Seasonal variability of the Bay of Bengal circulation inferred from TOPEX/Poseidon altimetry

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 105, NO. C2. PAGES , FEBRUARY 15, 2000 Seasonal variability of the Bay of Bengal circulation inferred from TOPEX/Poseidon altimetry Andreas Eigenheer and Detlef Quadfasel 2 Institut filr Meereskunde der Universit it Hamburg, Hamburg, Germany Abstract. The circulation in the interior of the Bay of Bengal and of its western boundary current, the East Indian Coastal Current, is inferred from historical ship drift data and from TOPEX/Poseidon altimeter data. The boundary current shows a strong seasonal variability with reversals twice per year that lead the reversal of the local monsoon wind field by several months. On the basis of model simulations it has been suggested that this unusual behavior can be explained by the influence of remotely forced planetary waves. Our data analysis confirms and refines this view by showing the role of topography in the northern bay. We also give an estimate of the relative importance of the different contributions. 1. Introduction type circulation that is closed through the coastal current at the western boundary; (2) planetary waves originating at the eastern The upper layer circulation of the Bay of Bengal is subjecto boundary through the radiation of coastal Kelvin wave energy strong seasonal variability. During the early northeast monsoon (These Kelvin waves are generated at the equator through the in November the large-scale flow pattern in the bay is cyclonic, impact of the equatorial Wyrtki Jet upon the coast of Sumatra.); and the western boundary current, the East Indian Coastal and (3) planetary waves generated at the eastern boundary Current (EICC), flows southward from the Bengal Shelf north of through fluctuations of the local alongshore monsoon winds. The to the east coast of Sri Lanka. In February the EICC phase shift between the fluctuations of the western boundary reverses and flows northward along the Indian coast, reaching its current and the monsoon winds can thus be explained by the time maximum strength during the early southwest monsoon in the planetary waves take to cross the Bay of Bengal. April/May [Shetye et al., 1993]. In the interior of the Bay of In this study we use two independent data sets, Bengal the large-scale flow is anticyclonic during this time. This TOPEX/Poseidon (T/P) altimeter data [Archiving, Validation and variability is associated with the Indian monsoon: dry Interpretation of Satellite Oceanographic Data (AVISO), 1995] northeasterly winds coupled with cooling and evaporation in and historical compilations of surface drifts [Cutler and Swallow, winter and southwesterly winds coupled with heating, 1984], to quantify the different mechanisms that have been precipitation, and an increased freshwater runoff into the suggested to contribute to the seasonal circulation of the Bay of northern bay in summer. Bengal and its western boundary current. The underlying concept The coupling of the oceanic circulation in the Bay of Bengal is a decomposition of the observed multivariate sea surface to the monsoon forcing is different from that in the Arabian Sea, height fields into principal modes of variability: seasonal where the development of the Somali Current closely follows the harmonics (SH) and complex empirical orthogonal functions cycle of the wind forcing, with a time lag of only a few weeks (CEOF). Since we are interested in the seasonal variability of the [Schott et al., 1990]. In contrast, the western boundary current in circulation, we restrict ourselves to the basic modes. the Bay of Bengal appears to lead the wind field, and the EICC flows againsthe local winds at the end of both monsoon seasons [Shetye et al., 1991, 1993]. 2. Data This unusual pattern lead various investigators to consider the The T/P sea level anomalies from September 1992 to October role of remote forcing effects from the eastern coast and from the 1994 (cycles 2-75) are taken from CD-ROM [AVISO, 1995]. interior of the bay on the western boundary current in addition to Along-track data are binned into a 2 ø longitude by 1 ø latitude the local forcing by alongshore winds [Yu et al., 1991; Potemra grid three times a month (every days) using a Gaussian et al., 1991; Prasanna Kumar and Unnikrishnan, 1995; Shankar interpolation scheme. This scheme applies a temporal and spatial et al., 1996; McCreary et al., 1993, 1996]. The remote forcing filter directly to the along-track data, thereby minimizing the effects identified include (1) interior Ekman pumping during the smearing of high-frequency, high-wavenumber fluctuations over height of the respective monsoon seasons, leading to a Sverdrupthe low-frequency, large-scale signal of interest. The procedure preserves the exact time of the sampling by the satellite and does not apply the widely used assumption of taking the single cycles Now at Swedish Meteorological and Hydrological Institute, NorrkOping, Sweden. as synoptic data. The time resolution of the grid is taken different 2Now at Niels Bohr Institute for Astronomy, Physics and Geophysics, from the length of the exact repeat cycles (9.92 days) in order to University of Copenhagen, Copenhagen, Denmark. reduce the systematicorrelation of the sampling locations of the satellite and the grid points. This procedure is an extension of the Copyright 2000 by the American Geophysical Union. method described by Nerem et al. [1994], who average the data Paper number 1999JC inside a defined radius (cutoff scale) using Gaussian weights (e /00/1999 JC $09.00 folding scale). 3243

2 3244 EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL 15øh 10øh 5o N 5o N 80'øE 90'øE 10 øe 80'øE 90'øE 10 ) ø -- 20ON ).0øN 80'øE 90 'E 10 øe 80 'E 90 'E 10 )øe N N N 80øE 90øE 100øE 80øE 90øE 100øE Figure 1. Near-surface circulation of the Bay of Bengal as derived (left) from ship drift data [Cutler and Swallow, 1984], and (fight) TOPEX/Poseidon (T/P),altimeter data. The ship drift data are averaged over intervals of 2 months, centered at the month shown; the T/P data are from 1992 and 1993, centered at the respective months and weight averaged over 1 month. See text for further details. For the present application to the Bay of Bengal the enough to resolve the structures of the sea level field discussed in geometrical parameters are taken as follows. The e-folding scales various model studies [Potemra et al., 1991; McCreary et al., are 4 ø in longitude, 2 ø in latitude, and 15 days in time, and the 1993, 1996]. cutoff scale is twice the e-folding scale for all dimensions. This The historical surface current data used were collected by the choice removes the impact of the temporal and spatial sampling British Meteorological Office for the period and are (satellite tracks) for waves propagating at zonal phase speeds compiled by Cutler and Swallow [1984] as 1øxl ø and 10 day <0.3 m s ' (4 ø in 15 days). On the other hand, the scales are small averages. For our study we used slightly different averages: 1 ø

3 EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL i ',-%,.;,'..,... 80OE 90OE 100øE 5ø N t ' '" ' ' x, ::::::::::::::::::::::::::::::::::: 80OE 90øE 100øE 80OE 90øE 100øE, 80OE 90øE 100øE Figure 1. (continued) latitude by 2 ø longitude over 2 months for the sparsely sampled were calculated from the sea level gradients derived from the interior of the bay and 2 ø latitude by 1.5 ø longitude monthly 1 ø latitude times 2 ø longitude grid of sea surface heights at the averages for the boundary current region. middle of the month indicated on the maps. The above mentioned smoothing of the T/P data gives independent data only about every 3 ø in latitude and 6 ø in longitude, which, however, is 3. Seasonal Development of the Circulation sufficient to resolve the large-scale circulation. The maps The seasonal cycle of the near-surface circulation in the Bay represent a time interval of 1 month. of Bengal north of based on the Cutler and Swallow [1984] When comparing the two sets of circulation maps, three differences inherento the measurement techniques have to be ship drift data and on the geostrophic estimates using T/P altimeter data from is shown in Figure 1 for 2 months kept in mind: (1) the T/P circulation solely represents the intervals. For the ship drift-derived currents we used 60 day geostrophic component of the flow, whereas the ship drift is means centered at the depicted months. The T/P surface currents really a combination of winds and currents (including also

4 ß 3246 EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL 2000,.OON 5øh 1500 J 0 km 80øE 85øE 90øE looo 8O 8O 5OO JAN MAR MAY JUL SEP NOV JAN MAR MAY JUL SEP NOV Figure 2. Time evolution of the coast parallel component of the East Indian Coastal Current (EICC) and the Monsoon Current from south of Sri Lanka to the Bengal Shelf from Cutler and Swallow's [1984] ship drift data. The data are binned in 2 ø latitude by 1.5 ø longitude boxes. The arrows in the inset map show the direction of the current component used. "windage," i.e., wind drag directly on the ship, and nongeostrophic Ekman currents), which, during the height of the monsoon seasons dominate the ship drift pattern in the interior of the bay; (2) because of the large uncertainties of altimeter measurements near coasts and because of the large spatial averaging applied to the data, the boundary currents are not resolved in the T/P-derived circulation. In contrast, the boundary currents are well sampled in the historical ship drift data set, and (3) the ship drifts represent an average over more than a century, but the T/P data are taken from only. We chose this particular interval out of the 2 years of data analyzed because an E1 Nifio occurred in 1994, that might have had a strong impact on the circulation of the eastern Indian Ocean [Meyers, 1996]. In contrast, 1993 is considered to be a normal year. The seasonal development of the EICC at the western boundary of the Bay of Bengal is shown in Figure 2 on the basis of overlapping 2 ø latitude by 1.5 ø longitude bin averages of the alongshore current component of the ship drifts plotted versus time. The inset shows the direction of the alongshore component taken at the positions where the data set has been subsampled. Both Figures 1 and 2 are the basis of the following discussion. In November, at the start of the northeast monsoon the Bay of Bengal has a basin-wide cyclonic circulation west of the arc of the Andaman Islands. The EICC flows southward from the Bengal Shelf north of to the east coast of Sri Lanka, increasing strength from 0.5 m s ' at to more than 1 m s ' at 8øN. Geostrophic estimates based on hydrographic data collected during December 1991 showed southward surface currents of 1.3 m s - at [Shety et al., 1996] and that transports in the upper 200 m of the water column reached 8 Sv (1 Sv = 106 m 3 s' ). The EICC continues southward and eventually feeds into the westward flowing Monsoon Current, connecting the Bay of Bengal with the Arabian Sea [Schott et al., 1994]. This current cames low-salinity water (S < 34) that spreads as a tongue into the Arabian Sea [Wyrtki, 1971]. In January, at the height of the northeast monsoon the strength of the EICC ceases rapidly, and the Monsoon Current south of Sri Lanka is now supplied from the east, from south of 8øN. The cyclonic circulation cell in the interior of the Bay of Bengal weakens and moves toward the west and north. Off the Bengal Shelf currents reverse and now have an eastward component. Along the Andaman Islands currents are directed toward the

5 .. EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL 3247 south. The numerical simulation of McCreary et al. [ 1996] of the 2 surface current field also shows this collapse of the basin-wide cyclonic circulation and the retreat of the cell toward the Indian coast, although somewhat more southward than is seen in the two observational data sets. 20ON Between the two monsoon seasons in March, with practically no local wind forcing, the EICC flows northward all along the Indian coast, picking up speed rapidly. The boundary current now forms the western part of two anticyclonic gyres, centered at and at 17øN, respectively. Between these gyres, near 14øN the historical ship drift data show a minimum in the northward coastal flow. During hydrographic surveys in March/April 1991 and March 1993, Shetye et al. [1993] and $anilkumar et al.?_ [1997] even observed a cyclonic eddy hugged against the coast near that latitude, which might be a remnant of the winter monsoon circulation. Evidence for the existence of the two anticyclonic gyres was also found in satellite infrared measurements made in February 1985 [Legeckis, 1987]. He observed two bands of warm water leaving the coast near 11 ø and 19øN, which coincide with the northern branches of the respective gyres. The now weakened Monsoon Current south of Sri Lanka is still being fed by the southern one of the two gyres. The northward EICC has its maximum speed around April, and in May, at the onset of the southwest monsoon the flow weakens again. The two gyres observed in March have now merged, forming an anticyclonic circulation cell that reaches from the Indian coast to the arc of the Andaman Islands. This cell is decoupled from the Monsoon Current in the south, which now flows eastward toward Sumatra. The sea surface salinity distribution at this time, shown in the Indian Ocean atlas [Wyrtki, 1971], confirms this picture and shows no intrusions of higher salinity water from the Arabian Sea into the Bay of Bengal. In July the southwest monsoon is at its full strength, but the EICC has weakened further and even reversed to a southward cm on the Bengal Shelf and off Madras at. The coastal maxima are separated by a sickle-shaped minimum extending direction in the northern part. Shetye et al.'s [1991] hydrographic from south of Sri Lanka at 4øN through the central Bay to 19øN observations during July/August 1989 show the near-boundary at India's east coast. To analyze the seasonal characteristics of the circulation to consist of two cyclonic eddies or gyres centered sea level variability, we follow two different approaches. near 15 ø and 18øN with southward flow near the coast and First, the amplitude and phase of the annual and semiannual northward flow in between the two features. In the T/P-derived current map this circulation pattern is shifted somewhat to the south and extends all the way across the Bay of Bengal. The southern gyre here is fed by an inflow from the Monsoon Current about halfway across the bay. It is interesting to note that the Monsoon Current south of Sri Lanka shows a slight break during harmonics of the sea level are determined by minimizing the squaredifference between the observed sea surface height and a cosine function. The sea surface height z at location x and at time t can then be written as z(t,x) -= aj(x)cos [(wjt - bi(x)], (1) July, picking up in speed again in August. A reduction of the transport and even a reversal of the near-coastal flow was also observed in shipboard current measurements and in records from moored instrumentation during 1991 and 1992 by Schott et al. [1994], who suspected Kelvin waves from the Bay of Bengal to be responsible for this current fluctuation. In September the southwest monsoon weakens, and the EICC has reversed almost along the entire length of the Indian coast. On the large scale the northern cyclonic gyre has spread farther south, and the southern cell is hardly visible in the T/P derived currents any more. 4. Analysis Methods The spatial distribution of the standard deviation for the period October 1992 to September 1993 of the T/P sea levels is displayed in Figure 3. Maximum values are found all along the coast surrounding the Bay of Bengal, reaching peak values of 9 :....=. i. :?.-':: ' -- : i ;;-? ii -U!:; ::-"'--.,:-;;::":. :'-.-:::-'; ---::...'... ß i i i i 80øE 85øE 90øE 95øE 100øE longitude Figure 3. Standard deviation (centimeters) of the sea level field from T/P altimeter data during the period October 1992 to September where c 6 =j2zc yr" j= 1...N. The functions a (x) and b (x) are amplitude and phase of the jth component of the annual (j = 1; 1SH), and semiannual harmonics (j = 2; 2SH). The method is analogous to the one described by Prasanna Kumar and Unnikrishnan [1995] where the phase and amplitude were determined for the seasonal cycle of the Levims [1982] temperature climatology in the upper layers of the Bay of Bengal. Second, principle component analysis (PCA) allows detection of propagating modes in the sea level fi ld [Barnett, 1983]. CEOFs are found from PCA of the complex analytical function (t) = z(t)+ i (t), which contains the real function z(t) as the real part and its Hilber transform (t) as the imaginary part [Bendat and Piersol, 1986]. According to '(t,x)= (t) (x); t=tl,..;t ; x= i,..;x (2)

6 3248 EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL i i i i i i i i i i i i /, :...' ','...,.,' /..' '..i, '. " 0 -',." '...¾,,' ". = /I 1" - '. 't " /,.' - "/s X."1 ' / '..' '-'...'.. O , ' I t".' I...' E 0'5... /... o / / ',,' b maximum off Madras), the first mode (ICEOF, explained 62% of the variance; Figure 4a) is associated with the annual cycle. The second mode (2CEOF, explained 23 % of the variance; Figure 4b) has components of the annual (-16%) and semiannual period (-7%). Both (envelope) amplitude functions show a distinct temporal drift, with opposing trends: the first component has a decreasing tendency and the second component has an increasing tendency. After April 1994 the second component is even more important than the first. As mentioned before, 1994 was a prominent E1 Nifio year, and the trends seen in the CEOFs are probably caused by the strong interannual variability of the sea level associated with this event. The spatial patterns of the ICEOF and the 1SH are very similar, and we only show the ICEOF (Figure 5a). Like the distribution of the standard deviation (Figure 3), it has maxima along the eastern and northern coasts, with amplitudes above 9 cm. The coastal maximum in the west, south of 17øN, is less high and rather broad, extending much farther into the open ocean. The two maxima are separated by a sickle-shaped minimum, stretching from the Monsoon Current south of Sri Lanka across the central bay to the northwest coast near 18øN. One might be tempted to associate the high amplitudeseen at the coasts with the presence of shelf edge waves or Kelvin waves. However, these coastal-trapped waves decay seaward with the deformation radius, which, in this area, is of order 50 km. They can therefore not be held responsible for the amplitude pattern observed in the 1CEOF. These two areas of maximum amplitude are approximately 8 months out of phase (Figure 5b). In the east and north, sea level is high in September, and in the west it is high during May. In the Andaman Sea east of 93øE the phase is almost constant. Its circulation is decoupled from the main part of the Bay of Bengal [Potemra et al., 1991]. Also to the southeast of the minimum amplitude line, to --3øN, the phase is fairly constant with a high -1.5 ' ' ' ' ' water level in September. This pattern most probably reflects the OCT DEC FEB APR JUN AUG OCT DEC FEB APR JUN AUG OCT path of the Monsoon Current, which, during the southwest monsoon, flows past the southern tip of Sri Lanka and then tums Figure 4. (a) Time evolution of the first principal component of northward to flow along 8øN (Figure 1). The water level at the T/P sea surface height for a location in the westem Bay of Bengal southern flank of this geostrophicurrent is high. In the central (, 82øE). Envelope (solid, absolute. in (2)), reconstructed time series (dashed, Re in (2)), and harmonic fit to the and western pan of the bay, just west of the Andaman Islands reconstructed time series (dotted, like acoswt-b in (1)). The and south of the Bengal Shelf, the phase of the annual signal First-Mode Complex Emperial Orthogonal Function (ICEOF) increases toward the west, reaching its maximum at the coast near explains 62% of the variance. (b) As in Figure 4a but for the. In the following we will explain this phase pattern in terms second principal component, which explains 23 % of the variance. of propagating Rossby waves Free Rossby Waves In the circulation models of Potemra et al. [1991] and the data set (t,x) can be expanded in complex harmonical McCreary et al. [1996], free Rossby waves are excited by the principle components,(t) and CEOFs,(x) [Preisendorfer, 1988]. As written here, CEOFs are unity vectors, and principle alongshore winds adjacento the northern and eastern coasts of components me dimensional time series with variance A. In the Bay of Bengal. The monsoon winds blow southerly during contrast to the standard EOF, which displays the spatial summer and northerly during winter. In summer, Ekman amplitude, the CEOF also contains spatial phase information. convergence leads to downwelling, and in winter, Ekman The same method has been applied to the annual sea level divergence forces coastal upwelling. Since the Coriolis parameter variations in the Indian Ocean by Petrigaud and Delecluse f increases with latitude, the free Rossby waves have a latitude- [1992]. dependent phase speed c = -/ C 2 if2 [Gill, 1982]. Here/ is the Rossby parameter, and c is the phase speed of baroclinic gravity waves. With a typical reduced gravity g' = 0.02 m S'2 and a mixed 5. Results layer depth of h = 100 m the phase speed of the gravity wave is c The first two CEOFs as calculated from the altimeter data = 1.4 m s '. For the free Rossby wave a phase speed of 8.5 cm s ' explain 85% of the total variance of the Bay of Bengal sea level results at a latitude of 9øN; at 14øN the speed is 3.5 cm s '. This field noah of 5øS. As illustrated by the reconstructed time series decrease of the phase speed with latitude leads, in the Northern and its seasonal harmonic for a position in the western bay (the Hemisphere, to a clockwise bend of the phase lines. In the

7 EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL c] I latitude range from 7 ø to 14øN we do indeed find this predicted bend of the phase lines (Figure 5b). Table I gives the phase speeds calculated from the data Modification by Topography Eq. 5øS 80øE I. 2 1: 20ON - i i i 85øE 90øE 95øE longitude 100øE i I The clockwise bent of the phase lines is only found south of 14øN (Figure 5b). To the north of that latitude the bent is anticlockwise and thus contrary to what one would expect from planetary effects alone. However, the increase of the phase speed north of 14øN, giving rise to the observed phase distribution, can be explained by the influence of the sloping topography. Here the latitude-dependent change of the Coriolis force is opposed by the decrease of the water depth in the northern part of the Bay of Bengal. In the extreme case of no stratification the potential vorticity gradient associated with the sloping bottom is dynamically equivalento the planetary vorticity gradient and simply adds to the latter. The phase speeds of the Rossby waves will thus be higher than for a flat bottom, as observed. However, this case cannot apply here since the northern Bay of Bengal is stably stratified, both in temperature and in salinity. Also, barotropic waves are much faster than the ones observed. In the other extreme case of very strong stratification the waves will be bottom-trapped [Pedlosky, 1984], and the currents at the surface will vanish, as do the associated signals in the sea surface topography. The same effect occurs when the bottom slope becomes very large and the planetary vorticity gradient can be neglected, i.e., at continental slopes. This also does not apply to the northern bay where the topography is only -3 times as important as the [3 effect. Thus, with reasonable confidence one can expect that the sloping bottom will make the waves faster but that there will still be a signal at the surface. Our observations supporthis. These baroclinic Rossby waves influenced by topography do not appear in the numerical simulation studies of Poternra et al. [1991] and McCreary et al. [1996]. The reason is simply that both models have a flat bottom, and the physics of topographic waves is thus not included in the models Sverdrup Circulation The spatial distribution of the projection of the 2CEOF (Figure 6) within the bay has the same pattern as the stream Eq. 5øS 80øE...?.:.265 ' '": :,..2-6i!3 ii! /..-:2:6 : i.: i: :: 27: : '279 -'2"81 85øE 90øE 95øE 100øE longitude Figure 5. Spatial projection of the 1CEOF from the T/P data: (a) amplitude (centimeters) (X, absolute, in (2)), and (b) phase (degrees) ( - atan Im! Re in (2)). i Table 1. Analysis of the TOPEX/Poseidon Sea Level Data: Latitude-Dependent Phase Difference of the Annual Seasonal Harmonic (1SH) First-Mode Complex Emperial Orthogonal Function (1CEOF) and Estimates of Phase Speed in the Southwestern Bay of Bengal. Latitude, Phase Difference Phase Speed, øn Over 10 ø Lon[situde cm s ø (186) 6.27 (6.67) ø (172) 6.92 (7.24) 10 ]66 ø (]6]) 7.53 (7.76) ø (149) 8.47 (8.41) ø (127) 9.74 (9.89) ø (115) (10.95)

8 3250 EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL.... SøN - Eq. - i i i i i 80øE 85øE 90øE 95øE 100øE longitude Figure 6. Spatial projection the amplitude (centimeters) of the 2CEOF from the T/P data ( A2absolute 2 in (2)). function of a typical subtropical gyre, with an amplitude maximum in the west off Madras. Long midlatitude Rossby waves forced by the wind stress curl in the interior of the oceans carry potential vorticity to the western boundary [Pedlosky, 1965]. Here these Rossby waves are reflected into short Rossby waves, which are trapped at the boundary, leading to an accumulation of energy. In the Bay of Bengal the 2CEOF mirrors the periodicity of the anticyclonic circulation in November and the cyclonic circulation in May (Figure 1). In the Atlantic and Pacific Oceans the Sverdrup circulation dominates the time mean anticycloni circulation forced by the trades and westefiies. In the Bay of Bengal, with the reversing winds of the monsoons the Sverdrup circulation explains a significant amount of the variance. In the latitude range from 12 ø to 18øN this annual mode contributes more than 60% to the variance of the observed sea levels in the EICC. The apparent phase shift of 3 months with the oceanic circulation leading the wind field can also be explained in this way. The long Rossby waves at have a typical phase speed of 0.05 m s". At this latitude the width of the Bay of Bengal between the Andaman and Nicobar Islands and the east coast of India is km, and it will thus take the waves -9 months to cross the bay. This phase lag of 9 months is equivalento a lead of 3 months, which is what the observationshow (Figure 1) Remote Forcing From the Equator The 2CEOF containes also a semiannual signal, which in the western bay is associated with downwelling in winter and summer and upwelling in spring and fall (Figure 4b). Strong semiannual sea level variability exists at the equator, linked to the occurrence of the eastward flowing Wyrtki Jet [Wyrtki, 1973]. This jet is driven by eastward winds in between the monsoon seasons and leads to high sea levels off the coast of Sumatra during May and November. From there, energy is radiated northward and southward in the form of coastal Kelvin waves [Quadfasel and Cresswell, 1992] that introduce sea level variability along the coasts on a 6 month timescale. These coastal waves, in turn, loose some of their energy to planetary waves that radiate out from the eastern coast. Their contribution to the variance of the EICC is about half that of the Sverdrup circulation and thus cannot be neglected. Free Rossby waves at a semiannual period can exist only south of. Only below that latitude can the northward traveling Kelvin waves contribute energy to the westward traveling Rossby waves at the semiannual period. It can be clearly seen in the 2SH and in the 2CEOF (Figure 6) that their impact is concentrated there. The propagation of these waves was successfully modeled by Jensen [1993] and later by McCreary et al. [1996], who showed the impact of the remote forcing from the equator on the EICC and found this mechanism to be even more importanthan our data suggest. In the simulation the variability was concentrated in the latitudes north of 12øN with northward flow (upwelling) in April and October and southward flow (downwelling) in June and December. This is the timing of the semiannual modulation that we find in the 2CEOF of the altimeter data (Figure 4b). 6. Conclusions We have decomposed the observed TOPEX/Poseidon (T/P) altimeter data into principle modes of variability (CEOFs). From this we find that the sea level variability in the East Indian loo lo o I. 2. CEOF ///1.CEOF--. ::...:.::i:?:'!' :. -" latitude [øn] Figure 7. Explained variance of the different modes in the area of the EICC. The annual seasonal harmonic (1SH; triangles) explains about the same amount of variance as the sum of the 1CEOF and the 2CEOF. Three different regions can be distinguished: To the north of 18øN, topographic Rossby waves explain more than 60% of the local sea level variance; from 12 ø to'18øn, trapped Rossby waves associated with the seasonally reversing Sverdrup circulation contribute the biggest part; and south of 12øN the variability is mainly determined by free Rossby waves.

9 3251 EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL Coastal Current (EICC) is mainly determined by remote forcing References through Rossby waves propagating from the interior of the Archiving, Validation, and Interpretation of Satellite Oceanographic Bay of Bengal. North of 14øN the EICC is influenced by Data (AVISO), AVISO User handbook: Sea-Level Anomaly Files, 1st seasonally upwelling and downwelling baroclinic topographic ed., CLS, Ramonville, Rossby waves forced by alongshore winds adjacento the Barnett, T.P., Interaction of the monsoon and Pacific trade wind system northern and eastern boundary. To the north of 18øN these waves at inter-annual time-scales, I, The equatorial zone, Mon. Weather Rev., III, , explain more than 60% of the local variability (Figure 7). In the Bendat, J.S., and A.G. Piersol, Random Data, 565 pp., John Wiley, New latitude range from 12øto 18øN the variability is mainly York, determined through Rossby waves trapped at the western Cutler, A.N., and J.C. Swallow, Surface currents of the Indian Ocean (to boundary, related to the Sverdrup circulation. This has previously 25øS, 100øE), 10S Tech. Rep. 187, 8 pp., Inst. of Oceanogr. Sci., been derived from in situ measurements [Shery et al., 1993] and Wormley, England, U.K., Gill, A.E., Atmosphere-Ocean Dynamics, 662 pp., Academic, San Diego, simulation studies [McCreary et al., 1996]. The interior Ekman Calif.,1982. pumping is upward in winter (cyclonic Sverdmp flow) and Jensen, T., Equatorial variability and resonance in a wind-driven Indian downward in summer (anticyclonic Sverdmp flow). Remote Ocean model, J. Geophys. Res., 98, 22,533-22,552, forcing effects from the equator modulate the coastal current on a Legeckis, R., Satellite observation of a western boundary current in the Bay of Bengal, J. Geophys. Res., 92, 12,974-12,978, semiannual timescale (Figure 4b). In the latitude range from 7 ø to Levitus, S., Climatological Atlas of the World Ocean, NOAA Prof Pap., 12øN, free Rossby wave propagation plays the most important 13, 173 pp., U.S. Gov. Print. Off., Washington, D.C., role for the sea level variability of the EICC (Figure 7). These McCreary, J.P., P.K. Kundu, and R.L. Molinari, A numerical waves are excited at the eastern margin by the seasonality of investigation of dynamics, thermodynamics and mixed-layer processes in the Indian Ocean, Prog. Oceanogr., 31, , alongshore winds. McCreary, J.P., W. Han, D. Shankar, and S.R. Shetye, Dynamics of the The picture derived from the T/P altimeter data is somewhat East Indian Coastal Current, 2, Numerical solutions, J. Geophys. more detailed than the one proposed by Prasanna Kurnar and Res., I01, 13,993-14,010, Unnikrishnan [1995], who used climatological hydrographic data Meyers, G., Variation of Indonesian Throughflow and the E1 Nino - [Levitus, 1982] to decompose the variability of the circulation in Southern Oscillation, J. Geophys. Res., I01, 12,255-12,263, Nerem, R.S., E.J. Schrama, C.J. Koblinsky, and B.D. Beckley, A the Bay of Bengal. Analysis of the thermocline topography on a preliminary evaluation of ocean topography from the monthly basis showed an annual wave with a westward TOPEX/Poseidon mission, J. Geophys. Res., 99, 24,565-24,583, propagation speed of twice the Rossby wave speed. The better spatial and temporal resolution of the T/P data compared to the Pedlosky, J., A note on the western intensification of the oceanic circulation, J. Mar. Res., 23, , hydrographic data set has now allowed for the identification of Pedlosky, J., Geophysical Fluid Dynamics, 624 pp., Springer-Verlag. the three different wave regimes. Likewise, the model New York, simulations, since they all work with a flat bottom, are not P6rigaud, C., and P. Delecluse, Annual sea-level variations in the capable of resolving the three regimes as they do not include the southern tropical Indian Ocean from GEOSAT and shallow-water simulations, J. Geophys. Res., 97, 20,169-20,178, topographic Rossby waves in the northern bay. Such an addition Potemra, J.T., M.E. Luther, and J.J. O'Brien, The seasonal circulation of to the models, however, should be rather simple and will lead to the upper ocean in the Bay of Bengal, J. Geophys. Res., 96, 12,667- more realistic simulation results. 12,683, 1991 The influence of planetary waves at semiannual timescales Prasanna Kumar, S.P., and A.S. Unnikrishnan, Seasonal cycle of originating from the equator appears to be overestimated in the temperature and associated wave phenomena in the upper layers of the Bay of Bengal, J. Geophys. Res., I00, 13,585-13,593, model [Potemra et al., 1991 ]. The reason might be that the island Preisendorfer, R.W., Principal Component Analysis in Meteorology and arc of the Andamans and Nicobars is approximated as a solid Oceanography, Dev. Atmos. Sci., vol. 17., 425 pp., Elsevier, New boundary in the models, which have a dissipation too low York, compared to reality where strong tidal currents modulate the flow. On the other hand, the T/P data near the coastal boundaries Quadfasel, D., and G. Cresswell, A note on the seasonal variability of the South Java Current, J. Geophys. Res., 97, , Salinkumar, K.V., T.V. Kuruvilla, D. Jogendranath, and R.R. Rao, are fairly unreliable, because of the undulations of the satellite's Observations of the western boundary current of the Bay of Bengal ground track combined with the rapidly changing geoid near the steep bottom topography. It may thus be that the near-coastal variability is not detected properly in the T/P data set. On the background of the present knowledge about the Bay of Bengal circulation and its variability we restricted the analysis to the lowest modes of variability, although higher frequencies and wave numbers are also covered by the T/P data. We also could have chosen shorter e-folding and cutoff scales, but then the impact of the temporal and spatial sampling of the satellite for the detection of propagating waves becomes more important, and it will be harder to compare the results from the data with the results from the models. Acknowledgments. We would like to thank two anonymous reviewers for the critical comments. Funding for this study was granted by the Bundesminister ftir Bildung, Wissenschaft, Forschung und Technologie within the German WOCE program (03 F0157A). from a hydrographic survey during March 1993, Deep Sea Res.,Part I, 44, , Schott, F., J.C. Swallow, and M. Fieux, The Somali Current at the equator: Annual cycle of currents and transports in the upper 1000 m and connection to neighboring latitudes, Deep Sea Res., Part A, 37, , Schott, F., J. Reppin, J. Fischer, and D. Quadfasel, Currents and transports of the Monsoon Current south of Sri Lanka, J. Geophys. Res., 99, 25,127-25,141, Shankar, D., J.P. McCreary, W. Han, and S.R. Shetye, Dynamics of the East India Coastal Current, 1, Analytic solutions forced by interior Ekman pumping and local alongshore winds, J. Geophys. Res., I01, 13,975-13,991, Shetye, S.R., S.S.C. Shenoi, A.D. Gouveia, G.S. Michael, D. Sundar, and G. Nampothiri, Wind-driven coastal upwelling along the western boundary of the Bay of Bengal during the southwest monsoon, Cont. ShelfRes., II, , Shetye, S.R., A.D. Gouveia, S.S.C. Shenoi, D. Sundar, G.S. Michael, G. Nampothiri, The western boundary current of the seasonal subtropical gyre in the Bay of Bengal, J. Geophys. Res., 98, , 1993.

10 3252 EIGENHEER AND QUADFASEL: SEASONAL VARIABILTY OF BAY OF BENGAL Shetye, S.R., A.D. Gouveia, S.S.C. Shenoi, D. Shankar, P.N. Vinayachandran, D. Sundar, G.S. Michael, and G. Nampothiri, Hydrography and circulation in the western Bay of Bengal during the northeast monsoon, J. Geophys. Res., 101, 14,011-14,025, Wyrtki, K., Oceanographic Atlas of the International Indian Ocean Expedition, 531 pp., Nat. Sci. Found., Washington, D.C., Wyrtki, K., An equatorial jet in the Indian Ocean, Science, 181, , Yu, L., J. O'Brien, and J. Yang, On the remote forcing of the circulation in the Bay of Bengal, J. Geophys. Res., 96, 20,449-20,454, A. Eigenheer, Swedish Meteorological and Hydrological Institute, Norrk6ping, D. Quadfasel, Department of Geophysics, Niels Bohr Institute for Astronomy, Physics and Geophysics, University of Copenhagen, Juliane Maries Vej 30, 2100 Copenhagen 65, Denmark (Received February 24, 1998; revised December 21, 1998; accepted April 20, 1999.)