The monsoon currents in the north Indian Ocean

D. Shankar a , P. N. Vinayachandran b, A. S. Unnikrishnan a , and S. R. Shetye a

a Physical

Oceanography Division, National Institute of Oceanography, Dona Paula, Goa 403 004, India.

b Centre

for Atmospheric and Oceanic Sciences, Indian Institute of Science, Bangalore 560 012, India.

Abstract The north Indian Ocean is distinguished by the presence of seasonally reversing currents that ow between the Bay of Bengal and the Arabian Sea. These currents are located between the equator and approximately 10 N. The Summer Monsoon Current (SMC) ows eastward during the summer monsoon (MaySeptember) and the Winter Monsoon Current (WMC) ows westward during the winter monsoon (NovemberFebruary), March April and October being months of transition between these well-dened current systems. We assemble data on ship drifts, winds and Ekman drift, geostrophic currents derive from TOPEX/Poseidon sea-level anomalies, and hydrography to dene the climatological currents in observations. An Oceanic General Circulation Model (OGCM) is used to simulate the climatology of these currents and estimate transports, and numerical experiments with a simpler model are used to investigate the processes that force these currents. The ship drifts show that the monsoon currents extend over the entire basin, from the Somali coast in the west to the Andaman Sea in the east. They do not, however, come into being, or decay, over this entire region at a given time. Different parts of the currents form at different times, and it is only in the mature phase that the currents exist as trans-basin ows. The westward WMC rst forms south of Sri Lanka in November and is initially fed by the equatorward East India Coastal Current (EICC); the westward WMC in the southern bay appears later. The WMC divides into two branches in the Arabian Sea, one branch continuing to ow westward, and the other turning around the Lakshadweep high off southwest India to ow into the poleward West India Coastal Current (WICC). The SMC in the Arabian Sea is a continuation of the Somali Current and the coastal current off Oman.

Preprint submitted to Elsevier Preprint

10 January 2001

It ows eastward and southeastward across the Arabian Sea and around the Lakshadweep low off southwest India. It continues as the eastward SMC south of Sri Lanka. In the Bay of Bengal, the SMC branches, one branch turning into the Bay of Bengal and the other owing eastward. Ekman drift driven by the monsoon winds overwhelms the geostrophic ow at the surface in the western Arabian Sea. During the summer monsoon, Ekman drift dominates over most of the Arabian Sea; it is only in the eastern Arabian Sea, in the eddies off Somalia, and in the Bay of Bengal that the geostrophic current makes a signicant contribution. During the winter monsoon, geostrophy dominates, and Ekman drift modulates the geostrophic current. The Ekman drift shows much less spatial structure than the geostrophic current. Signatures of westward propagation of sea-level anomalies are evident in the altimeter data in the regime of the monsoon currents. The OGCM simulations show that Ekman drift dominates in a shallow surface layer (about 20 m deep), but geostrophy dominates below this. The WMC is primarily a geostrophic current, with Ekman drift modulating it. The strong winds during the summer monsoon ensure that Ekman drift dominates at the surface, leading to a more complex vertical structure in the SMC than in the WMC. At the surface, the SMC in the Arabian Sea ows eastward and southeastward, feeding into the eastward SMC south of Sri Lanka. This ow branches east of Sri Lanka, one branch owing into the bay, the other continuing to ow eastward. The geostrophic component of the SMC is a continuation of the Somali Current. A part of the recirculation around the eddies off Somalia merges with the ow to the west of the Lakshadweep low off southwest India to form a curving SMC that ows into the eastward SMC south of Sri Lanka. The net transport due to the shallow monsoon currents is due to both Ekman drift and geostrophic ow. The WMC (SMC) transports

7 Sv ( 6 Sv) westward (eastward) in the top 100 m between 3-6 N at 80.5 E (south of

Sri Lanka) during the winter (summer) monsoon. Numerical experiments with a 1 1 2 -layer reduced-gravity model show that the dynamics of the north Indian Ocean on seasonal time scales is explicable by linear wave theory. The equatorial Rossby wave, the equatorial Kelvin wave, and the coastal Kelvin wave merge the Arabian Sea, the Bay of Bengal, and the equatorial Indian Ocean into a single dynamical entity, the north Indian Ocean, which must be modelled as a whole even to simulate circulation in its parts. Circulation at any point is decided by both local forcing and remote forcing, whose signals are carried by the equatorial and coastal waves. Superimposed on the currents associated with these waves is the local Ekman drift. The geostrophic component of the monsoon currents is forced by several processes. In the Bay of Bengal, the

currents are forced by Ekman pumping and by the winds in the equatorial Indian Ocean. To the west of Sri Lanka, in the eastern Arabian Sea, the major forcing is by winds along the east and west coasts of India and Sri Lanka. Ekman pumping in the central Arabian Sea and off the Somali coast are important processes in the central and western Arabian Sea, with the Rossby waves radiated from the Indian west coast also playing a role. Thus, the monsoon currents are actually composed of several parts, each of which is forced by one or more processes, these processes acting in concert to produce the continuous monsoon currents seen owing across the breadth of the north Indian Ocean. Key words: Tropical Oceanography. Summer Monsoon Current. Winter Monsoon Current. Coastal Kelvin wave. Equatorial Kelvin wave. Equatorial Rossby wave. Arabian Sea. Bay of Bengal.

Contents 1 1.1 1.2 2 2.1 2.2 2.3 3 3.1 3.2 3.3 4 4.1 4.2 4.3 5 Introduction Observational background Theoretical background Observations Ekman drift Geostrophic currents The net ow at the surface The monsoon currents in an OGCM Numerical model The model circulation Transport estimates Forcing mechanisms The numerical model and the control run Process solutions Dynamics of the north Indian Ocean Summary 12 13 17 23 23 31 39 48 48 49 56 62 62 68 86 91 95

Acknowledgements

List of Figures

Schematic representation of the circulation in the Indian Ocean during January (winter monsoon) and July (summer monsoon). The abbreviations are as follows. SC, Somali Current; EC, Equatorial Current; SMC, Summer Monsoon Current; WMC, Winter Monsoon Current; EICC, East India Coastal Current; WICC, West India Coastal Current; SCC, South Equatorial Counter Current; EACC, East African Coastal Current; SEC, South Equatorial Current; LH, Lakshadweep high; LL, Lakshadweep low; GW, Great Whirl; and SH, Socotra high. 18

Wind stress (dyne cm 2 ) from the climatology of Hellerman and Rosenstein (1983). 25

Surface Ekman drift (cm s

in the Indian Ocean. The drift is 26

computed from the wind-stress climatology of Hellerman and Rosenstein (1983) and is based on the Ekman spiral formula.

Monthly climatology of sea-level anomalies (cm, left panel) and geostrophic current (cm s 1 , right panel) in the Indian Ocean. Negative anomalies are indicated by dashed contours and the contour interval is 5 cm. The sea-level anomalies and geostrophic currents are derived from the TOPEX/Poseidon altimeter data for 19931997. 27

(continued)

28

(continued)

29

(continued) 5

30

Longitude-time plots of the monthly climatology of TOPEX/Poseidon sea-level anomalies (cm) at 10 N (top panel), 8 N (middle panel), and 5 N (lower panel). Negative anomalies are indicated by dashed contours and the contour interval is 5 cm. Westward propagation is evident in both Arabian Sea and Bay of Bengal at all latitudes. At 5 N, there is a break in the signal at 80 5 E (south of Sri Lanka), even though there is no land barrier there. 36

Latitude-time plots of the monthly climatology of zonal geostrophic current (cm s

1 ),

derived from TOPEX/Poseidon

altimetry, at 80 5 E (south of Sri Lanka). Westward ow is indicated by dashed contours and the contour interval is 5 cm s 7

1.

37

Depth-time plots of the meridionally averaged zonal geostrophic current (cm s

1)

derived from the climatologies of Levitus and

Boyer (1994) and Levitus, Burgett, and Boyer (1994). The left (right) panel shows the current averaged over 0 3 N (3 6 N). The depth is in metres. Westward ow is indicated by dashed contours and the contour interval is 5 cm s 1 . 8 The net ow (NF) at the surface (cm s 1 , left panel), computed as the sum of Ekman drift (Figure 3) and geostrophic ow (Figure 4), and ship drifts (cm s 1 , right panel). The source for the ship drifts are the Ocean Current Drifter Data CDROMs NODC-53 and NODC-54 (NODC, US Department of Commerce, NOAA). 8 8 8 9 10 11 (continued) (continued) (continued) OGCM currents (cm s OGCM currents (cm s OGCM currents (cm s 40 41 42 43

38

1) 1) 1)

at 5 m. at 35 m. averaged over the top 50 m. 6

50 51 52

12

Longitude-time plot of the OGCM meridional velocity (cm s 1 ) at 8 N. Southward ows are indicated by dashed contours and the contour interval is 5 cm s 1 .

54

13

Longitude-time plot of the OGCM meridional velocity (cm s 1 ) at 5 N. Southward ows are indicated by dashed contours and the contour interval is 5 cm s propagation at 80 E.

1.

Note the break in the westward 55

14

Depth-time plot of the OGCM zonal current (cm s 1 ) at 80.5 E (south of Sri Lanka). The upper (lower) panel shows the current averaged over 36 N (03 N). Westward ows are indicated by dashed contours and the contour interval is 5 cm s 1 .

57

15

Latitude-time plots of the depth-integrated zonal current (m2 s 1 ). The current is integrated over the top 100 m. Westward ows are indicated by dashed contours and the contour interval is 10 m2 s

1.

60

16

Sea-level deviation from the initial surface (cm, left panel) and upper-layer velocity (cm s 1 , right panel) for the nonlinear simulation. Negative sea level is indicated by dashed contours and the contour interval is 5 cm. 65 66

16 17

(continued) Longitude-time plot of sea-level deviation (cm) from the reduced-gravity model (nonlinear simulation). Negative sea level is indicated by dashed contours and the contour interval is 5 cm.

67

18

Sea-level deviation from the initial surface (cm, left panel) and upper-layer velocity (cm s interval is 5 cm.

1,

right panel) for the linear simulation. 69 70 7

Negative sea level is indicated by dashed contours and the contour

18

(continued)

19

Effect of winds along the western boundary of the Bay of Bengal (Process WB). Sea-level deviation (cm, left panel) and upper-layer velocity (cm s 1 ) is shown. Negative sea level is indicated by dashed contours and the contour interval is 5 cm. 72 73

19 20

(continued) Effect of winds along the eastern boundary of the Arabian Sea (Process EA). Sea-level deviation (cm, left panel) and upper-layer velocity (cm s 1 ) is shown. Negative sea level is indicated by dashed contours and the contour interval is 5 cm.

75 76

20 21

(continued) Effect of winds along the northern and western boundaries of the Arabian Sea, except the Somali coast (Process WA). Upper-layer velocity (cm s 1 ) is shown.

77

22

Effect of winds along the Somali coast (Process SA). Upper-layer velocity (cm s 1 ) is shown. 79

23

Effect of alongshore winds in the north Indian Ocean (Process CW). Sea-level deviation (cm, left panel) and upper-layer velocity (cm s 1 , right panel) are shown. Negative sea level is indicated by dashed contours and the contour interval is 5 cm. 80 81

23 24

(continued) Effect of ltering out forcing by alongshore winds in the north Indian Ocean (Process OP). Sea-level deviation (cm, left panel) and upper-layer velocity (cm s 1 , right panel) are shown. Negative sea level is indicated by dashed contours and the contour interval is 5 cm.

83 84

24 25

(continued) Ekman pumping (m day 1 ), derived from the wind-stress climatology of Hellerman and Rosenstein (1983). 8

85

26

Schematic illustrating the dynamics of the north Indian Ocean. The linear theoretical framework depicted here invokes the equatorial Kelvin wave, the equatorial Rossby wave, and the coastal Kelvin wave. These three waves merge the equatorial Indian Ocean, the Bay of Bengal, and the Arabian Sea into a single dynamical entity. The horizontal hatching indicates the equatorial waveguide, which extends about 2.5 on either side of the equator; the vertical hatching indicates the coastal waveguide. The coastal Kelvin wave is trapped at the coast poleward of a critical latitude; equatorward of this latitude, westward radiation of energy is possible, and the coastal Kelvin wave is inseparable from the westward propagating Rossby wave. The critical latitudes for Rossby waves at annual (semiannual) period is 42 ( 21 ); hence, annual and semiannual Kelvin waves are inseparable from westward propagating Rossby waves in the north Indian Ocean, and energy leaks at these periods from the eastern boundary into the open ocean (shown by arrows pointing out of the coastal waveguide). Shetye (1998) and Shankar (1998) called this the leaky waveguide of the north Indian Ocean. Energy is also generated by Ekman pumping (shown by the closed circles) in the interior of the basin, this signal also propagating westward as Rossby waves. 87

27

Monthly-mean geostrophic current (cm s 1 ), derived from TOPEX/Poseidon altimetry for January. The upper-left panel shows the 1993-1997 climatology, and the other panels show the geostrophic currents for the January of each of the years 19931997. 89

28

Monthly-mean geostrophic current (cm s 1 ), derived from TOPEX/Poseidon altimetry for July. The upper-left panel shows the 1993-1997 climatology, and the other panels show the geostrophic currents for the July of each of the years 19931997. 9 90

29

Geostrophic currents (cm s

1)

from TOPEX/Poseidon altimetry

for three cycles each during January (left panel) and July (right panel) of 1993. The plots show that the GWMC and GSMC can be traced even in individual TOPEX/Poseidon cycles, even though the currents are more noisy and meander more than in climatology (Figure 4) or in monthly averages (Figures 27 and 28). 93

10

List of Tables

Nomenclature used for currents in this paper. Many of the currents in the Indian Ocean have been referred to by different names in the literature. These are listed here. The rst column lists the name and acronym used by us. The second column lists names and acronyms used earlier; this column is blank if no other name is known to have been used by other authors. The names used here are based on the geographical location of the current, the common practice. Except in the case of the monsoon currents, no allowance has been made for a change in direction with season. 19

Currents and transports associated with the monsoon currents. The direction is given in parentheses in the rst column (N implies northward ow, S southward, etc.) 20 21

2 3

(continued) Observed and model zonal transports (in Sv; 1 Sv 106 m3 s

1 ) in

the top 300 m between 3 45 N and 5 52 N at 80.5 E (south of Sri

Lanka). Positive (negative) values indicate eastward (westward) ow, and the values listed are averages over the period indicated. All observations, except that marked (*), are for 1991; the marked observation is for 1992. The observed transports are derived from the direct current measurements of Schott, Reppin, Fischer, and Quadfasel (1994). The model was forced by climatological wind stress (Hellerman and Rosenstein, 1983). The last two values are average model transports for the winter and summer monsoons, respectively. 4 Zonal transport (in Sv; 1 Sv 58

106 m3 s 1 ) in the top 100 m in

the domain of the monsoon currents. Negative values indicate westward ows and the transports are averages over the periods mentioned. 5
1 Parameters for the 1 2 -layer reduced-gravity model.

61 63

11

1 Introduction

The winds over the Indian Ocean (see Figure 1) north of 10 S reverse direction twice during an year. Over the north Indian Ocean, they generally blow from the southwest during MaySeptember and from the northeast during NovemberFebruary, MarchApril and October being the months of transition with weak winds. In this paper, we refer to MaySeptember as the summer monsoon and November February as the winter monsoon. The winds are much stronger during the summer monsoon than during the winter monsoon. These seasonally reversing monsoon winds over the north Indian Ocean force a seasonally reversing circulation in the upper ocean. The best studied of the seasonally reversing currents are the Somali Current (SC), which ows poleward (equatorward) along the coast of Somalia during the summer (winter) monsoon (see the reviews by Schott, 1983; Shetye and Gouveia, 1998; Schott and McCreary, 2001, and the many references therein), and the current along the equator (called Equatorial Current (EC) in this paper), where eastward surface jets are observed during AprilMay and OctoberNovember (see, for example, Wyrtki, 1973a; OBrien and Hurlburt, 1974; Jensen, 1993; Han, McCreary, Anderson, and Mariano, 1999; Schott and McCreary, 2001, and the many references therein). In the last decade, however, other coastal currents have also received attention. These include the currents along the east coast of India, called the East India Coastal Current (EICC) (Shetye, Shenoi, Gouveia, Michael, Sundar, and Nampoothiri, 1991b; Shetye, Gouveia, Shenoi, Sundar, Michael, and Nampoothiri, 1993; Shetye, Gouveia, Shankar, Shenoi, Vinayachandran, Sundar, Michael, and Nampoothiri, 1996; Shankar, McCreary, Han, and Shetye, 1996; McCreary, Kundu, and Molinari, 1993; McCreary, Han, Shankar, and Shetye, 1996; Vinayachandran, Shetye, Sengupta, and Gadgil, 1996; Shetye and Gouveia, 1998; Schott and McCreary, 2001), the current along the west coast of India, called the West India Coastal Current (WICC) (Shetye, Gouveia, Shenoi, Sundar, Michael, Almeida, and Santanam, 1990; Shetye, Gouveia, Shenoi, Michael, Sundar, Almeida, and Santanam, 1991a; McCreary et al., 1993; Stramma, Fischer, and Schott, 1996; Shankar and Shetye, 1997; Shetye and Gouveia, 1998), and the current along the ArabianSea coast of Oman (McCreary et al., 1993; Flagg and Kim, 1998; Shetye and Gou12

veia, 1998; B ohm, Morrison, Manghnani, Kim, and Flagg, 1999; Shi, Morrison, B ohm, and Manghnani, 2000; Schott and McCreary, 2001). There have been no observational studies of the coastal current along the eastern boundary of the Bay of Bengal. Apart from these coastal currents, the most signicant large-scale currents known in the north Indian Ocean are the open-ocean, seasonally reversing monsoon currents. During the summer monsoon, the monsoon current ows eastward as a continuous current from the western Arabian Sea to the Bay of Bengal; during the winter monsoon, it ows westward, from the eastern boundary of the bay to the western Arabian Sea (see the schematic in Figure 1). We call these currents the Summer Monsoon Current (SMC) and Winter Monsoon Current (WMC), respectively. It is these currents, which transfer water masses between the two highly dissimilar arms of the north Indian Ocean, the Bay of Bengal and the Arabian Sea, that form the subject of this paper.

1.1 Observational background

The existence of seasonally reversing currents in the Arabian Sea has been known for long (see Warren, 1966, for references to medieval Arab sources), but the rst comprehensive study of the circulation in the Indian Ocean was made based on the hydrographic surveys conducted during the International Indian Ocean Expedition (IIOE) during 19591965. The IIOE led to a large number of papers, most of which, as noted above, were devoted to the Somali Current. Though the monsoon currents have not received as much attention, their importance to the circulation in the north Indian Ocean was recognized early. The rst major description of the currents followed soon after the IIOE (D uing, 1970; Wyrtki, 1971, 1973b). Based on these hydrographic data, Wyrtki (1973b) highlighted what he called the seasonally changing monsoon gyre as a gyre unlike those found in the other oceans. In his scheme of circulation, the monsoon gyre during the winter monsoon consists of the westward North Equatorial Current, a southward ow off the Somali coast, and the Equatorial Counter Current, which runs east between the equator and 8 S across the entire width of the ocean. During the summer monsoon, his monsoon gyre consists of the northern portions of the South Equatorial Current, which now extends almost to the equator, the strong So13

mali Current owing north as a western boundary current, and the monsoon current, into which the Counter Current has merged. There is no standard nomenclature for the monsoon currents. The SMC has been called Southwest Monsoon Current (or Drift) or the Indian Monsoon Current or just Monsoon Current, and the WMC has been called the Northeast Monsoon Current or the North Equatorial Current. In this paper, we shall stick to the terminology used above, i.e., Summer and Winter Monsoon Current, following a growing tendency among meteorologists to use the terms summer monsoon and winter monsoon (Sulochana Gadgil, personal communication, 2000). The nomenclature used in this paper is given in Table 1. Wyrtki (1973b) noted that the circulation in the Indian Ocean is complex; the winter monsoon gyre did not close cleanly in the east, with most of the ow from the South Equatorial Counter Current (SCC) owing into the South Equatorial Current (SEC), and a strong branch of the WMC turned north to ow along the Indian west coast, transporting low-salinity water from the Bay of Bengal into the eastern Arabian Sea. The circulation during the winter monsoon was shallow compared to that during the summer monsoon, when intense upwelling was observed in several places and the circulation penetrated deeper, affecting the movement of water masses below the thermocline, especially in the western Arabian Sea. The complexity of the circulation represented by the hydrographic data was seen in the large number of eddies (D uing, 1970; Wyrtki, 1973b), which were found to be connected intimately to the dynamics of the monsoon gyre. The most vigorous of these eddies lay about 300 km offshore of the Somali coast; large parts of the Somali Current were recirculated around this eddy, the Great Whirl. A different picture emerges from the ship-drift data (Defant, 1961; Cutler and Swallow, 1984; Rao, Molinari, and Festa, 1989) or surface-drifter data (Molinari, Olson, and Reverdin, 1990; Shenoi, Saji, and Almeida, 1999a), which tend to show broad eastward (westward) or southeastward ows across the Arabian Sea during the summer (winter) monsoon. Hastenrath and Greischar (1991) used ship drifts, hydrography, and Ekman drift computed from wind-stress climatologies to study the monsoon currents in the Arabian Sea. They concluded that the monsoon currents are essentially Ekman drifts forced by the monsoon winds, the geostrophic contribution to these ows being negligible. Shenoi et al. (1999a) compared hydrography based on the climatologies of Levitus and Boyer (1994) and Levitus 14

et al. (1994) to current estimates from surface drifters, and concluded that the role of geostrophic ows in representing the surface ows varies both geographically and seasonally. The agreement between the drifter data and hydrography was worst during the summer monsoon, when the winds are strong; at this time, the drifters showed southeastward ows all over the Arabian Sea, unlike in hydrography. The dynamic heights, however, do capture the drifter movement in the eastern Arabian Sea during the winter monsoon. Hydrographic data, however, were also used in later studies (Bruce, Johnson, and Kindle, 1994; Bruce, Kindle, Kantha, Kerling, and Bailey, 1998; Donguy and Meyers, 1995; Murty, Sarma, Rao, and Murty, 1992; Murty, Sarma, Lambata, Gopalakrishna, Pednekar, Rao, Luis, Kaka, and Rao, 2000; Gopalakrishna, Pednekar, and Murty, 1996; Vinayachandran, Masumoto, Mikawa, and Yamagata, 1999a), which showed strong geostrophic ows and transports associated with the monsoon currents. These estimates (Table 2) yield current strengths of ports of

10

106

m3 s 1

40 cm s

and trans-

in the upper 4001000 m, which implies that the geo-

strophic ows associated with the monsoon currents are not small, even if they are weaker than the surface Ekman ows in some regions during some seasons. The geostrophic ows estimated by Hastenrath and Greischar (1991) are weak probably owing to the averaging they did to obtain climatological currents and transports in a region; in contrast, the studies mentioned above usually used hydrographic data from individual cruises. The hydrographic data show that the monsoon currents are not found in the same location during a season or across different years; for example, Vinayachandran et al. (1999a) showed that the SMC in the Bay of Bengal intensies and shifts westward as the summer monsoon progresses. Despite these differences, all the observations using hydrography, ship drifts, and surface drifters show that the monsoon currents ow across the breadth of the north Indian Ocean. The branches of the SMC and WMC that ow around the Lakshadweep high and low in the southeastern Arabian Sea (McCreary et al., 1993; Bruce et al., 1994; Shankar and Shetye, 1997) link the circulations in the Arabian Sea and the Bay of Bengal (Figure 1). The SMC ows eastward south of Sri Lanka and into the bay. It is fed by a ow from the southwest near the equator and by the ow around the Lakshadweep low. East of Sri Lanka, the SMC ows northeastward into the Bay of Bengal. A part, however, appears to ow southeastward and crosses 15

the equator near Sumatra in the surface-drifter data (Shenoi et al., 1999a); recent hydrographic data (Unnikrishnan, Murty, Babu, Gopinathan, and Charyulu, 2001) also show that the SMC between 8088 E ows close to the equator and even to its south. The SMC transports high-salinity water (Arabian Sea High Salinity Water) into the bay (Murty et al., 1992; Gopalakrishna et al., 1996). The WMC ows westward south of Sri Lanka, where it divides into two branches, one owing westward into the southern Arabian Sea, and the other owing around the Lakshadweep high into the WICC. The WMC transports low-salinity water (Bay of Bengal Water) into the eastern Arabian Sea, where it is entrained into the Lakshadweep high and spread along the Indian west coast by the WICC (Bruce et al., 1994; Han, 1999; Shenoi, Shankar, and Shetye, 1999b; Shankar and Shetye, 1999). The passage between Sri Lanka and the equator is therefore signicant because the monsoon currents have to ow through it, making it the one location where the monsoon currents are geographically frozen, relatively speaking, unlike in the open ocean, where they meander a lot. It is also here, south of Sri Lanka, that the monsoon currents attain their maximum strength (D uing, 1970), probably because the currents are squeezed through a relatively narrow bottleneck. Hence, it is not surprising that the only direct current measurements of the monsoon currents have been made between Sri Lanka and the equator along 80 30 E (Schott et al., 1994; Reppin, Schott, Fischer, and Quadfasel, 1999). The current-meter and ADCP (Acoustic Doppler Current Proler) observations (Schott et al., 1994; Reppin et al., 1999) show that the SMC and WMC transport

10

106 m3 s

in

the upper 300 m. These direct measurements also conrm the observation in hydrography that the monsoon currents are shallow, with most of the variation being restricted to the upper 100 m. The moored array shows upward phase propagation, implying downward propagation of energy. Most striking is the difference between the Equatorial Current and the monsoon currents, even though both ow together in the same bottleneck. The Equatorial Current includes a large semiannual harmonic, unlike the monsoon currents, which are dominated by the annual harmonic; thus, the Equatorial Current reverses direction four times an year, but the monsoon currents reverse direction twice. Superimposed on these seasonal changes are large intraseasonal oscillations. Though the observations differ in their presentation of the monsoon-current sys16

tem in the north Indian Ocean, they show that the open-ocean currents in the north Indian Ocean extend all across the basin, reverse direction with season, and are relatively shallow compared to the deep western boundary current off Somalia. Given that there are many different interpretations of the monsoon currents (contrast, for example, Hastenrath and Greischar (1991) with Wyrtki (1973b)), it is not surprising that more than one hypothesis exists to explain the observations.

1.2 Theoretical background

Though the existence of the monsoon currents in observations has been known for long, the mechanism leading to their formation has been understood only during the last decade. Early ideas attributed the monsoon currents as seen in ship drifts to direct Ekman forcing by the monsoon winds (Defant, 1961; Hastenrath and Greischar, 1991), and as seen in hydrography to the local curl of wind stress (Murty et al., 1992). The north Indian Ocean is essentially a tropical basin with its northern boundary located south of

25 N. The pioneering work of Matsuno

(1966), Moore (1968), and Lighthill (1969) showed that baroclinic waves propagate fast in the tropics, and it is now appreciated that the open-ocean, equatorial, and coastal currents in the north Indian Ocean, all of which reverse seasonally, are manifestations of direct forcing (Ekman drift) by the monsoon winds, and of equatorial and coastal long, baroclinic waves generated by the seasonal winds (Cane, 1980; Potemra, Luther, and OBrien, 1991; Yu, OBrien, and Yang, 1991; Perigaud and Delecluse, 1992; McCreary et al., 1993; Shankar et al., 1996; McCreary et al., 1996; Vinayachandran et al., 1996; Shankar and Shetye, 1997; Shankar, 1998; Vinayachandran and Yamagata, 1998; Han, 1999; Han et al., 1999; Shankar, 2000). The small size of the basin implies that these waves can traverse the basin in a few months. This is unique to the north Indian Ocean. The framework that has evolved in the last decade suggests a unity of dynamics in the north Indian Ocean. The equatorial Rossby wave, the equatorial Kelvin wave, and the coastal Kelvin waves merge the Arabian Sea, the Bay of Bengal, and the equatorial Indian Ocean into a single dynamical entity, the north Indian Ocean, which must be modelled as a whole even to simulate circulation in its parts. Circulation at any point is decided by both local forcing and remote forcing, whose 17

Schematic of circulation in the Indian Ocean

30N

January
20N

Om

an

WICC

India
EICC

Arabian Sea

10N
ia So m al
WMC SC EC

LH

Bay of Bengal

Andaman Sea

WMC
m Su

0
EACC

Sri Lanka
SCC SEC

ra at

10S 40E 30N 50E 60E 70E 80E 90E 100E

July
20N

Om

an

WICC

India
EICC Bay of Bengal Andaman Sea

10N
m al
GW

Arabian Sea SMC

SH LL

SMC SMC
Su

So
SC

ia

0
EACC

EC SCC

Sri Lanka

m at ra

SEC

10S 40E 50E 60E 70E 80E 90E 100E

Fig. 1. Schematic representation of the circulation in the Indian Ocean during January (winter monsoon) and July (summer monsoon). The abbreviations are as follows. SC, Somali Current; EC, Equatorial Current; SMC, Summer Monsoon Current; WMC, Winter Monsoon Current; EICC, East India Coastal Current; WICC, West India Coastal Current; SCC, South Equatorial Counter Current; EACC, East African Coastal Current; SEC, South Equatorial Current; LH, Lakshadweep high; LL, Lakshadweep low; GW, Great Whirl; and SH, Socotra high.

18

Name of current (acronym)

Other commonly used names (acronyms)

Winter Monsoon Current (WMC)

Northeast Monsoon Current (NMC), North Equatorial Current (NEC)

Summer Monsoon Current (SMC)

Southwest Monsoon Current (SMC), Indian Monsoon Current (IMC), Monsoon Current

Equatorial Current (EC)

Equatorial jet or Wyrtki jet (when owing eastward)

South Equatorial Counter Current (SCC)

South Equatorial Counter Current (SCC), Equatorial Counter Current (ECC)

Somali Current (SC) West India Coastal Current (WICC) East India Coastal Current (EICC)

Table 1 Nomenclature used for currents in this paper. Many of the currents in the Indian Ocean have been referred to by different names in the literature. These are listed here. The rst column lists the name and acronym used by us. The second column lists names and acronyms used earlier; this column is blank if no other name is known to have been used by other authors. The names used here are based on the geographical location of the current, the common practice. Except in the case of the monsoon currents, no allowance has been made for a change in direction with season.

19

Current/Transport and direction

Location and period

Remarks

9 Sv (E)

78 E, 35 N; June 1992.

14 Sv (E) 9.3 Sv (E) 6 Sv (E) 14 Sv (W) 78 Sv (W)

68 E, 47 N; July 1995. 68 E, 26 N;
7279 E, 68 N; February 1993. 25 N; January 1996, February 1993, March 1992. 6 N, east of Sri Lanka; JanuaryFebruary climatology September 1993.

80 E, 35 N; July 1992.

Geostrophic transport with respect to 400 m, based on XBT data with salinity from T-S relation based on climatology (Levitus, 1982); Source Murty et al. (2000). As above. As above. As above. As above. As above, but average values for the given months over three different transects. Geostrophic transport with respect to 400 m, based on TOGA XBT data for 19851989, with salinity from T-S relation based on climatology (Levitus, 1982). Source Donguy and Meyers (1995). As above.

13 Sv (W)

11 Sv (NE)

19 Sv (N)

10 N, 6670 E; Winter monsoon, 1965.

West of Lakshadweep high (see Figure 1); JanuaryFebruary climatology.

15 Sv (S) 35 cm s  1 , 5.2 Sv (S)

10 N, 6772 E; Summer monsoon, 1963. 8 N, 72 10 75 E; August 1993.

Geostrophic transport with respect to 1000 m, evaluated from IIOE data; Atlantis II, Cruise 15. West of Lakshadweep high. Source Bruce et al. (1994). As above, but for IIOE data from Atlantis II, Cruise 8. Current from ADCP; geostrophic transport with respect to 1000 m from hydrography. West of Lakshadweep low (see Figure 1). Source Stramma et al. (1996).

Table 2 20 Currents and transports associated with the monsoon currents. The direction is given in parentheses in the rst column (N implies northward ow, S southward, etc.)

Current/Transport and direction

Location and period

Remarks

8 Sv (E), with a peak transport of 24 Sv

80 30 E; Summer monsoon, 1991.

12 Sv, 10 Sv (W), with a peak transport of 25 Sv 40 cm s 

1

80 30 E; Winter monsoon, 1991, 1992.

Transport estimated from current-meter moorings and ADCP. Moorings located south of Sri Lanka between 5 39 N and 4 10 N (transport estimated over top 300 m between 3 45 N and 5 52 N). Source Schott et al. (1994). As above. The rst value is for 1991, the second for 1992.

(N)

8789 E, 14 N; Summer monsoon, 1984.

Maximum of geostrophic current between 50100 m depth, estimated from hydrographic data with respect to 1000 m. Source Murty et al. (1992). Geostrophic current and transport with respect to 1000 m, estimated using hydrographic data. Source Gopalakrishna et al. (1996). As above. Geostrophic current and transport with respect to 400 m, estimated from TOGA XBT data during 19851996. The current is restricted to the top 200 m and moves westward as the season progresses. Source Vinayachandran et al. (1999a).

40 cm s  1 , 17 Sv (N)

8789 E, 11 N; July 1993.

40 cm s  1 , 14 Sv (N) 40 cm s  1 , 12 Sv (N)

8789 E, 12 N; August 1991. 8185 E, 6 N; July climatology.

Table 2 (continued)

21

signals are carried by the equatorial and coastal waves. This is seen in numerical simulations using both layered models and multi-level general circulation models. The layered models, in particular, emphasize the quasi-geostrophic dynamics that leads to the eddies and meanders so typical of hydrographic observations of the monsoon currents. In these models, the monsoon currents appear as the fronts of Rossby waves. For example, McCreary et al. (1993) and Shankar and Shetye (1997) emphasized the role of Rossby-wave radiation in forcing the Lakshadweep high and low, which are intimately connected to the monsoon currents seen in the hydrography of the southeastern Arabian Sea, and McCreary et al. (1993) and Vinayachandran et al. (1999a) showed that the westward movement of the SMC across the bay was a result of Rossby-wave radiation from the eastern bay and the generation of Rossby waves by Ekman pumping in the interior of the bay. Notwithstanding the success of numerical models in simulating the circulation in the north Indian Ocean, there remain several unanswered questions, especially with respect to the monsoon currents. The nature of currents associated with Rossby waves is strikingly different in places from the observed ship-drift and surfacedrifter data, this being more true of the SMC during the summer monsoon. Yet, all authors generally claim success for their respective models, attributing the differences between simulations and observations to Ekman ow. Given that both ship drifts and surface drifters show a circulation that differs signicantly from that seen in hydrography, a pertinent question is: what really are the monsoon currents? How do we describe them, and what are the causes for their existence? It is these questions that we seek to answer in this paper, and we begin with the commonly accepted denition of the monsoon currents as the open-ocean, seasonal currents that link the circulations in the Arabian Sea and the Bay of Bengal. In section 2, we assemble observations on ship-drifts, Ekman drift estimated from winds, geostrophic currents computed from sea-level anomalies obtained from satellite altimetry, and hydrography to dene the surface circulation associated with the monsoon currents. Numerical simulations with an Oceanic General Circulation
1 Model follow in section 3. In section 4, we use a 1 2 -layer reduced-gravity model

to analyze the forcing mechanisms. Section 5 concludes the paper.

22

2 Observations

To dene the climatological monsoon-current system and the associated circulation in the north Indian Ocean, we use climatological wind-stress data to estimate surface Ekman drift, and satellite altimeter data from TOPEX/Poseidon to estimate the geostrophic contribution to the surface currents. The Ekman drift and geostrophic current, and the net surface current due to them are compared to surface currents represented by ship drifts. We show that dening the monsoon-current system requires more than one observational method because each method accentuates certain aspects of the ow eld, thereby emphasizing a particular view of the surface currents.

2.1 Ekman drift

The Ekman drift is computed using the Ekman spiral method. The surface Ekman drift ows at 45 to the right (left) of the wind in the northern (southern) hemisphere and its magnitude is given by (Pond and Pickard, 1983) VE

 A f  

1 2 
2

(1)

where is the magnitude of the wind stress, A is the vertical eddy diffusivity and  f  is the magnitude of the Coriolis parameter. We use A 10 m2 s

and from the

wind-stress climatology of Hellerman and Rosenstein (1983) (Figure 2) to obtain the monthly climatology of the surface Ekman drift in the Indian Ocean north of 10 S (Figure 3), excluding the region within 2.5 of the equator, where (1) does not apply. The winter monsoon sets in during November, and the Ekman drift reverses direction to ow westward in the Arabian Sea and the Bay of Bengal. The drift is weak in the eastern Arabian Sea and the eastern Bay of Bengal. The winter monsoon strengthens in December; so does the Ekman drift. The magnitude of the drift is

15 cm s

in the western bay and

25 cm s

in the western Arabian Sea. The

winter monsoon peaks in January, with northeasterly winds over most of the north Indian Ocean, and the surface Ekman drift is westward. Current strengths approach 23

20 cm s

south of Sri Lanka and 30 cm s

in the southwestern Arabian Sea.

Ekman drift is weak in the eastern Arabian Sea and northern Bay of Bengal. The winter monsoon weakens in February; so does the Ekman drift. During March April, the months of transition between the winter and summer monsoons, Ekman ow is weak, except in the northern bay in April, where it has reversed direction since January to ow eastward. In both basins, a weak anticyclonic gyre is seen during March.

With the onset of the summer monsoon in May, the winds begin to blow from the southwest over most of the north Indian Ocean. The Ekman drift reverses to ow eastward over most of the Arabian Sea; it is southeastward in the eastern Arabian Sea, where the winds blow more from the west. In the bay, Ekman drift is eastward, except in the north and the west, where it tends to be oriented parallel to the coast. The current strengths are now

25 cm s

southeast of Sri Lanka and

off the Somali coast. The surface Ekman drift strengthens all over the north Indian Ocean in June. It is eastward and southeastward in the Arabian Sea, with a slight anticyclonic tendency; the drift is eastward in the bay. The direction remains the same through JuneSeptember, but the Ekman drift peaks in July, when current strengths approach

40 cm s

in parts of the bay and

100 cm s

off the

Somali coast. October is the month of transition between the summer and winter monsoons, with weak winds all over the north Indian Ocean. The Ekman drift is weaker than 5 cm s

over most of the basin, with currents of

20 cm s

seen

only southeast of Sri Lanka.

During the summer monsoon, the Ekman drift is strong in the western and central Arabian Sea and south of Sri Lanka; it is relatively weak in the eastern parts of the Arabian Sea and Bay of Bengal. During the winter monsoon, the spatial variation in magnitude is much less. Most striking is the spatial uniformity of the Ekman drift in comparison with the eddy-like circulations seen in hydrography (D uing, 1970; Wyrtki, 1971; Murty et al., 1992). It strengthens and weakens almost all over the north Indian Ocean at the same time, in harmony with the seasonally reversing winds. This lack of spatial structure in the Ekman drift implies that geostrophy must make a signicant contribution to the surface current in the north Indian Ocean. 24

Fig. 2. Wind stress (dyne cm  2 ) from the climatology of Hellerman and Rosenstein (1983).

25

Fig. 3. Surface Ekman drift (cm s  spiral formula.

in the Indian Ocean. The drift is computed from the

wind-stress climatology of Hellerman and Rosenstein (1983) and is based on the Ekman

26

Fig. 4. Monthly climatology of sea-level anomalies (cm, left panel) and geostrophic current

(cm s  1 , right panel) in the Indian Ocean. Negative anomalies are indicated by dashed

contours and the contour interval is 5 cm. The sea-level anomalies and geostrophic currents are derived from the TOPEX/Poseidon altimeter data for 19931997.

27

Fig. 4. (continued)

28

Fig. 4. (continued)

29

Fig. 4. (continued)

30

2.2 Geostrophic currents

Climatological hydrographic data (Levitus and Boyer, 1994; Levitus et al., 1994) are incapable of resolving, in both space and time, the geostrophic component of the rapidly changing monsoon circulation of the north Indian Ocean. Hence, to compute geostrophic currents, we use the sea-level anomalies from TOPEX/Poseidon altimetry, which are available on a 0 25  0 25 grid (Le Traon, Gaspar, Bouyssel, and Makhmara, 1995; Le Traon and Ogor, 1998; Le Traon, Nadal, and Ducet, 1998). We construct a monthly climatology of sea-level anomalies using the 10day repeat-cycle data for 19931997, and use this climatology to compute surface geostrophic currents in the Indian Ocean (Figure 4), excluding the region within 2.5 of the equator. Though the period of averaging is small for making a climatology, it is based on the best data available, and throws light on the monsoon-current system. Unlike the Ekman drift, which shows little spatial structure, the surface geostrophic ow is dominated by eddies. The geostrophic monsoon currents do not form, or decay, across the basin all at once. Instead, patches of the currents appear or decay at different times. The monsoon currents can be traced as continuous transbasin ows only in their mature phase. At other times, incipient or relic patches are identiable in the surface geostrophic ow. By November, with the onset of the winter monsoon, the geostrophic SMC (GSMC) breaks into separate currents in the Arabian Sea and the Bay of Bengal, this split into two relic currents being caused by the continuity of the ow along the coast of the Indian subcontinent, and by the relentless westward propagation of the sea-level anomalies associated with the GSMC. In the Arabian Sea, the relic GSMC is restricted west of 72 E, and it appears as a geostrophic ow around a low in sea level; this low has propagated westward from the Indian coast, where it appeared during the summer monsoon as the Lakshadweep low. The relic GSMC ows southwestward to the west of the low and eastward to its south. In the bay, the relic GSMC ows northeastward as the eastern arm of a geostrophic ow around a low in sea level off the east coast of India; its western arm is the equatorward EICC. The geostrophic WMC (GWMC) rst appears during November as a westward ow south of Sri Lanka. This incipient GWMC is fed by the equatorward EICC and it feeds, in turn, the poleward WICC. 31

The relic GSMC continues to shift westward, and by December, it is restricted to the west of 65 E in the Arabian Sea. The sea-level low abuts the Somali coast, and the southwestward GSMC to its west is now synonymous with the equatorward Somali Current. In the bay, what remains of GSMC is barely recognizable even as a relic. In December, the GWMC is fed by the EICC, but it also appears in the southern bay as a weak westward ow southeast of Sri Lanka. It ows west beyond 70 E before turning to ow around the Lakshadweep high, which forms by this time off southwest India and Sri Lanka. By January, the relic GSMC is restricted to a minor eastward ow west of 55 E, and the most signicant geostrophic ows in the north Indian Ocean are the GWMC, the WICC, and the recirculations around eddies in western Arabian Sea and western Bay of Bengal. The EICC reverses to ow poleward off Sri Lanka, and the GWMC now appears as a westward ow across the southern bay at 6 N; it ows southwestward in the eastern bay. The GWMC ows westward halfway across the southern Arabian Sea at 5 N, where it turns to ow northeastward around a high in sea level and into the WICC. A branch of the GWMC, however, turns to ow around the Lakshadweep high and into the WICC. The western high is distinct from the Lakshadweep high and retains its identity within a region of high sea level even later in the season. Thus, the GWMC ows westward to the south of the sea-level high and eastward to its north. Westward propagation of the sea-level anomalies is evident in the north Indian Ocean, and the relic GSMC nally disappears in February. The southwestward GWMC in the eastern bay has shifted west since January and is located

93 E.

The sea-level highs in the southern Arabian Sea have also spread and shifted westward, with the result that a continuous geostrophic ow exists around a sea-level high in the southern Arabian Sea. The GWMC ows westward (eastward) to the south (north) of the high. The eastward GWMC, as earlier in the season, feeds a current parallel to the Indian west coast; this poleward current was the WICC in January, but has since shifted offshore. Westward propagation of the sea-level anomalies continues during March. In the Bay of Bengal, the southwestward GWMC is located at 83 E and the GWMC no longer exists in the eastern bay. This part of the GWMC is coupled to a distorted 32

anticyclonic gyre in the western bay. The sea-level high in the southern Arabian Sea extends across the basin. The GWMC ows around the high, and at the western boundary, it is synonymous with the poleward Somali Current between 46 N. The eastward GWMC to the north of the high is also fed by an equatorward Somali Current. To the north of the eastward GWMC is a low in sea level, and the southwestward ow to its north feeds the equatorward Somali Current. The circulation during April is dominated by eddies. The GWMC in the bay has shifted westward and it now intersects the east coast of Sri Lanka; this breaks the GWMC into separate currents in the bay and the Arabian Sea. The relic GWMC in the bay is the eastern arm of an anticyclonic gyre. The relic GWMC in the Arabian Sea has weakened and meanders more, but its spatial structure is similar to that in March. During May, the relic GWMC in the bay ows southwestward from the central bay to Sri Lanka. In the Arabian Sea, it appears primarily as a westward ow at

7 N and is now fed by the equatorward WICC. The sea-level highs are now well

offshore, and the eastward ow to their north is weaker and meanders more than in April. An eastward ow appears in the southern bay between 48 N; it ows right across the basin, from the eastern boundary to Sri Lanka. This is the incipient GSMC, and it appears rst in the bay even as the summer monsoon sets in. In June, the relic GWMC is restricted to the central and western bay, owing southwestward from (90 E, 16 N) to the northern tip of Sri Lanka. In the Arabian Sea, the relic GWMC is traceable as a meander to the south and north of the sea-level highs in the southwest of the basin. The eastward relic of the GWMC, and the southeastward ow to the east of the high can be considered the incipient GSMC in the western and central Arabian Sea because the sea-level highs that propagated westward across the Arabian Sea lose their identity in the eddies off Somalia during July, when the anticyclonic geostrophic ow around them is part of the GSMC. By June, the GSMC is evident in the bay, owing northeastward from the southern tip of Sri Lanka to (90 E, 14 N); a branch of the GSMC also recirculates around a low in sea level to the east of Sri Lanka. The GSMC in the bay is pushed westward by the westward movement of a sea-level high from the eastern equatorial Indian Ocean. The GSMC also appears in the Arabian Sea as a ow around the 33

Lakshadweep low off southwest India, and it is fed by the equatorward WICC. In the southern Arabian Sea, there is eastward ow at 5 N between 6575 E, merging into the GSMC south of Sri Lanka. There is, however, no clear link yet between the GSMC in the eastern Arabian Sea and the incipient GSMC (or relic GWMC) in the west. The GSMC shifts westward in the bay during July, pushed by the sea-level high to its east. One branch appears as a northeastward ow from the southern tip of Sri Lanka almost to the northeastern boundary of the bay. Around 87 E, another branch of the GSMC ows due north and feeds into the relic GWMC, which has shifted westward and now appears as the eastern arm of an anticyclonic circulation around a high in sea level in the western bay. The GSMC south of Sri Lanka peaks in July with current strengths reaching 60 cm s 1 . It is fed by a ow around the Lakshadweep low, this being a branch of the GSMC that is fed by the WICC, and by an eastward ow across the southern Arabian Sea. This eastward GSMC in the south is the continuation of the GSMC in the western and central Arabian Sea; this branch of the GSMC ows geostrophically around the sea-level high (low) in the western (eastern) Arabian Sea. Though the GSMC now exists as a continuous transbasin current from the northern limits of the Somali Current to the southern tip of Sri Lanka and into the Bay of Bengal, the current consists of two branches, and the maximum inow into the GSMC south of Sri Lanka comes from the branch that ows around the Lakshadweep low, rather than from the eastward GSMC across the southern Arabian Sea. Westward propagation continues in the Bay of Bengal, and the GSMC is detached from the eastern boundary by August. It can be traced as a continuous current from the southern tip of Sri Lanka to the northern bay, though there are cyclonic eddies associated with it east of Sri Lanka and in the central bay. In the Arabian Sea, the Lakshadweep low propagates westward, taking the GSMC and WICC with it; the coastal current off the Indian west coast collapses. Westward propagation of the sea-level highs in the western Arabian Sea compresses them and strengthens the geostrophic ow around them. Except for this part of the basin, the GSMC is weaker in August than in July. With the monsoon winds weakening during September, so does the GSMC, ex34

cept in the western bay and, to a lesser extent, in the western Arabian Sea, where westward propagation compresses the sea-level lows and highs and strengthens the geostrophic ow associated with them. The Lakshadweep low is now detached from the coast and a part of the GSMC recirculates around it. The sea-level high that propagated westward across the southern bay is now almost at the southern tip of Sri Lanka, and the GSMC is squeezed against the coast. There is geostrophic ow around this high, and a westward current exists to its south, owing from the Andaman Sea to south of Sri Lanka. The sea-level high that pushed the GSMC westward through the bay abuts the Sri Lankan coast in October, breaking the connection between the GSMC in the bay and the Arabian Sea. In the bay, the relic GSMC ows from the east coast of Sri Lanka into the northcentral bay, and it is associated with, and is fed by the recirculations around, sea-level lows in the western bay. The EICC reverses in October to ow equatorward as the western arm of these cyclonic geostrophic ows. The Lakshadweep low is now well offshore and the WICC also reverses to ow poleward off southwest India. The eddies and sea-level highs in the western Arabian Sea begin disintegrating, but the relic GSMC is still traceable as a distinct southwestward current in the western Arabian Sea; this relic current turns to ow eastward in the southern Arabian Sea. Longitude-time plots of the sea-level anomalies (Figure 5) clearly show westward propagation at all latitudes, with the speed of propagation decreasing with increasing latitude. Similar westward propagation has been noted earlier in hydrography (Kumar and Unnikrishnan, 1995; Unnikrishnan, Kumar, and Navelkar, 1997; Rao, 1998) and altimetry (Perigaud and Delecluse, 1992), and has been attributed to westward propagating Rossby waves. Most striking, however, is the break in the westward propagating signal at 80 5 E, south of Sri Lanka, throughout the year, even though the GSMC and GWMC form continuous currents from the Arabian Sea to the Bay of Bengal. Latitude-time plots of the zonal component of the geostrophic current at 80 5 E (Figure 6) show that the signal south of Sri Lanka is dominated by the annual harmonic north of

3 N, and by the semiannual har-

monic in the vicinity of the equator; this is also seen in the zonal geostrophic current derived from the climatologies of Levitus and Boyer (1994) and Levitus et al. (1994) (Figure 7). This is in agreement with direct current measurements in this 35

Fig. 5. Longitude-time plots of the monthly climatology of TOPEX/Poseidon sea-level anomalies (cm) at 10 N (top panel), 8 N (middle panel), and 5 N (lower panel). Negative anomalies are indicated by dashed contours and the contour interval is 5 cm. Westward is a break in the signal at 80  5 E (south of Sri Lanka), even though there is no land barrier there. propagation is evident in both Arabian Sea and Bay of Bengal at all latitudes. At 5 N, there

36

Fig. 6. Latitude-time plots of the monthly climatology of zonal geostrophic current

ward ow is indicated by dashed contours and the contour interval is 5 cm s  1 .

(cm s  1 ), derived from TOPEX/Poseidon altimetry, at 80  5 E (south of Sri Lanka). West-

37

Fig. 7. Depth-time plots of the meridionally averaged zonal geostrophic current (cm s  1 ) derived from the climatologies of Levitus and Boyer (1994) and Levitus et al. (1994). The Westward ow is indicated by dashed contours and the contour interval is 5 cm s  1 . left (right) panel shows the current averaged over 0 3 N (3 6 N). The depth is in metres.

38

region (Schott et al., 1994; Reppin et al., 1999), and suggests that the monsooncurrent system south of Sri Lanka is distinct from the Equatorial Current farther south, even though it is difcult to distinguish between these two regimes during, say, DecemberFebruary, when the WMC ows westward, as does the Equatorial Current. The core of the climatological GWMC and GSMC south of Sri Lanka is not at the coast, but is separated from it. The core of the GWMC is at 5 N during NovemberDecember and at GSMC is at 4 75 N during JuneJuly and at 5 N during AugustSeptember.

4 5 N during JanuaryFebruary. The core of the

The geostrophic currents derived from altimetry show that excepting the GSMC and the GWMC in the vicinity of Sri Lanka, strong geostrophic ows in the north Indian Ocean are associated either with coastal currents, or with recirculations around eddies or highs and lows in sea level. These ows, however, are not weak. For example, south of Sri Lanka, the GWMC (GSMC) attains speeds of 50 cm s

during NovemberJanuary (JuneSeptember). These magnitudes are comparable to the Ekman drifts, except in the western Arabian Sea. Hence, we expect geostrophy to make a signicant contribution to the surface circulation in the north Indian Ocean.

2.3 The net ow at the surface

In the preceding subsections, we presented the monthly climatologies of Ekman drift and geostrophic ow at the surface. The sum of these two components constitutes, barring a residual, the net ow (NF) at the surface (Hastenrath and Greischar, 1991). We compare this estimated surface current with ship drifts (from the Ocean Current Drifter Data CDROMs NODC-53 and NODC-54, NODC, US Department of Commerce, NOAA) in Figure 8. The NF matches the ship drifts well in the regime of the monsoon currents. During November, geostrophy dominates in the Bay of Bengal and along the coast of the Indian subcontinent. The equatorward EICC, the westward WMC south of Sri Lanka, and the poleward WICC form a continuous current in both NF and ship drifts; the latter, however, are stronger than the former. Elsewhere, the picture is more confused. Both geostrophy and Ekman drift, however, combine to keep alive 39

Fig. 8. The net ow (NF) at the surface (cm s  1 , left panel), computed as the sum of Ekman drift (Figure 3) and geostrophic ow (Figure 4), and ship drifts (cm s  1 , right panel). The source for the ship drifts are the Ocean Current Drifter Data CDROMs NODC-53 and NODC-54 (NODC, US Department of Commerce, NOAA).

40

Fig. 8. (continued)

41

Fig. 8. (continued)

42

Fig. 8. (continued)

43

a weak eastward ow, a relic of the SMC, in the southern bay. Unlike in October, this current does not extend to Sri Lanka. The relic SMC in the Arabian Sea appears as a southwestward ow from ( 65 E, 12 N) to ( 55 E, 4 N), where it turns to ow eastward. Ekman drift strengthens in December, but it dominates geostrophy only in the northern Arabian Sea, where geostrophic currents are negligible. The drift, however, is crucial for producing good agreement between the NF and ship drifts. It combines with geostrophy to produce a strong westward WMC in the southern bay. The Ekman drift accentuates the westward geostrophic ow to the north of the sea-level low in southwestern Arabian Sea (Figure 4), and attenuates the eastward GSMC to its south; as a consequence, the WMC appears farther north in western Arabian Sea compared to eastern Arabian Sea. It branches around 65 E, one branch owing around the Lakshadweep high, the other continuing to ow westward. The latter branch of the WMC is synonymous with the relic SMC to the north of the sealevel low. It is the Ekman drift that enables the WMC to extend from the eastern bay to the western Arabian Sea as early as December, before the winter monsoon peaks. The NF in January is similar to that in December, but it is stronger. Both geostrophy and Ekman drift contribute to the WMC in the bay. In the Arabian Sea, the Ekman drift is responsible for extending the WMC west of 60 E, but geostrophic ow around the sea-level highs dominates in the east. There is gentle westward drift across the northern Arabian Sea, where geostrophic ow is weak. The NF is in excellent agreement with ship drifts both in the domain of the WMC and outside it, even though the ship drifts are noisier in the Arabian Sea. During February, Ekman drift is weak, and the NF is dominated by eddies, resulting in relatively poor agreement with ship drifts outside the domain of the strong WMC. Owing to the dominance of geostrophy, the WMC is almost identical to the GWMC. This dominance of geostrophy continues through to April. During March, Ekman drift contributes only in the southwestern Arabian Sea, where it accentuates the westward WMC to the south of the sea-level highs and attenuates the eastward WMC to their north. As in February, there is good agreement between the NF and 44

ship drifts only in the domain of the WMC. The major discrepancy occurs in the bay, where ship drifts show a basin-wide anticyclonic gyre, but the NF shows a distorted anticyclonic gyre restricted to the western bay. The poleward EICC is also stronger in the ship drifts. The winds are weakest in April, and the WMC is identical to the GWMC. The GWMC, as described earlier, appears in three distinct parts: a southwestward relic in the western bay, a meandering westward ow south of a high in sea level in the southern Arabian Sea, and a meandering eastward ow to the north of the high sea level. The two relics of the WMC in the Arabian Sea are connected by a poleward Somali Current. This is in agreement with the ship drifts, in which too relics of the WMC appear as three separate currents; the only difference is that the Somali Current ows poleward all along the coast in the ship drifts, but not in the NF. The EICC continues to ow poleward in both ship drifts and NF, but the latter exhibits a richer structure. The SMC appears in the bay in both ship drifts and NF during May, but it is broader in the latter because of the strong Ekman drift across the southern bay. The SMC can be traced as a continuous current from the northern limit of the poleward Somali Current to the eastern bay in both data sets. A primarily geostrophic WICC feeds into the SMC. Ekman drift and geostrophic ow combine to produce the SMC in the Arabian Sea, Ekman drift dominating in the west and geostrophic ow in the east. In the NF, these two components combine to form a curving ow across the Arabian Sea, but this current ows zonally across the basin at 10 N in the ship drifts. Ekman drift swamps geostrophic ow when the summer monsoon winds strengthen in June. This is more so in the Arabian Sea, where the currents cross the isolines of sea level, and a little less in the bay, where weaker winds compared to the Arabian Sea (Figure 2) combine with the multiplicity of eddies (Figure 4) to make the NF different from the uniform eastward Ekman drift. Though the ship drifts are a little noisier, the dominance of Ekman drift is seen in them too. In June, therefore, the SMC appears as an eastward or southeastward ow in the Arabian Sea. The summer-monsoon winds peak in July; so does the Ekman drift. Geostrophy makes a signicant contribution to the surface current only in the bay and south of 45

Sri Lanka. Again, there is good agreement between the NF and ship drifts, but the latter appear to be affected somewhat less by the direct Ekman forcing. Apart from a weakening of the SMC owing to slightly weaker winds, the NF and ship drifts during August are as in July. The summer-monsoon winds continue to weaken through September, making geostrophy more relevant to the surface ow eld. In the Arabian Sea, anticyclonic curvature of the Ekman ow increases compared to August, and Ekman drift dominates, except in the southeast, where geostrophic ow around the Lakshadweep low is important, and in the vicinity of the Great Whirl off Somalia, where there is strong geostrophic ow around a high in sea level. In the bay, Ekman drift dominates in the south, forcing broad eastward SMC, but geostrophy is strong enough to make a part of the ow turn northeastward into the central bay. In the rest of the bay, Ekman drift is weaker, and geostrophic ow around eddies dominates. Once again, there is good agreement between the NF and ship drifts, except in the western bay; the EICC is equatorward in the ship drifts, but the NF presents an eddy-dominated ow eld. October being a month of transition, Ekman drift is negligible, except in the southern bay and to the southwest of Sri Lanka. Nevertheless, it combines with geostrophy to sustain the SMC as a continuous current from the central Arabian Sea to the eastern bay. The SMC is split in the bay, a separate relic existing as a geostrophic northeastward ow from eastern Sri Lanka to the central bay. These features are also evident in the ship drifts, except that the relic SMC is farther east. The ship drifts also show strong currents in the western Arabian Sea, but this is not seen in the NF. The Ekman drift eliminates the westward propagation seen in the zonal geostrophic currents (Figure 5). It is dominated by the annual harmonic south of Sri Lanka, and therefore, though weaker than the geostrophic ow here, accentuates the annual signal in the NF. Ekman drift dominates the surface circulation in the north Indian Ocean during the summer monsoon, with geostrophic ow being signicant only south of Sri Lanka, around the eddies off Somalia, and in the Bay of Bengal. Geostrophic ow and Ekman drift combine to produce the surface ow depicted in the ship-drift data during the winter monsoon. During the transitions between 46

the monsoons, in MarchApril and October, the geostrophic ow is still strong, especially along the Indian coasts and in the domain of the monsoon currents, and it dominates the circulation. Thus, both Ekman drift and geostrophic currents are important components of the surface circulation associated with the monsoon currents.

47

3 The monsoon currents in an OGCM

The foregoing description of the monsoon currents based on observations shows that both Ekman drift and geostrophy contribute to the surcial circulation in the north Indian Ocean. The former decays rapidly with depth, while the latter is expected to be signicant even in the subsurface layers. To ascertain the vertical structure of the monsoon currents and to make estimates of the transports due to them, we use a multi-level Ocean General Circulation Model (OGCM) to simulate the circulation associated with the monsoon currents.

3.1 Numerical model

The OGCM is based on the Modular Ocean Model (Pacanowski, 1996). The model domain covers the tropical Indian Ocean (30 S30 N, 30115 E). The model has realistic coastline and topography based on the ETOPO5 data set. The horizontal the top 100 m. Horizontal eddy viscosity and diffusivity are 2 107 cm2 s 107 cm2 resolution is 0 33 0 33 and there are 25 levels in the vertical, of which 8 are in

and

respectively, and vertical mixing is parameterized using the scheme of

Pacanowski and Philander (1981). The model is spun up for 5 years from a state of rest and climatological temperature (Levitus and Boyer, 1994) and salinity (Levitus et al., 1994) using the wind-stress climatology of Hellerman and Rosenstein (1983). The model reproduces the monsoon circulation in the Indian Ocean reasonably well. Vinayachandran et al. (1999a) compared the SMC along 6 N in a similar model with TOPEX/Poseidon altimetry and geostrophic currents. They found that the model SMC near Sri Lanka compares well with that derived from TOPEX/Poseidon altimetry, though the core of the SMC east of Sri Lanka was weaker in the model compared to that derived from XBT data. Vinayachandran, Saji, and Yamagata (1999b) used the model to investigate the unusual conditions in the equatorial Indian Ocean in 1994 and noted that the model Equatorial Current is consistent with direct current measurements (Reppin et al., 1999). 48

3.2 The model circulation

The ow in the rst 3 model levels is dominated by Ekman ow, geostrophy dominating below this. At each level, however, both contribute to the model ow eld. Hence, we present the model circulation at 5 m (Figure 9) and 35 m (Figure 10) depths, comparing them with the Ekman drift (Figure 3) and geostrophic ow (Figure 4). Then we present the depth-averaged ow in the top 50 m (Figure 11), which is more representative of the shallow monsoon currents than the surface ow alone. As the model ow is a composite of both Ekman and geostrophic ows, there are several additional features in the model ow at 5 m (Figure 9) compared to the estimated Ekman drift (Figure 3). For example, unlike the estimated Ekman drift, the model ow at 5 m contains strong coastal currents. These include the Somali Current, the coastal current off Oman, the EICC, and the WICC. The EICC ows equatorward during November and feeds the westward WMC south of Sri Lanka, this ow standing out from the eastward jet at the equator. In the western Arabian Sea, there is a westward drift at 9 N; this is the relic SMC or the incipient WMC. The trans-basin WMC is seen during JanuaryMarch. Most of the ow is westward at

5 N, but a branch of the WMC turns to ow around the Lakshadweep high

in the southwestern Arabian Sea; this branch of the WMC propagates westward. Apart from the WMC, the outstanding feature during the winter monsoon is the anticyclonic gyre in the bay. Thus, during the winter monsoon, when the winds are relatively weak, geostrophy dominates even at the surface, Ekman drift modulating the geostrophic currents. Ekman drift dominates the model surface ow during the summer monsoon. This is more so in the Arabian Sea, where the winds are strongest (Figure 2). Nevertheless, the coastal currents, the intrusion of the SMC into the bay, and the ow around the eddies off Somalia, which are due to geostrophy, are evident even at 5 m. The model SMC south of Sri Lanka is fed by the general southeastward drift across the Arabian Sea; unlike in the NF, the contribution from the ow due west is much less. The model currents at 35 m (Figure 10) are dominated by geostrophy because the Ekman drift decays exponentially with depth; this ow compares well with the geostrophic current estimated from TOPEX/Poseidon altimetry (Figure 4). The 49

Fig. 9. OGCM currents (cm s  1 ) at 5 m.

50

Fig. 10. OGCM currents (cm s  1 ) at 35 m.

51

Fig. 11. OGCM currents (cm s  1 ) averaged over the top 50 m.

52

genesis and decay of the (G)WMC and (G)SMC follow the pattern established from satellite altimetry. There does not, however, seem to be a clear link between the recirculation in the eddies off Somalia and the SMC south of Sri Lanka; the latter is fed more by the branch of the SMC that ows around the Lakshadweep low, this in turn being fed by the WICC and the ow across the central Arabian Sea. The recirculation in the eddies off Somalia instead ows mostly into the eastward current in the western equatorial Indian Ocean. This eastward current feeds the SMC south of Sri Lanka in ship drifts (Figure 8) and in the simulations of McCreary et al. (1993); in the OGCM, however, this ow crosses the equator, as it does in the surface-drifter data (Shenoi et al., 1999a). In the depth-averaged ow (DAF) in the top 50 m (Figure 11), the WMC appears primarily as a geostrophic ow modulated by Ekman drift. During the summer monsoon, the SMC in the Arabian Sea appears as a strong southeastward drift modulated by geostrophy; in the bay, the SMC is primarily geostrophic. Unlike in the 35 m ow eld, the DAF around the eddies off Somalia merges with the Ekman drift to form a single broad southeastward SMC across the Arabian Sea; this current forms the broad western arm of the ow around the Lakshadweep low off southwest India, and feeds into the SMC south of Sri Lanka. A part of the recirculation around the eddies off Somalia also feeds the eastward Equatorial Current in the western Indian Ocean. Since geostrophy makes a signicant contribution to the DAF, it does not compare well with the NF (Figure 8) during the summer monsoon, when the winds are strong and the Ekman drift swamps the geostrophic current at the surface. At subsurface levels, however, geostrophic ow is signicant, as is evident from the DAF in the OGCM. Hence, the comparison between the DAF and the NF is good during the winter monsoon. Westward propagation associated with Rossby waves forms an essential component of the monsoon circulation in the north Indian Ocean, as seen in the longitude-time plots of OGCM meridional velocity at 5 N and 8 N (Figures 12 and 13). This propagation is best seen in the ow at 35 m. In the bay, however, westward propagation is seen at all depths over the entire year, indicating the dominance of geostrophy there. A prominent example is the westward propagation of the SMC (Figures 9 11); this, however, shows a break south of Sri Lanka (Figure 13), like the sea-level anomalies in altimetry (Figure 5). In the Arabian Sea, westward propagation is 53

ows are indicated by dashed contours and the contour interval is 5 cm s  1 .

Fig. 12. Longitude-time plot of the OGCM meridional velocity (cm s  1 ) at 8 N. Southward

54

in the westward propagation at 80 E.

ows are indicated by dashed contours and the contour interval is 5 cm s  1 . Note the break

Fig. 13. Longitude-time plot of the OGCM meridional velocity (cm s  1 ) at 5 N. Southward

55

seen at all depths during the winter monsoon; a prominent example is the westward propagation of the WMC associated with the Lakshadweep high (Figures 911). During the summer monsoon, the dominance of Ekman drift eliminates the Rossby wave signal at 5 m, but westward propagation is seen at 35 m; this signal, however, is weak, possibly because the strong eastward Ekman drift slows down the Rossby waves (Vinayachandran and Yamagata, 1998).

3.3 Transport estimates

The monsoon currents are shallow, unlike the deep currents observed in the western boundary current off Somalia and the associated eddies during the summer monsoon (see, for example, Figure 7 and Schott and McCreary (2001)). The shallowness of the monsoon currents in the OGCM, especially the SMC, is seen in the depth-time of the zonal current at 80.5 E (Figure 14). The SMC was shown by Vinayachandran et al. (1999a) to be trapped close to the surface because of the downwelling Rossby wave propagating westward from the eastern boundary of the equatorial Indian Ocean; this signature is seen in the westward ow below the eastward SMC during the summer monsoon. Upward (downward) propagation of phase (energy) is evident, indicating the existence of free propagating waves. The heat advected by the SMC has been shown (Shenoi, Shankar, and Shetye, 2001) to be important for the heat budget of the near-surface Arabian Sea. Given the large difference in salinity between the Bay of Bengal and the Arabian Sea, the contribution of the SMC and WMC to advection of salt is even greater. Since subsurface observations of these seasonal, open-ocean currents are limited (see Table 2), numerical models that simulate well the observed surface circulation are an important source of information on the transports associated with them. Given the signicant contribution of Ekman ow to the depth-averaged ow in the top 50 m, a comparison between the model transport and estimates based on hydrography is not meaningful. Hence, we choose to compare the model transports with those estimated from direct current measurements south of Sri Lanka (Schott et al., 1994) (Table 3). Latitude-time plots of the depth-integrated zonal current (over the top 100 m) are 56

Fig. 14. Depth-time plot of the OGCM zonal current (cm s  1 ) at 80.5 E (south of Sri Lanka). The upper (lower) panel shows the current averaged over 36 N (03 N). Westward ows are indicated by dashed contours and the contour interval is 5 cm s  1 .

57

Period

Observed

Model

10 Jan to 15 Feb 10 Jan to 15 Feb  1 Jun to 5 Jul 10 Jul to 15 Aug 1 Jun to 15 Aug 1 Nov to 30 Mar 1 May to 30 Sep

-12.8 -10.4 8.4 4.1 7.8

-10 -10 9.1 6.2 7.9 -7.1 5.4

Table 3

Observed and model zonal transports (in Sv; 1 Sv

106 m3 s  1 ) in the top 300 m between 3 45 N and 5 52 N at 80.5 E (south of Sri Lanka). Positive (negative) values indicate east-

ward (westward) ow, and the values listed are averages over the period indicated. All observations, except that marked (*), are for 1991; the marked observation is for 1992. The observed transports are derived from the direct current measurements of Schott et al. (1994). The model was forced by climatological wind stress (Hellerman and Rosenstein, 1983). The last two values are average model transports for the winter and summer monsoons, respectively.

58

shown in Figure 15. In the Arabian Sea (65 E), the core of the westward (eastward)

of the westward WMC is at 5 N. The WMC is relatively narrow and strong over the entire basin. In contrast, the SMC stands out only south of Sri Lanka and in the bay, where its eastward ow is seen to shift poleward with time, marking the westward propagation of the northeastward SMC across the bay. In the Arabian Sea at 65 E, the depth-integrated SMC is weak, but broad, the eastward ow extending from

WMC is at 4 N (8 N); in the bay (85 E) and south of Sri Lanka (80.5 E), the core

46 N and the weak eastward ow between 04 N show the complicated spatial structure of the geostrophic ow associated with the SMC. The model transports associated with the SMC and WMC are listed in Table 4. These transport estimates are comparable to the estimates for the other major currents in the north Indian Ocean (see Shetye and Gouveia (1998) and Schott and McCreary (2001) for the estimates for the Equatorial Current, the EICC, the WICC, and the Somali Current). Thus, the OGCM simulations compare well with the Ekman, geostrophic, and net surface currents estimated from wind-stress and sea-level climatologies. The transports estimated from the model also compare well with those estimated from direct current measurements. However, though the model is capable of simulating well the observed monsoon currents, it is too complex to yield insight into the dynamics underlying the monsoon-current system. Hence, we turn to a simpler numerical model of the Indian Ocean to isolate the processes that force the monsoon currents.

1220 N during the summer monsoon; the weak westward ow between

59

Fig. 15. Latitude-time plots of the depth-integrated zonal current (m2 s  1 ). The current is integrated over the top 100 m. Westward ows are indicated by dashed contours and the contour interval is 10 m2 s  1 .

60

Current

Location

Period

Transport

WMC WMC WMC WMC SMC SMC SMC SMC

65 E, 2.56 N 65 E, 06 N 80  5 E, 36 N 85 E, 37 N 65 E, 1220 N 80  5 E, 36 N 85 E, 48 N 85 E, 711 N

February February January January July JuneJuly June August

-13.1 -19.0 -10.1 -8.1 6.8 8.1 15.3 7.4

Table 4 Zonal transport (in Sv; 1 Sv periods mentioned.

106 m3 s  1 ) in the top 100 m in the domain of the monsoon

currents. Negative values indicate westward ows and the transports are averages over the

61

4 Forcing mechanisms

4.1 The numerical model and the control run

To isolate the processes responsible for forcing the monsoon currents, we need a simple model that has only the minimum physics required for simulating the wind1 forced circulation in the upper ocean. We use a dynamical 1 2 -layer reduced-gravity

model (Shankar, 1998), in which the density of the model layers does not vary in space or time. The equations for the upper layer are

 Hv t !

"# vH v

f k $ H v

gH H

% 1 ! 2  H v 2 H

'&

i "# H v

(2a) (2b)

Ht ! "# H v

&

(H where H f

H)

h the deviation of the layer thickness from the initial value, v 0 u v the velocity,

 

(2c)

H ! h is the instantaneous layer thickness, H the initial layer thickness,

y the Coriolis parameter, 1 x y the wind stress, g the acceleration due

to gravity, and

% the reduced-gravity parameter, with

&

1 . is

an average density that is representative of the ocean. and are the Laplacian mixing coefcients for momentum and thickness, the latter being included to damp the small-scale noise in the H eld, and is a Rayleigh friction coefcient. is the deviation of model dynamic height, computed with respect to the motionless deep ocean, from the initial state. In the discussion that follows, we refer to as the model sea level rather than as dynamic height. The model parameters are listed in Table 5. Equations (2) are integrated numerically on a staggered Arakawa-C grid using the leapfrog scheme; diffusive terms are evaluated at the backward time level and all other terms at the central time level. To inhibit time-splitting instability, the elds are averaged between successive time levels every 41 time steps. The model domain is as in McCreary et al. (1993). The upper-layer thickness is not allowed to shallow beyond 10 m or deepen beyond 190 m. The no-slip condition is applied at continental boundaries and the gradient boundary condition is applied at the 62

Parameter (units)

Symbol

Value

Laplacian mixing coefcient for momentum (cm2 s  1 ) Laplacian mixing coefcient for thickness (cm2 s  1 ) Thermal expansion coefcient ( C  1 ) Haline contraction coefcient (PSU  1 ) Reduced-gravity parameter Initial upper layer thickness (m) Minimum upper layer thickness (m) Maximum upper layer thickness (m) Linear Kelvin wave speed for H 100 m (cm s  1 )

T S H Hmin Hmax c cR x 4 y t

5 2 107 1 2 107

3 0  00025
0.00125 0.0035 100 10 190 185 12.2 55 72

Rossby wave speed at 10 N (cm s  1 ) Grid size (km) Time step (minutes)

Table 5
1 -layer reduced-gravity model. Parameters for the 1 2

63

open southern boundary at 29 S. The open boundary condition and the restrictions on upper-layer thickness force changes in the total mass in the basin; to conserve mass, a uniform correction is applied to the upper-layer thickness after each timestep. A linear damper (Rayleigh friction) is applied on the zonal velocity eld near the southern boundary; it is required to inhibit the development of large-scale instability along the boundary caused by the application of the gradient boundary condition on u. The damper is present only near the boundary, with

1 day

within 150 km of the boundary, and decreasing to zero linearly in the interval from 150 km to 300 km. There is no corresponding damper on v and H , so that uid can pass freely through the boundary. The forcing is derived from the wind-stress climatology of Hellerman and Rosenstein (1983) (Figure 2), their wind stress being interpolated linearly to the model grid and then smoothed (McCreary et al., 1993). The model is spun up from a state of rest, the winds being ramped up from zero to the appropriate level over ve days to damp inertial oscillations. Results discussed below (Figure 16) are from the tenth year of the simulation, by when the solution approaches stationarity. The solution emphasizes the geostrophic contribution to the ow. The contribution of Ekman drift is weak, except in the western Arabian Sea, because the currents are averaged over the model upper layer. Westward propagation of patterns in sealevel deviation is evident in the solution, especially in the eastern Bay of Bengal and southeastern Arabian Sea (Figure 17). This solution compares well with the DAF of the OGCM (Figure 11), except that the recirculation in the eddies off Somalia does not cross the equator, the eastward ow in the western equatorial Indian Ocean instead feeding the SMC south of Sri Lanka, as in the simulations of McCreary et al. (1993). The observations are also not clear in this regard, there being differences between the ow in the ship drifts (Figure 8) and surface drifters (Shenoi et al., 1999a). Though there is much less detail in the solution compared to the geostrophic ow estimated from TOPEX/Poseidon altimetry (Figure 4), the model does capture the essential features of the monsoon currents on the seasonal time scale. To examine to what extent these are attributable to nonlinear effects, we linearize the model equations to obtain (Shankar, 1998) vt ! f k v ! gh % 1 ! 2 v 64

&

i " v

(3a)

Fig. 16. Sea-level deviation from the initial surface (cm, left panel) and upper-layer velocity contours and the contour interval is 5 cm.

(cm s  1 , right panel) for the nonlinear simulation. Negative sea level is indicated by dashed

65

Fig. 16. (continued)

66

Fig. 17. Longitude-time plot of sea-level deviation (cm) from the reduced-gravity model (nonlinear simulation). Negative sea level is indicated by dashed contours and the contour interval is 5 cm.

67

ht ! " v 2 h

(3b)

(3c)

The solution to these linear equations (Figure 18) reproduces all the features seen in the nonlinear simulation; the one major exception is the Great Whirl off Somalia, which appears to owe its existence to nonlinear dynamics, and appears in the linear solution as part of a large anticyclonic gyre in the western Arabian Sea. Notwithstanding this discrepancy, the linear solution shows the strong poleward Somali Current, its recirculation in the gyres offshore, and the continuity of this current across the Arabian Sea to the southern tip of Sri Lanka.

4.2 Process solutions

That a linear system is capable of describing the seasonal, open-ocean monsoon currents implies that we can split the solution into components forced by different processes, the sum of these parts then being equal to the whole, the linear solution, which we call the control run. To isolate the effect of the processes on the monsoon currents, we apply two sets of boundary conditions along continental boundaries (McCreary et al., 1996; Shankar, 1998). One set is the usual no-slip condition u v 0 The other set, u n" v (4a)

&

n" k

f

v k n" v

(4b)

is applied to the boundaries of the Bay of Bengal and the Arabian Sea in some of the test solutions. In (4b), k is a unit vector directed out of the -plane and n is a unit vector normal to the boundary that points out of the bay (inshore) along its eastern and northeastern margins, into the Bay of Bengal (offshore) along its northern and western margins, out of the sea (inshore) along the southern boundaries of India and Sri Lanka and along their west coast, and into the sea (offshore) along the northern and western boundaries of the Arabian Sea; v 68

5 u v , where u and v 

Fig. 18. Sea-level deviation from the initial surface (cm, left panel) and upper-layer velocity contours and the contour interval is 5 cm.

(cm s  1 , right panel) for the linear simulation. Negative sea level is indicated by dashed

69

Fig. 18. (continued)

70

are the velocity components normal to and along the boundary; and 6 x y is the wind stress. Conditions (4b) allow Ekman ows to pass through boundaries, and therefore circulations driven by coastal Ekman pumping are ltered out of the solutions. This implies that coastal Kelvin waves are not generated along the coasts where this condition is applied. Condition (4a) is applied at continental boundaries for the control run. The modications made to obtain the process solutions are described below. To isolate the effect of the winds that blow along the eastern and northern boundaries of the Bay of Bengal (process EB here, process RA in McCreary et al. (1996); eastern boundary to (87 E, 20 N), thereby eliminating the effect of the alongshore winds there. The difference between the control run and this run gives the process solution forced by EB. These winds force coastal Kelvin waves, which propagate anticlockwise along the perimeter of the bay, radiating Rossby waves from the eastern boundary in the process. The Kelvin waves forced by these winds are weak, though the winds in the eastern bay are not signicantly weaker than those in the western bay (Figure 2); it is the alignment of the coast, almost normal to the wind vector, that results in a weak alongshore component, and hence, in a weak response (not shown). To isolate the effect of the strong winds that blow along the western boundary of the bay, or the east coast of India and Sri Lanka (process WB here, process LA in McCreary et al. (1996) and Shankar (1998)), we apply conditions (4b) along this coast (from 87 E, 20 N to 82 E, 6 5 N), thereby eliminating the effect of the alongshore winds there. The difference between the control run and this run gives the process solution forced by WB (Figure 19). The winds along the western boundary of the bay are southwesterly during the summer monsoon and northeasterly during the winter monsoon. The southwestnortheast alignment of the coast makes the wind vector parallel to the coast and these winds generate strong coastal Kelvin waves; the EICC closely follows the wind eld. Since the coastal Kelvin waves propagate with the coast on their right in the northern hemisphere, process WB affects only the EICC and has no effect on the circulation elsewhere in the bay. It does, however, have a strong effect on the circulation in the southeastern Arabian Sea (McCreary et al., 1993; Shankar and Shetye, 1997). By itself, it can force both the 71 Shankar (1998)), we apply conditions (4b) along these coasts, from 2 5 N at the

Fig. 19. Effect of winds along the western boundary of the Bay of Bengal (Process WB). Sea-level deviation (cm, left panel) and upper-layer velocity (cm s  1 ) is shown. Negative sea level is indicated by dashed contours and the contour interval is 5 cm.

72

Fig. 19. (continued)

73

Lakshadweep high and low, and the WICC, GWMC, and GSMC associated with them. The high forced by process WB, however, is much stronger than that forced in the control run.

To isolate the effect of the winds that blow along the eastern boundary of the Arabian Sea, or the south and west coasts of Sri Lanka and India (process EA here, process WLA in Shankar (1998)), we apply conditions (4b) along this coast (from 82 E, 6 5 N to 67 5 E, 25 N), thereby eliminating the effect of the alongshore winds there. The difference between the control run and this run gives the process solution forced by EA (Figure 20). Along this coast, the winds blow equatorward throughout the year, the winds along the west coast of Sri Lanka being the exception; here, the winds are similar to those along the Indian east coast. The alongshore winds generate coastal Kelvin waves that propagate poleward along the eastern boundary of the Arabian Sea, radiating westward propagating Rossby waves into the interior in the process. During the summer monsoon, the southwesterlies over the Arabian Sea turn around in the central Arabian Sea to blow from the northwest along the Indian west coast. These winds favour coastal upwelling and are stronger than the equatorward winds during the winter monsoon. The alignment of the coast also ensures that the winds during the winter monsoon generate but a weak coastal Kelvin wave. Therefore, though the winds along the eastern boundary of the Arabian Sea do not contribute signicantly to the Lakshadweep high, they force a low off southwest India and strong upwelling off western India and in the northern Arabian Sea during the summer monsoon. Hence, process EA contributes to the curving ow of the GSMC in the eastern and central Arabian Sea.

The effect of the winds along the northern and western boundaries of the Arabian Sea is evaluated in two parts. First, we consider the effect of winds along the boundary from the northeastern corner of the model Arabian Sea (67 5 E, 25 N) to the northern tip of Somalia (process WA), applying (4b) only over this part of the continental boundary; the difference between the control run and this run gives the effect of process WA. Second, we consider the effect of the winds along the coast of Somalia, from the northern tip of Somalia to 2 5 N (process SA), applying (4b) only over this part of the continental boundary; the difference between the control run and this run gives the effect of process SA. 74

Fig. 20. Effect of winds along the eastern boundary of the Arabian Sea (Process EA). Sea-level deviation (cm, left panel) and upper-layer velocity (cm s  1 ) is shown. Negative sea level is indicated by dashed contours and the contour interval is 5 cm.

75

Fig. 20. (continued)

76

Fig. 21. Effect of winds along the northern and western boundaries of the Arabian Sea, except the Somali coast (Process WA). Upper-layer velocity (cm s  1 ) is shown.

77

Process WA has only a local effect (Figure 21). The winds force the local coastal current, with some of the energy propagating equatorward along the coast via coastal Kelvin waves to the Somali coast. Process SA is the major forcing mechanism for the strong Somali Current, but its effect is also felt in the equatorial Indian Ocean via Kelvin wave propagation, and in the Bay of Bengal (Figure 22). The winds off Somalia force strong local upwelling during the summer monsoon, and this signal is carried into the equatorial Indian Ocean, forcing a westward, upwelling-favourable Equatorial Current. This, together with the effect of the Rossby wave reected from the coast of Sumatra in May, ensures a westward Equatorial Current during JuneSeptember in the eastern equatorial Indian Ocean and weakens the eastward current in the western equatorial Indian Ocean. This separates the eastward GSMC north of Sri Lanka from the westward ow at the equator. Applying condition (4b) along all the continental boundaries of the north Indian Ocean (from 2 5 N in the eastern bay to 2 5 N in the western Arabian Sea) eliminates the effect of these coastal winds, ltering out the effect of coastal Kelvin waves generated by them; this leaves the effect of winds in the equatorial waveguide and of Ekman pumping in the interior of the basin (process OP). The difference between the control run and this run gives the effect of all the continental winds together (process CW). The alongshore continental winds in the north Indian Ocean (process CW) contribute signicantly to the circulation (Figure 23) through the excitation of coastal Kelvin waves and the westward propagating Rossby waves that these Kelvin waves radiate when propagating poleward along an eastern ocean boundary. Most important for the monsoon currents are the winds along the coasts of India and Sri Lanka because it is the forcing by these winds that forces the GSMC and GWMC south of Sri Lanka and in the eastern Arabian Sea, linking the circulation between the two basins. The GWMC and GSMC, however, do not form in the Bay of Bengal in the absence of other processes. The GSMC also ows too far north in the Arabian Sea during the summer monsoon before turning to ow into the WICC, which, in turn, feeds the GSMC around the Lakshadweep low. In the absence of forcing by interior Ekman pumping, the sea-level high in the western Arabian Sea is weak and 78

Fig. 22. Effect of winds along the Somali coast (Process SA). Upper-layer velocity (cm s  1 ) is shown.

79

Fig. 23. Effect of alongshore winds in the north Indian Ocean (Process CW). Sea-level deviation (cm, left panel) and upper-layer velocity (cm s  1 , right panel) are shown. Negative sea level is indicated by dashed contours and the contour interval is 5 cm.

80

Fig. 23. (continued)

81

stretched out, and the GSMC that ows to its south and east does not turn to ow eastward across the southern Arabian Sea. Process OP, in contrast, is important for the monsoon currents in the Bay of Bengal, where the currents forced by the alongshore winds (process CW) oppose the SMC and WMC, and in the western and central Arabian Sea, where it forces the sea-level highs and the recirculations around them during the summer monsoon (Figure 24). It is also the only process solution that contains the Ekman drift. The WMC and SMC in the bay are the result of Ekman pumping in the southern bay, especially in the southwest, and of the westward propagating Rossby wave radiated from the eastern boundary owing to the reection of the equatorial Kelvin wave off Sumatra (Vinayachandran et al., 1999a; McCreary et al., 1993). Not only are these currents forced by Rossby waves, their elimination in the eastern bay and propagation to the west is also due to Rossby waves, but of the opposite sign (Vinayachandran et al., 1999a). During the summer monsoon, the SMC forced by Process OP ows westward in southeastern Arabian Sea, breaking the connection between the SMC in the two basins. This is due to the strong cyclonic Ekman pumping in the vicinity of Sri Lanka (Figure 25). The continuity of the SMC south of Sri Lanka is due to processes WB and EA. This abrupt change in the mechanisms forcing the SMC south of Sri Lanka results in the sharp break in the sea level and zonal current there in both observations (Figure 5) and models (Figures 13 and 17). The WMC, in contrast, shows greater continuity because both processes WB and OP force a westward zonal current (Figures 19 and 24). It is more difcult to separate the effect of winds in the equatorial Indian Ocean. To do this, it is necessary to force the model with only these winds, but this distorts the forcing towards the northern and southern edges of the equatorial waveguide. These winds, however, have been shown to be important forcing mechanisms for the EICC and the WICC (Yu et al., 1991; McCreary et al., 1993, 1996), and the GWMC and GSMC in the Bay of Bengal (McCreary et al., 1993; Vinayachandran et al., 1999a). The forcing from the equatorial Indian Ocean consists of two parts. First, there is the effect of direct forcing, or the generation of equatorial Kelvin and Rossby waves. Second, there is the effect of the reection of these waves at the eastern and western boundaries. Both processes are crucial to the circulation in the north Indian Ocean, the reection and resulting resonance being responsible for the 82

Fig. 24. Effect of ltering out forcing by alongshore winds in the north Indian Ocean (Proare shown. Negative sea level is indicated by dashed contours and the contour interval is 5 cm.

cess OP). Sea-level deviation (cm, left panel) and upper-layer velocity (cm s  1 , right panel)

83

Fig. 24. (continued)

84

Fig. 25. Ekman pumping (m day  1 ), derived from the wind-stress climatology of Hellerman and Rosenstein (1983).

85

large semiannual harmonic at the equator (Jensen, 1993).

4.3 Dynamics of the north Indian Ocean

The numerical experiments described above show that the GSMC and GWMC are complex currents, forced by several processes. The effect of the processes varies in time and space, each process being important at some time and in some part of the north Indian Ocean. The continuity of the GSMC and GWMC across the north Indian Ocean is due to the way these processes interact, connecting the various strands of these currents into a continuous trans-basin current. Despite the complexity of the monsoon currents, they are explicable by a simple linear theoretical framework, which has evolved over the last decade. The dynamics of the north Indian Ocean in this framework can be summarized by the leaky waveguide (Shetye, 1998; Shankar, 1998) depicted in the schematic in Figure 26. The framework depicted here invokes the equatorial Kelvin wave (which propagates eastward along the equator with a speed of

2ms

and is trapped within

2 5 of the

equator), the equatorial Rossby wave (which propagates westward with a speed that decreases with increasing latitude, the speed at the equator being 0 5 m s 1 ), and the coastal Kelvin wave (which propagates with the coast on its right in the northern hemisphere with a speed of 2 m s

1 , and has an offshore e-folding length scale of

100 km). These three waves merge the equatorial Indian Ocean, the Bay of Ben-

gal, and the Arabian Sea into a single dynamical entity, which must be modelled as a whole even to simulate the circulation in its parts. The process solutions show that the Kelvin waves forced by the winds along the east and west coasts of India and Sri Lanka are crucial for maintaining the continuity of the monsoon currents south of Sri Lanka. The coastal Kelvin wave forced by the winds along the east coast bends around Sri Lanka to propagate poleward along the west coast. This is possible because the sum of the e-folding scales of wave ( 2 5 ) is less than the distance between the southern tip of Sri Lanka and the coastal Kelvin wave south of Sri Lanka ( 2 25 ) and the equatorial Kelvin

the equator ( 6 ), forcing a separation between the equatorial waveguide and the coastal waveguide. This allows the Kelvin waves from the Bay of Bengal to pass

unhindered into the Arabian Sea. The coastal and equatorial waveguides overlap 86

30N

20N

10N

10S 40E 50E 60E 70E 80E 90E 100E

Fig. 26. Schematic illustrating the dynamics of the north Indian Ocean. The linear theoretical framework depicted here invokes the equatorial Kelvin wave, the equatorial Rossby wave, and the coastal Kelvin wave. These three waves merge the equatorial Indian Ocean, the Bay of Bengal, and the Arabian Sea into a single dynamical entity. The horizontal hatching indicates the equatorial waveguide, which extends about 2.5 on either side of the equator; the vertical hatching indicates the coastal waveguide. The coastal Kelvin wave is trapped at the coast poleward of a critical latitude; equatorward of this latitude, westward radiation of energy is possible, and the coastal Kelvin wave is inseparable from the westward propagating Rossby wave. The critical latitudes for Rossby waves at annual (semiannual) period is

42 ( 21 ); hence, annual and semiannual Kelvin waves are inseparable from

westward propagating Rossby waves in the north Indian Ocean, and energy leaks at these periods from the eastern boundary into the open ocean (shown by arrows pointing out of the coastal waveguide). Shetye (1998) and Shankar (1998) called this the leaky waveguide of the north Indian Ocean. Energy is also generated by Ekman pumping (shown by the closed circles) in the interior of the basin, this signal also propagating westward as Rossby waves.

87

at the western and eastern boundaries of the equatorial Indian Ocean. Energy from the coastal waveguide leaks into the equatorial waveguide at the western boundary via Kelvin waves, and energy from the equatorial waveguide leaks into the coastal waveguide at the eastern boundary via the reection of the equatorial Kelvin wave. This reection at the eastern boundary results in a part of the energy propagating poleward along the eastern boundary of the basin as coastal Kelvin waves (Moore, 1968; Yu et al., 1991). These coastal Kelvin waves are inseparable from the westward propagating Rossby waves almost all over the north Indian Ocean at the semiannual and annual periods (Figure 26). Energy also leaks out of the coastal waveguide into the interior of the basin from the eastern boundaries via radiation of westward propagating Rossby waves. Westward propagating signals in the form of Rossby waves are also generated in the interior by Ekman pumping. Westward propagating Rossby waves are not restricted to the equatorial waveguide at semiannual and annual periods. Hence, unlike the equatorial Kelvin wave, they can interact with the coastal Kelvin wave forced by process WB and inuence the monsoon currents south of Sri Lanka. An example of this was seen during 1994 and 1997, when easterly anomalies in the winds over the equatorial Indian Ocean not only forced changes in the equatorial circulation (Vinayachandran et al., 1999b; Saji, Goswami, Vinayachandran, and Yamagata, 1999; Webster, Moore, Loschnigg, and Leben, 1999; Unnikrishnan et al., 2001), but also inuenced the monsoon currents in the Bay of Bengal and south of Sri Lanka (Figures 27 and 28). Thus, complex circulation patterns can be generated even by the simple linear framework described above, which seems to be remarkably good at describing the wind-forced, seasonal circulation in the north Indian Ocean.

88

Fig. 27. Monthly-mean geostrophic current (cm s  1 ), derived from TOPEX/Poseidon altimetry for January. The upper-left panel shows the 1993-1997 climatology, and the other panels show the geostrophic currents for the January of each of the years 19931997.

89

Fig. 28. Monthly-mean geostrophic current (cm s  1 ), derived from TOPEX/Poseidon altimetry for July. The upper-left panel shows the 1993-1997 climatology, and the other panels show the geostrophic currents for the July of each of the years 19931997.

90

5 Summary

We have described the climatological seasonal cycle of surface circulation associated with the open-ocean monsoon currents in the north Indian Ocean. Both Ekman drift, computed using climatological wind stress, and surface geostrophic currents, derived from TOPEX/Poseidon altimetry, contribute to the net surface current. The monsoon currents extend over the entire basin, from the Somali coast in the west to the Andaman Sea in the east. They do not, however, come into being, or decay, over this entire region at a given time. Different parts of the currents form and decay at different times, and it is only in the mature phase that the monsoon currents exist as continuous, trans-basin currents. Information on subsurface ow comes from simulations with an OGCM. The WMC is primarily a geostrophic current, with Ekman drift modulating it. It ows westward from the eastern Bay of Bengal to Sri Lanka and beyond. In the Arabian Sea, the WMC branches, one branch owing westward and the other turning around a high in sea level in the southern Arabian Sea to ow eastward into the poleward WICC. The Ekman drift, in contrast to the geostrophic ow, is unidirectional (westward); hence, at the surface, it accentuates the westward WMC to the south of the sea-level high, and attenuates the eastward WMC to its north. The curving ow of the WMC is, therefore, evident only below

20 m. The SMC ows eastward,

from the western Arabian Sea to the eastern Bay of Bengal. The strong winds during the summer monsoon ensure that Ekman drift dominates at the surface, with geostrophy dominating below 20 m. This leads to a more complex vertical structure than is associated with the WMC. At the surface, the SMC in the Arabian Sea ows eastward and southeastward, feeding into the eastward SMC south of Sri Lanka. This ow branches east of Sri Lanka, one branch owing into the bay, the other continuing to ow eastward. The geostrophic SMC, which dominates below

20 m, is a continuation of the Somali Current. A part of the recirculation around

the eddies off Somalia merges with the ow to the west of the Lakshadweep low off southwest India to form a curving SMC that ows into the eastward SMC south of Sri Lanka. Since both Ekman drift and geostrophy make signicant contributions to the monsoon currents, the transport associated with them is also due to both components. 91

Numerical simulations with a reduced-gravity model show that the monsoon currents are composed of several parts, each of which is forced by one or more processes. In the Bay of Bengal, the major processes are Ekman pumping and remote forcing from the equatorial Indian Ocean. The continuity of the SMC south of Sri Lanka is due to remote forcing by the winds along the east coasts of India and Sri Lanka, which force the Lakshadweep high off southwest India and contribute signicantly to forcing the monsoon currents in the eastern Arabian Sea. In the rest of the Arabian Sea, Ekman pumping combines with Rossby waves radiated from the west coast of India to produce the curving ows associated with the geostrophic monsoon currents. Superimposed on these geostrophic ows is the local Ekman drift. Thus, a simple linear framework based on three linear waves the equatorial Kelvin wave, the equatorial Rossby wave, and the coastal Kelvin wave and Ekman drift explains the seasonal cycle of the climatological monsoon currents. That the monsoon currents, despite their complexity, are seen in climatology is due to the regular seasonal cycle associated with the monsoon. This regularity, bordering on predictability is one facet of the monsoon. The second facet of the monsoon is ckleness: the monsoon varies a lot from one year to another. This results in considerable interannual variability in the monsoon currents (Figures 27 and 28). There are also variations on the sub-monthly time scale superimposed on the seasonal monsoon, forcing sub-monthly variability in the monsoon currents (Figure 29). The sea-level anomalies and geostrophic currents from each TOPEX/Poseidon cycle show a much richer structure than do the monthly averages (Figures 27 and 28) or climatology (Figure 4). The monsoon currents, though still identiable as continuous trans-basin ows in these snapshots, meander much more in owing past the many eddies that dominate the surface ow eld. The actual currents are stronger and noisier than the currents in climatology or in monthly averages, and the climatological monsoon currents are therefore the weaker ensembles of these meandering currents. The multiplicity of eddies in the sea-level anomalies in each TOPEX/Poseidon cycle is reminiscent of the eddy-dominated ow seen in the hydrographic data of the IIOE (D uing, 1970; Wyrtki, 1971). Those eddies, however, were different and were due to the sparsity of data, which prevented the mean currents from standing out. The eddies seen in altimetry are, on the other hand, due to the better resolution that 92

Fig. 29. Geostrophic currents (cm s  1 ) from TOPEX/Poseidon altimetry for three cycles each during January (left panel) and July (right panel) of 1993. The plots show that the GWMC and GSMC can be traced even in individual TOPEX/Poseidon cycles, even though the currents are more noisy and meander more than in climatology (Figure 4) or in monthly averages (Figures 27 and 28).

93

the satellite data affords in both space and time. The vast amount of such nearsynoptic data being made available by satellites reveals more complex structures in the domain of the monsoon currents as the resolution of the data becomes ner in both space and time. That eddies are common in the north Indian Ocean is obvious from these data, which reveal a picture of the surface circulation that is quite different from the smooth currents seen in climatological or monthly-mean data, or in the linear simulations described in in this paper. It remains to be seen if the linear theoretical framework described herein will hold even for the eddying motion evident at these spatial and temporal scales, or if its applicability is restricted to the ensemble of these eddies that is represented by the seasonal cycle.

94

Acknowledgements

The ship-drift data are from the Ocean Current Drifter Data CDROMs NODC53 and NODC-54, which are due to the National Oceanic Data Center (NODC), US Department of Commerce, National Oceanic and Atmospheric Administration (NOAA), USA. The altimeter products have been produced by the CLS (Collecte Localisation Satellites) Space Oceanography Division as part of the European Unions Environment and Climate projects AGORA (ENV4-CT9560113) and DUACS (ENV4-CT96-0357), with nancial support from the CEO (Centre for Earth Observations) programme and Midi-Pyr en es regional council. We thank B. S. Beena for helping with these data. This was carried out under projects funded by the Department of Ocean Development, New Delhi. The simulations with the OGCM were carried out when P. N. Vinayachandran was at the Institute for Global Change Research, Frontier Research System for Global Change, Tokyo, Japan. The gures were made using FERRET, GMT, and Xg. This is NIO contribution xxxx.

95

References E. B ohm, J. M. Morrison, V. Manghnani, H.-S. Kim, and C. N. Flagg. The Ras al Hadd Jet: Remotely sensed and acoustic Doppler current proler observations in 19941995. Deep-Sea Research II, 46:15311549, 1999. J. G. Bruce, D. R. Johnson, and J. C. Kindle. Evidence for eddy formation in the eastern Arabian Sea during the northeast monsoon. Journal of Geophysical Research, 99(C4):76517664, April 1994. J. G. Bruce, J. C. Kindle, L. H. Kantha, J. L. Kerling, and J. F. Bailey. Recent observations and modeling in the Arabian Sea Laccadive high region. Journal of Geophysical Research, 103(C4):75937600, April 1998. M. A. Cane. On the dynamics of equatorial currents, with application to the Indian Ocean. Deep-Sea Research, 27A:525544, 1980. A. N. Cutler and J. C. Swallow. Surface currents of the Indian Ocean (to 25 S, 100 E): Compiled from historical data archived by the Meteorological Ofce, land, 1984. 36 charts, 8pp. A. Defant. Physical Oceanography, volume I. Pergamon Press, Oxford, 1961. J. R. Donguy and G. Meyers. Observations of geostrophic transport variability in the western tropical Indian Ocean. Deep-Sea Research I, 42:10071028, 1995. W. D uing. The monsoon regime of the currents in the Indian Ocean. International Indian Ocean Expedition Oceanographic Monographs, Number 1. EastWest Center Press, Honolulu, 1970. C. N. Flagg and H.-S. Kim. Upper ocean currents in the northern Arabian Sea from shipboard ADCP measurements collected during the 19941996 U.S. JGOFS and ONR Programs. Deep-Sea Research II, 45:19171959, 1998. V. V. Gopalakrishna, S. M. Pednekar, and V. S. N. Murty. T-S variability and volume transport in the central Bay of Bengal during southwest monsoon. Indian Journal of Marine Sciences, 25:5055, 1996. W. Han. Inuence of salinity on dynamics, thermodynamics and mixed-layer physics in the Indian Ocean. PhD thesis, Nova Southeastern University, Dania, Florida, USA, 1999. W. Han, J. P. McCreary, D. L. T. Anderson, and A. J. Mariano. On the dynamics of the eastward surface jets in the equatorial Indian Ocean. Journal of Physical Oceanography, 29:21912209, 1999.

Bracknell, UK. Report 187, Institute of Oceanographic Sciences, Wormley, Eng-

96

S. Hastenrath and L. L. Greischar. The monsoonal current regimes of the tropical Indian Ocean. Observed surface ow elds and their geostrophic and winddriven components. Journal of Geophysical Research, 96:1261912633, 1991. S. Hellerman and M. Rosenstein. Normal monthly wind stress over the world ocean with error estimates. Journal of Physical Oceanography, 13:10931104, 1983. T. G. Jensen. Equatorial variability and resonance in a wind-driven Indian Ocean model. Journal of Geophysical Research, 98:22,53322,552, 1993. S. P. Kumar and A. S. Unnikrishnan. The seasonal cycle of temperature and associated wave phenomena in the Bay of Bengal. Journal of Geophysical Research, 100:13,58513,593, 1995. P. Y. Le Traon, P. Gaspar, F. Bouyssel, and H. Makhmara. Using TOPEX/Poseidon data to enhance ERS-1 data. Journal of Atmospheric and Oceanic Technology, 12:161170, 1995. P. Y. Le Traon, F. Nadal, and N. Ducet. An improved mapping method of multisatellite altimeter data. Journal of Atmospheric and Oceanic Technology, 25: 522534, 1998. P. Y. Le Traon and F. Ogor. ERS-1/2 orbit improvement using TOPEX/Poseidon: The 2 cm challenge. Journal of Geophysical Research, 103:80458057, 1998. S. Levitus. Climatological atlas of the world ocean. NOAA Professional Paper 13, 173 pp., U.S. Government Printing Ofce, Washington D.C., 1982. S. Levitus and T. P. Boyer. World Ocean Atlas 1994, Volume 4: Temperature. NOAA Atlas NESDIS 4. 117 pp., 1994. S. Levitus, R. Burgett, and T. P. Boyer. World Ocean Atlas 1994, Volume 3: Salinity. NOAA Atlas NESDIS 3. 99 pp., 1994. M. J. Lighthill. Dynamic response of the Indian Ocean to the onset of the southwest monsoon. Philosophical Transactions of the Royal Society of London, Series A, 265:4592, October 1969. T. Matsuno. Quasi-geostrophic motions in the equatorial area. Journal of the Meteorological Society of Japan, 44:2543, 1966. J. P. McCreary, W. Han, D. Shankar, and S. R. Shetye. Dynamics of the East India Coastal Current, 2. Numerical solutions. Journal of Geophysical Research, 101 (C6):13,99314,010, June 1996. J. P. McCreary, P. K. Kundu, and R. L. Molinari. A numerical investigation of the dynamics, thermodynamics and mixed-layer processes in the Indian Ocean.

97

Progress in Oceanography, 31:181244, 1993. R. L. Molinari, D. Olson, and G. Reverdin. Surface current distributions in the tropical Indian Ocean derived from compilations of surface buoy trajectories. Journal of Physical Oceanography, 95:72177238, 1990. D. W. Moore. Planetary-Gravity Waves in an Equatorial Ocean. PhD thesis, Harvard University, Cambridge, Massachusets, February 1968. V. S. N. Murty, M. S. S. Sarma, B. P. Lambata, V. V. Gopalakrishna, S. M. Pednekar, A. S. Rao, A. J. Luis, A. R. Kaka, and L. V. G. Rao. Seasonal variability of upperlayer geostrophic transport in the tropical Indian Ocean during 19921996 along TOGA-I XBT tracklines. Deep-Sea Research I, 47:15691582, 2000. V. S. N. Murty, Y. V. B. Sarma, D. P. Rao, and C. S. Murty. Water characteristics, mixing and circulation in the Bay of Bengal during southwest monsoon. Journal of Marine Research, 50:207228, 1992. J. J. OBrien and H. E. Hurlburt. Equatorial jet in the Indian Ocean: Theory. Science, 184:10751077, 1974. R. C. Pacanowski. MOM2 Version 2.0 (Beta): Documentation, users guide, and reference manual. GFDL Technical Report 3.2, GFDL, Princeton, New Jersey, 1996. R. C. Pacanowski and S. G. H. Philander. Parameterization of vertical mixing in numerical models of the tropical ocean. Journal of Physical Oceanography, 11: 14421451, 1981. C. Perigaud and P. Delecluse. Annual sea level variations in the southern tropical Indian Ocean from Geosat and shallow-water simulations. Journal of Geophysical Research, 97(C12):20,16920,178, December 1992. S. Pond and G. L. Pickard. Introductory dynamical oceanography. Pergamon Press, second edition, 1983. J. T. Potemra, M. E. Luther, and J. J. OBrien. The seasonal circulation of the upper ocean in the Bay of Bengal. Journal of Geophysical Research, 96(C7): 12,66712,683, July 1991. A. S. Rao. Rossby waves in the Bay of Bengal: Observations and simulations. PhD thesis, Andhra University, Vishakhapatnam, India, March 1998. R. R. Rao, R. L. Molinari, and J. F. Festa. Evolution of the climatological nearsurface thermal structure of the tropical Indian Ocean, 1. Description of mean monthly mixed-layer depth and sea-surface temperature, surface-current, and

98

surface meteorological elds. Journal of Geophysical Research, 94:10,801 10,815, 1989. J. Reppin, F. A. Schott, J. Fischer, and D. Quadfasel. Equatorial currents and transports in the upper central Indian Ocean: Annual cycle and interannual variability. Journal of Geophysical Research, 104:1549515514, 1999. N. H. Saji, B. N. Goswami, P. N. Vinayachandran, and T. Yamagata. A dipole mode in the tropical Indian Ocean. Nature, 401:360363, 1999. F. Schott. Monsoon response of the Somali current and associated upwelling. Progress in Oceanography, 12:357381, 1983. F. Schott and J. P. McCreary. The monsoon circulation of the Indian Ocean. Progress in Oceanography, 2001. F. Schott, J. Reppin, J. Fischer, and D. Quadfasel. Currents and transports of the Monsoon Current south of Sri Lanka. Journal of Geophysical Research, 99 (C12):25,12725,141, December 1994. D. Shankar. Low-frequency variability of sea level along the coast of India. PhD thesis, Goa University, Goa, India, December 1998. D. Shankar. Seasonal cycle of sea level and currents along the coast of India. Current Science, 78:279288, 2000. D. Shankar, 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. Journal of Geophysical Research, 101(C6):13,975 13,991, June 1996. D. Shankar and S. R. Shetye. On the dynamics of the Lakshadweep high and low in the southeastern Arabian Sea. Journal of Geophysical Research, 102(C6): 12,55112,562, June 1997. D. Shankar and S. R. Shetye. Are interdecadal sea level changes along the Indian coast inuenced by variability of monsoon rainfall? Journal of Geophysical Research, 104:2603126042, 1999. S. Shenoi, P. Saji, and A. Almeida. Near-surface circulation and kinetic energy in the tropical Indian Ocean derived from Lagrangian drifters. Journal of Marine Research, 57(6):885907, 1999a. S. S. C. Shenoi, D. Shankar, and S. R. Shetye. On the sea surface temperature high in the Lakshadweep Sea before the onset of the southwest monsoon. Journal of Geophysical Research, 104:15,70315,712, July 1999b.

99

S. S. C. Shenoi, D. Shankar, and S. R. Shetye. Differences in heat budgets of nearsurface Arabian Sea and Bay of Bengal: Causes and implications to the summer monsoon. Journal of Geophysical Research, 2001. S. R. Shetye. West India Coastal Current and Lakshadweep high/low. Sadhana, 23:637651, 1998. S. R. Shetye and A. D. Gouveia. Coastal circulation in the north Indian Ocean. Coastal segment (14,SW), volume 11 of The Sea, chapter 18, pages 523556. John Wiley and Sons, Inc, 1998. S. R. Shetye, A. D. Gouveia, D. Shankar, S. S. C. Shenoi, P. N. Vinayachandran, D. Sundar, G. S. Michael, and G. Nampoothiri. Hydrography and circulation in the western Bay of Bengal during the northeast monsoon. Journal of Geophysical Research, 101(C6):14,01114,025, June 1996. S. R. Shetye, A. D. Gouveia, S. S. C. Shenoi, G. S. Michael, D. Sundar, A. M. Almeida, and K. Santanam. The coastal current off western India during the northeast monsoon. Deep-Sea Research, 38(12):15171529, December 1991a. S. R. Shetye, A. D. Gouveia, S. S. C. Shenoi, D. Sundar, G. S. Michael, A. M. Almeida, and K. Santanam. Hydrography and circulation off the west coast of India during the southwest monsoon 1987. Journal of Marine Research, 48: 359378, 1990. S. R. Shetye, A. D. Gouveia, S. S. C. Shenoi, D. Sundar, G. S. Michael, and G. Nampoothiri. The western boundary current of the seasonal subtropical gyre in the Bay of Bengal. Journal of Geophysical Research, 98(C1):945954, January 1993. S. R. Shetye, S. S. C. Shenoi, A. D. Gouveia, G. S. Michael, D. Sundar, and G. Nampoothiri. Wind-driven coastal upwelling along the western boundary of the Bay of Bengal during the southwest monsoon. Continental Shelf Research, 11(11):13971408, November 1991b. W. Shi, J. M. Morrison, E. B ohm, and V. Manghnani. The Oman upwelling zone during 1993, 1994, and 1995. Deep-Sea Research II, 47:12271247, 2000. L. Stramma, J. Fischer, and F. Schott. The ow eld off southwest India at 8N during the southwest monsoon of August 1993. Journal of Marine Research, 54 (1):5572, January 1996. A. S. Unnikrishnan, S. P. Kumar, and G. S. Navelkar. Large-scale processes in the upper layers of the Indian Ocean inferred from temperature climatology. Journal

100

of Marine Research, 55:93115, 1997. A. S. Unnikrishnan, V. S. N. Murty, M. T. Babu, C. K. Gopinathan, and R. J. K. Charyulu. On the anomalous current structure in the eastern equatorial Indian Ocean during southwest monsoon of 1994. Marine and Freshwater Research, 2001. P. N. Vinayachandran, Y. Masumoto, T. Mikawa, and T. Yamagata. Intrusion of the Southwest Monsoon Current into the Bay of Bengal. Journal of Geophysical Research, 104:1107711085, 1999a. P. N. Vinayachandran, N. H. Saji, and T. Yamagata. Response of the equatorial Indian Ocean to an unusual wind event during 1994. Geophysical Research Letters, 26:16131616, 1999b. P. N. Vinayachandran, S. R. Shetye, D. Sengupta, and S. Gadgil. Forcing mechanisms of the Bay of Bengal circulation. Current Science, 71(10):753763, November 1996. P. N. Vinayachandran and T. Yamagata. Monsoon response of the sea around Sri Lanka: Generation of thermal domes and anticyclonic vortices. Journal of Physical Oceanography, 28:19461960, 1998. B. A. Warren. Medieveal Arab references to the seasonally reversing currents of the north Indian Ocean. Deep-Sea Research, 13(2):167171, 1966. P. J. Webster, A. M. Moore, J. P. Loschnigg, and R. R. Leben. Coupled oceanatmosphere dynamics in the Indian Ocean during 199798. Nature, 401:356 359, 1999. K. Wyrtki. Oceanographic Atlas of the International Indian Ocean Expedition. National Science Foundation, Washington, D.C., November 1971. 531 pp. K. Wyrtki. An equatorial jet in the Indian Ocean. Science, 181:262264, 1973a. K. Wyrtki. Physical oceanography of the Indian Ocean. In B. Zeitschel and S. A. Gerlach, editors, Biology of the Indian Ocean, volume 3 of Ecological Studies: Analysis and Synthesis, pages 1836. Chapman and Hall Limited, London, 1973b. L. Yu, J. J. OBrien, and J. Yang. On the remote forcing of the circulation in the Bay of Bengal. Journal of Geophysical Research, 96(C11):20,44920,454, November 1991.

101