Dynamics of the intraseasonal oscillations in the Indian Ocean South Equatorial Current

Transcription

1 Dynamics of the intraseasonal oscillations in the Indian Ocean South Equatorial Current Lei Zhou and Raghu Murtugudde Earth System Science Interdisciplinary Center, College Park, Maryland Markus Jochum National Center for Atmospheric Research, Boulder, Colorado Lei Zhou Address: Computer & Space Sciences Bldg. 33, University of Maryland, College Park, MD 74 Phone:

2 Abstract The spatial and temporal features of intraseasonal oscillations in the southwestern Indian Ocean are studied by analyzing model simulations for the Indo-Pacific region. The intraseasonal oscillations have periods of 4-to-8 days, with a wave length of ~65 km. They propagate westward as Rossby waves, with a phase speed of ~5 cm s -1 in boreal winter and spring. The baroclinic instability is the main driver for these intraseasonal oscillations. The first baroclinic mode dominates during most of the year but during boreal winter and spring, the second mode contributes significantly and often equally. Consequently, the intraseasonal oscillations are relatively strong in boreal winter and spring. Whether the atmospheric intraseasonal oscillations are also important for forcing the oceanic intraseasonal oscillations in the southwestern Indian Ocean needs further investigation.

3 1. Introduction: The distinguishing feature of the Indian Ocean is the seasonally reversing Indian monsoon, especially in the northern Indian Ocean where the winds are generally southwesterly in boreal summer and northeasterly during boreal winter. Driven by the monsoons, both ocean circulation and sea surface temperatures (SSTs) in the northern Indian Ocean display a dominant semi-annual variability. The monsoon driven circulation and related climate variability have been reviewed recently in Schott and McCreary (1) and Annamalai and Murtugudde (4). Oceanic intraseasonal oscillations (OISOs) and internal variabilities are found to be important part of the ocean dynamics and heat budget in the mixed layer. They can affect the ocean-atmosphere interactions by changing the local SSTs and the horizontal heat transports due to non-linear effects (Waliser et al. 3, 4; Jochum and Murtugudde 5). They can also influence the atmospheric boundary layer by modifying surface heat fluxes and the deep convection in the tropical atmosphere (Thum et al. ; Raymond et al. 4). In the northern Indian Ocean, OISOs in the Bay of Bengal during 1998 summer monsoon were detected from in-situ observed SSTs and were shown to cause differences as large as C between the in-situ observations and standard SST reanalysis products (Sengupta and Ravichandan 1). In the tropical Indian Ocean, there is strong mesoscale variability due to Ekman pumping and the thermocline variability. In addition, the main mode of atmospheric intraseasonal oscillations (AISOs), viz., Madden-Julian Oscillations (MJOs) propagate eastward in the tropics during boreal winter, engendering OISOs in the equatorial Indian Ocean as a response (Waliser et al. 3). Reppin et al. (1999) used the moored ADCP data to estimate that OISOs in the equatorial Indian Ocean 3

4 were driven by the local winds with a similar period. Sengupta et al. (1) used a model driven by NCEP data to show that the 3-5 day oscillations appeared due to the westward propagating Rossby waves and were enhanced due to the oceanic instability in the western equatorial Indian Ocean. As for the southern Indian Ocean, there have been some observational and modeling studies in the southeastern Indian Ocean (SEIO). The OISOs in SEIO are closely related to the intraseasonal variation of the Indonesian through flow (ITF, Potemra et al. ) with a close relationship to OISOs in the western Pacific. Moreover, because ITF is intensified near the thermocline (Song and Gordon 4), it generates strong vertical velocity shears. As a result, baroclinic instability is another important driver for OISOs in SEIO (Feng and Wijffels ) with the barotropic instability presumed to be insignificant. Yu and Potemra (6) found that the OISOs in the Indo-Australian basin are strong in the second half of the year but the barotropic and the baroclinic instabilities contribute almost equally to their genesis. Schiller and Godfrey (3) simulated OISOs in the southern Indian Ocean with an OGCM and showed that the local surface heat flux of an intraseasonal event is an important part of the heat budget in the mixed layer. They found that due to mean temperature inversion in the subsurface, the entrainment can warm the sea surface rather than cool it. Using TOPEX/POSEIDON data, Iskandar et al. (5) found intraseasonal sea level variability along the southern coast of Sumatra and Java, which were attributed to both remote winds and the equatorial Kelvin waves (also see Sprintall et al. ). Iskandar et al. (6) analyzed a high-resolution OGCM simulation to show that the first two baroclinic modes are the leading modes in the South Java Coastal Currents. In summary, OISOs in SEIO are subject to the combined influence 4

5 of the AISOs and the planetary waves in the ocean. Generally, there are two reasons for the genesis and evolution of OISOs; one is external forcings and the other is internal processes. External forcings mainly consist of the AISOs, precipitation and river run-offs in the intraseasonal band. Internal processes mainly include local barotropic and baroclinic instabilities, as well as the effects of the instabilities radiated from a remote ocean region. Of course, the observed OISOs are often a result of the combination of all of the above. For example, in the Pacific Ocean, Kessler et al. (1995) confirmed with a coupled model that OISOs are a delayed response to the AISOs. In the eastern Pacific Ocean, from 9 o -13 N, the observed OISOs are caused by baroclinic instability and propagate westward as the baroclinic Rossby waves. OISOs in this latitude band can also be triggered by the mountain gap winds in the eastern equatorial Pacific Ocean (Farrar and Weller 6). Han (5) showed with HYCOM (HYbrid Coordinate Ocean Model) results that the peaks at intraseasonal periods (-9 days) in the equatorial Indian Ocean are directly caused by the response to 9-day oscillation in the winds, and enhanced by the Rossby waves. In the Atlantic Ocean, the waves in the intraseasonal band are found in the North Brazil Current rings in the high-resolution OGCM, caused by the barotropically unstable North Equatorial Counter Currents (Jochum and Malanotte-Rizzoli 3). The OISOs in Bay of Bengal during 1998 summer were attributed to the riverine and precipitation-related fresh water forcing (Sengupta and Ravichandan 1). To our knowledge, there have been neither systematic observations nor modeling studies of OISOs thus far in the southwestern Indian Ocean (SWIO) which is the upwelling region in the southern Indian Ocean (e.g., Murtugudde and Busalacchi 1999). 5

6 It has large seasonal-to-interannual variations in SSH and SST due to the wind-stress curl variability between the monsoons to the north and the stable southeasterly trade winds to the south (Annamalai et al. 3). Moreover, SWIO is the western half of the Indian Ocean Zonal/Dipole mode (Murtugudde et al. 1998; Saji et al. 1999; Webster et al. 1999; Murtugudde et al. ). The variations in SWIO greatly influence rainfall over Eastern Africa and may also have a feedback to ENSO in the Pacific Ocean (Xie et al. ; Annamalai et al. 5). Therefore, OISOs in SWIO need to be better understood to better exploit the predictability of Southeast African climate. The purpose of this paper is to describe the spatial and temporal dynamic properties of OISOs in SWIO, and then to provide the dynamic reasons for their development. Although the thermodynamic properties, especially the SSTs, are critical to the ocean-atmosphere interactions, they are not the focus of this paper but will be reported in our future work. In Section, the model is described and the model results are compared with the observations. In Section 3, the features of OISOs in SWIO are described. In Section 4 the dynamic reasons for their genesis and enhancement are explored by analyzing the stability and the vertical modes of the ocean currents. The conclusions and discussions are presented in Section 5.. Model description and comparison The model used in this study is a reduced gravity, sigma-coordinate, primitive equation OGCM, with a horizontal resolution of 1/3º in latitude and 1/º in longitude over the Indo-Pacific domain covering 3 o E-76 o W, 3 o S-3 o N (Murtugudde et al. 1996, 1998). There are 15 sigma layers in the vertical below the variable depth mixed layer with a resolution of ~ m in the thermocline in the SWIO, so that the vertical oscillation in the interior ocean can be adequately resolved. Surface mixed layer is determined by 6

7 hybrid mixing scheme of Chen et al. (1994) which explicitly accounts for the entrainment due to surface turbulent kinetic energy, shear-induced dynamic stability mixing, and convective mixing to remove static instabilities. The last sigma-layer is a prognostic variable whereas the other sigma layers are specified constant fractions of the total depth below the mixed layer to the motionless abyssal layer. The model is driven by the climatological weekly NCEP Reanalysis winds (see Murtugudde et al. for details). The model outputs for all the analyses presented here are weekly mean fields from the last years of a 7 year simulation. Note that we only focus on OISOs in the climatological run to avoid contributions from interannual forcings (also see Jochum and Murtugudde 5)..1 Model-data intercomparisons The OGCM has been reported in many previous applications demonstrating its ability to simulate the ocean dynamics and thermodynamics reasonably well in the tropical oceans. It resolves all major features of the surface ocean currents (defined as a mean over the variable depth mixed layer), which are described below. Since the stable southwesterly trade winds are dominant, the westward south equatorial current (SEC) stretches from 8 S to S. When SEC reaches the east coast of Madagascar, it splits into two branches at about 18 S. One branch flows northward, crosses the northern tip of Madagascar, and then merges into the western boundary current. During boreal summer, part of the Somali current turns southward at around 5 S, returning to the SEC and forming a cyclonic gyre there. The remaining SEC continues northward across the equator to join the Somali current, but goes unstable as low-latitude jets are wont to do merging into the eastward Equatorial Current (Nof and Olson 1993), with some of this 7

8 water returning southward via the unstable equatorial roll (Miyama et al. 3). The second branch of SEC turns southward along the eastern coast of Madagascar and most water in this branch leaves the Indian Ocean through the southern boundary, while small part of water goes into eastern gyral current (EGC). All of the features resolved in our model are consistent with the observations in Schott et al. (). Obviously, what we are concerned most with is the model s ability to resolve OISOs in the Indian Ocean. Since we intend to study the internal intraseasonal variability in the ocean, the model is driven by the climatological winds as in Jochum and Murtugudde (5). This also means that we do not expect perfect matches between the model results and interannual observations. The root-mean-square (RMS) of the intraseasonal SSHAs, which are high-pass filtered with a cut-off frequency of 1 days, and the uncertainty of its annual mean values from the model simulation and the satellite observations are shown in Fig. 1. The eddy kinetic energy (EKE) as determined from SSHA is well captured in the model, especially the zonal bands of high EKE at 5 S and 1 S. Moreover, the high EKE along the Somali coast and in the southern Bay of Bengal are also resolved. The main shortcomings of the model are the too weak EKE along the Sumatra coast and along 1 S west of 8 E.. Comparisons with previous studies Since there are no systemic observations of OISOs in SWIO, the intraseasonal features of the model results are compared with the previous observations reported in Feng and Wijffels (, FW hereafter), which examined the intraseasonal variability in SEIO by analyzing TOPEX/POSEIDON altimeter data and WOCE data. The STDs of SSHAs are large in boreal autumn and winter, but small in the other two seasons. This 8

9 seasonal variability is also obvious in the wavelet spectra of SSH at 1 S, 11 E (Fig. ), with high energy for 4-to-8 day period. The zero-lag spatial correlation of high-pass filtered SSHA with that at 1 S, 11 E are shown in Fig. 3. This is the same location that was examined in FW, hence direct comparisons are possible. There is an obvious northwestward trend of the correlation pattern, indicating a northwestward propagation of wave. The trend is steep to the east of 11 E but flat to the west of 11 E. The wave length (distance between two maxima in positive correlations) is about 7 km to the east of 11 E and extends to about 1 km to the west of this longitude. These estimations of wave length are almost identical to the observations of FW, although there is a large uncertainty band for both altimeter observations and model results. Therefore, with these comparisons between the model data and observations in SEIO, we conclude that the model ably resolves both the spatial and temporal properties of OISOs in the Indian Ocean. 3. Properties of ISOs in the SWIO All calculations that follow are based on the weekly-mean model outputs. Before describing the properties of OISOs in SWIO, we would like to confirm that they are a significant part of the variability in the SWIO. The STDs of the intraseasonal SSHAs account for %-4% of the total STDs of SSHAs in the zonal belt. The maximum ratio can be as large as 5%. Therefore, OISOs are indeed a non-negligible component of the variability in the region and are important for understanding the oceanic processes in the SWIO. Fig. 4 shows the wavelet spectrum of SSHAs at 6 S, 66 E. This point is chosen because the STDs of the intraseasonal SSHAs at this point are large. The obvious peak is 9

10 from 4 days to 9 days, centered at 8 days. The OISOs in SWIO strengthen in boreal winter and spring. Their autocorrelation length is ~65 km (Fig. 5). There is a westward propagation, which is more clearly shown in Fig. 6. After about 4 days, the SSHAs are significantly negatively correlated with themselves. This indicates a main period of 8 days, which is consistent with the high values in the wavelet spectrum (Fig. 4). The westward phase speed is estimated to be 5 cm s -1 in boreal winter and spring (Fig. 6). For example, there is a strong positive signal starting from 77 E in early November and ending at 65 E in the late December of the first year. The phase speed reduces to about 17 cm s -1 in other seasons, as seen by a positive anomaly at 8 E in late March of the second year reaching 6 E in late August. The seasonal change of phase speed is attributable to the variability of ocean stratification. According to the linear theory of = d Rossby waves, the phase speed is c β /( k + L ), where β is the planetary vorticity gradient, k is the zonal and meridional wavenumber respectively, NH L d = f is the Rossby radius of deformation, N is the Brunt-Väisälä frequency, H is the scale height, and f is the Coriolis parameter. In boreal winter, the warm advection from the SEIO (mainly caused by ITF) reaches the SWIO, leading to the strong stratification (Song and Gordon 4), i.e. large N. As a result, the phase speed in boreal winter is larger than it is in other seasons. We also note that the phase speed changes with longitude. For instance, there is a negative signal around 75 E in February of the second year which speeds up at about 67 E in May. This phase speed change is caused by the interactions between the local forcing and the Rossby waves (Wang et al. 1). The range of estimated phase speed is comparable to the theoretically expected range. In addition, the dispersion relation of the westward propagation signals is consistent with that of the linear theoretical Rossby 1

11 waves (Fig. 7). The differences between model results and theoretical results may be reduced by taking the meridional wave number or finite depth effect into account. In summary, it can be concluded that the westward propagations of the signals are caused by the Rossby waves (Masumoto and Meyers 1998). 4. Dynamic processes responsible for OISOs in the SWIO Both external forcing and internal instability are explored in this section, to determine the dynamic reasons for OISOs in the SWIO. High frequency wind stresses are one of the leading candidates. However, the wind stresses used to force the model do not have pronounced peaks in the intraseasonal band (Fig. 8); it is close to a red noise instead. Thus the peaks in wind stress spectrum and the ocean current spectrum do not match each other. In addition, the low-frequency currents are significantly correlated with the geostrophic flow, which indicates that they are generally driven by the winds and satisfy the Sverdrup relation. On the other hand, the correlations between the intraseasonal currents (i.e. OISOs) and the geostrophic flow are statistically negligible over the entire south Indian Ocean indicating that the OISOs in SWIO in our model simulation are not caused by the local response to external wind forcing. Naturally, we assume that OISOs in SWIO are attributable to the internal instabilities of the ocean currents. The prerequisite for the instability analysis is the determination of the basic flow (Pedlosky, 1987), where by definition, the basic flow is the flow that is not disturbed by the eddy flow. Thus it is not simply equivalent to the mean flow. In the model, the low-pass filtered velocity has a large STD for the whole period of years, but small STDs in each month. On the contrary, the high-pass filtered velocity has large STDs in each month, but a relatively small STD for the year period. 11

12 Since the model is driven by the climatological winds, the STD of the wind stresses is large while the wind stress in each month is identically the same (STD of interannual variability is zero). The low-frequency velocity is mainly a response to external wind forcing, hence it has a large annual cycle but changes very little in the same month of different years (interannual variability is small; see Jochum and Murtugudde, 5). However, the high frequency velocity is mainly caused by the internal oceanic processes, especially eddies, hence changes significantly in the same month of different years, even though the wind forcing is the same. With these considerations, the low-pass filtered velocity with a cut-off period of 15 days is considered to be the basic flow. 4.1 Barotropic and Baroclinic Instabilities in the SWIO dq U The necessary condition for barotropic instability is that = β, where q is dy y the quasi-geostrophic potential vorticity (QGPV) and U is the basic flow, should change dq sign in the domain under consideration. Fig. 9(a) shows in the southern Indian Ocean dy dq in November. Although OISOs should be large in November (Fig. 4), is almost dy U always positive in SEC, because the relative vorticity gradient y is too small to dq overcome the planetary vorticity gradient β. Since the condition that change sign is a dy necessary condition for barotropic instability, it is very unlikely that the barotropic instability can be triggered in SWIO, because other seasons have even weaker SEC and a weaker relative vorticity gradient. 1

13 13 The necessary condition for baroclinic instability is also that the meridional gradient of QGPV, viz. dy dq should change sign in the domain. However, in the baroclinic situation, = z U N f z y U dy dq β. Fig. 9(b) shows the distribution of the annual mean dy dq in the SWIO. Since the areas with positive and negative signs are similar and the positive and negative magnitudes are comparable, the baroclinic instability is very likely to occur in the SWIO. However, since satisfying the necessary condition alone cannot guarantee the triggering of an instability, the energy conversions are examined with the model fields below. The energy conversions caused by the barotropic and the baroclinic instability are calculated with the equation in Weisberg and Weingartner (1988), ( ) [ ] ( ) ( ) ( ) ( ) ( ) ( ) [ ] ( ) ) ( ) ( ) ( / ) ( ) ( / / e p w p v p u d v u w v u v v u u c v u g b wv v wu u U V v u vv v uu u a g v u Dt D z y x z y x z y x z z y x y x z = + + ρ ρ ρ ρ ρ ρ ρ ρ ρ ρ where term b stands for the energy conversion caused by barotropic instability and term c represents the energy conversion due to baroclinic instability. Because the vertical velocity in the model results may have large uncertainties, and horizontal gradient of horizontal velocities are the main cause of the barotropic instability, the last two terms are neglected in term b. Thus, the barotropic (EBT) and baroclinic (EBC) energy conversions are calculated by

14 EBT = ρ EBC = g [ < u u > U x+ < u v > ( Vx + U y ) + < v v > Vy] [ < u ρ > ρx+ < v ρ > ρ y ]/ ρz The barotropic energy conversion is small in the southern Indian Ocean, as shown in Fig. 1(a). South of 1 S, the barotropic energy conversion is almost zero. Between 5 S and 1 S, the barotropic energy conversion is slightly larger, because the horizontal shears are large. However, the barotropic energy conversion is always smaller than 1-4 W m -3, since the zonal shears are not very strong. The baroclinic energy conversion are at least orders of magnitude larger than the barotropic energy conversion (Fig. 1b), and can be as large as W m -3. Thus, the energy conversion between eddy flow and mean flow in the southern Indian Ocean should be dominated by the baroclinic energy conversion. This is consistent with the above calculations of the necessary conditions for the instability that the barotropic instability is not likely to contribute, while the baroclinic instability dominates in the SWIO. In order to show the relations between baroclinic energy conversion and OISOs in SWIO, the zero-lag correlation between the two are presented in Fig. 11. Meridional velocities are chosen because they are less contaminated by the low-frequency oscillations than the zonal currents. Actually, a similar pattern can be obtained by calculating the correlations between the baroclinic energy conversions and the intraseasonal zonal currents (not shown). Though the correlations are not high in SWIO, they are statistically significant and much higher than the correlations between intraseasonal currents and wind stresses. Therefore, the oceanic internal process, viz., baroclinic instability, is the main dynamic reason for OISOs in the SWIO in our model. 4. Seasonality of Baroclinic Energy Conversion and OISOs 14

15 Monthly mean baroclinic energy conversion for four months representing the four seasons are shown in Fig. 1. In SWIO, the seasonal changes are pronounced. In April and July, the baroclinic energy conversion is small with only a few patches of moderate conversion rate of 1-3 W m -3. At the end of boreal summer, the conversion rates begin to increase and in October, they are larger than 1-3 W m -3 in almost the entire SWIO exceeding W m -3 in some areas. During boreal winter, the baroclinic conversion rates decrease, but there are still large areas with conversion rates of 1-3 W m -3. The seasonal changes in Fig. 1 are consistent with the seasonal changes of intraseasonal zonal currents shown in Fig. 4 which confirm that the strong OISOs in the SWIO are caused by baroclinic instability. Figure 13 shows the longitude-depth section of baroclinic energy conversion, averaged between 8 S to 1 S for the four months. The conversions in SWIO are strong in January and October, but negligible in April and July, which is consistent with the seasonal variability shown in the horizontal plain (Fig. 1). However, the strong conversions in the SWIO are not at the surface, but in the depth range from 5 m to 1 m, which coincides with the upper part of the thermocline. This thermoclineintensification is examined further in Section 4.3. Generally, baroclinic instability is mainly caused by the vertical velocity shear. Fig. 14 shows the vertical velocity shears around the thermocline (~1 m) where they reach maximum, along with the corresponding baroclinic energy conversions. In boreal winter and spring, the vertical velocity shear increases due to the thermocline-intensified zonal advection from the SEIO (see Section 4.3). Thus, the strong baroclinic energy conversion is caused by a larger density gradient around the thermocline and by the vertical shear 15

16 between the SEC and the eastward currents underneath. Moreover, the patterns in Fig. 14 also show a similar westward propagation as in Fig. 6, especially the vertical velocity shear from 8 E to 7 E. This indicates that the large vertical shears around the thermocline are caused by the strong currents propagating from the SEIO. 4.3 Dominant baroclinic modes In order to explain why the baroclinic energy conversions or the vertical shears of horizontal currents are intensified around the thermocline (Fig. 13), a modal decomposition method is applied. The horizontal velocities, u and v, are decomposed into vertical modes, where n is the mode number, u n and vn u v N ( z t; x, y) = u ( t; x, y) ( z; x, y), ψ, n= 1 N n ( z t; x, y) = v ( t; x, y) ( z; x, y), ψ, n= 1 ψ n is the vertical eigenfunction. Following Gill (198), ψ n with boundary conditions d dz 1 N dψ n n n are the coefficients of mode decomposition, and ( z) ψ ( z) n n + = dz cn, satisfies dψ n dz ( z) dψ ( z) = z = z= H n dz =, where N is the Brunt-Väisälä frequency, cn is the phase speed for mode n, and H is the water depth. The vertical profile of Brunt-Väisälä frequency and the first three baroclinic modes at 1 S, 6 E are shown in Fig. 15. The first baroclinic mode reaches maximum at the surface and has a zero crossing at around 1 m, which is the position of the 16

17 thermocline in our model, while the second baroclinic mode reaches maximum around 1 m. The projections of the intraseasonal zonal currents onto the first two modes are shown in Fig. 16. The motions in the first mode are strong in the layers above the thermocline (around 1 m). Generally, they are larger than the motions in other modes. Only during boreal winter and spring, the motions in the second mode are comparable to the first mode. In most areas in SWIO, the first mode explains more than 5% of the intraseasonal variability while the second mode explains about 4% of the total variability. However, it turns out that the currents of the second baroclinic mode are mainly responsible for the enhancement of the velocity shear around the thermocline, i.e. for the genesis of OISOs in the SWIO, as shown below. The baroclinic energy conversion caused by various baroclinic modes is calculated. If the -year average is taken, the baroclinic conversion of the first mode dominates (Fig. 17). However, as shown above, the OISOs in SWIO are strong in boreal winter and spring. If the average in December is examined, one would find that the second baroclinic mode is the dominant cause for the OISOs, because the baroclinic energy conversion caused by the second mode is a slightly larger than that of the first mode (Fig. 17). Therefore, the intensification of baroclinic energy conversion around the thermocline (Fig. 13) is caused by the second baroclinic mode currents (Fig. 16). Furthermore, enhancement of the currents of the second baroclinic mode indicates the relation between the OISOs in the SWIO and the SEIO. Iskandar et al. (6) found that the OISOs forced by winds over SEIO off Sumatra were mainly captured by the second baroclinic mode. These wind-driven OISOs propagate westward as baroclinic 17

18 Rossby waves (Masumoto and Meyers 1998; Xie et al. ), leading to baroclinic instability in the SWIO. Thus the reinforcement of OISOs in SWIO should be significantly influenced, or even determined, by the OISOs in the SEIO. The in-depth analyses of these remote forcings are part of our future study but beyond the scope of this paper. 5. Conclusions and discussions The OISOs in the SWIO are studied by analyzing a model simulation to demonstrate that they have a period of 4 to-8 days and propagate westward as Rossby waves, with a phase speed of about 5 cm s -1 in boreal winter and spring and a wave length of nearly 65 km. The main dynamic reason for OISOs in our model is the baroclinic instability caused by the vertical shear and the density gradients. They are strong in boreal winter and spring, due to the second baroclinic mode waves propagating from the SEIO. It is indicated that the OISOs in the SWIO have a close relation with those in SEIO. According to our model, the baroclinic instability is confirmed to be important for OISOs in the SWIO even though we cannot completely exclude the effect of AISOs on the OISOs. The first reason is that in the wavenumber spectrum, the high-frequency part of the NCEP Reanalysis data decreases much faster than the satellite observations. Thus the NCEP wind data may not be able to resolve the wind fields with horizontal scale smaller than 8-1 km (Milliff et al. 4). Secondly, parts of the AISOs are excluded in the climatological winds, even though we use weekly mean winds. Therefore, by comparing the model results driven by the interannual wind stresses and the present model results, the effects of the external forcing and the internal instabilities on OISOs in the SWIO are assumed to be separable for evaluation, which will be reported elsewhere. 18

19 In the SWIO, ocean dynamic processes contribute more to the SST variability than the surface heat fluxes do (Klein et al. 1999; Chambers et al. 1999; Murtugudde and Busalacchi 1999). Xie et al. () pointed out that the SST variability in SWIO is not locally forced but due to the Rossby waves that propagate from SEIO. Since OISOs in SWIO are enhanced due to the second baroclinic mode waves propagating from the SEIO, they are potentially critical to the local heat budget. As for the local influence, some recent studies have shown that the meso-scale eddies in the intraseasonal band can significantly contribute to the energy budget of the low-frequency variability (e.g., Jochum and Murtugudde 5). For example, in the Atlantic Ocean, the barotropically unstable North Equatorial Current carried South Atlantic water into the subtropical gyre by shedding nonlinear eddies (Jochum and Malanotte-Rizzoli 3). In the Pacific, due to nonlinear dynamics, the influence of intraseasonal variability on the mixed layer heat budget is non-negligible even under seasonal forcing (Jochum et al. 4). Jochum and Murtugudde (5) argued that the internal variability with a period of 1 days is important and may be critical for predicting SSTs in the Indian Ocean, because it may significantly affect the Indian monsoon. In addition to the modification of the local heat budget, OISOs also have an impact on the onset of ENSO (Slingo et al. 1999) and have feedbacks to MJOs (Waliser et al. 1999). Therefore, it is reasonable to assume that OISOs in the SWIO may have remote effects on the low-frequency variations, such as Indian monsoons and ENSO, adding to the growing evidence for the importance of this region to regional climate variability. The thermodynamic properties of the OISOs and their feedbacks to the atmosphere are a focus of our future explorations. 19

20 Acknowledgements This work was supported by NASA Indian Ocean Mesoscale Funding and NOAA Coupled Mesoscale funding. We deeply appreciate the assistance of James Beauchamp and Eric Hackert for preparations of the model data, as well as Dudley Chelton and Ernesto Munoz for helpful suggestions. We also thank two anonymous reviewers for their comments and suggestions.

21 Reference Annamalai, H., P. Liu, and S. P. Xie, 5: Southwest Indian Ocean SST variability: Its local effect and remote influence on Asian Monsoons. J. Clim., 18, Annamalai, H. and R. Murtugudde, 4: The role of the Indian Ocean in regional climate variability. Air-Sea Interactions and Climate, AGU Monograph. Annamalai, H., R. Murtugudde, J. Potemra, S. P. Xie, P. Liu, and B. Wang, 3: Coupled dynamics over the Indian Ocean: spring initiation of the Zonal Mode. Deep-sea Res. II, 5, Chambers, D. P., B. D. Tapley, and R. H. Stewart, 1999: Anomalous warming in the Indian Ocean coincident with El Nino. J. Geophys. Res., 14, Chen, D., L. M. Rothstein and A. J. Busalacchi, 1994: A hybrid vertical mixing scheme and its application to tropical ocean models. J. Phys. Oceanogr., 4, Farrar, J. T. and R. A. Weller, 6: Intraseasonal variability near 1 N in the eastern tropical Pacific Ocean. J. Geophys. Res., 111, doi:1.19/5jc989. Feng M. and S. Wijffels, : Intraseasonal Variability in the South Equatorial Current of the East Indian Ocean. J. Phys. Oceanogr., 3, Gill, A. E., 198: Atmosphere-Ocean Dynamics, Academic Press, New York, NY. Han W., 5: Origins and Dynamics of the 9-Day and 3 6-Day Variations in the Equatorial Indian Ocean. J. Phys. Oceanogr., 35, Iskandar, I., W. Mardiansyah, Y. Masumoto, and T. Yamagata, 5: Intraseasonal Kelvin waves along the southern coast of Sumatra and Java. J. Geophys. Res., 11, C413, doi:1.19/4jc58. Iskandar, I., T. Tozuka, H. Sasaki, Y. Masumoto and T. Yamagata, 6: Intraseasonal 1

22 variations of surface and subsurface currents off Java as simulated in a high-resolution ocean general circulation model. J. Geophys. Res., 111, doi:1.19/6jc3486. Jochum, M. and P. Malanotte-Rizzoli, 3: On the generation of North Brail Current rings. J. Mar. Res., 61, Jochum, M., P. Malanotte-Rizzoli, and A. Busalacchi, 4: Tropical instability waves in the Atlantic Ocean. Ocean Modell., 7, Jochum, M. and R. Murtugudde, 5: Internal variability of Indian Ocean SST. J. Clim., 18, Kessler, W. S., M. J. McPhaden, and K. M. Weickmann, 1995: Forcing of intraseasonal Kelvin waves in the equatorial Pacific. J. Geophys. Res., 1, Klein, S. A., B. J. Soden, and N.-C. Lau, 1999: Remote sea surface temperature variations during ENSO: Evidence for a tropical atmospheric bridge. J. Clim., 1, Masumoto, Y., and G. Meyers, 1998: Forced Rossby waves in the southern tropical Indian Ocean. J. Geophys. Res., 13, Milliff, R. E., J. Morzel, D. B. Chelton, and M. H. Freilich, 4: Wind stress curl and wind stress divergence biases from rain effects on QSCAT surface wind retrievals. J. Atmos. Oceanic Technol., 1, Miyama, T., J. P. McCreary Jr., T. G. Jensen, J. Loschnigg, S. Godfrey, and A Ishida, 3: Structure and dynamics of the Indian-Ocean cross-equatorial cell. Deep-sea Res. II, 5, Murtugudde, R., J. P. McCreary Jr., and A. J. Busalacchi, : Oceanic processes associated with anomalous events in the Indian Ocean with relevance to J. Geophys. Res., 15,

23 Murtugudde, R. and A. J. Busalacchi, 1999: Interannual variability of the dynamics and thermodynamics of the Indian Ocean. J. Clim., 1, Murtugudde, R., A. J. Busalacchi, and J. Beauchamp, 1998: Seasonal-to-interannual effects of the Indonesian throughflow on the tropical Indo-Pacific basin. J. Geophys. Res., 13, Murtugudde R., R. Seager, and A. Busalacchi, 1996: Simulation of tropical oceans with an ocean GCM coupled to an atmospheric mixed layer model. J. Clim., 9, Nof, D. and D. Olson, 1993: How do western abyssal currents cross the equator. Deep- Sea Res. I, 4, Pedlosky, J., 1987: Geophysical Fluid Dynamics, Springer-Verlag, New York, nd edition. Potemra, J. T., S. L. Hautala, J. Sprintall, W. Pandoe, : Interaction between the Indonesian Seas and the Indian Ocean in Observations and Numerical Models. J. Phys. Oceanogr., 3, Raymond, D. J., S. K. Esbensen, C. Paulson, M. Gregg, C. S. Bretherton, W. A. Petersen, R. Cifelli, L. K. Shay, C. Ohlmann, and P. Zuidema, 4: EPIC1 and the coupled ocean-atmosphere system of the tropical east Pacific. Bull. Am. Meteorol. Soc., 85, Reppin, J., F. A. Schott, and J. Fischer, 1999: Equatorial currents and transports in the upper central Indian Ocean: Annual cycle and interannual variability. J. Geophys. Res., 14, Saji N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 41, Schott, F., M. Dengler, and R. Schoenefeldt, : The shallow overturning circulation of 3

24 the Indian Ocean. Prog. Oceanogr., 53, Schott, F. A. and J. P. McCreary Jr., 1: The monsoon circulation of the Indian Ocean. Prog. Oceanogr., 51, Schiller, A. and J. S. Godfrey, 3: Indian Ocean Intraseasonal Variability in an Ocean General Circulation Model. J. Clim., 16, Sengupta, D. and M. Ravichandan, 1: Oscillations of the Bay of Bengal SST during the 1998 summer monsoon. Geophys. Res. Lett., 8, Sengupta, D., R. Senan, and B. N. Goswami, 1: Origin of intraseasonal variability of circulation in the tropical central Indian Ocean. Geophys. Res. Lett., 8, Slingo, J. M., D. P. Rowell, K. R. Sperber, and F. Nortley, 1999: On the predictability of the interannual behaviour of the Madden Julian oscillation and its relationship to El Nino. Quart. J. Roy. Meteor. Soc., 15, Song, Q. and A. Gordon, 4: Significance of the vertical profile of the Indonesian Throughflow transport to the Indian Ocean. Geophys. Res. Lett., 8, Sprintall, J., A. L. Gordon, R. Murtugudde, and R. D. Susanto, : A semi-annual Indian Ocean forced Kelvin waves observed in the Indonesian Seas in May J. Geophys. Res., 15, Thum, N., S. K. Esbensen, D. B. Chelton, and M. J. McPhaden, : Air-sea heat exchange across the northern equatorial sea surface temperature front in the eastern tropical Pacific. J. Clim., 15, Waliser, D. E., R. Murtugudde, and L. Lucas, 4: Indo-Pacific Ocean Response to Atmospheric Intraseasonal Variability. Part II: Boreal Summer and the Intraseasonal Oscillation. J. Geophys. Res., 19, C33, doi:1.19/3jc 4

25 Waliser, D. E., R. Murtugudde, and L. Lucas, 3: Indo-Pacific Ocean Response to Atmospheric Intraseasonal Variability. Part I: Austral Summer and the Madden-Julian Oscillation. J. Geophys. Res., 18, C5, 316, doi:1.19/jc16. Waliser, D. E., K. M. Lau, and J. H. Kim, 1999: The influence of coupled sea surface temperatures on the Madden Julian oscillation: A model perturbation experiment. J. Atmos. Sci., 56, Wang, L., C. J. Coblinsky, S. Howden, 1: Annual Rossby wave in the Southern Indian Ocean: Why does it appear to break down in the middle ocean? J. Phys. Oceanogr., 31, Webster, P. J., W. M. Andrew, J. P. Loschnigg, and R. R. Lebon, 1999: Coupled oceanatmoshspere dynamics in the Indian Ocean during Nature, 4, Weisberg R. H. and T. J. Weingartner, 1988: Instability Waves in the Equatorial Atlantic Ocean. J. Phys. Oceanogr., 18, Xie, S. -P., H. Annamalai, F. A. Schott and J. P. McCreary Jr, : Structure and mechanisms of South Indian Ocean climate variability. J. Clim., 15, Yu Z., J. Potemra, 6: Generation mechanism for the intraseasonal variability in the Indo-Australian basin. J. Geophys. Res., 111, C113, doi:1.19/5jc33. 5

26 Captions Figure 1 RMS of intraseasonal SSHAs of -year model simulations (a) and of TOPEX/POSEION data from (b) and its uncertainty defined as the RMS of its annual mean values for the model (c) and the TOPEX/POSEIDON data (d). All fields are cm. Figure Intraseasonal SSHAs at 1 S, 11 E for six years and the corresponding wavelet spectrum (both in cm). Figure 3 zero-lag correlations of SSHAs with that at 1 S, 11 E. All correlations shown are statistically significant at the 95% confidence level. Figure 4 Intraseasonal SSHAs at 6 S, 66 E for four years and the corresponding wavelet spectrum (both in cm). Figure 5 Correlations of intraseasonal SSHAs with that at 6 S, 66 E, with -time lag, two-week lag, four-week lag, and six-week lag, respectively. All correlations shown are statistically significant at the 95% confidence level. Figure 6 Longitude-time plot of intraseasonal SSHAs along 8 S. Figure 7 Wave number-frequency plot of intraseasonal SSHAs along 8 S, normalized with the mean energy of SSHAs. The unit is cm / m -1 / s -1. The color codes show the natural logarithm ( log 1 ) of the real spectrum values. The black curve shows the dispersion relation of linear Rossby waves. Figure 8 Power spectrum of zonal wind stress at 6 S, 66 E. The unit is Pa / day -1. The dashed line indicates the spectrum of a red noise. dq U Figure 9 = β for November (a) and annual mean of dy y dq dy U = β y 1 1 m -1 s -1. f z N U (b). The black contour is the contour of zero. The units are z Figure 1 Barotropic (a) and baroclinic (b) energy conversions averaged over twenty years and above the thermocline in the SWIO. Figure 11 Zero-lag correlations between the baroclinic energy conversions and intraseasonal meridional velocities. All colored areas have statistically significant correlations at the 95% confidence level. Figure 1 Baroclinic energy conversion (in 1-3 W m -3 ) averaged over twenty years and above the thermocline in 4 months. Figure 13 Longitude-depth plots of baroclinic energy conversions for January, April, July, and December. The contour interval is W m -3. Regions with positive values are shaded. Figure 14 Vertical shears of zonal currents around the thermocline (~1 m) along 8 S 6

27 for 6 random years (contours) superposed with the baroclinic energy conversions (shaded area). The units for velocity shear are 1-3 s -1. The light gray area shows the energy conversion of W m -3 and the dark gray shows that of W m -3. Figure 15 Vertical profiles of squared Brunt-Väisälä frequency (a) and the eigenfunctions of the first three baroclinic modes (b). The eigenfunctions are normalized so that ψ ( ) = 1. n Figure 16 Projections of intraseasonal velocities onto the first two baroclinic modes in four years. Figure 17 Baroclinic energy conversion of the first baroclinic mode (a), of the second baroclinic mode (b), of the first baroclinic mode in December (c), and of the second baroclinic mode in December (d). The unit is 1 3 W m -3. 7

28 1 5 (a) Model simulations 1 (b) TOPEX/POSEIDON 5 3 Latitude (c) 1 (d) Latitude Longitude Longitude

29 1 S S HA in 6 years, 1HI, 1 S, 11 E S S HA /c m Y ear 1 Y ear Y ear 3 Y ear 4 Y ear 5 Y ear Wavelet s pec trum of S S HA, Morlet P eriod /day Y ear 1 Y ear Y ear 3 Y ear 4 Y ear 5 Y ear 6

30 -4 C orref with (1 S 11 E ), 1HI, time lag L atitude L ongitude

31 1 S S HA in 4 years, 1HI, 6 S, 66 E S S HA /c m Y ear 1 Y ear Y ear 3 Y ear Wavelet s pec trum of S S HA, Morlet 6 5 P eriod /day Y ear 1 Y ear Y ear 3 Y ear 4 1

32 .6. lag -.. two weeks lag -... L atitude L atitude four weeks lag s ix weeks lag (about 4 days ) L ongitude L ongitude

33 S S HA along 8 S (c m) Month L ongitude 3 4

34 1 x F requenc y (1/s ) Wave number (1/m)

35 1 P ower s pec trum of τ x R ed noise, 1/ f P eriod (day)

36 β - U yy in November -6 (a) L atitude L ongitude (b) A nnual mean dq/dy L atitude L ongitude -1

37 B arotropic energy c onvers ion ( 1-5 W/m -3 ) L atitude -6 (a) L ongitude (b) B aroc linic energy c onvers ion ( 1-3 W/m -3 ) 4 3 L atitude L ongitude -4

38 L atitude L ongitude

39

40 J an A pr Depth (m) J ul Dec Depth (m) L ongitude L ongitude

41 .4.8 Y ear Y ear Y ear Y ear Y ear Y ear L ongitude

42

43 5 c m/s 5 Depth (m) Y ear 1 Y ear Y ear 3 Y ear Depth (m) Y ear 1 Y ear Y ear 3 Y ear 4-5

44