Strong subsurface meridional current forced by monsoon intraseasonal oscillation in the southern Bay of Bengal during summer 2020

The meridional current in the southern Bay of Bengal (BOB) exhibits prominent intraseasonal variability (ISV), which exerts a critical influence on meridional mass and energy exchange. However, its relationship with the northward propagating monsoon intraseasonal oscillation (MISO), which is the predominant variability in the tropical Indian Ocean during summer, is not well understood. Using a one-year mooring deployed at 5.5 °N, 90 °E, a strong ISV of the meridional current is observed between 150 and 250 m during summer, exhibiting amplitudes exceeding 0.4 ms−1 and periods of 30–70 d. Further analysis shows that the summer ISV is forced by a strong MISO event with the following dynamic processes. The MISO first drives the zonal wind stress at the equator, leading to the equatorial Kelvin wave and the reflected equatorial Rossby wave at 5 °N. Then, the MISO propagates northward and generates local wind stress curl-induced Ekman pumping near the mooring site, leading to the subsequent off-equatorial Rossby wave at 8 °N. As a result, the synergy of the remotely- and locally-forced Rossby wave causes strong subsurface ISV in southern BOB. This study provides a new insight into the dynamic relationship between the MISO and the meridional current in the southern BOB, which has important implications for regional climate research and prediction.


Introduction
The meridional current in the southern Bay of Bengal (BOB) is critical to meridional mass and energy exchange and thus influences local and large-scale air-sea interactions (Sengupta et al 2001, Masumoto et al 2005, Rao et al 2017, Phillips et al 2021. In particular, the meridional current exhibits prominent intraseasonal variability (ISV) (Han 2005, Schott et al 2009, Chen et al 2017, Zhang et al 2022, which is forced by atmospheric ISVs of the Madden-Julian Oscillation (MJO; Julian 1971, 1972) during boreal winter (hereafter, the word boreal will be omitted) and the Indian monsoon intraseasonal oscillation (MISO) during summer (Suhas et al 2013, Zhou et al 2017a. However, the relationship between meridional current and the MISO is less understood than that with the MJO, especially in the subsurface layer of 150-250 m. Several studies (e.g. Schott and Mc Creary Jr 2001, Ogata et al 2008, Chen et al 2017 have attributed the winter ISV of meridional current in the southern BOB to the eastward propagating MJO, which dominates the ISV along the equator. For example, based on a mooring deployed at 5.0 • N, 90.5 • E in 2013, Chen et al (2017) reported a strong ISV of the meridional current in the upper 150 m, with a magnitude of 0.4 ms −1 and period of 30-50 d. Their analysis suggested that the ISV of the meridional current is mainly associated with the remotely forced equatorial wave at the equator, which is forced by the MJO-induced zonal wind stress along the equator.
In comparison, the mechanism of summer ISV of meridional current in the southern BOB is relatively complicated because the MISO not only induces significant ISVs at the equator through eastward propagating, but also causes prominent ISV in the BOB through northward propagation (Krishnamurti et al 2017, Zhang et al 2018. Based on moored current measurements deployed at 5.0 • N, 90.5 • E in 2016, Huang et al (2021) reported an enhanced summer ISV in meridional currents in the upper 500 m of the southern BOB. Similar to Chen et al's (2017) result, they associated the summer ISV with the remotely forced equatorial wave, which is caused by the MISO-induced zonal wind at the equator. In contrast, Girishkumar et al (2013) and Li et al (2017a), Li et al (2017b) showed that the ISV in the upper layer of the southern BOB is mainly caused by the MISO-induced local wind stress and the subsequent ocean Rossby wave and other processes.
Thus far, it is known that the MISO can induce strong ISV in the southern BOB through remotely forced equatorial wave or locally forced wave in the BOB. However, the dynamic relationship between local and remote factors is unclear; hence, how the local-remote interaction impacts the meridional current in the southern BOB has not been determined. To address these two issues, a subsurface mooring was deployed at 5.5 • N, 90 • E (figure S1), which is very close to the location from two mooring studies in 2013 and 2016 (Chen et al 2017, Huang et al 2021. In particular, a strong MISO event was observed during summer 2020, with a larger northward propagation amplitude than that in 2013 and 2016 (figure S2). As compared with the sea surface height anomaly (SSHA) in 2013 and 2016 (figure S2), the strong MISO event in 2020 is accompanied by the strongest magnitude of SSHA along 5 • N and 8 • N, which resembles the oceanic Rossby wave. Such synergy of the two Rossby waves under the influence of strong MISO event is absent in previous studies, and thus provides an opportunity to examine the relationship between MISO-induced local and remote factors and their influence on the ISV of the meridional current in the southern BOB.

Data and method
2.1. Long-term mooring, multivariable satellite and reanalysis data To investigate the relationship between the meridional current in the southern BOB and the MISO, a subsurface mooring, which was supported by an international cooperation project named the 'Joint Advanced Marine and Ecological Studies (JAMES)' , was deployed at 5.5 • N, 90 • E from 15 February 2020 to 15 February 2021. An upward-looking 75 kHz acoustic-Doppler-current profiler was equipped to monitor the current velocity in the upper 100-500 m. The vertical resolution is 15 m, and the sampling time frequency is 1 h. Due to lack of data from 100 to 150 m during summer, the vertical range of the data was kept at 150-390 m to ensure the completeness of spatial coverage of data. The current velocities are linearly interpolated onto uniform 10 m intervals, and hourly measurements are averaged into daily data. In addition, six conductivity-temperature-depth profilers were deployed at the same site to measure the temperature data at depths of 110 m, 130 m, 150 m, 170 m, 200 m and 300 m at hourly intervals. The temperature data are linearly interpolated onto uniform 20 m and daily intervals. In this study, only isotherm depths of 14 • C and 13 • C are used to express vertical fluctuations caused by meridional current.
Furthermore, we collected multivariable satellite and reanalysis data to investigate the dynamic process associated with the ISV of the moored meridional current. The oceanic data comprising five daily interval total surface current fields (geostrophic plus Ekman components) from the Ocean Surface Current Analysis-Real Time data (OSCAR, Bonjean and Lagerloef 2002) and daily surface geostrophic current and SSHA data from the Archiving, Validation, and Interpretation of Satellite Oceanographic (AVISO, Ducet et al 2000) were utilized to understand ISV of surface current and SSHA in the Indian Ocean. Additionally, the daily isotherm depth calculated from temperature data from the Hybrid Coordinate Ocean Model reanalysis (Bleck 2002) was used to examine the subsurface ISV in the Indian Ocean. It is noted that the HYCOM data agrees well with the moored V in the subsurface layer, and the correlation is 0.85 for 150-250 m averaged V and 0.83 for 250-350 m averaged V. Atmospheric forcing comprising daily sea surface wind data from the Cross-Calibrated Multi-Platform (Atlas et al 2011) as well as daily precipitation data from the Global Precipitation Climatology Project (Huffman et al 1997) were collected to understand the relationship between MISO-induced atmospheric ISVs and moored meridional currents.

Spectral analysis and wave decomposition methods
To extract the ISV from observations, the Butterworth filter with fourth order is used to obtain the intraseasonal signal (30-70 d) in meridional current and other related fields. The power spectrum and cross-wavelet transform method (Torrence and Compo 1998) are applied to investigate the dominant period of the meridional current and its relationship with other surface fields at mooring site in timefrequency domain. A meridional decomposition of SSHA into Kelvin and Rossby modes is performed by projecting the analytical solutions from a 1.5layer shallow water model under longwave approximation (detailed information can be found in papers by Matsuno 1966, Zheng et al 1995 onto the meridional SSHA in the Eastern Indian Ocean (EIO) to understand the wave component associated with the observed ISV.

ISV of subsurface meridional current
To emphasize the relationship between the strong meridional current (V) and MISO event. The intraseasonal (hereafter, intraseasonal denotes the variability of 30-70 d) precipitation and zonal wind stress anomaly averaged between 80 • E and 95 • E are firstly shown in figures 1(a) and (b). The result shows a significant northward propagating feature during 1 July-15 September 2020. In the meantime, strong ISV of V (figure 1(c)) was also observed during the same period (red vertical line), whereas the ISV of zonal current (U, figure 1(d)) is much weaker than its meridional counterpart. Specifically, V exhibits strong ISV during MISO period, with the largest value exceeding 0.4 ms −1 from 150-250 m, whereas U mainly exhibits semiannual variability within 150-200 m. This difference can be seen more clearly in the corresponding power spectra of U and V in figures 1(g) and (h). As expected, the 150-250 m-averaged V shows prominent spectral power at intraseasonal bands from 30-70 d. In particular, the largest power spectrum is at 52 d, which is much stronger than that at other frequencies. Significant spectral power at the same intraseasonal band is also observed below the thermocline from 250-390 m but with weaker amplitude. In comparison, the power spectra of U are much weaker near the same intraseasonal band. Instead, U shows significant power spectra at semiannual period (>120 d) in both layers. Huang et al (2019) reported this semiannual variability of U at 5 • N, 90 • E and associated it with equatorial wave guide.
The intraseasonal current and isothern depth are shown in figures 1(e) and (f). Figure 1(f) clearly shows the prominent subsurface ISV of V during summer. It is highly correlated with the significant ISV of isotherm depth with a phase difference, and the largest correlation is −0.64 with isotherm depth leading 13 d. This indicates a lagged correlation between the horizontal pressure gradient and V, which could be related to the geostrophic adjustment between the Rossby wave and local wind stress. Meanwhile, U shows much weaker ISV (figure 1(e)). It is less correlated with the strong ISV of isotherm depth and the correlation peaks at −0.40 with U leading 100 d. Note that weaker ISV of V and isotherm depths are also observed from October to January (figure 1(f)), which could be related to the eastward propagating MJO during the same period (figures 1(a) and (b)). Figure 1 shows that the strong summer ISV of meridional current in the subsurface layer is associated with the atmospheric oscillations of the MISO. Therefore, the intraseasonal relationship between subsurface V and surface oceanic and atmospheric quantities at the mooring site is examined. First, power spectra analysis is performed on SSHA, surface V current from OSCAR and AVISO, zonal and meridional sea surface wind stress and precipitation to investigate the dominant frequency (figures 2(a)-(d)). In total, all spectra show significant variability in the intraseasonal band of 30-70 d. In particular, both the SSHA and surface V (figures 2(a) and (b)) exhibit two power spectrum peaks of 40 and 52 d, of which the larger peak of 52 d is consistent with that of subsurface V (figure 1(d)). Surface V from both OSCAR and AVISO exhibit very similar distributions of power spectra, indicating the geostrophic feature of this ISV ( figure 2(b)). On the other hand, the power spectra in the atmospheric variables (figures 2(c) and (d)) also show significant oscillations within similar intraseasonal band of 30-70 d.

ISV at the ocean surface
Furthermore, the phase relationship between subsurface V and the surface oceanic and atmospheric quantities in the intraseasonal band of 30-70 d is explored using cross-wavelet transform analysis. As figures 2(e)-(j) show, subsurface V exhibits a significant correlation with oceanic and atmospheric quantities at the intraseasonal band during the observation period, with the largest power spectra in the summer period. During summer, subsurface V and surface V (figures 2(f) and (g)) are almost in phase, whereas subsurface V lags the SSHA by approximately 5 d (figure 2(e)). Meanwhile, subsurface V lags the zonal and meridional wind stress by approximately 20 d (figures 2(h) and (i)) and lags surface precipitation by approximately 10 d (figure 2(j)). The phase difference of different atmospheric variables could be explained that the ocean current is directly forced by the wind stress and the precipitation alters the wind through divergence and convergence of wind.

Local and remote processes
The results in section 3 show that the subsurface V is significantly correlated with surface air-sea quantities at the mooring site during MISO period, hence the associated local and remote air-sea processes in the horizontal plane of Indian Ocean are investigated in this section. A positive phase during 5 July and 29 July 2020 is selected when both the moored V and MISO are significant (figure 1). As figure 3 shows, the most distinctive feature is the ocean dynamics associated with the remote and local wind forcing.
The first event is a westerly wind stress at the equator on 5 July (figure 3(k)), which is associated with an eastward current and elevated dynamical height in both surface (figure 3(a)) and subsurface layers (150-250 m, figure 3(f)). This intraseasonal signal at the equator reaches the Sumatra coast and travels along the coast, and then reflected near the northern coast as westward propagating eddy, with anti-cyclonic current and elevated dynamical height in the upper ocean (figures 3(b) and (g)). This westward anti-cyclonic eddy along 5 • N reaches mooring site on 11 July ( figure 3(d)). Similarly, a negative phase initiates on 23 July with easterly wind, westward current and negative dynamical height (figures 3(n) and (d)).
The second event is the positive local wind stress north of mooring site on 5 July (black box, figure 3(p)), which is associated with a local anticyclonic eddy (black box, figure 3(a)). This eddy propagates westward and mainly influences the area west of mooring site between 5 • N and 8 • N. Note that this anti-cyclonic eddy continuously interacts with the eddy from the first event during 5 July and 23 July both in surface and subsurface layers. Similarly, a negative phase reoccurs with negative local wind stress and cyclonic eddy on 23 July (figures 3(d) and (s)).
The northeastward movement of wind stress anomaly from the equatorial area to the off-equatorial area in the BOB is associated with that of precipitation anomaly. This relationship alternates about 20 d, which is consistent with previous analysis (figures 1 and 2). Such coherent phase change of wind stress and precipitation oscillations is consistent with the classical relationships for these variables in the MISO (Suhas et al 2013, Zhou et al 2017b. On the other hand, the ocean dynamics associated with remote and local atmospheric processes resemble the wind-forced equatorially trapped Kelvin and the subsequent reflected Rossby waves, and the off-equatorial Rossby wave. To verify this, we have selected three latitudinal bands along the equator, 5 • N and 8 • N (red line, figure 3). The corresponding wave phase speed can be calculated based on the observed SSHA (figures 4(a)-(c)). Then the speed is compared with that from the theory. The phase speed of Kelvin wave is estimated by fitting the observed mean meridional structure of SSHA (η) to the theoretical cross-equatorial structure of the equatorial Kelvin wave solution (η (y) = η (0) e −βy 2 2c ), using the least-squares method (Pujiana and McPhaden 2020). Where η (0) is the reconstructed SSHA at the equator, β = 2.310 −11 , y is the distance from the equator, c is the phase speed. The observed η is obtained by averaging the intraseasonal SSHA along 85 • E during an eastward Kelvin wave event from 15 June-10 July 2020. Figure S3(a) shows that the best least-squares fit exhibits an exponential function with η (0) of ∼0.016 m and c of ∼1.49 ms −1 . This estimated phase speed is close to the theoretical phase speed of the second baroclinic mode Kelvin wave (1.5 ms −1 , Nagura and McPhaden 2010), and is also consistent with Pujiana and McPhaden (2020)'s estimate based on multiyear in situ current velocity data.
The same method is difficult to apply on the Rossby wave due to the dynamical difference between equatorial and off-equatorial Rossby waves. Instead, it is calculated as the ratio of distance and the time delay. As shown in figure 4(b), the equatorial Rossby wave along 5 • N exhibits two different phase speed which is separated by 90 • E, where the mooring locates. The first one east of 90 • E shows phase speed of 0.47 ms −1 (the lagged correlation between 90 • E and 95 • E along 5 • N shows maximum correlation of 0.86, when SSHA at 95 • E lags the SSHA at 90 • E by 13 d). This result agrees well with the equatorial wave theory that the phase speed of Rossby wave is about one third of Kelvin wave, and is close to the theoretical propagation speed (0.5 ms −1 ) of the equatorial Rossby wave at 5 • N in the BoB (Webber et al 2012). The second one west of 90 • E shows phase speed of 0.34 ms −1 (88 • E to 84 • E, 5 • N, 19 d), which is close to the phase speed of 0.36 ms −1 for the off-equatorial Rossby wave at 8 • N (90 • E to 84 • E, 17 d). The wave phase speed at 8 • N agrees well with the theoretical propagation speed of the off-equatorial Rossby wave at 8 • N in the BOB (Killworth and Blundell 2005). for the off-equatorial Rossby wave, where c n is the nth baroclinic mode speed, f 0 is Coriolis parameter at given latitude). Based on the equation, an overlap of phase speed is found between the equatorial Rossby wave and the off-equatorial Rossby wave from 5 • N to 8 • N ( figure S3(b)), indicating the possibility of interaction between them. This is validated by the observed SSHA in the area west of 90 • E at 5 • N and at 8 • N, where the estimated phase speed is very similar. For now, it is difficult to discriminate the overlapped influence from those two waves between 5 • N and 8 • N under current theoretical framework, but this problem may be solved by another theory proposed by Aiki et al (2017), in which the equatorial and off-equatorial waves are unified based on group-velocity-based energy flux. In next study, we will apply the numerical experiment to further investigate the interaction between the equatorial and off-equatorial Rossby waves associated with the new theory.

Local versus remote mechanisms
Last section we have shown that the equatorial and off-equatorial waves coincide with remote and local wind forcing. In this section, the possible causes for those waves and their contribution to the moored V are further analyzed. To exhibit the complete wave guide related to the moored V during 15 June-25 August the intraseasonal SSHA along the equator and Sumatra coast, 5 • N and 8 • N is presented in figures 4(a)-(c). The equatorial wave guide is consist of the equatorial Kelvin wave ( figure 4(a)) and the reflected equatorial Rossby wave from 95 • E to 90 • E at 5 • N ( figure 4(b)). The off-equatorial wave guide is the Rossby wave from 90 • E to 83 • E at 8 • N ( figure 4(c)). The SSHA from 90 • E to 83 • E at 5 • N is influenced by both the equatorial and off-equatorial Rossby waves ( figure 4(b)). Figure 3 shows that the westerly (easterly) wind stress in the equatorial area leads to positive (negative) SSHA in the same region. This relationship is further verified by the lead-lag correlation between zonally averaged wind stress and SSH along equator, with the highest correlation of 0.75 in the EIO ( figure 4(d)). Meanwhile, the SSHA at 5 • N is significantly correlated with that in the upper stream where the Kelvin wave reaches the eastern boundary (black box, figure 3(a)). The slope of the largest correlation is consistent with the phase speed of Rossby wave from 95 • E to 90 • E. This is consistent with equatorial wind-forced Kelvin wave and the subsequent reflected Rossby wave from previous studies (Iskandar andMcPhaden 2011, Nagura andMcPhaden 2012). The off-equatorial wave and local Ekman pumping velocity around 90 • E, 8 • N are not only well correlated in space (figure 3), but also are highly correlated in time ( figure 4(f)). The slope of the largest correlation is consistent with the phase speed of Rossby wave from 90 • E to 83 • E. This locally wind-forced Rossby wave at 8 • N has been reported and discussed by Girishkumar et al (2013) and Webber et al (2018).
Lastly, the relationship between moored V and the SSHA along equator, 5 • N and 8 • N is investigated. As figures 4(g)-(i) shows, the SSHA along three latitudinal bands exhibit high correlation with the moored V during the same period in figure 3. In addition, the slope of the largest correlation is similar to that of SSHA (figures 4(a)-(c)), meaning that the moored V is associated with the equatorial and off-equatorial wave propagation. It is observed from figure 3 that the equatorial Kelvin wave and the subsequent Rossby wave at 5 • N are directly connected to the moored V. However, the equatorial wave alone cannot cause such persistent and strong variation in moored V. On one hand, it normally takes 4-7 d for a Rossby wave (wavelength of 200-300 km, phase speed of 0.5 ms −1 ) to pass a mooring site (Chen et al 2017). But the Rossby wave in this study takes almost twice of that normal time (13 d during 11 July-23 July, figures 3(b)-(d)). The possible reason would be the equatorial Rossby wave is blocked by the anti-cyclonic eddy west of mooring site, which could be resulted from the interaction between equatorial and offequatorial Rossby waves as discussed in section 4.1 (figures 3(d) and (i)). On the other hand, the current velocity of equatorial Rossby wave-related anticyclonic eddy at mooring site may be strengthened by the off-equatorial Rossby wave-related cyclonic eddy above the mooring site ( figure 3(d)). It is possible because the dynamical height at the 90 • E, 5 • N and at 90 • E, 8 • N is almost out of phase (−0.75) in the surface during summer, and such relationship also exists in the subsurface layer (−0.65), indicating the persistent interaction of those two eddies.

Summary and discussions
The ISV of V in the southern BOB is studied using a mooring deployed at 5.5 • N, 90 • E from February 2020 to February 2021. During summer, a strong ISV of V, which is absent in U, is observed in the subsurface layer of 150-250 m. The amplitude and structure of V during summer are different from those of the two moored V at 5.0 • N, 90.5 • E in 2013 (Chen et al 2017) and 2016 (Huang et al 2021). Note that we have also checked the surface meridional current at 5.0 • N and the slightly difference in latitude does not affect the conclusion much. In particular, the former shows a very weak ISV of V in the subsurface layer (∼0.1 ms −1 , their figure 2(b)), while the latter shows a strong ISV in both U and V in the subsurface layer (∼0.4 ms −1 , their figure 2(c)). Huang et al (2021) proposed that a strong eastward current during summer could increase the amplitude of ISV near the mooring site. A similar analysis is also performed, and the result suggests that the zonal current from this study is much weaker than theirs, with a westward current in late summer ( figure S4). This indicates that the eastward current is not a major contributor for the strong V during summer 2020. In this study, the SSHA along 5.0 • N is not only determined by the equatorial Rossby wave, but also is influenced by the interaction between the equatorial and off-equatorial Rossby waves.
In this study, the equatorial wave during summer shows an asymmetry meridional structure of SSHA with larger magnitude at 5 • N. Based on meridional decomposition method, previous studies have found that the sum of Kelvin and first two meridional mode Rossby waves mainly determines the observed asymmetry structure of SSHA in the equatorial Indian Ocean, which could be related to the slanted coastline (Han 2011, Huang et al 2019. Same analysis has been performed to the SSHA along three sections of 85 • E, 90 • E and 93 • E on 5 July, 11 July, and 23 July 2020. Figure S5 shows that the sum of Kelvin and first three Rossby waves is in better agreement with observed SSHA along 85 • E and 93 • E, especially at 5 • N. Compared with previous studies (Chen et al 2017, Huang et al 2021, such difference in meridional mode of Rossby wave may be caused by the MISO-related Ekman pumping at 93 • E (figure 3(a) and (p)) and the off-equatorial Rossby wave at 83 • E ( figure 3(b)). Note that the decomposition method fails to reproduces the meridional structure of SSHA along 90 • E north of equator. This could be caused by two kinds of Rossby waves at 5 • N and 8 • N (figure S5(d)), which does not fit the theoretical meridional structure of equatorial Rossby wave around 5 • N. This situation cannot be simply solved by projection of wave solution, and numerical model such as reducedgravity model will be established to tackle this problem in our next study.
Based on multiyear statistics of intraseasonal SSHA near the mooring site, Chen et al (2017) identified a similar number of events with strong ISV in winter and summer, meaning that the strong ISV could be caused by either MJO or MISO. In this study, the significant MISO variability and its association with strong ISV in southern BOB suggest the possibility of a statistical relationship between them. Therefore, the multivariate MISO index (figure S1) is used to construct composite maps for the intraseasonal SSHA in the eastern Indian Ocean for eight phases. The result shows a meridional structure of SSHA similar to that of this study (figure S6). Essentially, two sets of Rossby wave-like SSHA are presented at 5 • N and 8 • N, with significant magnitude. This means that the MISO-induced local and remote impacts on the ISV in the southern BOB could be statistically significant, and how the ocean responds in patterns, strength and period under climate forcing needs further research.

Data availability statement
The authors Wang to acknowledge the OSCAR for providing the total surface current at www.oscar. noaa.gov/datadisplay/oscar_datadownload.php, the AVISO for providing the surface geostrophic current and SSHA data at www.aviso.altimetry.fr/en/ data/data-access.html, the Hybrid Coordinate Ocean Model reanalysis for temperature data at www. hycom.org, the Cross-Calibrated Multi-Platform for providing the sea surface wind vector data at http:// podaac.jpl.nasa.gov/dataset/CCMP_MEASURES_ ATLAS_L4_OW_L3_0_WIND_VECTORS_FLK, and the Global Precipitation Climatology Project for providing the precipitation data at www.ncei.noaa. gov/data/global-precipitation-climatology-projectgpcp-daily/access/.
The data that support the findings of this study are openly available at the following URL/DOI: https:// zenodo.org/record/7029285#.YwtfTHZByUk.