Variability in the southeast Indian Ocean from altimetry: Forcing mechanisms for the Leeuwin Current

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. C9, PAGES 18,529-18,544, AUGUST 15, 1998 Variability in the southeast Indian Ocean from altimetry: Forcing mechanisms for the Leeuwin Current Rosemary Morrow and Florence Birol Laboratoire des Etudes G6ophysiques et Oc6anographiques Spatiales, Toulouse, France. Abstract. Seasonal and interannual variability in the southeastern Indian Ocean is investigated with the aid of Topex/Poseidon (T/P) altimeter data for the 3-year period The annual Rossby wave signal around 10øS is clearly marked, and consistent with modelling and Geosat results from earlier periods. A band of higher mesoscale variability between 20 ø and 35øS extends across the entire Indian Ocean, with characteristic timescales between 120 and 180 days and length scales of order 500 km. Sea level anomalies are shown to propagate at around 1.5 to 2 times the theoreticalinear Rossby wave speed, with an associated signal in sea surface temperature (SST) anomalies, and the propagation suggests that the variability is not locally forced but originates near the eastern boundary. Altimeter data is also used to examine variations in the alongshore pressure gradient, thoughto be the principal mechanism forcing the Leeuwin Current poleward againsthe prevailing equatorward winds. The T/P data confirms that the alongshore pressure gradient is maximum in May when the Leeuwin Current is strongest, but we find a consistent secondary peak in November which is not evident in the climatological data. There is also significant interannual variation, related to large interannual variations on the Australian Northwest Shelf. The seasonal and interannual variations also influence the thermal structure of the instabilities associated with the Leeuwin Current, which may be the source of the westward propagating anomalies between 20øS and 35øS. The results compare well with satellite SST and in situ XBT data at the eastern boundary and suggesthat satellite data can be used to monitor the variability of the Leeuwin Current. 1. Introduction The eastern boundary of the Indian Ocean is a region of unusually high seasonal and interannual variability, which can play an important role on the coupled ocean atmosphere system. The eastern Indian Ocean is directly forced by the strong seasonal monsoons, which provide local and remote wind forcing; but is also influenced by the remote ocean forcing between the Pacific and Indian Oceans via the Indonesian Throughflow. Indeed, recent model results show that increasing the Pacific wind forcing can strengthen the Indonesian Throughflow, which then modulates the entire southern Indian gyre [Reason et al., 1996]. The southeast Indian Ocean is also noted for its anomalous eastern boundary current, which flows poleward against the prevailing winds. This boundar3, current, the Leeuwin Current, is highly variable in terms of seasonal transport and also its associated mesoscale variability. Finally, the southeastern boundary includes the source of the annual westward propagating signal at 10øS [Perigaud and Delecluse, 1992] and significant Rossby wave activity between 20 ø and 35øS, which contribute to carrying the interannual variations from the eastern boundary into the ocean interior. However, we are a long way from understanding the dynamics of these ocean variations. The Indian Ocean is rather more complex than the Pacific Ocean in this regard Copyright 1998 by the American Geophysical Union. Paper number 98JC /98/98 JC $09.00 because the dynamical processes are essentially nonlinear, and simple linear models of the tropics are not as effective in the Indian Ocean. The description of the ocean-atmosphere interaction is also not clear [Godfrey et al., 1995], especially related to the monsoon variability. For example, we do not know whether the ocean response is primarily due to local or remote forcing. The presence of the Indonesian Throughflow determines the atypical dynamics of the southeast Indian Ocean. The combination of the Indonesian Throughflow and the monsoonal wind regime creates a warm water pool near the equatorial eastern boundary, contrary to the cold equatorial upwelling which exists at the other eastern boundaries with continental barriers in the tropics (Pacific and Atlantic Oceans). This warm water pool is associated with a large meridional ocean heat flux toward the poles: the lowered thermocline is propagated southward by coastal Kelvin waves [Godfrey et al., 1995] and' warmer water advected by the anomalous poleward eastern boundary current. The net climatic effect is an anomalously large heat loss from the ocean to the atmosphere, extending to 40øS along the eastern boundar3' [Oberhuber, 1988], with warm water tropical marine species found as far south as 35øS off Western Australia [Pearce and Griffiths, 1991]. The driving mechanism for the poleward Leeuwin Current is thought to be the strong alongshore density gradient. Although all eastern boundary currents show a negative alongshore steric height slope, the Leeuwin Current is by far the strongest [see Godfrey and Ridgway, 1985]. Thompson [1987] proposed that this poleward pressure gradient generates an eastward geostrophic flow that is sufficiently 18,529

2 18,530 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT large to overcome any coastal wind-forced upwelling; the eastern flow downwells, deepening the surface mixed-layer near the shelf edge, and sets up a poleward flow there which is balanced by bottom friction. This theory has also been proposed by other studies [e.g., McCreary and Kundu, 1989]; the effects of the poleward pressure gradient are estimated in various numerical studies [Batteen and Rutherford, 1990; Weaver and Middleton, 1989], and to some extent validated Most of these improvements have now been included in the updated merged T/P data (AVISO, 1996), which include using the new JGM3 orbit calculated by the University of Texas; applying a SSH drift correction for the Topex altimeter of 3 mm/yr (since identified as the Topex oscillator drift); correcting the backscatter co-efficient of the two altimeters, and then recalculating the wind speed; inserting a relative bias between the two altimeters coherent with the previous by a year-long Leeuwin Current Interdisciplinary Experiment corrections; ocean tides and tidal loading from the CSR3 (LUCIE) in [Smith et al., 1991]. The Leeuwin Current is known to be seasonally variable, reaching a model; solid earth tide applied; EM biases based on the BM4 model for the two altimeters; ionosphere corrections based on maximum in May. LUCIE results suggesthat this seasonal a spline filtered bifrequency correction for Topex, and DORIS variation is due to variations in the wind stress, as they found correction for Poseidon; TMR radiometer for the wet little seasonal dependence in the alongshore pressure gradient. troposphere correction and ECMWF model for the dry The interannual variability in the forcing terms has never been investigated, mainly due to the lack of observational data. Altimetry provides a means of accessing the seasonal and interannual variability in the region, including variations of troposphere; and the inverse barometer correction is replaced by a correction relative to the global mean pressure over the ocean. The edited and corrected data were interpolated onto fixed the alongshore pressure gradient. Previous altimetric alongtrack positions every 7 km, and residuals calculated missions such as GEOSAT and even ERS-1 had large longwavelength orbit errors, which effectively masked any ocean surface variations at long wavelength, such as the meridional pressure gradient. Considerable progress has now been made with the accurate dynamic height data available from the Topex-Poseidon altimeter, with excellent coverage to determine the long-wavelength ocean variability. Weekly and monthly wind field products are also available from the ERS- 1/-2 scatterometers. For the first time we have accurate data relative to the mean at each position. The alongtrack data is then Lanczos filtered with a cutoff wavelength of 50 km. The variance of these T/P alongtrack residuals is shown in Plate 1 for the southeast Indian Ocean. Two bands of high variance are noted in the extra-tropicalatitudes: a large maximum band around 10 ø to 15øS, with local maximums at 90øE and 105øE exceeding 180 cm 2, and a second less energetic band from 25 ø to 30øS with offshore values of around 120 cm 2. The first of these bands relates to energetic annual Rossby with extensive spatial and temporal coverage to investigate waves at 10øS, which have previously been studied using a the seasonal and interannual forcing in the southeastern Indian Ocean. shallow-water model and Geosat data by Perigaud and Delecluse [ 1992], and will be discussed further in the section on annual harmonics. The second band has not previously been studied in detail; we will follow this up later in the paper. Note also in Plate 1 the very high variability off the west Australian coast, with maximum variability greater than The present paper analyzes the variability in the southeastern Indian Ocean for the 3-year period , using Topex/Poseidon data and ERS-1 scatterometry winds. The first part of the paper concentrates on a descriotion of the variability in this domain. Section 2 presents the data 200 cm 2, values more typical for western boundary regimes processing and a description of the spatiotemporal than an eastern boundary domain. This mesoscale variability characteristics of the variability from spectral analysis over is associated with instabilities of the Leeuwin Current [e.g., 15 ø bins in the southeastern domain. This analysis highlights Pearce and Griffiths, 1991], and as we will see, has a fairly the importance of signals at annual, semiannual and strong seasonal and interannual signal, which needs to be interannual timescales, which will be investigated in more taken into account when calculating the alongshore pressure detail in section 3. In section 4, the meridional pressure gradient. gradient is calculated from T/P data (the variable component) and Levitus climatology (for the mean slope); this pressure 2.1. Spectral Analysis gradient forcing is then compared with the meridional wind Before progressing to a detailed analysis of the dynamics stress to calculate the balance of forcing for the Leeuwin in a particular region, we would like to know how the spatial Current over these 3 years. Although there are no in situ and temporal characteristics of the variability vary in the current records during this period to verify the strength and different parts of the domain. Thus a spectral analysis is direction of the resulting Leeuwin Current, the forcing can be necessary, which will help us choose the different spatial and compared with in situ XBT data from a repeat XBT line north temporal decorrelation scales, and identify the important of 30øS and satellite SST data, all presented in section 5. Conclusions are given in section Topex/Poseidon Data, Spectral Analysis, and Data Mapping The T/P mission has a 9.95-day repeat orbit with a distance of 315 km between neighbourin groundtracks at the equator. The data set we use spans 3 years from cycle 8 to cycle 120 (December 1, 1992 to December 29, 1995). The T/P data set we used for this study was derived from the merged T/P data (AVISO, 1992), with further improvements made by the Centre de Topographie des Oc6ans at LEGOS, Toulouse. harmonics to study. We have calculated variance-preserving frequency-wavenumber spectra from the alongtrack data in 15 ø bins for the southeast Indian region 5ø-35øS, and 85 ø- 115øE, presented in Plate 2. For the band from 5 ø to 20øS (top panels), the annual signal at around 1500 km wavelength is the most striking feature. The annual signal continues across the Indian Ocean basin (not shown), with maximum amplitude around 90øE. This signal corresponds to annual Rossby waves which will be discussed further in section 3.1. In the band 20 ø to 35øS (bottom panel), the annual signal is present but with reduced amplitude and reduced spatial scales at higher latitude, with a maximum peak at 500 km wavelength in the western-most bin. At this latitude band, the

3 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWlN CURRENT 18, I i! -5-2O -30 B -351 =ll,, longitude >200 cm 2 Plate 1. Variance of the T/P alongtrack residuals (cycles 8-120; December 1992 to December 1995) for the southeast Indian Ocean. annual signal weakens further west (not shown). In contrast, propagating waves, the harmonic signals and the meridional the largest peaks in this band are the semiannual peaks around pressure gradient. The interpolation uses a sub-optimal 150 and 180 days, with wavelengths around km. The larger amplitudes continue across the south Indian Ocean at this latitude band, and have propagation characteristicsimilar to semiannual Rossby waves (see section 3.2). space-time objective analysis package (P. De Mey, personal communication, 1992), to produce maps every 10 days starting from December 16, 1992 (cycle 9). The covariance function for the analysis is defined Close to the coast (100ø-115øE), both latitude bands show significant energy at the annual and semiannual timescales, with wavelengths of km, however there is also F(r,t) = 1 + ar - greater energy at the mesoscale, with peaks at 60 and 90 days [ +l(ar7 6 I ) (]) - ct and wavelengths decreasing at higher latitudes from 800 to 250 km, consistent with the reduced Rossby radius. The very Here dt is the time lag, rct is the temporal decorrelation radius, a is constant = 3.337, and the nondimensional large amplitude signals in the northeastern bin at 50, 60, and 75 day periods are probably coastal tidal errors, from the few function, r, is given in terms of the spatial lags, dx, dy, and tracks that cross the large tidal zone on the Australian the spatial decorrelation radii rcx, rcy, as Northwest Shelf. Finally, the interannual signal (with periods greater than 1.5 years) is quite significant, especially in the bins close to the shelf and in the band 20 ø to 35øS. Thus the annual and semiannual signals are fairly r = I (rcx)2 dx (rcy)2 dy 2 (2) significant in their respective latitude bands, with energy that The spatial and temporal decorrelation radii rcx, rcy, rct, were appears coherent across each latitude band. The mesoscale chosen as 320 km, 250 km and 15 days, respectively, based and interannual signals are also particularly energetic in the on the satellite sampling pattern and the alongtrack spectral bins closesto the eastern boundary. These signals will be analysis (see Plate 2). The a priori noise variance was chosen analyzed separately through a harmonic decomposition in at 20%. This value is large compared with the alongtrack section 3, and their propagation characteristics will also be error level of T/P data and represents the filtering of smallerexamined. scale variance that is not adequately measured between the satellite groundtracks. Indeed, the mapping reduces the total 2.2. Optimal Interpolation Mapping of the Data Onto a variance by around 30%, which represents the variance at Regular Grid. scales less than 300 km wavelength. Thus the mapped T/P The 3-year Topex/Poseidon sea level anomalies (SLA) are interpolated onto a regular 0.5øx0.5 ø grid for calculations of residuals will only represent the larger-scale circulation features. (ar)3]e-are

4 18,532 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT (a) Spectrum (crn^2) (b) Spectrum (crn^2) ' "... ' ' ' 1 O0 100 (c) 1000 ' 1 ) o o Wcwe th Ocm) Wcwe th (km) Plate 2. Variance-preserving frequency-wavenumber spectra in 15 ø bins for the southeast Indian region for a) 2 ' contours 5ø-20øS; are 85ø0øE; at cm b)5ø-20øs; intervals from 100ø-115c E ; red c)20ø-35øs contours ß are 85ø0øE:, at cm z intervals d)20ø-35øs; from ø-115ø1.. cm Blue 3. Annual, Semiannual, and Interannual Signals A large annual signal is also associated with the South Java Current south of Indonesia, which appears in-phase along the 3.1. Annual Harmonics entire south coast of Indonesia, with maximum amplitude in February centered around 114øE. This strong annual signal Plate 3 shows the amplitude and phase of the annual harmonic signal, based on 3 years of the mapped T/P data in the eastern Indian Ocean. The most striking annual signal is centered around 10øS, with a maximum amplitude of 15 cm occurring at 90øE. This signal corresponds to annually forced Rossby waves, previously described by Perigaud and Delecluse [1992] using Geosat data for and a shallow water tropical model. These waves appear to be forced east of 100øE by the seasonal cycle of the trade winds may be related to the seasonal reversing of the South Java Current, which has a strong westward flow in August during the southeast monsoon, and a minimum westward transport during the north monsoon in February [Tomczak and Godfrey, 1994]. The maximum amplitude in this analysis occurs in February when the Indian Equatorial Countercurrent is fully developed and supplies equatorial water to the eastern basin raising sea level along the Java coast. This reduces the pressure gradient from the Pacific to the Indian Ocean, and so [Woodberry et al., 1989], although it is not well understood the Throughflow is also minimal in February [Tomczak and why the maximum amplitude then occurs where the wind stress curl is weakest. The waves propagate west-southwest in the eastern domain (Plate 3b), but with greatly reduced amplitude west of 80øE (not shown). Godfrey, 1994]. Localized maximum annual amplitudes are also noted along the western Australian coast, with positions correlated with the local bathymetry (Plate 3a). Maximum amplitudes

5 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT 18,533 (a) O O 3O0 L (b) cm -5-2O _._... _ 0, ". -: Dec Feb Apr Jun Aug Oct Dec Plate 3. a) Amplitude and b) timing of the maximum in the annual harmonic signal, based on 3 years of the mapped T/P data from December December 1995.

6 18,534 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT (a) '-'-"" (b) o 5 lo ClTI O 7oo :: - I ooo , Dec Jan Feb Mar Apr May Plate 4. a) Amplitude and b) timing of the maximum in the semiannual harmonic

7 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT 18,535 occur offshore around 22ø-24øS, 29øS and 34øS and along the coast between 27 ø and 32øS, with small spatial scales of km. These annual signals are relatively in-phase; maxima occur during April-May when the Leeuwin Current is generally the strongest. These signals may be influenced by mesoscale meanders, forming from instabilities in the densitydriven Leeuwin Current when the density gradient is maximum. These meanders can remain fairly stationary for periods up to 3 months in satellite SST images [Pearce and Griffith, 1991 ]. Offshore, some eddies may separate, as noted in the SST images, and the phase diagram suggests they drift westward, although the amplitude is greatly reduced. The annual appearance of these offshore meanders, although perhaps due to the annual variations in alongshore density gradients, may also influence the actual calculation of meridional sea-surface height variations. Their presence needs to be taken into account when calculating the alongshore pressure gradient in section 4. and the direction is west-south-westward; a comparison of neighboring latitude lines show that when eddies "appear" or "disappear" at 29øS, they have in fact moved in from the north, or drifted further south. This is supported by a separate directional spectral analysis which shows most of the energy in the range days propagates westward, with slightly more energy at 180 days propagating to the southwest. These westward propagating features appear to have the characteristics of semiannual Rossby waves. We can calculate the maximum phase speed for the first baroclinic Rossby wave at this latitude, assuming nondispersive waves from [Kundu, 1990], Cx= fø 2 (3) where [3 is the beta plane constant, fo is the mean latitude, and c is the wave speed for the first internal mode. Using values from D. B. Chelton (personal communication, 1997) for the first internal Rossby radius, ri=c/fo, we calculate the average 3.2. Semiannual Harmonics westward phase speed as 3 cm/s east of 80øE, increasing to The amplitude of the semiannual harmonic in the eastern 3.9 cm/s around 40øE. Our T/P phase speeds (around 5 cm/s) Indian Ocean in Plate 4 is much less than the annual signal, are faster than these theoretical maximum value for linear though still significant in certain regions. In the South Java Current region, for example, there is a significant semiannual signal with amplitude of 8 cm, which is maximum in May and November, and may relate to the semiannual forcing by the warm Equatorial Jet [Tomczak and Godfrey, 1994]. However, this coastal region may also be affected by tidal aliasing of the K1 constituent, and at present we have no accurate tidal model for the Indonesian region which allows waves, as already noted in other T/P Rossby wave studies [e.g., Chelton and Schlax, 1996; P. Rogel et al., Propagation of the dominant sea level signals in the North Atlantic from Topex/Poseidon altimeter data, submitted to Journal of Geophysical Research, 1996, hereinafter referred to as submitted manuscript]. However, this simplified theory does not take into account any variations due to sloping bottom topography, nonlinear interactions or the presence of a mean us to define the amplitude of K1 in the coastal domain.. In current. For example, the mean current around 29øS the deep ocean these errors are minor, so we have more comprises the westward return flow of the South Indian confidence in analyzing the ocean signal. There is a Subtropical Gyre at this latitude in both Sverdrup models significant semiannual signal in the band from 20 ø to 30øS which is patchy and appears "mesoscale" with spatial scales of around 300 to 500 km wavelength. The interesting part of this signal is that it appears to originate in the eastern part of the basin, with maximum amplitude close to the shelf in May [Godfrey, 1989] and global ocean models [Serntner and Chervin, 1992]. Mean current values of around 5 cm/s are noted in the Semtner and Chervin model, although from theory, the presence of a mean westward current of 2 cm/s would be sufficient to explain our T/P values. and November. Near the coast the large amplitude The bathymetry (shown at the bottom of Plate 5) plays an semiannual signals extend from 20 ø to 35øS, but offshore there is near-zero amplitude over the Broken Plateau ridge at 30øS. Although there is a clear westward componento the important role in the westward propagation. At the eastern boundary on the shelf edge, the features remain stationary for periods of up to 3 months and evidence of eastward propagation at all latitudes, the presence of the ridge appears propagation suggests meandering structures. This is to block the propagating anomalies so that 30øS appears to be the southern limit for the band of large-amplitude anomalies. This band of westward propagating semiannual signal continues right across the Indian Ocean in both the semiannual harmonic maps and with a wider range of frequencies in the frequency wavenumber spectra (not shown). For a clearer view of this westward propagation, we have plotted the sea level anomalies across the basin at latitude 29øS, for the 3-year period of T/P data (Plate 5). Latitude 29øS was chosen as the central position of the large annual harmonic in the eastern boundary region. There is a clear propagation of eddy features across the basin from 110øE until they pass over the Madagascar Ridge around 46øE and interact with the higher eddy activity in the Mozambique Basin. These individual eddies have different amplitudes, but consecutive positive (negative) eddies are separated by around 180 days - the dominant period in this band from our spectral analysis. The propagation velocity is around 5 cm/s consistent with satellite SST analysis of the Leeuwin Current region which shows large meanders and occasional separation of eddies with temporal decorrelation of around 90 days [Pearce and Griffiths, 1991]. At 29øS, propagating signals pass to the north and parallel to the Broken Plateau between 85 ø and 100øE. The theoretical phase speed drops to cm/s across this region as the Rossby radius is reduced to near 25 km. The T/P data also show that propagating features do tend to slow down near the meridional Ninety East Ridge (at this latitude nearer 86øE!), as evidenced in the latter half of Finally, the amplitude and propagation rate increase enormously after the features pass over the Madagascar Ridge at 46øE and interact with the eddies and southwest mean flow of the East Madagascar Current. Semiannual Rossby waves have also been noted as far south as the Crozet Basin at 40øS from Geosat data by Park [1990]; here the semiannual signal has wavelengths from 300 to 600 km and westward propagation. To our knowlegde, the near semiannual signal has not been explicitly described from

8 .. 18,536 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWlN CURRENT Dec 95 (a) Jul Jan 95 Jul Jan 94 Jul Jan O0 110 Longitude cm O0 110 Plate 5. Longitude-time plot of T/P sea level anomalies at 29øS, for the 3-year period of T/P data (December, 1992 to December, 1995). Ocean depths along this latitude are given in the bottom panel.

9 MORROW AND B IROL: FORCING MECHANISMS FOR LEEUWIN CURRENT 18,537 Dec 95 Jul Jan 95 Jul. I- Jan 94 Jul Jan Longitude I 1.5 deg. C Plate 6. Longitude-time plot of Reynolds satellite SST at 29øS, for the 3-year period from December, 1992 to December, 1995.

10 18,538 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT numerical models of the Indian Ocean, as most models have concentrated on the tropics north of 20ø-25øS, or the 4. Alongshore Pressure Gradient and Wind Stress Agulhas/Southern Ocean section south of 30øS. Similar semiannual Rossby waves have also been detected in the To better understand the forcing of the poleward flowing South Atlantic at the same latitude band [Le Traon and Leeuwin Current, we calculate the relative balance of the Minster, 1993]. Their westward phase speed from Geosat data alongshore pressure gradient and the alongshore wind stress was calculated as 3 cm/s, around twice as large as the using our 3-year series of satellite data. Following Godfrey theoretical value for Rossby waves at that latitude in the south and Ridgway [1985], we consider the depth integral of the Atlantic; this difference was explained by the presence of a meridional (alongshore) momentum equation at the mean zonal westward flow as part of the South Atlantic continental shelf edge, relative to a given level of no motion as2 subtropical gyre and sloping bathymetry. Finally, we note that these propagating anomalies have an associated SST signature. We have calculated the DVdz_Y+jU=_gdp Dt y+-- xy P0 (4) propagation of SST at 29øS in Plate 6, for comparison with the T/P calculation. These signals have had the large annual where U, V are the depth integrals of the velocity components cycle, the long-term drift and the zonal average temperature (u,v) in the eastward, northward directions, respectively; p is removed to highlight the propagating signal. Propagating the depth-integrated steric height; 'l; y is the northward wind SST anomalies are evident, especially in the central part of the basin, with amplitudes of +/- IøC. Although they are stress; P0 is the mean density; D/Dt is the Lagrangian derivative; and Y is the depth integral of all the Reynolds visually correlated with the T/P residuals, their point-by-point stress terms. The terms of the fight-hand side can be regarded correlations are not significant, perhaps due to the large spatial smoothing by the SST interpolation procedure. In general, the propagating features maintain their SST anomalies a long time after their formation Interannual Signal In addition to the semiannual signal, a number of other as the forcing terms for the meridional (alongshore) flow. Our calculation of the meridional gradient in steric height is based on the 3 years of T/P data in this domain. Altimetric heights contains both the baroclinic and barotropic components of the signal, but we assume that the barotropic component is small in the eastern Indian, following comparisons of altimetry, hydrography and an OGCM [C. features are shown in Plate 5. The annual variations in steric Perigaud et al., unpublished manuscript, 1998]. The height over the ocean interior are shown as the bands of higher dynamic height centred around March-April every year. A clear interannual signal is also noted in the amplitude of propagating features in Plate 5. For example, at the eastern boundary the large positive anomaly which is apparent in altimetric sea surface height would then representhe surface steric height, but the depth-integrated steric height requires some knowledge of the vertical structure of the upper ocean. We assume that the geopotential anomalies decrease exponentially with depth over the top 500 m as an April/May of is absent in This has approximation to the vertical structure obtained from Levitus important repercussions for the amplitude of the propagating profiles, and we use this structure to calculate depthsignal. If we remove the annual steric height signal, we find that a period with mostly negative anomalies at the eastern integrated steric height from surface data alone. More importantly, we have removed the mean from our altimeter boundary (such as 1993) leads to a period of reduced dynamic data, a process which also removes the time-invariant height in the ocean interior around 70ø-90øE in Thus interannual changes in the mesoscale variability at the eastern Indian boundary can have large effects on the ocean interior via these propagating features. meridional pressure gradient. To replace this, we have calculated the mean depth-integrated steric height from the annual climatological data (Levitus et al., 1994) relative to 500 db (Plate 8), and added that to our T/P residuals. The The cause of these interannual variations at the eastern Levitus data clearly show the large mean meridional pressure boundary are currently being investigated. The interannual gradient at the eastern boundary, and the band of less dense trend over the 3 years of T/P data is shown in Plate 7 for the INdian and western Pacific Oceans. Throughouthis period, the eastern INdian Ocean shows a net sea level rise of mm/yr east of 90øE, with the maximum centered off the water between 10 ø and 25øS which originates in the Indonesian region, and continues across the INdian Ocean reaching a maximum off Madagascar. Figure 1 a shows the forcing terms on the fight-hand side of western Australian coast at 22øS. This net sea level drift is equation 4 calculated from the T/P-Levitus data and monthly not well correlated with the local wind stress curl forcing on the NW Shelf, although there is a net negative drift in the wind stress curl (t/j) for the region 20ø-30øS from ERS-1 scatterometry data (not shown). This is downwelling favourable, which helps to raise the SSH and enhance locally the effect from the NW Shelf. Variations in remote wind forcing in the tropical and subtropical Pacific give rise to interannual changes in the INdonesian Throughflow [e.g., Reason et al., 1996] and the formation of a warm pool on the Australian NW Shelf a few months later (C. Perigaud, personal communication, 1997). These large interannual variations in the eastern Indian Ocean will also play a role in the forcing of the density-driven Leeuwin Current, as we will see in the next section. gridded ERS-1 scatterometry data (available to March 1995). The meridional pressure gradient term is calculated monthly between 22øS and 32øS from latitudinal averages between 110 ø and 115øE. This averaging is done to improve the statistical sample and to minimise the effects of small-scale eddies, though eddies of around 500 km wavelength and 90 days duration remain and influence the calculation. (The calculation was repeated along a single line following the shelf edge at 1000 m as an estimate of the true alongshore direction; the pressure gradient terms was slightly noisier but the amplitude and phase remained the same). These forcing terms can be compared to a similar calculation by Godfrey and Ridgway [1985, Figure 11] based on climatological hydrographic and wind data (Figure l e). The first point to

11 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT 18,539 30, ; Longi.ude 180-6{ O mrn / ]year Plate 7. Long-term trend in sea level anomalies from 3 years of T/P data from December, 1992 to December, 1995, as a measure of net interannual change. 2O 10-2O -3O -4O O ms/kg Plate 8. Mean dynamic height (m2/s 2) at 0 m relative to 500 m from the annual Levitus data.

12 18,540 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT note is that the 3-year wind field remains very close to the climatology in terms of amplitude and phase - winds remain northward (upwelling favorable) throughouthe year, with maximum winds in austral summer (November to March), weakening to near zero in June-July. The meridional pressure gradient opposes the wind forcing, with a maximum peak in April - June in both the climatology and the 3-year average from T/P data. The 3-year T/P gradients are slightly For an idea of the spatial distribution of the anomalies that govern the forcing, we have mapped the monthly T/P anomalies (Plate 9a) and wind stress curl (Plate 9b) in May and November for The role of wind forcing on these sea level anomalies is currently being investigated, but it is clear that the wind stress curl fields are quite different, particularly between May 1993 and May The increased positive wind stress curl during May 1994 on the NW Shelf is stronger than the climatological data which spans many downwelling-favourable - so local wind-forcing here decades, and tends to average out large anomalous features enhances a deeper thermocline and may contribute to the that would remain in the shorter T/P record. What is higher T/P sea level anomalies. For T/P, the large-scale significant in our calculation is the large peak in November, structure in May 1993 is very different from the two suggesting a semiannual component in the alongshore successive years; the large negative anomaly due to the gradient at these latitudes. annual Rossby wave signal is centered closer to the Australian The three separate years show very strong interannual coast around 10ø-20øS, 100 ø- 110øE in 1993; whereas in variations in the alongshore pressure gradient (Figure lb, l c, 1995, it is centered around 10øS, 90øE. This lowering of the ld) has a lower maximum value in April-June which sea level in the northern domain has a direct effect of only just balances the equatorward wind forcing and the reducing the alongshore slope in May 1993 (Figure lb). The record is dominated by the large gradient centered around November. In 1994 the pressure gradient continues to large-scale positive anomaly on the NW Shelf in May 1994 is increase reaching an absolute maximum for the 3-year record unusually high, and strengthens as it develops westward over in April, with a smaller peak occurring in December the next few months. This anomaly directly influences the again has large peaks in May-June and November-December, strong alongshore gradient in early The rapid decrease though with reduced magnitude in comparison with All in the gradient around July-August in all years is due to the 3 years show the second peak in November-December, so the westward propagation of the northern positive anomaly into 3-year average in Figure l a is representative of each the ocean interior, and simultaneously the development of a individual year. However, the peaks in November 1993 and 1994 appear to be a forerunner of the larger maximums in strong positive eddies to the south. The strong gradient in November/December each year reflects the early development May of the following year. The alongshore wind stress of the large-scale positive anomaly in the north and a forcing has the same phase structure in the separate years, but simultaneous negative eddy at 30øS. Thus our complex the amplitude of 'c y is reduced in Thus, in 1994, the gradient signal is influenced by both the annual anomalies to decreased winds allow the forcing to be dominated by the already strong alongshore pressure gradient. the north and the development of higher frequency quasistationary eddies to the south. -5 -lo o (a) May 1993 Nov 1993,.,--, -,_.-,, 0,,. -,-,. 0 ß, ;; -lo -lo May 1994 Nov _ -20, ~ -20 _ cm (b) May Nov 1993,, -,, 0-, <;' "- -20 "' I May 1994 Nov 1994 q, o ß " '! -5 [ ' , I _ o ' ' N/rn 2 Plate 9. Mapped monthly (a) T/P anomalies and (b) ERS-1 wind stress curl in May and November for

13 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT 18,541 3 year average i... i i i i l ß _õ O i i i i i i i i i i u_ J F M A M J J A S O N D 5 b) J F M A M J J A S O N D O i i i i i i i i i i " J F M A M J J A S O N D d)... I 5. Comparison With SSTs and XBTs In the previous section we have measured the forcing terms for (4); here we test this against the resulting alongshore coastal currents. The LUCIE Experiment in 1985/1986 has shown that the Leeuwin Current is typically maximum in April/May within 100 km of the shelf break, with a smaller maximum in November [Smith et al., 1991]. Unfortunately, during the T/P period, there are no direct current measurements in the Leeuwin Current, but we can measure the temperature change for the variable southern penetration of the warm coastal current from high-density repeat XBT sections and satellite SSTs Satellite Sea Surface Temperatures (SSTs) We first compare the satellite forcing with the weekly optimal-interpolated Reynolds SST data, which includes mixed satellite and in situ data [Reynolds and Smith, 1994]. At the shelf edge, the Leeuwin Current should be very coherent through the water column [Smith et al., 1991], so SST data can be an indicator of the southward extent of the warm Leeuwin Current. Two time series of SST are presented at the shelf edge on the NW Shelf at 22øS and in the core of the Leeuwin Current at 32øS, for the 3-year period (Figure 2a). These points are chosen as the extremities of the T/P slope calculation. Intermediate time series have not been presented, as the temperature gradients are quite consistent along the coast between 22øS and 32øS. The dominant feature is the large annual signal in SST, both 3O 28 (a) Sea Surface Temperature,,,,,,,,,,,, 22S 115E 26 i i i i i i i i i i J F M A M J J A S O N D O 1 03,,. E e)' Climatology i,,,,,,,, ß /", F M A N F M A N F M A N o u_ j i i i i i i i i i i F M A M J J A S O N D (b) SST Difference,. i.,..,., i. Figure 1. (a) Average forcing terms for equation (4) for the 3-year period : winds (dashed line) are based on the monthly gridded ERS-1 scatterometry data (available to March 1995); pressure gradient (solid line) is based on T/P SLAs added to the Levitus 500 db mean dynamic height - the meridional slope is calculated between 22øS and 32øS from monthly and latitudinal averages between 110 ø and 115øE. The forcing terms for each separate year are shown for (b) 1993; (c) 1994; and (d) (e) Forcing terms calculated by Godfrey and Ridgway [1985] based on climatological hydrographic and wind data. i i i i i i i i i i i i F M A N F M A N F M A N Figure 2. (a) Time series of SST near the west Australian coast shelf edge at 22øS and 32øS, for the 3-year period (b) The SST difference between 22øS and 32øS.

14 18,542 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT location showing maximum SST in February-May, with a slower onset of heating from November-December, and a rapid decrease in SST in June. There is also strong interannual variability in the SST. At 22øS, the maximum SST is around 25øC for 1993, reaching 26øC for only 2-3 weeks in May. In contrast, the later years show maximum SSTs greater than 27øC for 2 months. The minimum SST is also 1 o warmer in 1994 than This general warming on the Australian NW Shelf in was also noted in the 3-year trend from T/P data (Plate 7), and in numerical modeling results which show a direct link with warming in the westem Pacific (C. Perigaud, personal communication, 1997). To estimate the response of the Leeuwin Current, we are interested in the SST difference between 22øS and 32øS (Figure 2b). The mean SST difference is 4.5øC, and for most of the record, the two signals are relatively in phase. The maximum differences along the shelf edge during the onset of larger depth-integratedensity gradients (April-May in 1993; February-June in 1994; February-April in 1995). The minimum SST differences occur after both the May and November peaks in the alongshore density gradient. Thus, the months characterised by weaker meridional SST gradients are typical of periods with a large density gradient, a strong Leeuwin Current, and a rapid southward advection of warm surface water XBT Results The SST results are consistent with our alongshore pressure gradient forcing, in regard to the interannual variations, and the importance of the semiannual November maxima for increasing the warm southward flow. However, the altimetric anomalies represent a depth-integrated quantity, whereas the SST values are directly influenced by the surface forcing, without the depth-integrated "memory." The poleward flowing Leeuwin Current is coherent with depth over the top 250 m, with an equatorward undercurrent below 300 m, so is best represented by an integrated calculation. This is possible with XBT data, and we have available a highdensity XBT line which passes a minimum of one time per month between Fremantle and Java. The mean position of this line is shown in Figure 3, and runs parallel to the shelf edge between 22øS and 29øS, i.e., essentially along the Leeuwin Current axis. Dynamic heights calculated from XBTs rely on climatological salinities, which may not be reliable in this 6. Discussion and Conclusion region of variable temperature-salinity characteristics. The question of whether satellite data can be used to Instead, we have chosen to analyze the depth of the 20øC isotherm, as a representation of the thermocline depth and as monitor the forcing and response of the Leeuwin Current still remains open. We have shown that the combination of an indicator of upper ocean thermal content. Figure 4 shows accurate satellite altimetry data and scatterometry wind fields the 20øC isotherm depth along the Fremantle-Sumatra XBT line for the 3 years , extracted at the latitudes 22øS and 29øS. Missing points are for months when the XBT line can be used to evaluate the variability of certain forcing terms for the Leeuwin Current. The alongshore pressure gradient calculated from altimetry and Levitus data provides a much was too far from the shelf edge, did not extend far enough stronger forcing in terms of amplitude and variability than the south, or where outcropping of the isotherm occurred. At the northern end (22øS), the 1993 and 1995 records show a gradual deepening of the 20øC isotherm to its winter maximum (August), with a second maximum occurring in November. However, there is a marked interannual difference in the mean 20øC isotherm depth: 1993 is m shallower than 1995, and 1994 shows the same deep 20øC isotherm as in 1995, although without the maximums in August and wind stress forcing. The wind stress forcing has the same phase every year, but the maximum amplitude is slightly reduced in January-April 1994, coinciding with the maximum alongshore pressure gradient. This combination of increased geostrophic flow onshore and weaker wind-forced upwelling, leads to a regime where we expect the maximum in the southward flowing Leeuwin Current. The satellite SST and XBT data do indicate that this 1994 May maximum flow was -2O -25-3O -35 0Ooo -4O Longitude Figure 3. Position of the mean XBT transect between Fremantle and Sunda Strait, Sumatra (asterisk line). November. This complements our SST results: that the jump of 2øC warmer SSTs in at 22øS is not surface limited but also relates to a general warming of the upper 150 m, which translates into an increased dynamic height for these years as measured by T/P. At 29øS, the annual cycle in 20øC isotherm depth is very clear, with gradual deepening of the isotherms after summer in March-July, with the isotherm surfacing around September/October after the winter cooling and wind-forced convection. The depth of the 20øC isotherm at 29øS is very well correlated with the satellite forcing terms plotted in Figure 1, in both amplitude and phase. When the alongshore pressure gradient is fairly weak in 1993, the depth of 20øC isotherm remains around 100 m; with a stronger gradient in April-June 1994, the 20øC isotherm depth increases to 180 m. At this latitude, the southward geostrophic transport of the Leeuwin Current is maximum [Smith et al., 1991], and so there is considerable southward heat transport during April- June 1994.

15 ß MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT 18,543-5O "" 0 0-2OO J -5O "" 0 0 (a) (b) 22S,,,,,,,,,, I I I I I I I I I I F M A M J J A S O N D Months 29S,,,,,,,,.., / ß ß /..'.,' -200 ' i i i i i i i i i J F M A M J J A S O N D Months Figure 4. Depth of 20øC isotherm depth along the Fremantle- Sumatra XBT line for the 3 years , extracted at latitudes (a) 22øS and (b) 29øS. ß / monsoon periods in April/October, roughly one month before the maximum density gradient response. The dynamical reponse to this localised wind forcing is currently being investigated [Birol and Morrow, manuscript in preparation, 1998]. Our analysis of T/P data in the latitude bands from 20 ø to 35øS clearly shows evidence of propagating Rossby wave signals originating near the eastern boundary. These waves appear to have a strong semiannual signal at the eastern boundary, and continue mainly westward with dominant periods between 120 and 180 days, and spatial scales of around 500 km. Their phase speed varies slightly in the eastern domain, influenced by the interactions with bathymetry and the change in stratification, but between 45øE and 90øE the phase speed is fairly consistent and approximately 1.5 to 2 times faster than predicted by linear theory, in agreement with other T/P studies of Rossby wave propagation [Chelton and Schlax, 1996, Rogel et al., submitted manuscript, 1996]. It is possible that the measured variability may be forced or modified by local wind forcing over the basin interior, but our analyse suggests that most of the variability from 20 ø to 35øS propagates freely from the eastern boundary. Is this variability locally forced at the eastern boundary? There is some evidence of a semiannual component in the wind stress fields at the eastern boundary, but north and south of the 20ø-30øS band. Rather, our calculations of the pressure gradient versus wind stress forcing suggesthat the semiannual component of the pressure gradient forcing may be more importanthan the local wind forcing, which suggest remote forcing from outside the 20 ø to 35øS band. If the alongshore density gradient is maximum in May and Nov, then these periods are when the density driven current is likely to be most unstable. Pearce and Griffiths [1991] used warmer than normal, with a large increase in the southward transport of heat, even though its surface geostrophic velocity calculated from alongtrack T/P data was not greatly increased. Note that apart from the spatially interpolated SST data, the other data sets used in this verification can be strongly influenced by eddies infringing on the shelf edge, or the strong current meanders, so it is difficult to determine the SST images of the Leeuwin Current and laboratory experiments to show that a current with steady thermal forcing can develop unstable waves in the form of large meanders, which remain fairly stationary with time and either decay or form eddies after periods of around 3 months. With variable density forcing, it is likely that the generation of these meanders and eddies is correlated to the periods of absolute Leeuwin Current strength. However, the maximum density gradient. Indeed, we find regions of higher independent verification suggests that satellite data can help monitor the forcing and response of variable Leeuwin Current. Although a November maximum was noted during LUCIE in , the consistently strong November signal during semiannual amplitude in the T/P gridded data occur along the coast in May/November (Plate 4). Thus our satellite analysis tends to supporthe proposition of Pearce and Griffith [ 1991 ] that the alongshore thermal/density gradient can be an important driving force for the generation of instabilities in was unexpected since it was not present in the Leeuwin Current domain. historical hydrographic data [Godpey and Ridgway, 1985]. Furthermore, the vertical structure and dynamic The fact that the signal appears in the alongshore pressure topography of these meanders and eddies will depend on the forcing, the SSTs and the XBT results indicate that it is a significant dynamic response during this T/P period. A check with the decadal record of Leeuwin Current at 25øS [Figure 4b of Meyers, 1996] also shows that secondary November maxima are common in most years, and that maximum annual transports occurred in November in , 1987, The reasons for the strong semiannual signal in the calculation of the alongshore gradient (and in the Leeuwin Current response) for certain years is not clear. It is possible that wind forcing on the Australian Northwest Shelf could play an important role, since the T/P SLAs and ERS-1 wind stress curl fields have a positive correlation (0.5) north of 20øS. The minimum wind speeds occur during the intercharacteristics of the density gradient during formation, for example, deeper thermoclines with maximum dynamic heights in May We find that the amplitude and sign of the propagating waves in Plate 5 have a clear interannual signature, related to the interannual variability we have measured at the eastern boundary. The period in late was marked by the formation of mainly negative (shallow thermocline) anomalies on the eastern boundary, which propagated westward and produce reduced SLAs around 65 ø to 85øE in late Similarly, the large positive (deep thermocline) anomalies generated on the eastern boundary around May 1994 eventually separate and produced raised SLAs from 95 ø to 110øE in late 1994.

16 18,544 MORROW AND BIROL: FORCING MECHANISMS FOR LEEUWIN CURRENT This can have an important effect on the ocean-atmosphere Bentamy A., N. Grima, Y. Quilfen, V. Harscoat, C. Maroni, and S. interactions in the ocean interior, since these propagating Pouliquen, An atlas of surface wind from ERS-1 scatterometry measurements, IFREMER publication, 229 pp. IFREMER, warm/cold anomalies carry a slightly increased/decreased DRO/OS, B P 70, Plouzan6, France, SST signature (Plate 6). In late 1994-early 1995 the eddy Chelton, D.B., and M. Schlax, Global observations of oceanic paths are somewhat masked in the interior by a larger-scale Rossby waves, Science, 272, , warming over the same longitude range lasting 4-5 months. Godfrey, J.S., A Sverdrup model of the depth-integrated flow for the (This is after the seasonal steric effect has been removed.) It is world ocean allowing for island circulations. Geophys. Astrophys. likely that increased warming along the entire eastern FluM Dyn.,45, , Godfrey, J.S., A. Alexiou, A.G. Ilahude, D.M. Legler, M.E. Luther, boundary in 1994 increases the tendency for deeper J.P. McCreary, Jr., G.A. Meyers, K. Mizuno, R.R. Rao, S.R. thermocline anomalies to form at all latitudes between 20 ø and 35øS. These anomalies then propagate westward with similar phase speeds (progressively decreasing at higher latitudes), producing a large-scale propagating region of increased temperature anomalies. The larger-scale warming event appears to coincide with the arrival of these temperature anomalies at 90ø0øE, combined with the additional steric effect from summer heating in early The large-scale warming may be surface-contained, since the dynamic response measured by T/P is less marked (Plate 5). However, this 0.5øC SST anomaly with a scale of 1000 km square lasting 4-5 months appears to influence the atmospheric conditions in the ocean interior; for example, an anomalous negative wind stress curl occurred at this location during April-May 1995 [Bentamy et al., 1996], giving upwellingfavorable winds coinciding with the end of this warm SST anomaly. This suggests larger-scale coupled forcing back on the ocean surface, and these coupled processes clearly need further study. The interesting point is that large-scale interannual changes in the tropical oceans can directly influence the subtropical mesoscale variability on the eastern boundary, and propagating mesoscale features then carry the interannual signal into the ocean interior where it can influence the local atmospheric conditions after a number of months or years. Thus getting the mesoscale variability right in the subtropical zone may be a key factor in future coupled ocean-atmosphere models. We have shown that T/P data can provide a good description of the variations in the alongshore density gradient and the formation of instabilities in the Leeuwin Current Sheyte, J.H. Toole, and S. Wacogne, The role of the Indian Ocean in the global climate system: recommendations regarding the Global Ocean Observing System. Report of the Ocean Observing System Development Panel, Texas A&M University, College Station, TX, USA, 89 pp., Godfrey, J.S., and K.R. Ridgway, The large-scalenvironment of the poleward-flowing Leeuwin Current, Western Australia: longshore steric height gradients, wind stresses and geostrophic flow. J. Phys. Oceanogr., 15, , Kundu, P.K., Fluid Dynamics, 638 pp., Academic, San Diego, Calif., Smith, R.L., A. Huyer, J.S. Godfrey, and J.A. Church, The Leeuwin Acknowledgments. Our thanks to Pierre-Yves Le Traon, Jean- Current off Western Australia ,/. Phys. Oceanogr., 21, Franqois Minster, Gilles Reverdin and our three reviewers for helpful , comments on this manuscript. The preliminary processing of T/P Thompson, R.O.R.Y., Continental-shelf-scale model of the Leeuwin data and the SST data was performed by Bertrand Ferret and Claude Current.,/. Mar. Res., 45, , Brossier of the Centre de Topographie des Oc6ans, LEGOS, Tomczak, M., and J.S. Godfrey, Regional Oceanography: An Toulouse. introduction, Elsevier Sci., New York, Weaver, A.J., and J.H. Middleton, On the dynamics of the Leeuwin Current. Phys. Oceanogr., 19, , References Woodberry, K.E., M.E. Luther, and J.J. O'Brien, The wind-driven seasonal circulation in the southern tropical Indian Ocean, AVISO, AVISO User Handbook: Merged TOPEX-POSEIDON Geophys. Res., 94, 17,985-18,002, products, 3 ra ed., 3.0, Publ. AVI-NT-O21-CN, 200 pp., Cent. Natl. D'Etudes Spatiales, Toulouse, France, F. Birol and R. Morrow, Laboratoire des Etudes G6ophysiques et AVISO, AVISO User Handbook: Merged TOPEX/POSEIDON Oc6anographiques Spatiales (LEGOS), UMR5566/GRGS, 18, av. products, 2 ed., Publ. AVI-NT-O21-CN, 84 pp., Cent. Natl. Edouard Belin, Toulouse Cedex 4, France. ( D'Etudes Spatiales, Toulouse, France, Rosemary. Batteen, M., and M. Rutherford, Modeling of eddies in the Leeuwin Current: The role of thermal forcing. J Phys. Oceanogr., 20, (Received December 13, 1996; revised December 1, 1997; , accepted March 9, 1998.) Le Traon, P.-Y., and J.-F. Minster, Sea level variability and semiannual Rossby waves in the South Atlantic subtropical gyre. J. Geophys. Res., 98, 12,315-12,326, Levitus, S., T.P. Boyer, and R. Burgett, World Ocean Atlas 1994, NOAA Atlas NESDIS 3 and 4, Washington, D.C., McCreary, J.P., and P.K. Kundu, Thermohaline forcing of eastern boundary cun'ents: With application to the circulation off the west coast of Australia. J. Mar. Res., 44, 71-92, Meyers, G., Variation of Indonesian throughflow and the E1-Nino Southern Oscillation. J. Geophys. Res., 101, 12,255-12,263, Oberhuber, J.M., An atlas based on the "COADS" data set: the budgets of heat, buoyancy and turbulent kinetic energy at the surface of the global ocean, Rep. 15, Max-Planck-Inst. far Meteorol., Hamburg, Germany, Park, Y.H., Mise en 6vidence d'ondes plan6taires semi-annuelles dans le sud de l'oc6an Indien par altim6trie satellitaire, C. R. Acad. $ci. Paris, 310, , Pearce, A.F., and R.W. Griffith, The mesoscale structure of the Leeuwin Current: a comparison of laboratory models and satellite imagery. J. Geophys. Res., 96, 16,739-16,757, Perigaud, C., and P. Delecluse, Annual sea level variations in the southern Indian Ocean from Geosat and shallow-water simulations. J. Geophys. Res., 97, 20,169-20,178, Reason, C.J.C., R.J. Allan, and J.A. Lindesay, Evidence for the influence of remote forcing on interdecadal variability in the southern Indian Ocean, J. Geophys. Res., 101, 11,867-11,882, Reynolds, R.W., and T.M. Smith, Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7, , region, which can directly influence the remote and local ocean conditions. To further understand the dynamical processes governing these variations, we intend to continue with a more detailed analysis of the wind and thermohaline forcing in the western Pacific and the eastern Indian, in Semtner, A.J., Jr., and R.M. Chervin, Ocean general circulation from a global eddy-resolving model. J. Geophys. Res., 97, , particular with the aid of numerical models