Interannual variability of sea level in the South Indian Ocean: Local versus remote forcing mechanisms

. The subtropical South Indian Ocean (SIO) has been described as one of the world's largest heat accumulators due to its remarkable warming during the past two decades. However, the relative contributions of the remote (of Pacific origin) forcing and local wind forcing to the variability of heat content and sea level in the SIO have not been fully attributed. Here, we combine a general circulation model, an analytic linear reduced gravity model, and observations to disentangle the spatial and temporal inputs of each forcing component on interannual to decadal timescales. A sensitivity experiment is conducted with artificially closed Indonesian straits to physically isolate the Indian and Pacific Oceans, thus, intentionally removing the Indonesian throughflow (ITF) influence on the Indian Ocean heat content and sea level variability. We show that the relative contribution of the signals originating in the equatorial Pacific versus signals caused by local wind forcing to the interannual variability of sea level and heat content in the SIO is dependent on location within the basin (low vs. mid latitude; western vs. eastern side of the basin). The closure of the ITF in the numerical experiment reduces the amplitude of interannual-to-decadal sea level changes compared to the simulation with a realistic ITF. However, the spatial and temporal evolution of sea level patterns in the two simulations remain similar and correlated with El Nino Southern Oscillation (ENSO). This suggests that these patterns are mostly determined by local wind forcing and oceanic processes, linked to ENSO via the ‘atmospheric bridge’ effect. We conclude that local wind forcing is an important driver for the interannual changes of sea level, heat content, and meridional transports in the SIO subtropical gyre, while oceanic signals originating in the Pacific amplify locally-forced signals. The sum weighted model-data misfits model The data constraints include temperature and salinity profiles (from floats, CTD, XBT, and ice-tethered profilers), satellite altimetry and gravimetry measurements, surface temperature fields from passive microwave radiometry, and satellite observations of sea-ice concentration. The control parameters include the initial temperature and salinity fields, the 3D parameters of Gent-McWilliams/Redi mixing and the with a lower amplitude in the ITF-off experiment compared to the ITF-on simulation. As expected, the results suggest the importance of upwelling/downwelling off the coast favored by southerly/northerly wind anomalies. Note that the area close to the West Australia coast is characterized by the strongest positive meridional wind stress in the Indian Ocean and elevated standard deviations (not shown). The reconstruction of SLA using the obtained regression coefficients demonstrates that the along-shore winds over the 1992-2018 time period explain 15% and 11% of the SLA variance in the ESIO in the ITF-on and ITF-off simulations, respectively. It should be noted that this relationship is time dependent, and becomes significant if we focus on the 2004-2018 time period. Over this time period, the along-shore winds explain 25% and 36% of the SLA variance in the ESIO in the ITF-on and ITF-off simulations, respectively.

modulate the ITF transport (Sprintall et al., 2009;Drushka et al., 2010;Lu et al., 2018;Pujiana et al., 2019). Although sometimes ENSO and IOD are regarded as a single entity (Allan et al., 2001), there is more evidence that the two processes are independent from each other (Ashok et al., 2003). Finally, the Southern Annular Mode (SAM) was diagnosed as the ultimate large-scale climatic forcing over the SIO (Schott et al., 2009). The SAM has significant effects on the supergyre , which links all the three subtropical gyres of the Southern Hemisphere (Cai et al., 2005;Ridgway and Dunn, 2007). SAM further contributes to modulating the teleconnections between ENSO and IOD by enhancing sea surface temperature (SST) gradients within the SIO (Cleverly et al., 2016).
The signals of the Pacific origin that enter the SIO with the ITF and reach the West Australia coast radiate westward as eddies and Rossby waves (Cai et al., 2005, Zhuang et al., 2013Zheng et al., 2018) . This process is called hereafter remote or eastern boundary forcing. On seasonal to interannual time scales, local wind forcing and related Ekman pumping over the ocean interior can also generate oceanic Rossby waves and/or alter those emanating from the eastern boundary (Masumoto and Meyers, 1998;Birol and Morrow, 2001). Menezes and Vianna (2019) found that the westward propagation observed in the Eastern SIO (ESIO - Figure 1b) basin and in the Western SIO (WSIO - Figure 1b) basin reveals a superposition of Rossby waves generated by processes near the eastern boundary on top of Rossby waves generated by Ekman pumping in the mid-basin. Thus, westward propagation provides the primary mechanism for both the "ocean tunnel" and the "atmospheric bridge" effects to transfer energy and seawater properties across the entire SIO (Morrow and Birol, 1998) On the basis of satellite altimetry, previous studies have examined the relative importance of the eastern boundary and the local wind forcing on the interannual variability of sea level and heat content in the SIO. Volkov et al. (2020) showed that the signals of the Pacific origin dominate sea level variability in the ESIO, while the local wind-induced variability is more influential in the WSIO, indicating a dependence on longitude. They also noted that the relative importance of local and remote drivers in the WSIO is time-dependent: in some periods remote and local forcing mechanisms appear to have similar magnitudes, while in other periods local wind forcing may become dominant. The latter happened in 2016-2018 when the OHC and sea level in the SIO recovered after an abrupt drop associated with the 2014-2016 El Niño (Figure 1a). Nagura and McPhaden (2021) extended the analysis of the eastern boundary and local wind forcing mechanisms by also demonstrating a latitudinal dependence. They confirmed that the influence of the eastern boundary forcing is confined to the ESIO but only at low latitudes (from 10° and 17°S). Between 10°S and 35°S, they identified two regimes: one in the tropics where local forcing has a bigger effect, and another in the mid latitudes (subtropics) where remote forcing has a greater influence. These results are in agreement with earlier studies that suggested the dominance of local wind forcing in driving sea level variability at low latitudes (from 11° to 13°S - Masumoto and Meyers, 1998;Zhuang et al., 2013), and the prevalence of variability radiated from the eastern boundary at mid latitudes (from 20° to 25°S - Zhuang et al., 2013;Menezes and Vianna, 2019). 3 It is usually assumed that the interannual variability of OHC and sea level along the West Australia coast is strongly linked to remote wind forcing in the tropical Pacific and, therefore, represents the "ocean tunnel" effect ( e.g., Nagura and McPhaden, 2021, and references therein). However, the impact of the along-shore wind forcing that also drives sea level variability along the coast is often disregarded. The along-shore winds are part of the large-scale atmospheric circulation over the SIO dominated by southeasterly trades, the variability of which is modulated by ENSO via the "atmospheric bridge" effect.
The objective of this study is to revisit the question of the interplay between remote and local drivers causing interannual-todecadal sea level variability in the SIO with a different approach. Specifically, we perform an ocean model sensitivity experiment, in which we physically isolate the Indian and Pacific Oceans by closing the Indonesian straits, thus removing the ITF ("ocean tunnel") influence on the Indian OHC and sea level variability. In this experiment, any variability observed along the SIO eastern boundary is primarily due to local forcing by construct. The solutions of this experiment were investigated by comparing them to a simulation with open Indonesian passages. By comparing these two runs, we can better quantify the relative contribution of the remote vs local drivers. In addition, numerical simulations are combined with a linear reduced gravity model and observations to disentangle the spatial and temporal dominance of each forcing component.

Ocean Model
The ocean model used in this study is a global ocean and sea-ice state estimate based on the Massachusetts Institute of Technology General Circulation Model (MITgcm, Marshall et al., 1998) and produced by Estimating the Circulation and Climate of the Ocean (ECCO) consortium (https://ecco-group.org). The ECCO consortium aims to create accurate, physically consistent, time-evolving estimates of ocean circulation by combining MITgcm with selected in situ and satellite observations (Menemenlis et al., 2005;Wunsch and Heimbach, 2007;Forget et al., 2015;Fukumori et al., 2017). The estimate uses the adjoint method to iteratively minimize the squared sum of weighted model-data misfits and to adjust the model control parameters (Wunsch et al., 2009;Wunsch and Heimbach, 2013). The data constraints include temperature and salinity profiles (from Argo floats, CTD, XBT, and ice-tethered profilers), satellite altimetry and gravimetry measurements, sea surface temperature fields from passive microwave radiometry, and satellite observations of sea-ice concentration. The control parameters include the initial temperature and salinity fields, the 3D parameters of Gent-McWilliams/Redi mixing scheme (Redi, 1982;Gent and McWilliams, 1990), and the time-varying atmospheric boundary conditions. Presently, all ECCO models employ a so-called Lat-Long-Cap (LLC) grid, ranging from LLC90 (~1°), LLC270 (~1/3°), to LLC4320 (~1/48°), which allows for an improved representation of the Arctic (no polar singularity and fine grid for small deformation radius). The LLC grid has five faces covering the whole globe, with a simple, locally isotropic latitudelongitude grid between 70°S and 57°N and an Arctic cap (Forget et al., 2015). In this study, we use the ECCO LLC270 configuration (Zhang et al., 2018), which provides a better representation of mesoscale variability compared to the latest ECCO LLC90 release (ECCO-V4r4, Forget et al., 2015). The horizontal resolution of the LLC270 grid varies spatially from 12 km at high latitudes to 28 km at midlatitudes. The vertical grid has 50 vertical layers, with the spacing increasing from 10 m near the surface to 457 m near the maximum model depth set to 6145 m.
The ECCO LLC270 solution used in this study is obtained by a free forward model integration from January 1992 to June 2018 using the adjusted control parameters and forced by the adjusted ERA-Interim atmospheric fields (Dee et al., 2011). To separate the influence of the Pacific Ocean on the variability of sea level and OHC in the Indian Ocean, we performed a numerical experiment with artificially closed Indonesian passages and the Torres Strait between Australia and New Guinea.
The closure of the Indonesian passages is achieved by modifying the model bathymetry ( Figure 2a). To distinguish between the optimized, realistic model run with open Indonesian passages and the experiment with closed Indonesian passages, we refer to them hereafter as the ITF-on and the ITF-off experiments, respectively. The ITF-off experiment was run twice. The first run was performed to let the model reach a stable state. The second run was initialized using the output of the first run in January 2018, and it was integrated again from January 1992 to June 2018. From now on, the ITF-off experiment refers to the second ITF-off run only. Because the ECCO LLC270 is a volume-conserving Boussinesq model, it does not reproduce the global mean sea level change. Therefore, the global mean sea level is subtracted prior to the analysis (Greatbatch, 1994).

Model for sea level variability
The westward propagation of Rossby waves and eddies provide the primary mechanism for both the "ocean tunnel" and the "atmospheric bridge" effects to transfer energy across the SIO. Under the long-wave approximation, these processes can be quantified by a 1.5-layer reduced-gravity model (e.g., Qiu, 2002), which has also been widely used to investigate the variability of sea level in the SIO (e.g., Zhuang et al., 2013;Jin et al., 2018;Menezes and Vianna, 2019;Volkov et al., 2020;Nagura and McPhaden, 2021). This reduced-gravity model, hereafter called RG model, is governed by the following linear vorticity equation, which separates the impacts of sea level signals originating at the eastern boundary and those generated by local wind forcing: where η is the baroclinic component of sea level anomaly (SLA), x is the longitude, t is the time, cR is the zonal phase speed of long baroclinic Rossby waves, g is the gravitational constant, g' is the reduced gravity, τ' is the wind stress anomaly vector, ρ0 is the mean sea water density, f is the Coriolis parameter, and ε is the damping coefficient.
The first term on the right side of Eq.
(2) represents the SLA signal that originates at the eastern boundary. The second term represents the SLA signal generated by the local wind stress curl. Both signals propagate westward and decay at a rate determined by the damping coefficient, ε. The RG model was initialized with the low-pass-filtered SLA at 110°E, η(xe, t), obtained from the two numerical simulations (ITF-on and ITF-off) at 13°S and 25°S (similar to Nagura and McPhaden, 2021 obtained empirically, and they are summarized in Table 1. Specifically, the phase speeds (cR) were set to 6.5 cm s −1 at 25°S and 13 cm s −1 at 13°S, based on the Hovmöller diagrams of the altimetry and model SLA in the SIO (see Section 3.1). These phase speeds are consistent with those estimated by Nagura and McPhaden (2021). The optimal values of g' and ε were obtained iteratively using a linear regression. The zonal mean g' varies from 0.05 m s -2 at 25°S to 0.07 m s -2 at 13°S for the ITF-on experiment and from 0.04 m s -2 at 25°S to 0.07 m s -2 at 13°S for the ITF-off experiment, reflecting the meridional variation in stratification (Zhuang and al., 2013). A regression analysis yielded the following optimum damping coefficients: ε -1 =2.5 years at 13°S and ε -1 =1.9 years at 25°S for the ITF-on experiment, and ε -1 =3.3 years at 13°S and ε -1 =1.2 years at 25°S for the ITF-off experiment (Table 1).

Data
Steric or density-driven changes dominate large-scale sea level variability on seasonal-to-interannual time scales (e.g., Gill and Niller, 1976). The monthly maps of altimetry SLA from 1993 to 2019 provided by the Copernicus Marine Environment Monitoring Service (CMEMS; Ducet et al., 2000) are used to validate the sea level changes simulated by the ECCO LLC270 model. The global mean sea level rise was subtracted from SLA maps to focus on regional variations. Since density variations in low and mid-latitudes are mainly caused by temperature variations, satellite measurements of sea level can be used as a proxy for OHC (Roemmich and Gilson, 2006 10°S to 30°S (green and red lines in Figure 1a, respectively). The time series are strongly correlated (R=0.90) confirming that SLA is a good indicator of OHC changes in the SIO.
We use the MEI, produced by the National Oceanic and Atmospheric Administration (NOAA) Earth System Research Laboratory's Physical Sciences Division (www.esrl.noaa.gov/psd), which incorporates both oceanic and atmospheric variables to provide a single index of ENSO intensity. This monthly index integrates the impact of five factors over the tropical Pacific basin (sea level pressure, sea surface temperature, zonal and meridional components of the surface wind, and outgoing long-wave radiation). Positive MEI events are related to warm, El Niño conditions and negative MEI events to cold, La Niña conditions ( Figure 1a). As indicators of IOD, we employed the DMI, a monthly product of the NOAA Earth System Research Laboratory (Saji and Yamagata, 2003). DMI is defined as the difference between averaged sea surface temperature over the western equatorial Indian Ocean (50°E-70°E; 10°S-10°N) and the south eastern equatorial Indian Ocean (90°E-110°E; 10°S-0°). Finally, we used the monthly Marshall SAM index based on available station zonal pressure observations between 40°S and 65°S (Marshall, 2003) obtained from British Antarctic Survey's website (www.nercbas.ac.uk/icd/gjma/sam.html).

Statistical analysis
To identify the dominant spatio-temporal modes of SLA variability in the Indian Ocean, an Empirical Orthogonal Function (EOF) analysis was carried out using the Matlab Climate Data Toolbox (Monahan et al., 2009;Greene et al., 2019). The obtained spatial patterns of the variability are referred to as EOFs, and their temporal evolutions are shown by the Principal Component time series (PCs). Prior to computing the EOFs, the global mean sea level, the seasonal cycle, and the linear trend were subtracted from the data. Then the data were low-pass filtered with a cutoff period of 1 year to focus on the interannual-to-decadal variability. The spatial patterns of EOFs are represented as regression maps obtained by projecting SLA onto the standardized (divided by standard deviation) PC time series. Thus, the regression coefficients are in centimeters (local change of sea level) per 1 standard deviation change of the PC. The linear regression analysis was also performed to relate the SLA to atmospheric circulation patterns (i.e., the meridional surface wind stress). Using these regression coefficients, we can reconstruct a SLA time series at each grid point associated with the PC or the meridional wind stress variability.

Variability of sea level in the Indian Ocean
Before exploring the ocean processes related to sea level changes in the ITF-on and ITF-off experiments, we first validate the model's performance by comparing the modeled upper-ocean ITF transport with that obtained from moored velocity It has been shown that the relative importance of remote and local drivers for sea level variability in the SIO depends on longitude and latitude (Volkov et al., 2020;Nagura and McPhaden, 2021). Based on these findings, we examined the SLA averaged over three different areas within box A (see insert in are mostly due to remote forcing at mid-latitudes, while local forcing is more important at low-latitudes. While investigating the relative importance of each mode of atmospheric variability goes beyond the scope of this paper, it has been shown that all these climate modes can influence the interannual-to-decadal variability of sea level in the SIO but none of them alone is able to explain the whole complexity of the local wind forcing (Volkov et al., 2020). Along with the signals coming from the equatorial Pacific, the variability of sea level and OHC along the West Australia coast is also driven by along-shore winds. These winds are modulated by ENSO via the "atmospheric bridge" effect. The regression of the monthly modeled SLA on the monthly wind stress averaged over 110-115°E and 20-35°S (rectangles in Fig. 6 a,b) is presented for both the ITF-on (Fig. 6 a,c) and ITF-off ( Fig. 6 b,d) simulations. The regression shows that northward/southward wind anomalies along the West Australia coast favor lower/higher sea level along the coast, with a lower amplitude in the ITF-off experiment compared to the ITF-on simulation. As expected, the results suggest the

Local versus remote forcing mechanisms
We quantify and assess here the combined and relative contributions of the eastern boundary forcing and the local wind stress curl to the interannual variability of the SIO using the 1.5-layer RG model (see Section 2.2) at 13°S and 25°S for both the ITF-on and the ITF-off simulations.
In the ITF-on experiment, the RG model reproduces SLA reasonably well (Figure 7a,b). The local wind forcing appears to dominate over the eastern boundary forcing at 13°S in the WSIO (Figure 7b We also applied the 1.5-layer RG model at 13°S and 25°S for the ITF-off simulation (Figure 9). Once again, the RG model is able to reproduce the simulated ECCO LLC270 SLA (Figure 9a,b). The signals simulated in the ITF-off experiment ( Figure   9) generally have the same characteristics as those simulated in the ITF-on experiment but with smaller amplitudes at 25°S (Figure 7). The local wind forcing appears to be the main driver for SLA variability at 13°S in the WSIO (Figure 9b,d), and the eastern boundary forcing dominates at 25°S in the ESIO (Figure 9f,g).

Meridional transport of the subtropical gyre
The large-scale sea level variability in the subtropical SIO reflects the meridional transports associated with the Meridional Overturning Circulation in the Indian Ocean (Zhuang et al., 2013;Nagura, 2020). Due to the connection between the Pacific and Indian Oceans via the Indonesian passages, these transports play a fundamental role in the inter-ocean exchange of mass, heat, salt, and carbon within the global climate system. A diagnosis of the Indian Ocean meridional transports is thus important for our understanding of the global climate and its variability (e.g., Wang, 2019). The zonal gradient of SLA between the western and eastern regions of the subtropical SIO is a proxy for this zonally integrated meridional transport (Zhuang et al., 2013;Nagura, 2020). Displayed in Figure 11 (c, d) are the zonal differences between SLA averaged over 50°-55°E (SLAwest) and 110°-115°E (SLAeast) as a function of latitude (15-30°S) and time in both the ITF-on and the ITF-off simulations. The differences between these two simulations are used to diagnose the relative importance of the ocean tunnel and atmospheric bridge effects on the interannual variability of the meridional transport.
In the ITF-on simulation (Figure 11c), the zonal (west-east) differences of SLA are clearly related to ENSO (see MEI in Figure 11a,b): the positive differences (in 1997-1998, 2002-2007, 2009-2010 and 2014-2016) are generally associated with El Niño conditions (cold anomalies in the ESIO) and the negative differences (in 1999-2001, 2008, 2011-2013)  To verify the relationship between the zonal gradient of SLA and the meridional transport in the SIO subtropical gyre, we computed the zonally-integrated transports across 15°S in both the ITF-on and the ITF-off simulations. This latitude is a key position where the meridional transports are used to measure the subtropical cell variability (Zhuang et al., 2013;Nagura, 2020). The transports are integrated from coast to coast thus including contributions from the interior and boundary currents. is amplified/reduced by ± 1 Sv from 300 m to the bottom in the ITF-on experiment. In the ITF-off experiment, the transport anomaly is observed mainly between 500 and 3000 m with a maximal value of 0.7 Sv. The consistency between the stream functions demonstrates that the meridional geostrophic transport is modulated by ENSO via the "atmospheric bridge" effect.

Discussion and Concluding remarks
In this study, we revisit the interplay between the remote and local drivers for the OHC and sea level variability in the SIO using an eddy-permitting ECCO LLC270 solution. The data-model comparisons demonstrate that the model performs well in simulating the circulation of the Indian Ocean including the ITF transport. In order to better quantify the relative contribution of the remote vs local drivers, we conducted a sensitivity experiment with closed Indonesian and Torres straits to physically separate the Indian and Pacific Oceans and eliminate the ocean tunnel effect on sea level variability in the SIO (ITF-off experiment).
Effects of the ITF on the circulation and thermal structure of the Indian Ocean were investigated by comparing solutions of an ocean general circulation model with open and closed Indonesian passages in previous studies (e.g., Hughes et al., 1992;Lee et al., 2002;Song and Gordon, 2004). These studies focused on seasonal to inter-annual variability and showed that the ITF warms the upper Indian Ocean and deepens its thermocline. Subsequently, the full role of ITF was analyzed with coupled general circulation models (e.g., Wajsowicz and Schneider, 2001;Song et al., 2007). The changes were similar to those reported by ocean-only model experiments but with larger amplitude. In addition, they showed that the closure of the ITF significantly increased interannual variability in the eastern tropical Indian Ocean. Closing the ITF shoals the eastern tropical Indian Ocean thermocline, resulting in greater cooling episodes due to enhanced atmosphere-thermocline coupled feedback.
The observed SLA variability in the SIO is well simulated by the ITF-on experiment showing a persistent increase from 2004 to 2014, while no pronounced increase of SLA is observed in the ITF-off experiment (Figure 3a). This result suggests that the observed decade-long heat accumulation is due to the ocean tunnel effect, linked to the ENSO variability. Shutting the ITF off removes heat supply from the Pacific, slows down the SEC, and raises the thermocline in the SIO (Lee et al., 2002). The fact that the observed heat accumulated affected nearly the entire SIO is somewhat inconsistent with numerical tracer experiments showing that the ITF waters entering the Indian Ocean are carried mainly by the SEC (e.g., Song et al., 2004). Nevertheless, Valsala and Ikeda (2007) showed that the ITF waters can also flow along the west coast of Australia, via LC, and then spread westward across the SIO.
As evidenced by satellite altimetry and Argo measurements, the decade-long accumulation of heat content in the SIO subtropical gyre ended with a strong cooling in 2014-2016. An abrupt decrease of SLA from 2014 to 2016 was well simulated in the ITF-on simulation, with an amplitude similar to satellite observations. This time period was marked by a strong El Niño, the associated reduction of southeasterly trade winds and a cyclonic wind anomaly across the SIO (Volkov et al., 2020). This resulted in Ekman divergence favoring the observed upper-ocean heat content decrease. In addition, the 2014-2016 El Niño was in phase with a weak positive IOD in 2015 and the lowest on record negative IOD in 2016. These phenomena, along with a deepening of the thermocline and increased sea levels along the Maritime continent, resulted in the reduction of the ITF transport and upper-ocean heat content (Figures 1a, 2b). In 2014-2016, the ITF-off experiment also revealed a decrease in SLA, albeit the magnitude of this decline was two times less than in the ITF-on simulation. This sensitivity experiment confirms that neither the atmospheric bridge nor the ocean tunnel effect alone can fully explain the observed basin-wide cooling.
The EOF analysis revealed that the spatio-temporal structures of the SLA variability in the ITF-on and ITF-off experiments are similar. The difference between the two simulations is mainly limited to the magnitude of the signals. This suggests that the spatio-temporal structure of the regional SLA and OHC variability is determined by local processes and wind forcing, Winds along the West Australia coast are part of the large-scale atmospheric circulation over the SIO dominated by southeasterly trades and modulated by ENSO via the "atmospheric bridge" effect. We find that these winds are able to explain only 15% and 11% of the SLA variability averaged over 110-115°E and 20-35°S in 1995-2018 in the ITF-on and ITF-off simulations, respectively. The rather small impact of the along-shore wind stress on the regional SLA variability is in agreement with the study of Nagura and McPhaden (2021), who showed that the lead-lag correlation between SLA and wind forcing does not exceed the 95% confidence level at any lag within ±12-month window. This relationship becomes stronger and significant between 2004 and 2018, where the along-shore winds explain greater SLA variance in the ESIO in the ITFoff simulation compared to the ITF-on simulation. This result shows the importance of the along-shore wind forcing that also drives sea level variability along the coast via the "atmospheric bridge" effect.
The processes driving the interannual variability of sea level and OHC in the SIO were further analyzed in the context of westward-propagating Rossby waves. Confirming previous observation-based studies (Volkov et al., 2020;Nagura and McPhaden, 2021), we showed that the relative importance of signals propagating from the eastern boundary and the local wind forcing varies with latitude (low vs. mid latitude) and longitude (WSIO vs. ESIO). The time series of the simulated SLA averaged over the WSIO at low latitudes (13°S) are closely aligned in both numerical experiments. It appears that the closure of the Indonesian passages does not significantly change the interannual variability of SLA in this area. Our analysis further supports the idea that the local wind forcing appears to be the main driver for SLA variability in the tropical WSIO.
Decadal tendencies are also similar in the two simulations over the ESIO at low latitudes, although the interannual variability is somewhat different. Overall, the results showed that SLA variability in the ESIO is dominated by signals radiated from the eastern boundary. At low latitudes, the interannual sea-level variability is driven by the local wind forcing in the WSIO and by the eastern boundary forcing in the ESIO. This highlights the longitudinal dependence causing interannual SLA variability in the south Indian Ocean demonstrated by Volkov et al. (2020). The variability of SLA is quite different in the two simulations at mid latitude, although the westward propagation can be easily traced in many cases. Results showed that the interannual variability of SLA at mid latitudes is mainly driven by waves radiated from the eastern boundary related to the ocean tunnel effect, i.e., the ITF and coastally trapped waves that rapidly transfer anomalies generated in the Pacific along the west Australian coast. This highlights the latitudinal dependence demonstrated in earlier studies (Masumoto and Meyers, 1998;Zhuang et al., 2013;Menezes and Vianna, 2019;Nagura and McPhaden, 2021).
Considering the role of meridional overturning circulation in the mass and heat transport, it is important to examine the year- to-year variability in this circulation. We confirmed in our study that the interannual variability of the meridional transport is exhibit similar amplitudes and patterns. Nagura and McPhaden (2021) showed that the meridional transport of the subtropical gyre is primarily driven by variability radiated from the eastern boundary. Our analysis shows that local wind forcing modulated by the "atmospheric bridge" effect is an important driver for the meridional transport in the SIO subtropical gyre.
Overall, these findings advance our understanding of regional heat content and sea level variability in this key region. To study the far-reaching impacts of the heat accumulation in the SIO, more research employing ongoing observations as well as ocean and climate models is necessary.

Data availability
The altimetry products were provided by the Copernicus Marine Environment Monitoring Service (https://marine.copernicus.eu/www.esrl.noaa.gov/psd/). The IOD index is provided by NOAA/PSL using the HadISST1.1 SST dataset (https://psl.noaa.gov/gcos_). The SAM index is obtained from British Antarctic Survey's website (www.nercbas.ac.uk/icd/gjma/). The ECCO LLC270 solution is available for download at https://ecco.jpl.nasa.gov/Version5/Alpha. All data needed to evaluate the conclusions in the paper are present in the paper Additional data related to this paper may be requested from the authors.