Intraseasonal vertical velocity variation caused by the equatorial wave in the central equatorial Indian Ocean



[1] Intraseasonal vertical velocity variation in the central equatorial Indian Ocean was investigated using observations from the field experiment “Mirai Indian Ocean Cruise for the Study of the MJO-convection Onset” and from the Research Moored Array for African-Asian-Australian Monsoon Analysis and Prediction, in October-November 2006. Using an array of four subsurface moored acoustic Doppler current profilers, we estimated vertical velocity by applying the continuity equation. Results indicated alternating downwelling and upwelling episodes at around thermocline depth, with maximum amplitudes larger than 8.0 × 10−5 m s−1, or about 7 m per day. The vertical velocity variation was mainly produced by the divergence/convergence of meridional currents, with a quasi-biweekly period of 11–16 days. The temporal changes in temperature around thermocline depth were consistent with variations in vertical velocity, whereas upwelling had less impact on the surface layer temperature. Intraseasonal variations in the ocean may be a part of biweekly fluctuations by remotely forced mixed Rossby-gravity waves, which have a meridional current maximum at the equator, accompanied by divergence/convergence in the surface layer a few degrees from the equator.

1. Introduction

[2] Most of the equatorial region of the tropical Indian Ocean is occupied by warm sea surface temperatures (SST) >28°C and includes part of the so-called Indo-Pacific warm pool. Steady easterly winds do not occur, and the ocean current system differs from those of the other two tropical oceans. One distinguishing character is the lack of climatological equatorial upwelling structured as north-south surface overturning cells [Schott et al., 2002; Miyama et al., 2003]. The thermocline ridge, an indication of upwelling, is located in the southwestern tropical Indian Ocean [Xie et al., 2002], unlike the shallow thermocline located in the central eastern part of the equatorial Pacific and Atlantic. In the tropical ocean, equatorial circulation alters horizontal and vertical heat exchange and thereby modulates SST through ocean-atmosphere interactions. However, currently, our understanding of three-dimensional circulation in the tropical Indian Ocean is based mainly on numerical models [e.g., McCreary et al., 1993; Miyama et al., 2003] because of the lack of in situ observations. Thus, a better quantitative presentation of ocean current structures in the equatorial Indian Ocean would be beneficial, even with limited data in time and space [e.g., Schott et al., 2002].

[3] Ocean surface currents show significant intraseasonal variability (ISV) in the equatorial Indian Ocean. Major atmosphere and ocean ISVs reported in past studies are (1) 30–60 day oscillations created by eastward propagating Madden-Julian Oscillations [Madden and Julian, 1972, 1994; McPhaden, 1982] or northward propagating summertime intraseasonal oscillations [Lawrence and Webster, 2002]; (2) 20–30 day variations in the western equatorial Indian Ocean caused by oceanic instabilities [Kindle and Thompson, 1989; Woodberry et al., 1989]; and (3) biweekly variability related to monsoonal wind variations [Chen and Chen, 1993; Reppin et al., 1999; Chatterjee and Goswami, 2004; Fukutomi and Yasunari, 2005]. These ISVs produce large-amplitude SST changes [e.g., Shinoda and Hendon, 1998; Han et al., 2006a; Han et al., 2007], and they can further affect interannual climate variations, such as the Indian Ocean dipole [Waliser et al., 2003; Rao and Yamagata, 2004; Shinoda and Han, 2005; Han et al., 2006b]. The ISVs can also have some effect on seasonal-to-interannual meridional heat transport in the tropical Indian Ocean [Halkides et al., 2007]. Therefore, study on the ISVs would lead to a further understanding of longer time scale phenomena in the ocean and atmosphere.

[4] This study focuses on the biweekly variability in the equatorial surface current, which is the most outstanding signal of the meridional current in the central eastern equatorial Indian Ocean. This biweekly variability has been suggested to have characteristics of the mixed Rossby-gravity wave (Yanai wave) and is generated by meridional wind stresses associated with atmospheric ISV [Reppin et al., 1999; Sengupta et al., 2001, 2004; Miyama et al., 2006; Ogata et al., 2008]. According to the theoretical solution of equatorial waves [Matsuno, 1966] and the phase speed of the low-order baroclinic mode, the wave that has meridional current on the equator with a quasi-biweekly period is the antisymmetric mixed Rossby-gravity wave. In the wave, the phase propagates westward while the energy eastward. Using an oceanic general circulation model (OGCM), Sengupta et al. [2004] reproduced the three-dimensional structure of the biweekly variation, with upwelling/downwelling on either side of the equator. They concluded that the biweekly variation is from equatorially trapped mixed Rossby-gravity waves generated by atmospheric forcing. The characteristics have been further investigated in detail using a linear ocean model and OGCMs [Miyama et al., 2006; Ogata et al., 2008]. Utilizing a linear ocean model and OGCM, Miyama et al. [2006] demonstrated that low-order baroclinic modes of mixed Rossby-gravity waves forced by intraseasonal winds and resonance created the biweekly variability in the Indian Ocean. Recently, the biweekly signals associated with the mixed Rossby-gravity waves were also observed at deeper depth (450–1000 m) in the central equatorial Indian Ocean [David et al., 2011].

[5] To observe atmospheric conditions and variability associated with intraseasonal disturbances and the resulting ocean responses in the central equatorial Indian Ocean, the Mirai Indian Ocean Cruise for the Study of the MJO-Convection Onset (MISMO) field experiment was conducted in the central equatorial Indian Ocean from late October to late November 2006 [Yoneyama et al., 2008]. Together with data from the Research Moored Array for African-Asian-Australian Monsoon Analysis and Prediction (RAMA) [McPhaden et al., 2009], intensive observation of oceanic and atmospheric variables during MISMO provided an invaluable data set for studies of air-sea interactions and associated variability on an intraseasonal time scale.

[6] A moored subsurface acoustic Doppler current profiler (ADCP) array was deployed to capture the oceanic response to atmospheric ISV during MISMO. The array consisted of three ADCPs from the MISMO and an ADCP from the RAMA, and each ADCP recorded vertical profiles of horizontal currents. The array data enabled us to estimate vertical velocity, applying a continuity equation by vertically integrating horizontal divergence. Here, we aim to elucidate wave-induced equatorial circulation, using these data to present features of intraseasonal vertical velocity variation in the central equatorial Indian Ocean.

[7] In section 2, we explain the field experiment and describe the data and method used to estimate vertical velocity. Section 3 briefly describes the major features of oceanic variations at the study site. Section 4 examines the horizontal velocity components and vertical velocity variations and confirms the validity of our vertical velocity estimate. Discussion and conclusions are given in section 5.

2. Field Experiment, Data, and Methods

2.1. MISMO Ocean Observation

[8] The goal of MISMO was to observe atmospheric conditions and variability associated with intraseasonal disturbances and resulting ocean responses in the central equatorial Indian Ocean. These observations were conducted in the Indian Ocean warm pool region where SST is greater than 28.5°C (Figure 1). At the end of the observation period, large-scale intraseasonal convective systems developed over the central Indian Ocean and moved eastward [Yoneyama et al., 2008]. For an overview of preliminary results from MISMO, see Yoneyama et al. [2008], and for ocean observations, refer to Masumoto et al. [2008].

Figure 1.

Location map of the R/V Mirai observations and mooring buoys during the field experiment “Mirai Indian Ocean cruise for the Study of the MJO-convection Onset (MISMO)” from 27 October to 21 November 2006. (left) SST during the MISMO period, taken from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) [Wentz, 1997]. (right) The triangular region surrounded by four ADCPs (0°, 79°E; 0°, 80.5°E; 0°, 82°E; and 1.5°S, 80.5°E) was used to estimate horizontal velocity divergence/convergence and vertical velocity.

[9] During MISMO, a subsurface array of upward looking subsurface ADCPs was configured in a triangle to observe vertical profiles of horizontal currents and their divergence/convergence (Figure 1). The Japan Agency for Marine-Earth Science and Technology (JAMSTEC) deployed three ADCPs at (0°, 79°E), (0°, 82°E), and (1.5°S, 80.5°E). An ADCP at (0°, 80.5°E) had been deployed previously by the U.S. Pacific Marine Environmental Laboratory (PMEL)/National Oceanic and Atmospheric Administration (NOAA) and the National Institute of Oceanography (NIO), India, in October 2004. Table 1 gives the mooring locations, periods, and properties of the instruments. Because our focus was on time scale variations greater than several days, we prepared daily averaged time series of the data from each ADCP. Ocean current meters (10 m) installed on miniTRITON (m-TRITON) [Ueki et al., 2010] and ATLAS buoys also observed surface currents near the locations of the ADCPs. Surface current data were also available at a buoy at (1.5°N, 80.5°E). We used these surface current data to interpolate the velocity profiles of ADCPs from 30 or 40 m (Table 1) to the surface, as ADCP near-surface layer data were contaminated by acoustic signals reflected at the surface. As the surface current data at 82°E had the shortest period (28 October to 21 November), the length of our analysis was 25 days.

Table 1. Specifications of ADCPs in the MISMO Array
PositionDeployed ByPeriodObserved Depth (m)BinTime ResolutionAccuracy
0°, 79°EJAMSTEC25 Oct to 24 Nov 200640–340 m8 mHourly±1% or 0.5 cm s−1 ±2°
0°, 82°EJAMSTEC27 Oct to 22 Nov 200640–340 m8 mHourly±1% or 0.5 cm s−1 ±2°
1.5°S, 80.5°EJAMSTEC26 Oct to 1 Dec 200640–270 m8 mHourly±1% or 0.5 cm s−1 ±2°
0°, 80.5°ENOAA/PMEL and NIO27 Oct 2004 to present30–180 m5 mHourly±1% or 0.5 cm s−1 ±2°

[10] Hydrographic data were also acquired by the R/V Mirai and mooring buoys during the MISMO period. The R/V Mirai maintained a stationary position at (0°, 80.5°E) from 28 October to 21 November 2006 and observed temperature and salinity using a conductivity-temperature-depth (CTD) system. The resolution of CTD casts was 1 m, with 3 h intervals. Water sampling for chlorophyll-a was also conducted once a day from 28 October to 21 November, at seven levels of vertical depth (5, 20, 40, 50, 60, 80, 100 m). Two m-TRITON buoys deployed by JAMSTEC (0°, 79°E and 0°, 82°E) and three ATLAS buoys deployed by NOAA/PMEL and NIO (1.5°S, 80.5°, 0°, 80.5°E, and 1.5°N, 80.5°E) also observed ocean temperature and salinity. Using these data, we were able to demonstrate the vertical motion of oceanic properties to evaluate our calculation of vertical velocity variations.

2.2. Estimated Vertical Velocity

[11] In general, climatological vertical velocity in the equatorial ocean is estimated at about 10−5 m s−1 or a few meters per day [e.g., Bryden and Brady, 1985; Johnson et al., 2001]. As in past studies [e.g., Halpern et al., 1989; Weisberg and Qiao, 2000; Helber and Weisberg, 2001], an indirect method of calculating from ocean divergence/convergence is appropriate to estimate small-scale vertical velocities. We estimated the vertical velocity in the MISMO triangle region using the continuity equation: ∂w/∂z = − (∂u/∂x + ∂v/∂y). The left hand side is the divergence of the vertical velocity “w,” and an upward direction is defined as positive in the z coordinate. The right hand side gives the zonal and meridional components of the horizontal divergence, which we calculated from the ADCP array.

[12] To apply the continuity equation from the surface to the subsurface, we analyzed horizontal current profile data at 1 m intervals from the surface to 180 m as follows. First, we set the 10 m current data of the m-TRITON/ATLAS buoys in the 10 m layer of ADCP profiles at the locations. Sea surface current (0 m) was assumed to be the same as the 10 m current. We discuss the interpolation error in section 2.3. Next, the profiles were interpolated vertically at every 1 m using the Akima spline method [Akima, 1970]. Then, we calculated the mass flux crossing each side of the triangle by averaging the velocity profiles and estimated horizontal divergence in the region by summing the three mass fluxes. Finally, assuming that w was zero at the surface, we integrated horizontal divergence vertically from the surface downward using the trapezoidal method, thus obtaining w as a function of depth.

2.3. Error

[13] In estimating w at a certain depth, errors in ∂w/∂z, deriving from the estimate of total horizontal divergence in each layer, accumulated from the surface to the depth. Possible sources of ∂w/∂z error include (1) measurement error in ADCP observations; (2) interpolation error, mostly in the surface layer between the surface and uppermost layer (30 or 40 m) of ADCP observations (see Table 1); (3) error from the effect of ADCP tilt; and (4) finite difference error occurring from finite samplings in space and time. We quantified the accumulation of these errors in estimating w.

[14] The nominal measurement error of each ADCP is reported to be ±1% or 0.5 cm s−1 in magnitude and ±2° in rotation for each hourly ensemble average [McPhaden et al., 2009]. This error can be estimated to be less than 1.1 cm s−1 in both zonal and meridional currents at each ADCP and every bin. We must also account for observation errors derived from various field measurement sources. To provide a measure of this value, we directly compared the ADCP data with measurements by the Sontek Argonaut current meter installed at 40 m depth at the ATLAS buoy at (0°, 80.5°E). The comparison used daily averaged values during the data-available period from 6 September to 31 December 2006 (117 points). From the root mean square (RMS) differences between the data, the observation errors in zonal and meridional currents were estimated as 4.4 cm s−1. Although this value is not the actual observation error, they are four times as large as the nominal measurement error. As a conservative estimate, we applied the 4.4 cm s−1 as the observation error.

[15] Since we used ocean current data at 10 m to estimate currents at the near-surface layer, the interpolation error is expected to be relatively small, compared with cases of linear extrapolation from values below the surface [e.g., Weisberg and Qiao, 2000; Helber and Weisberg, 2001]. However, error still remains from the surface to 10 m. To quantify the interpolation error, we carried out comparison experiments, preparing other profiles using simple linear extrapolation (0–10 m) and interpolation (10–40 m). After comparing these profiles with the present case, we estimated the interpolation errors in each layer to be less than 5.0 cm s−1 (0–10 m) and 0.7 cm s−1 (10–40 m).

[16] We also estimated the effect of ADCP tilt. The bin numbers of each set of ADCP data were stable during the observation period, indicating that the tilt of the ADCP mooring was less than 15°, which corresponds to a horizontal shift of less than 100 m. In that the distance between each ADCP was 235–333 km, this error is negligible in calculating ocean current divergence/convergence.

[17] Finite difference error occurs because data are recorded at finite intervals in space and time. While an hourly observation interval is fine enough for the time scale of the phenomena we focused on, we expected to see a problem in spatial sampling. Observations were conducted using only four ADCPs, and the spatial curvature of the phenomenon would produce error when the spatial variability of horizontal currents was large relative to the size of the array. Although a qualitative estimate of this error in our analysis was difficult, we discuss this point in section 4, with reference to our results.

[18] To estimate w, we combined observation and interpolation errors, using the sampling theory for error propagation of Emery and Thomson [1998]. We treated these errors as systematic errors. The error sources increased with depth in integrating ∂w/∂z. For example, the amplitudes of w errors at depths of 10 m, 50 m, 100 m, and 150 m were 0.6 × 10−5, 2.6 × 10−5, 5.0 × 10−5, and 7.4 × 10−5 m s−1, respectively. Although the errors are quite large, the amplitude of the errors does not negate our conclusion (see section 4).

3. Horizontal Current

[19] Here, we describe the major features of horizontal current variability observed in the MISMO region. During the MISMO period, large-scale sea surface signals from the 2006 Indian Ocean dipole (IOD) event [Vinayachandran et al., 2007; Horii et al., 2008] dominated in the tropical Indian Ocean, as demonstrated by the cooler than normal SST spread in the southeastern Indian Ocean (Figure 1). Although the Indian Ocean warm pool migrated westward, warm SST still enclosed the MISMO region. Thus, SST around the region was almost the same as that of the seasonal climatology. Easterly wind episodes, which are in the opposite direction to the climatology, appeared from mid-September to mid-November, representing the large-scale anomalous atmospheric condition during the IOD (Figure 2a).

Figure 2.

(a) Time series of QuikSCAT zonal (red) and meridional (blue) winds averaged within a 5° longitudinal and latitudinal window centered at (0°, 80.5°E). Time-depth sections of (b) zonal and (c) meridional currents observed by the ADCP at EQ, 80.5°E. Contour intervals are 10 (cm/s). Note that the color shadings are different in Figures 2b and 2c. Vertical lines indicate the MISMO period. The 10 m ocean current meter data installed at Atlas buoy at (0°, 80.5°E), available from 6 September, are used to interpolate currents from 30 m upward.

[20] The horizontal currents shown in Figures 2b and 2c are from the ADCP at (0°, 80.5°E) from September to December 2006, including the MISMO period. This ADCP had been previously deployed by NOAA/PMEL and NIO before the MISMO experiment. Similar patters of horizontal current variations were seen at other ADCPs at (0°, 79°E), (0°, 82°E), and (1.5°S, 80.5°E), although these observations were limited to the MISMO period. Notice that the color shading scale differs for zonal and meridional currents, with the amplitude of zonal current variation being roughly twice as large as that of the meridional current.

[21] Zonal current during MISMO was characterized by a westward surface current at around 0–40 m and an eastward subsurface current below (Figure 2b). The surface current was likely forced by the easterly wind above. The westward current during 23–28 September and 4 October to 27 November most likely corresponded to that of the easterly wind on 15–26 September and 3 October to 18 November, with a lag of several days. The peak correlation between the surface wind and the ocean 10 m current (r = 0.68) occurred with the current lagging the wind by 3 days. Note that with the easterly wind, the surface current flowed in the opposite direction to that of the climatological condition; in local climatology, eastward surface currents associated with westerly surface winds appear in the central equatorial Indian Ocean in this season. From late September to late November 2006, an eastward current under the westward surface current occurred in the subsurface. The eastward current, also reported by Masumoto et al. [2008], occurred at about thermocline depth, with a maximum speed of 79 cm s−1 around 31 October. The vertical structure of the zonal current during this period resembled the climatological pattern of the equatorial Pacific and Atlantic oceans. This suggests that a pressure gradient generates the subsurface eastward current, as occurs in the equatorial undercurrent (EUC) dynamics [Philander and Pacanowski, 1980]. The easterly wind, the westward surface current, and the subsurface eastward current all showed a reduction in amplitude from the beginning to the end of the MISMO period. The anomalous zonal currents and their evolutions in the peak phase of the 2006 IOD can be explained by the second baroclinic waves [Nagura and McPhaden, 2010], in which zonal currents forced by the winds then reduce the amplitude and reverse the direction, caused by Rossby waves reflected from the coasts of Sumatra, about a month after the wind forcing.

[22] To observe the general structures of zonal currents at the four ADCPs, we averaged each current time series during the MISMO period (Figure 3). Inversion structures of zonal currents from the surface westward to the subsurface eastward are clearly seen in all of the profiles. Subsurface eastward currents occurred from 50 m to 180 m in three profiles on the equator and from 70 m to 250 m in the profile at (1.5°S, 80.5°E). The off-equatorial current (1.5°S, 80.5°E) was weaker in magnitude, both at the surface and subsurface. Along the equator, the subsurface eastward currents accelerated from 79°E to 82°E, indicating zonal divergence and resembling the EUC in the central equatorial Pacific [Weisberg and Qiao, 2000].

Figure 3.

Mean zonal current profiles from four ADCPs during the MISMO period.

[23] The meridional current exhibited a strong quasi-biweekly variability from the surface to below 150 m, with upward phase propagation (Figure 2c). The southward and northward episodes during MISMO must have been part of a wave-like oscillation prevailing in the region. Local meridional winds, however, showed no corresponding variability (Figure 2a). Instead, the meridional wind exhibited a variation of less than 10 days, with a mean positive wind (southerly) before and during the MISMO period. Yasunaga et al. [2010] indicated that the short-period signal in the meridional winds came from atmospheric mixed Rossby-gravity waves. The different features between meridional winds and currents imply that the quasi-biweekly oceanic variability during MISMO was produced by remote forcing.

[24] To confirm the dominant period of meridional variability, we computed the power spectra for a 3 year time series of local meridional winds and 30–100 m mean meridional currents at (0°, 80.5°E) (Figure 4). The power spectra showed that the quasi-biweekly signal was a prominent feature of the variability both in winds and currents. The significant period of the meridional winds (currents) around the biweekly band was days 11–15 and 17–19 (11–16), which exceeded the 95% confidence level of red noise background (Figures 4a and 4b). A significant peak of squared coherency was found at 14–15 days (Figure 6c). Note that the phase difference of 54°–76° for the period corresponds to a lag of about 2–3 days for ocean currents leading winds (Figure 6d). This indicates that the meridional variability in the currents is not local responses but produced by remote wind forcing. Such variability in the meridional current with a quasi-biweekly period has been reported in past studies at (0°, 83°E) [Sengupta et al., 2004] and (0°, 90°E) [Masumoto et al., 2005; Ogata et al., 2008]. These studies proposed the cause of the quasi-biweekly period to be the equatorial mixed Rossby-gravity wave, with phase propagation from east to west. Upward phase propagation (Figure 2c) is also consistent with modeling studies of the mixed Rossby-gravity wave [Sengupta et al., 2004; Miyama et al., 2006].

Figure 4.

Variance-preserving spectra of (a) QuikSCAT meridional winds averaged within a 5° longitudinal and latitudinal window centered at (0°, 80.5°E) and (b) 30–100 m mean meridional currents observed by the ADCP at (0°, 80.5°E). A Hanning filter was applied five times to smooth the spectra. The dashed line denotes the 95% confidence limits of the theoretical red noise background (degree of freedom is 10). (c) Squared coherency and (d) phase between the meridional wind with the current. The 95% confidence limits for the coherency are shown by the dashed horizontal lines. Positive phase indicates that the currents lead the winds. The spectra are calculated for the period from 27 October 2004 to 16 August 2007 (1024 days).

[25] Profiles of the meridional currents of ADCPs are presented in each phase, owing to the variability during MISMO. Figure 5 shows the structures of meridional currents separated into three phases: 27 October to 4 November (Phase 1); 5–14 November (Phase 2); and 15–22 November (Phase 3). The phases are defined by averaging the meridional velocities in the upper 100 m. In Phase 1, southward currents dominated from the surface to 50–80 m, decelerating from the equator to 1.5°S. This indicates a meridional convergence at the depth. From Phase 1 to Phase 2, the currents transitioned from southward to northward, roughly in the layer shallower than 150 m, except for some southward flow around 0–40 m (0°, 79°E) and 60–120 m (0°, 82°E), with no clear indication of meridional divergence/convergence at this phase. Surface currents shallower than 50 m switched direction zonally from 79°E to 82°E along the equator, as seen in the surface layer and at around 100 m depth in Phase 2. In Phase 3, southward currents dominated again, from the surface to 140 m. On average, the southward current reduced speed from the equator to 1.5°S at around 50–90 m, indicating a return to convergence.

Figure 5.

Mean meridional current profiles from four ADCPs during (a) 27 October to 4 November, (b) 5–14 November, and (c) 15–21 November.

[26] Figure 6 shows each time series from the ADCPs averaged in the upper 100 m during MISMO, with longer time series at (0°, 80.5°E) and (1.5°N, 80.5°E) indicating the continual biweekly variability. The wave appears to propagate from east (82°E) to west (79°E) along the equator, at an interval of about 2 days (Figure 6a). A rough estimate assuming a 3° propagation on the equator in 2 days yielded a westward phase speed of 1.9 m s−1, which is consistent with the phase speed of mixed Rossby-gravity waves in the equatorial Indian Ocean produced in OGCMs with real wind forcing [Ogata et al., 2008], although our observation period was too short to be definitive.

Figure 6.

(a) Time series of 0–100 m mean meridional currents from four ADCPs. (b) Same as Figure 6a, but the period is from 1 September 2006 to 1 March 2007. The superimposed long time series are 30–100 m mean meridional currents from the ADCP at (0°, 80.5°E; gray line) and the 10 m meridional current from the ATLAS buoy at (1.5°N, 80.5°E; dashed line).

[27] The southward and northward episodes of meridional currents observed in the MISMO period must be a part of the quasi-biweekly oscillation (Figures 2c and 6b). Similar quasi-biweekly oscillations with same amplitude appeared in the 30–100 m mean meridional current at (0°, 80.5°E) before and after the MISMO period. This variability was also found in ocean 10 m current data at (1.5°N, 80.5°E), with a somewhat large amplitude. Note that the meridional currents have same direction on either side of the equator, as seen in the data at 1.5°S and at 1.5°N. This is consistent with the structure of currents on the antisymmetric mixed Rossby-gravity waves. Also note that the 10 m current (1.5°N, 80.5°E) and 30–100 m mean current (0°, 80.5°E) are not completely in phase. The 10 m current tended to lag several days, mainly after the MISMO period, possibly due to the upward phase propagation from subsurface to surface (Figure 2c).

[28] In section 4, we use these current profiles from the ADCP array to estimate vertical velocity and its variation.

4. Vertical Velocity

[29] In this section, we calculate the horizontal and vertical divergence of the ADCP array, and estimate w, the vertical velocity. Figure 7 shows the evolutions of ∂u/∂x (zonal divergence), ∂v/∂y (meridional divergence), and −∂w/∂z (horizontal divergence), calculated as described in section 2.2. Cool and warm colors denote divergence and convergence, respectively. In the zonal component, divergence mainly dominated from the beginning to mid-November, except in the surface layer (Figure 7a). The time-averaged structure (Figure 3) corroborates this point. The divergence peaked at about 70 m around 29 October to 6 November and then decreased to near zero around 16 November. This reduction could be closely related to the evolution of the zonal current: the zonal divergence changes its amplitude concurrent with the weakening of the surface westward and subsurface eastward currents (Figure 2b).

Figure 7.

Time-depth sections of (a) zonal, (b) meridional, and (c) vertical divergence/convergence estimated over the triangular region (see Figure 1) by the four ADCPs. Positive and negative values denote divergence and convergence, respectively. Contour intervals are 0.5 × 10−6 s−1. Calculations are limited to 28 October to 21 November because the m-TRITON buoy at (0°, 82°E) did not record 10 m current on 27 October and 22 November.

[30] The meridional components (∂v/∂y) showed alternating convergence and divergence episodes (Figure 7b), clearly reflecting the meridional current variation (Figure 2c). Convergence dominated from the beginning to 3–6 November, with upward phase propagation, consistent with the southward currents during this period. Then, divergence appeared from 3 to 16 November, with another upward propagation, concurrent with the northward current. The signal might have reached to the surface, although there was a short gap at around 30 m depth on 6–12 November, as also seen in the meridional current. These meridional convergence and divergence signals remained in the total horizontal divergence (−∂w/∂z; Figure 7c). Overall, the wave-like feature in the meridional current and its divergence was prominent in the horizontal current variability during MISMO.

[31] The w values estimated from integrating the horizontal divergence exhibit alternating downwelling and upwelling episodes (Figure 8a). The amplitude of w was large (>5.0 × 10−5 m s−1) from 50 m downward. The three phases of meridional currents described in section 3 agree substantially with the w variations: two downwelling episodes during 27 October to 3 November, followed by an upwelling episode from 4 to 10 November, and downwelling episodes during 13–19 November. The upwelling signal propagated upward ruing 5–15 November, although the feature was not as clear as that of the meridional current and its divergence. As expected in the variation in −∂w/∂z, the variation in w depended on the meridional components.

Figure 8.

(a) Time-depth section of vertical velocity estimated by the integration of horizontal divergence/convergence. Black line denotes the 20°C isotherm of the MISMO region. Gray line denotes mixed-layer depth (MLD), defined as the depth at which the potential density change from the surface is equivalent to a temperature change of 0.5°C [Sprintall and Tomczak, 1992]. (b) Time-depth sections of temperature profile (contour) and time tendency (color) (°C d−1). The profile was calculated using CTD observations from the R/V Mirai and m-TRITON/ATLAS buoys at the corners of the MISMO triangle (Figure 1). (c) Time series of vertical velocity at 20°C isotherm (blue) and vertical movement of the 20°C isotherm (red) (×10−5 m s−1). Light blue shading indicates vertical velocity error ranges.

[32] Although the period was too short to obtain a significant “mean” structure, the time-averaged w during the MISMO period showed upwelling around the thermocline depth. The mean ± one standard deviation of the w was 1.2 ± 5.4 × 10−5 at 100 m and 1.6 ± 6.4 × 10−5 m s−1 at 150 m. The mean upwelling was consistent with the local easterly wind at the period, since easterly wind forcing in the area could create equatorial upwelling through Ekman divergence in the surface layer.

[33] To assess our estimate of w, we prepared a time series of the temperature profile of the MISMO region (Figure 8b). The profile was calculated using CTD observations from the R/V Mirai and three moored buoys at the corners of the MISMO triangle (Figure 1). The deepening or shoaling of the isotherms around thermocline depth, which result in warming or cooling, respectively, is related to the vertical velocity variation. For instance, a cooling episode during 5–9 November was concurrent with upwelling. Two short warming signals around the 16°C–26°C isotherm on 28–29 October and 2–4 November were also accompanied by downwelling episodes. Notably, we obtained no clear signal of temperature related to upwelling in the surface layer. Due to salt stratification during MISMO [Masumoto et al., 2008], the mixed layer depth was relatively shallow, from 11 to 43 m. Heat flux dominated surface temperature variation (figure not shown), as will be further detailed in a future report.

[34] To expand the quantitative evaluation, we compared temperature changes at the 20°C isotherm with vertical velocity (Figures 8a and 8c). The temporal changes in the 20°C isotherm roughly agree with the vertical velocity at the depth (Figure 8c). The close correspondence between the vertical movement of the isotherm and the vertical velocity suggests that vertical and horizontal diffusion are small compared with vertical advection around the thermocline depth.

[35] Note that the vertical velocity tends to overestimate the temperature change, whereas both amplitude and its variations are comparable, on the order of 10−5 m s−1 (Figure 8c). One possible cause of the large error in the estimate of w is that velocity component gradients are not homogeneous in space. To evaluate the finite difference errors, we further estimated w in the western triangle (0°, 79°E∼0°, 80.5°E∼1.5°S, 80.5°E) and the eastern triangle (0°, 80.5°∼0°, 82°E∼1.5°S, 80.5°E). Although these estimates were not totally independent, as the ADCPs along 80.5°E appear in both calculations, the two calculations yield similar patterns of w, but with a large RMS difference of 2.12 × 10−5 m s−1 (37% of actual amplitude) at thermocline depth. In addition, the error could be caused by the meridional curvature of the observed phenomena, which could not be resolved by the ADCP array. Figures 3 and 5 clearly indicate that the horizontal currents on the equator are about twice as large as that at 1.5°S, but no data establish the spatial change between them. Given these points, the finite error should be quite large in our estimate. Nevertheless, the observed movement at the 20°C isotherm notably falls into the error range of estimated w at this depth for most of the MISMO period.

5. Discussion and Conclusions

[36] In this study, we examined the intraseasonal vertical velocity variation in the central equatorial Indian Ocean during October–November 2006, using data from the field experiment MISMO and from the RAMA mooring. Vertical velocity was estimated by a triangle array of four deployed subsurface moored acoustic Doppler current profilers (ADCPs). We found both downwelling and upwelling episodes around thermocline depth, with large amplitude exceeding 8.0 × 10−5 m s−1 (about 7 m per day) at approximately thermocline depth of the Indian Ocean warm pool region. Meridional currents exhibited strong quasi-biweekly oscillation from the surface to 180 m, with upward phase propagation, with the meridional divergence/convergence being the main producer of the vertical velocity variation. Considering the period and property of the observed signals, we suggest that the vertical velocity variation was driven by the mixed Rossby-gravity waves prevailing in the equatorial Indian Ocean.

[37] To assess whether the mixed Rossby-gravity waves can have sufficient amplitude to generate the vertical velocity variation, we examined the properties of waves appearing in a shallow water model [Matsuno, 1966]. Past studies have reported that biweekly oscillations in the warm pool region are due to mixed Rossby-gravity waves of the first and/or second baroclinic mode [Zhu et al., 1998; Sengupta et al., 2004; Ogata et al., 2008]. Miyama et al. [2006] showed that low-order baroclinic modes of mixed Rossby-gravity waves forced by intraseasonal winds and resonance can create the biweekly variability in the Indian Ocean. In our case, following Miyama et al. [2006], we adopted the phase speed of the first (second) baroclinic mode of 2.82 (1.75) m s−1 to check the wave structures. In the theoretical dispersion relationship [Matsuno, 1966], the wavelength of the first (second) baroclinic mixed Rossby-gravity wave with a 14 day period is about 2450 (4370) km, with a westward phase speed of 2.0 (3.6) m s−1. In the wave, a divergence and convergence pair occurs on both sides of the equator. When the peak northward current is 50 cm s−1 on the equator, as in observed currents (Figure 2), the wave of the first (second) baroclinic mode has a peak divergence of 3.39 (6.93) × 10−7 s−1 around 2.5°S (2.3°S). This can be converted to a vertical velocity of 4.1 × 10−5 m s−1 in the first mode and 8.3 × 10−5 m s−1 in the second mode at 120 m depth. These theoretical values are roughly consistent with the observed values calculated from the ADCP array (Figure 8), although the array (0°∼1.5°S) is rather closer to the equator, compared with the theoretical divergence peak.

[38] For further evidence of mixed Rossby-gravity waves, we also analyzed data from an upward looking ADCP at (0°, 90°E) which was located about 10° east of the MISMO region and deployed by JAMSTEC from November 2000 to March 2009 [Masumoto et al., 2005]. The surface meridional current had a strong signal representing mixed Rossby-gravity waves [Ogata et al., 2008]. We examined the phase relationship between 40 and 100 m averaged meridional currents of the ADCP (0°, 90°E) and that at (0°, 80.5°E) from October 2004 to August 2008. When the meridional current at 90°E leads that at 80.5°E by 3.5 days, the currents have a maximum correlation (0.37) that is statistically significant at the 99.9% level. Given the distance of 1055 km between 80.5°E and 90°E, the westward phase speed expected from the observation is 3.5 m s−1, which differs from the phase speed presumed at the ADCP array (1.9 m s−1; see section 3), although each phase speed can be explained by the first and second baroclinic waves. We also find a maximum correlation (0.41) when the current at 80.5°E leads that at 90°E by about 10 days. Considering the biweekly period and westward phase propagation, the 10 day lag can be interpreted as the time between one crest at 80.5°E and the next crest at 90°E. The higher correlation of 0.41 may imply energy propagations from west to east, as this is a property of biweekly mixed Rossby-gravity waves. In addition, data from both ADCPs show a temporally upward phase propagation feature, consistent with the wave feature found by Miyama et al. [2006]. Finally, we note that the waves are possibly forced by the meridional wind. The propagation of northward wind events from the east (about 100°E) to west (about 70°E) around the MISMO period is in line with the biweekly atmospheric variation explained by Fukutomi and Yasunari [2005].

[39] The impact of the vertical variation in SST, and hence its possible influence on ocean-atmosphere interactions, is unclear. The equatorial wave changed the temperature around the thermocline depth at an amplitude of about 1.0°C/d, whereas vertical velocity variation appeared to have less impact on surface layer temperature (Figure 8). The mixed layer during MISMO was relatively shallow (11∼43 m; 22.9 m on average), partly due to the density stratification of the so-called barrier layer. Temperature related to upwelling in the surface layer showed no clear signal. We also examined SeaWiFS satellite data of chlorophyll-a at the sea surface during MISMO (, but found no signal consistent with the vertical velocity variation in this study. One yet-to-be-demonstrated but exciting hypothesis is that the biweekly variation might have an impact on SST at the eastern edge of the Indian Ocean (off the coast of Sumatra), when the thermocline is shallow enough to influence the mixed layer, as in the case of a positive Indian Ocean dipole event. We need further research to verify this point.

[40] We have suggested that the mixed Rossby-gravity wave produces vertical motion in the thermocline, as well as controls subsurface temperature. Whereas this analysis focused on a short-term field experiment, our next goal in this observational study will be to diagnose the long-term mean of equatorial three-dimensional circulation in the Indian Ocean. This will further contribute to validating and advancing coupled ocean-atmosphere models for climate prediction. To this end, we need to keep constructing, maintaining, and properly extending ongoing Indian Ocean observation systems [McPhaden et al., 2009], which are truly effective in understanding and predicting the Indian Ocean climate.


[41] We thank the officers and crew of the R/V Mirai and the data processing team for cruise operations and data management efforts related to the MISMO project. Special thanks go to chief scientist Kunio Yoneyama. Yasushi Ishihara, Takeo Matsumoto, Masayuki Yamaguchi, and Nobuhiro Fujii greatly contributed to the development, deployment, and data management of the m-TRITON buoy. We also thank NOAA/PMEL and NIO for providing ADCP and ATLAS buoy data. TMI SST data and QuikSCAT are produced by Remote Sensing Systems and sponsored by NASA. These data are available at We thank two anonymous reviewers who provided helpful comments. Comments and suggestions from Michael J. McPhaden, Kelvin Richards, and Hideaki Hase are also acknowledged. This study was partly supported by the Japan Earth Observation System (EOS) Promotion Program, sponsored by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan. S.P.K. acknowledges NIO, Goa, and Council of Scientific and Industrial Research (CSIR), New Delhi, for the encouragement and support. NIO contribution number is 4991.