Niño events in the tropical Pacific. As a response to El Niño, positive SLPA with anticyclonic atmospheric circulation is observed over the Maritime Continent region and the southeastern TIO (as also shown in Yu and Rienecker 1999), and these anomalous easterly winds off Sumatra/Java can persist into SON. The El Niño accompanied Australian high anomaly is favorable for the strengthened easterly wind anomalies, which in turn are conducive to the development of the eastern and western pole of autumn IOD via oceanic processes. However, these atmospheric responses to ENSO are absent in summer IOD years. During the summer IOD, owing to the lack of strong positive SLPA over the Australian continent in SON, the easterly winds generated by air-sea interactions intrinsic to the Indian Ocean are weaker and short-lived. Therefore, the summer IOD events gradually decline in SON. The climate dynamic processes responsible for the two distinctive types of IOD events have been illustrated in Fig. 11. Local air-sea interactions intrinsic to the Indian Ocean are vital for the formation of the IOD with an earlier peak season, while ENSO can be regarded as the predominant external forcing that promotes and prolongs the growth of IOD events. Yang et al. (2015) reached a similar conclusion that the internal IOD starts to develop in June, while the ENSO-related IOD lags about two months. Thus, the climate dynamics underlying the evolution of the eastern and western poles of the summer IOD and the autumn IOD are different. Our results can help to deepen the understanding of IOD events through comprehensive analyses of IOD diversity.

Fig. 11
figure 11

Schematic diagram of the mechanisms for the development of (a) the summer IOD and (b) the autumn IOD. (a) WES and LWR respectively indicate wind-evaporation-SST feedback and longwave radiation reflected, contributing to the SST warming in the WIO of the summer IOD. (b) As a response to El Niño, positive SLPA with anticyclonic atmospheric circulation is favorable for the intensification of easterly wind anomalies, therefore dominates the formation of the western and eastern poles via oceanic downwelling Rossby waves and Bjerknes feedback. The black vectors denote the easterly wind anomalies. The red curve denotes the Rossby waves

Full size image

We have also preliminarily attempted to apply the CMIP6 historical simulation data to study the spatial–temporal characteristics of the two types of IOD events. The CMIP6 simulations results show that some IOD events tend to peak in JJA, and the spatial–temporal characteristics of the summer IOD are consistent with the conclusion in observations. While the warming of the WIO during autumn IOD tends to present in summer, which is slightly different from the feeble response of the WIO in observation. It might be due to an overestimation of the responses of Indian Ocean air–sea coupling strength to ENSO in climate model simulations (Zhang et al. 2021; Zhong et al. 2023). Thus, it is necessary to make further studies on the performance of CMIP models in simulating different types of IOD events.

Previous studies have shown that the mid-high latitude climate signals in the southern Indian Ocean can trigger the onset of the IOD independently of ENSO (Fischer et al. 2005). The Southern Annular Mode (SAM) is an influential climate signal in the Southern Hemisphere (Rogers and Loon 1982; Thompson and Wallace 2000). Some studies suggest that SAM can promote IOD-conducive wind anomalies via extratropical SLP fluctuations (Lau and Nath 2004), while others found that SAM cannot significantly impact the IOD (Cai et al. 2011a). It seems that the interaction between the high-latitude SAM and tropical IOD is still under debate. Moreover, a mode resembling the SIOD in the subtropical southern IO can be observed in the initial phase of the autumn IOD (Fig. 2d). The SIOD associated with the Mascarene high is supposed to be an important trigger for the IOD occurring several months later (Terray et al. 2007; Feng et al. 2014; Zhang et al. 2018). Zhang et al. (2020) further proposed that the Southern Hemisphere Mechanism (SHM), connecting the phase transition of SAM and SIOD, could induce the early onset of IOD events in the absence of ENSO. However, the autumn pIOD, which accompanies the SIOD in the preceding boreal spring, is shown to concurrently occur with El Niño events (Fig. 2f). Therefore, further investigation is needed to understand how the mid-high latitude climate signals modulate the tropical ocean–atmosphere couplings accounting for the two types of IOD presented in this study.