UPPER LAYER CIRCULATION IN THE BANDA SEA IN RESPONSE TO THE ONSET OF MONSOON WINDS

Aspects of the upper layer circulation in the interior of the Banda Sea, Indonesia, associated with local forcing by monsoon winds are examined numerically through the use of a reduced gravity model. The basin is located between approximately 4°S and 8°S and is partially enclosed by chains of islands. The primary emphasis is an evaluation of the free wave response which contributes to the steady or slowly varying circulation. Basin response appears to be characterized by interacting Kelvin waves and Rossby at low frequencies, and by evanescent Poincare waves of higher frequencies. Passages between islands along the perimeter of the basin appear to be nearly, impermeable to Rossby waves, which contribute to a pattern of westward propagating quasi geostrophic eddies. This pattern would persist during periods of wind transition.


INTRODUCTION
The Banda Sea is located in the eastern Indonesian Archipelago, between approximately 4°S and 8°S.It is a very deep basin partially enclosed by chain of islands (Figure 1).The basin is approximately 900 km long and 500 km wide; the maximum depth is nearly 7400 m.The Banda Sea study region.
Southeast Asian monsoon dominates the large-scale meteorological forcing.Strong east winds blow across the archipelago from May ti early September, and west winds from November to March.Winds are very steady with characteristic speeds from 6 to 10 ms -1 (WYRTKI, 1961).During transition periods, the wind is weak and variable.Knowledge of the circulation in this region is primarily from studiesby WYRTKI (1961,1987).Wyrtki recognized that monsson winds produce a direct locally forced response, as well as fluctuations in the persistent flow from Pacific to Indian Oceans through the Indonesian archipelago associated with the Pacific-Indian Ocean pressure gradient.RAO (1966), have described free wave motion in a uniformly rotating basin.He has shown that the excited waves are of Kelbin-type and that these waves are progessively more trapped along the boundary as the rate of rotation increases.Previous studies of motions in non uniformly rotating closed basin have shown that the baroclinic oscillations are confined to the response adjustment to the boundaries (RATTRAY and CHARNELL, 1966).
The oscillations are two distinct of types : gravity waves which are either dynamically similar to Kelvin waves or Poincare waves and Rossby waves (MOFJELD and RATTRAY, 1971).The importance of Rossby waves in the adjustment processes has been pointed out by LIGHTHILL (1969), and the interplay of Kelvin waves and Rossby waves have been described in more detail CANE and SARACHIC (1976) in low latitude ocean.
In this paper we will examine some of the characteristics of the upper layer circulation in the interior of the Banda Sea produced by the local wind forcing associated with monsoon winds.The focus of our study is an evaluation of the free waves response which contributes to the steady or slowly varying circulation.Effects of the iarga scale pressure gradient between the Pacific Ocean and Indian Ocean and the important effects of advection of water with distinct temperature and salinity characteristics into the basin (i.e., low salinity water entering from the Flores Sea) are not considered.
Numerical results for wind induced motion in the Banda Sea was interpreted in the light of the previous results for wave motions in uniformly and non uniformly rotating basins.

BRIEF DESCRIPTION OF THE MODEL
In this study we use a reduced gravity model which can incorporate a variable Coriolis parameter.The model is based on the following equations, The average upper layer thickness is choses to be 200 m.A density difference xxxx of 0.005 kgm is specified.As discussed below, a simplified rectangular basin configuration is first considered; then a more realistic basin configuration is introduced which is determined from the 200 m bathymetric contour from the Snellius-Expedition (1927) survey.Equations ( 1) are solved numerically based on KOUTITONSKY et al, (1987).The extended study area has a 44 x 24 finite difference mesh.The equation of motion is integrated in time using a leapfrog finite difference scheme.The model time step is approximately 1 hour.The size of the grid is 30 km which is chosen to be significantly smaller than the internal Rossby radius (170 km).Since the fluid model is inViscid, we apply the free slip condition at the solid boundaries.An ORLANSKI (1976) radiation boundary condition is imposed at passages to allow the gravity waves and Rossby waves to be radiated out from the interior of the domain.The model was started from rest and forced by an impulsive uniform wind stress with a value of 1.5 dyne cm" 2 .Winds are applied for 3 months until a approximate oscillatory steady state was achieved, and then allowed to relax.Simulations were continued long enough so that the resulting free wave motion could be described through spectral analysis.

RESULTS AND DISCUSSION
Here we present results for the response of a bounded and partially bounded rotating basin to impulsive wind forcing.The basin scales are chosen to be characteristic of the Banda Sea.Because of the complexity of wave motion in a basin at this latitude, our strategy is to describe first the response of a bounded uniformly rotating rectangular basin, second, the response of a bounded basin with non uniformly rotation, and third the response of a partially bounded basin.

The bounded uniformly rotating basin.
Examples of transport vectors and layer thickness deviation are shown in Figure 2. Strong free wave motion dominates the response even during active wind forcing.The pattern precess about the basin in a clockwise and show clear evidence of basinwide gravity wave motion.
Spectral analysis applied, for example, to a time series for components of velocity at a point in the basin (Figure 3) can provide insight into the types of gravity wave motion present.Distinct peaks appear at approximately 12.8 days, 4.8 days, 3.8 days and 3.2 days in the spectrum for east-west component.An especially strong peak at 3.8 days appears in the spectrum for the north-south component.These spectral peaks can be compared to the inertial period which is 5.7 days and to the lowest seiche mode of 11.7 days as determined by the relation (Rao, 1966) : The general effect of rotation on the excited gravity waves in a basin is to split the gravity waves at the seiche frequencies, producing lower frequency Kelvin-type motion and higher frequency Poincare-type motion.In our basin the ratio of the lowest seiche period to the inertial period (also the ratio of basin scale to internal Rossby radius) is approximately 2.0 which implies moderate rotation.
The spectral peak at 12.8 days (Figure 3) represents motion with a period intermediate between the lowest seiche mode and the lowest Kelvin mode which is approximately 18 days.The high frequency peaks represent superinertial Poincare waves.For stronger rotation the splitting would be stronger and the low frequency peak would approach the lowest Kelvin mode.

The bounded non uniformly rotating basin
Examples of transport and layer thickness deviation (Figure 4) indicate complex patterns of wave motion.In addition to basin-wide gravity wave motion, there is ividence for strong short wavelength Rossby wave motion in the region of western boundary.The relationship between transport and layer thickness deviation suggest that this motion is quasi geostrophic.
Spectrum for the velocity components (Figure 5) indicate the presence of very low frequency motion; distinct peaks occur at approximately 80 days and longer  Mar.Res. Indonesia Vol.28, 1992: 81-95 The 7.6 days spectral peak which is especially strong for the north-south velocity component could be interpreted in term of Poincare waves.In our non uniformly rotating basin the local inertial period varies from approximately 14 days to 4 days.We therefore attribute this spectral peaks to evanescent Poincare waves trapped along the northern boundary of the basin.
We can, hencefore, summarize that the basin response in terms of oscillation which consist of combination of planetary waves and lower mode of gravity waves

The partially confined non uniformly rotating basin
In this section, we examine some aspects of response to the onset of winds in a partially confined basin.Numerical experiments were conducted to evaluate the influence of a passage of a different width and position in the perimeter of the domain on the characteristics of wave motion in the interior.
In light of previous result (Me Kee, 1972) on the scattering of Rossby waves near a gap, we expected that our gaps would be -essentially impermeable to basin scale Rossby waves.We expected this to be true as long as gap width was significantly less than the wavelength of the incoming waves.Results of our experiments seem to confirm this suggestion.The most definitive results (Figure 6) was that gaps of the order of one Rossby radius in width (short period gravity waves) is in order, especially if the gaps was positioned contiguous to a left bounded corner.
We conclude that Kelvin-type waves do radiate from the basin through the gaps of the order of one larger Rossby radius wide, and that basin scale Rossby do not.The low frequency basin scale quasigeostrophic motion could be expected to persist long after the wind has relaxed.Because the Kelvin waves appear to interact with the Rossby waves, the gaps affect the Rossby waves if only indirectly.
Example of the results for simulation which incorporate a more realistic basin configuration characteristic of the Banda Sea (Figure 7) including gaps in the eastern and western perimeter are shown in Figure 8.These results emphasize the initial Kelvin wave radiation: from the basin and the persistent low frequency basin scale Rossby wave motion.This motion is characterized by quasigeostrophic eddies which tend to propagate westward.Certain aspects of this flow pattern agree with the flow described by Wyrtki (1961), particularly during the transitional period in April.Mar.Res. Indonesia Vol.28, 1992: 81-95 Figure 1.The Banda Sea study region

Figure 2 .Figure 3 .
Figure 2. Layer thickness deviation (cm) and transport (cm 2 /s) in a rectangular bounded basin for 3.5 and 7.0 days after relaxation of wind.Basin parameters are length, L = 1320 km, and wide, W = 720 km.The Coriolis parameter f = 1.27 x 10 -5 /s is constant.
Figure 4. Layer thickness deviation (cm) and transport (cm /s) in a rectangular bounded basin for week 1 and 12 after relaxation of wind.Basin parameters are length, L -1320 km. and wide, W -720 km km.The Coriolis parameter is f =βy, where β= 2.3 x 10 -11 /s.

Figure 5 .
Figure 5.The spectral density of velocity components (east-west component U2, north-south component V2) at an interior point of the non uniformly rotating basin.Basin parameters were described if Figure 4. Bandwidth = 2 x l 0 -3 / d .

UpperFigure 6 .
Figure 6.Energy ratio in rectangular partially confined basin with a gap in the perimeter as a function of gap width and distance of gap from a bounded corner.Eo is initial of total energy upon opening the gap, and E is final of total energy after 500 days.