1838 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
Interaction between the Indonesian Seas and the Indian Ocean in Observations and Numerical Models*
JAMES T. P OTEMRAϩ AND SUSAN L. HAUTALA School of Oceanography, University of Washington, Seattle, Washington
JANET SPRINTALL Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California
WAHYU PANDOE# Department of Oceanography, Texas A&M University, College Station, Texas
(Manuscript received 11 January 2001, in ®nal form 2 November 2001)
ABSTRACT Recent measurements from six bottom-mounted gauges are used with numerical model results to study the exchange of water between the Indonesian seas and the Indian Ocean via the Lesser Sunda Islands known collectively as Nusa Tengarra. The observations are approximately three years in length, from late 1995 to early 1999, and include measurements of bottom pressure, temperature, and salinity. The locations of the gauges are at the boundaries of three straits connecting the southern Indonesian seas with the eastern Indian Ocean: the Lombok Strait, the Ombai Strait, and the Timor Passage. The magnitude of intraseasonal variations in the pressure data dominates over that of the seasonal cycle. Intraseasonal variability appears most frequently and largest in magnitude at the westernmost strait (Lombok) and decreases along the coastline to the Timor Passage. Comparison to wind data shows these intraseasonal variations to be due to Kelvin wave activity in the Indian Ocean, forced by two distinct wind variations: semiannual monsoon reversals and Madden±Julian oscillation (MJO) activity. Sea level variations from both forcing mechanisms are then adjusted by local, alongshore winds. Longer-duration model results show the observation period (1996 through early 1999) to be a time of increased ENSO-related interannual variability and of suppressed annual cycle. MJO activity is also increased during this time. These factors explain the dominance of the higher frequency signals in the pressure data and the relative lack of a distinct annual cycle. An optimal ®t of model sea level variations to model through-strait transport variations is used to estimate transport variability from the observed pressure records. At each strait the optimal ®t is consistent with a cross-strait geostrophic balance for transport variations in the upper 250 m.
1. Introduction region. One previous measurement was made in the Lombok Strait (Murray and Arief 1988) using current For a variety of reasons, estimates of Paci®c to Indian meters deployed for one year. Similar observations were Ocean exchange, particularly at the out¯ow straits in made for one year in the Ombai Strait (Molcard et al. Nusa Tengarra, is very challenging. Aside from the 2001) and for two other years in the Timor Passage Shallow Pressure Gauge Array (SPGA) program that (Molcard et al. 1994, 1996). Several hydrographic sur- was responsible for the data used in the present study veys were made as part of the Java Australia Dynamics and a repeat XBT line from Australia to Java (Meyers Experiment (JADE) and World Ocean Circulation Ex- et al. 1995), there are few long term observations in the periment (WOCE) programs as well, but these provide only single time measurements. A complete review of measurements is given in Lukas et al. (1996). * School of Ocean and Earth Science and Technology Publication 5956 and International Paci®c Research Center Publication 129. More recently, bottom mounted gauges that measure ϩ Current af®liation: SOEST/IPRC, University of Hawaii, Hon- pressure, temperature, and salinity were placed at nine olulu, Hawaii. locations in Nusa Tengarra (see Fig. 1) to study the # Additional af®liation: BPPT, Jakarta, Indonesia. exchange of water between the Indonesian seas and the Indian Ocean (via the geostrophic relationship between Corresponding author address: Dr. James T. Potemra, SOEST/ cross-strait sea level differences and shallow ¯ow). The IPRC, University of Hawaii, 2525 Correa Road, Honolulu, HI 96822. gauges were initially deployed in December of 1995, E-mail: [email protected] and the data records terminate in early 1999. These
᭧ 2002 American Meteorological Society
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC JUNE 2002 POTEMRA ET AL. 1839
FIG. 1. The pressure gauges span ®ve straits and are indicated with a star. Land is shaded and the 1000-m and 100-m isobaths are drawn. observations now provide more than three years of con- from the pressure data alone. Hautala et al. (2001) use tinuous measurements in the region. a geostrophic approximation and ADCP data to level Tides were removed using a least squared ®t of 15 the cross-strait pressure gradient. Here, results from a harmonics, and long-term changes due to anchor settling large-scale, 2½-layer reduced gravity model with real- were identi®ed and removed. Data processing and pre- istic coastlines are used to determine a relationship be- liminary presentation of the ®rst two years of data is tween sea level variations (that are observed) and var- given in Chong et al. (2000). Further analysis, including iations in through-strait transport. This relationship is leveling the pressure variations with acoustic Doppler then applied to the observed pressure signal to provide current pro®ler (ADCP) data taken during cross-strait estimates of transport variability based on the obser- transects is reported in Hautala et al. (2001). Here, the vations. pressure signals at the three main out¯ow straits (Lom- bok, Ombai, and Timor) are examined in greater detail. 2. Out¯ow observations Wind data from the European Centre for Medium- Range Forecasts (ECMWF) along with numerical model Gauges at Bali and Lombok measure ¯ow in the Lom- results are used to address the following questions. First, bok Strait that connects the Java Sea to the Indian what are the pressure variations at the out¯ow straits Ocean. The strait is very narrow, and the gauges are and how do they compare to each other? What forcing located to the north (upstream) of a 225-m sill [from mechanisms are responsible for the measured pressure Smith and Sandwell (1997); see Fig. 2]. The mean depth variations? Are the variations during the observed pe- across the strait between the two gauges is 700 m. riod (1996 through 1998) typical of longer term vari- The center passage is Ombai Strait between Timor ations? Finally, what inferences can be made to trans- (South Ombai gauge) and Alor (North Ombai gauge) port? Implications for heat and freshwater budgets is a Islands. This strait is the deepest of the three, reaching key issue but will be addressed in a subsequent study. a maximum of 2700 m between the two gauges, al- In the following section, pressure data from Lombok though to the north there is a sill of about 1900 m. Water Strait, Ombai Strait, and Timor Passage are presented passing from the Banda Sea between the two gauges and described. Next, the pressure data are analyzed in comes from either west or south of Wetar, an island that comparison with ECMWF wind stress. Finally, an es- lies to the east of Alor. timation of transport variations is made using the ob- The southernmost observed strait is Timor Passage, served pressure. It is dif®cult to get absolute transport measured by gauges at Roti Island (southwest of Timor)
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC 1840 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
FIG. 2. The three principle out¯ow straits are shown in the left panels. Approximate pressure gauge locations are indicated by asterisks. Bathymetry along a section between the two gauges at each strait (from Smith and Sandwell 1997) is shown in the right panels.
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC JUNE 2002 POTEMRA ET AL. 1841
TABLE 1. Station locations and strait depths. Lat Lon Width Avg depth Station ID Location Strait (ЊS) (ЊE) (m) (m) BAL Bali Lombok 8.4 115.7 36 484 708 LBK Lombok 8.4 116.0 ROT Roti Timor 10.8 122.7 136 235 816 ASH Ashmore 12.2 123.0 NOM N. Ombai Ombai 8.4 125.1 35 005 1689 SOM S. Ombai 8.7 125.1 and Ashmore Reef (northwestern reef of Australia). This covered in 1999, so there are no data at these sites for passage is broad, about 136 km, but very shallow on early 1998 to early 1999. both sides. The maximum depth between the two gauges Pressure variations at the straits can be caused by a is about 1700 m. The present analysis is limited to the variety of mechanisms including thermodynamic ef- ¯ow between the shelf and Timor, but there could be fects, tides, and wind. For this study only the direct, ¯ow on the shallow Australian shelf (e.g., Cresswell et wind-driven effect will be considered (tides were re- al. 1993) to the south of the Ashmore gauge that cannot moved from the pressure data as described in Chong et be addressed here. al. 2000). Both local (alongshore) and remote (Paci®c The precise locations of the gauges are given in Table and Indian Oceans) winds can have an affect on sea 1. The South Ombai and Ashmore gauges were not re- level (and therefore bottom pressure) along the coast.
FIG. 3. Three-day surface wind stress from ECMWF at two locations are given: zonal mean in the central Indian Ocean (3ЊSto3ЊN, 60Њ to 80ЊE) with the dashed line and zonal mean along the south shore of Java with shading. For each, positive values indicate downwelling favorable, westerly winds. Each was ®t with a 60-day running-mean ®lter, given by the heavy dashed line for the equatorial winds and the heavy solid line for the S. Java winds.
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC 1842 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
Local alongshore wind stress along the southern coast annual and higher frequency changes. Further, the ef- of the archipelago drives Ekman ¯ow perpendicular to fects from the various wind forcing interact with each the shoreline that leads to changes in sea level. During other within the archipelago. For example, local winds April through October, winds along the southern coast are upwelling favorable along south Java during the of the archipelago are generally to the west (Fig. 3), southeast monsoon (April±October), and downwelling and the resulting offshore Ekman ¯ow creates coastal from semiannual Kelvin waves generated in the equa- upwelling. During the rest of the year the winds are torial Indian Ocean is diminished at this time. During toward the east, and Ekman ¯ow creates downwelling the northwest monsoon (November±March) downwell- along the southern coasts of Nusa Tengarra. ing favorable alongshore winds would enhance the re- Remote winds from both Paci®c and Indian Oceans motely forced downwelling Kelvin waves. In the fol- can also contribute to observed pressure variations in lowing section, the observed pressure signal at each the out¯ow straits. Winds in the Paci®c drive large-scale strait will be presented and variations will be related to circulation that could force low-frequency changes in the wind ®eld. ¯ow through the out¯ow straits (e.g., Godfrey 1989), and local sea level would adjust geostrophically. Low a. Lombok Strait frequency changes in Paci®c winds may also modulate the pressure head between the Paci®c and Indian Oceans Aside from the present study, there is only one pre- (Wyrtki 1987) that would also cause low-frequency ¯ow vious observational effort made in Lombok Strait. Five variability in the straits. moorings (currents measured at 35 m, 75 m, 300 m, and Winds in the Indian Ocean can affect sea level along 800 m) were deployed for one year during 1985 (Murray the archipelago via the generation of coastal Kelvin and Arief 1988). The estimated seasonal cycle in 1985 waves. These coastal Kelvin waves can arise when was a broad minimum of 1 Sv (Sv ϵ 106 m3 sϪ1) from equatorial Kelvin waves, forced by winds in the central February through May, and a maximum of 4 Sv in July Indian Ocean, impinge on the west coast of Sumatra, through September (long-term mean of about 2 Sv). or they can be forced by local winds along the west Most of the estimated transport (80%) was con®ned to coast of Sumatra. There are two main frequencies as- the upper 200 m (Murray and Arief 1988). The measured sociated with the near-equatorial winds that give rise ¯ow was predominantly southward, and discrete occa- to Kelvin waves in the Indian Ocean. Intraseasonal sions of ¯ow reversal (observed in February, March, atmospheric oscillations, so-called Madden±Julian os- and April) may be related to downwelling coastal Kelvin cillations (MJO) (Madden and Julian 1972), are char- waves that would act to increase the cross-strait pressure acterized by westerly anomalies in the surface winds gradient by raising sea level on the western side of the near the equator throughout the Indian Ocean and ap- strait. Arief and Murray (1996) also deployed pressure pear as high frequency (30 to 50 day) bursts in the gauges in Lombok Strait and suggested that they indi- alongshore winds off south Java, for example, Decem- cated only a tendency toward geostrophic balance in ber±February 1996/97 (Fig. 3). Semiannual variations this strait. However, their results are based on short in westerly winds occur in the central equatorial Indian timescales, as only four months of data were recovered during the monsoon transition (Fig. 3), roughly in May for the mooring on the western side of the strait. In and November. general, sea level in the strait was observed to match The coastal Kelvin waves formed by these wind var- the ¯ow; during times of intense southward ¯ow, sea iations propagate eastward along the southern coast of level dropped and, when ¯ow reversed to the north, sea the archipelago and pass the out¯ow straits. The gaps level rose. in the coastline are suf®ciently small (less than the ®rst The SPGA observed pressure record, long-term mean mode Rossby radius of 130 km) so that the signal can removed, at Lombok Strait is given in Fig. 4. The signal theoretically propagate to Ombai Strait before having at Bali (dashed line), on the western side of the strait, to turn north and then west (the gap at Timor Passage dominates the cross-strait gradient, consistent with in- is larger than 130 km). These waves, then, play a role tensi®cation of ¯ow on the Bali side in ADCP sections. in adjusting sea level at the out¯ow straits. Coastal The cross-strait pressure gradient correlates at 0.87 to Kelvin waves also affect the seasonally reversing South the pressure variations at Bali, but it is not signi®cantly Java Current (SJC) that generally ¯ows eastward in re- correlated to Lombok (r ϭ 0.26). There is a correlation sponse to the downwelling waves (Quadfasel and Cres- (r) of 0.70 between pressure variations at Bali and those swell 1992; Sprintall et al. 1999). at Lombok (for the measured time period). Most of the In summary, the causes of wind-driven variations in energy is in variations of monthly and longer timescales sea level (and therefore variations in the observed bot- (see spectra in Fig. 5), which limits the degrees of free- tom pressure) can be local (alongshore) or remote. Re- dom in the signal, but a correlation of r ϭ 0.7 is sig- mote Paci®c winds, through resulting equatorial Rossby ni®cant at the 90% level. and coastal Kelvin waves (Potemra 1999), can lead to Variations in the pressure gradient (long-term mean low frequency variations in sea level in the Indian removed from each pressure record) gives a measure of Ocean, while remote Indian Ocean winds cause semi- variations in the surface geostrophic ¯ow through the
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC JUNE 2002 POTEMRA ET AL. 1843
FIG. 4. Three day mean pressure variations are plotted for each strait: the upper panel is for Lombok Strait, the center panel is for Ombai Strait, and the lower panel is for Timor Passage. The dashed lines are pressure at the northern or western side (Bali, North Ombai, and Roti), and the thin dotted lines are the pressure at the southern or eastern side (Lombok, South Ombai, and Ashmore). The heavy solid line is the cross-strait pressure gradient plotted such that positive values are indicative of an increase in ¯ow toward the Paci®c (lower through¯ow). strait, and Fig. 4 suggests maximum southward ¯ow to Lombok, and neither one is signi®cantly correlated through Lombok Strait in June±October. Interestingly, to local winds off south Java. when sea level is lower than average, the cross-strait The highest correlation of pressure at both sides of gradient in sea level is also lower than average (see Fig. Lombok Strait with equatorial wind in the Paci®c occurs 4). This could explain why Murray and Arief (1988) at a 120 day lag (lower panels of Fig. 6). This is con- found a correlation between ¯ow through the strait and sistent with Potemra (1999), where it was shown nu- sea level variations at their Bali site [since the cross- merically that equatorial Rossby waves generated in the strait gradient is correlated to the variations at the Bali eastern equatorial Paci®c would affect through¯ow var- side and the Murray and Arief (1988) pressure data was iability as they generate poleward propagating coastal predominantly from the Bali side]. Kelvin waves along the eastern side of the Indonesian To investigate the effects of local and remote forcing seas. A ®rst baroclinic mode equatorial Rossby wave, on pressure variations at Lombok Strait, correlations generated at 140ЊW, would take about 120 days to reach were made between zonal ECMWF wind stress and the the Indonesian seas. observed pressure signals at Bali and Lombok (see Fig. Variability from remote Indian Ocean forcing would 6). At zero lag (upper panels of Fig. 6), the highest take less time to reach the out¯ow straits due to the correlations are with zonal wind off Sumatra. The cor- relatively fast phase speed of Kelvin waves. If winds at relation to pressure at Bali is larger than the correlation 70ЊE generate ®rst mode Kelvin waves with an ap-
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC 1844 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
FIG. 5. Power spectral density (variance preserving) is shown for each strait: the upper panel is for Lombok Strait, the center panel is for Ombai Strait, and the lower panel is for Timor Passage. The dashed lines are the pressure at the northern or western side (Bali, North Ombai, and Roti) and the dotted lines are the pressure at the southern or eastern side (Lombok, South Ombai, and Ashmore). The solid lines are the power spectra of the cross-strait pressure gradients at each site.
proximate phase speed of 2.8 m sϪ1, the waves would in the local wind stress but are in the equatorial winds take approximately 15 days to reach Sumatra and an- in the Indian Ocean. other 9 days to reach the Lombok Strait. Consistent with Speci®c variations in pressure at Bali and Lombok this, the region of highest correlation between pressure can be traced back to wind events in the central Indian at Lombok Strait and wind stress at a 24 day lag (wind Ocean (Fig. 7). Bursts of positive sea level can be seen stress leads pressure variations) occurs in the equatorial at both Lombok and Bali in May 1996, November 1996, Indian Ocean (center panels of Fig. 6). May/June 1997, May/June 1998, and November of A substantial part of the variability in the pressure 1998. Consistent with the phase speed of ®rst-mode data occurs on intraseasonal timescales (Fig. 4), and it Kelvin waves, these waves were generated by westerly is hypothesized that these intraseasonal variations are winds approximately 21 days earlier in the central equa- the result of Kelvin waves forced remotely in the Indian torial Indian Ocean during the monsoon transitions (Fig. Ocean. Semiannual monsoon wind changes and 30 to 3), and they are also coincident with reversals in the 50 day MJO activity are two such forcing mechanisms. surface pressure gradient at Lombok Strait. Occasion- Similar to the modeling work of Qiu et al. (1999), there ally there are short-term variations in local winds during are spectral peaks in the Bali and Lombok pressure sig- these events, but they are small compared to those in nals at 50 and 75 days (Fig. 5). There are no such peaks the equatorial Indian Ocean. In 1996 the May down-
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC JUNE 2002 POTEMRA ET AL. 1845
FIG. 6. Correlation (r) between ECMWF zonal wind stress and observed pressure variations at Bali (left panels) and Lombok (right panels) at three different lags. The upper panels are correlations at zero lag, the center panels are for pressure lagging wind stress by 24 days, and the lower panels are for pressure lagging wind stress by 120 days. welling event is small compared with 1997 and 1998, 1999). In fact, these easterlies produce upwelling Kelvin as the local wind stress was upwelling favorable. In waves that may have acted to reduce the through¯ow 1997 and 1998 the upwelling favorable winds occur a in late 1997. few weeks later. There is also a reduction in the bottom In late 1998 and early 1999, the alongshore winds off pressure at Bali just prior to the arrival of downwelling Java are westerly and downwelling favorable, leading Kelvin waves in March of 1996, March/April 1997, and to increased positive pressure variations. Local and In- March/April 1998 (Fig. 7). To some extent, local winds dian Ocean winds are not suf®cient to account for the account for this, with easterly, upwelling favorable magnitude, and a Paci®c in¯uence is likely. winds in March/April (Fig. 7). Since there is no asso- Westerly winds associated with the MJO generated ciated pressure reduction on the Lombok side, however, in the central Indian Ocean also cause downwelling, and these could also be indicative of an increase in the Pa- these are evident as positive pressure variations at the ci®c to Indian Ocean geostrophic ¯ow, forced on the Bali site in December, January, and February. Paci®c Ocean side of the through¯ow, or as response to local, upwelling-favorable winds. Finally, the Octo- b. Ombai Strait ber/November Kelvin wave is typically weaker than the April/May wave, perhaps due to the deeper Indian Ombai Strait lies between Alor and Timor islands. Ocean thermocline later in the year (Han et al. 1999). The pressure gauges were located at 8.4ЊS (North Om- In addition, the local alongshore winds south of Java bai) and 8.7ЊS (South Ombai) at 125.1ЊE. The gauge at are upwelling favorable through December (Fig. 7), South Ombai was not recovered in 1999, so the joint which probably also accounts for the reduced down- difference record spans December 1995±March 1998 welling signal from the November Kelvin waves. In (Fig. 4). Like the Lombok Strait, pressure measured at November 1997 a downwelling wave was not generated the two sides of Ombai Strait are highly correlated (r (Sprintall et al. 2000) due to prolonged easterlies that ϭ 0.9), but in Ombai Strait the pressure difference be- existed in the equatorial Indian Ocean during the later tween the two gauges is not correlated to either gauge. half of 1997 (Yu and Rienecker 1999; Webster et al. This may be due to the coastline geometry at Ombai.
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC 1846 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
FIG. 7. Observed pressure variations (in db) are plotted as shading. The upper panel is for the Bali gauge, the center panel is for Lombok, and the lower panel is the cross-strait pressure difference. Mean winds (in N m Ϫ2) at three locations are also shown: mean zonal wind stress in the central Indian Ocean (3ЊSto3ЊN, 60Њ to 80ЊE) with the dashed lined, alongshore mean along the west coast of Sumatra with the thin solid line, and zonal mean along the south shore of Java with the heavy line. In each case the wind is plotted such that downwelling favorable winds are positive (i.e., eastward in the central Indian Ocean and along the Java coast, and southeastward along the Sumatra coast). As described in the text, the central Indian Ocean winds are lagged by 24 days, and the Sumatra winds are lagged by 9 days.
At Lombok, coastal waves encounter Bali ®rst then in Ombai Strait the downwelling events associated with Lombok while at Ombai the orientation of the coast and the November Kelvin waves are less apparent. the southern boundary of Timor Island cause sea level to be similar at both sides of Ombai simultaneously. c. Timor Passage The propagation of coastal waves is also evident in the high correlation of sea level variability at Bali and The southernmost out¯ow strait is Timor Passage. both sides of the Ombai Strait (r ϭ 0.86 between Bali The northern gauge was located offshore of the island pressure and Ombai Strait pressure ®ve days later). The of Roti, off the southern tip of Timor. The southern two straits are separated by 1034 km. Assuming the gauge was at Ashmore Reef, on the northwest Australian pressure signal at Bali propagates to Ombai as a ®rst shelf. The channel between the two reaches about 1700 mode coastal Kelvin wave (c ϭ 2.8msϪ1), it would m deep, and is the pathway for wave energy to reach take about 4.3 days for the signal to reach Ombai. At the Indian Ocean from the Paci®c and Indonesian seas Ombai there are extended periods of lower sea level in (Potemra 1999). Analysis at Timor Passage is hampered June±October 1996 and June 1997 through February by missing data from late 1996 through early 1997 at 1998 (when the record ends) and three episodes of pos- the Ashmore Reef and, since the gauge was not recov- itive sea level in December 1996, January 1997, and ered in 1999, the record ends in late 1998. March 1997. The Kelvin wave signals driven by mon- The coastal waves evident at the Ombai and Lombok soon changes observed at Bali in April/May are also Straits are much diminished at Timor Passage. As dis- apparent at the Ombai gauges in 1996 and 1997. Again, cussed previously, the gaps between the islands are suf-
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC JUNE 2002 POTEMRA ET AL. 1847
FIG. 8. Correlation (r) between ECMWF zonal wind stress and observed pressure variations at Roti at three different lags. The upper panel is correlation at zero lag, the center panel is for pressure lagging wind stress by 24 days, and the lower panel is for pressure lagging wind stress by 120 days.
®ciently small so that Kelvin wave energy can propagate stress to the observed pressure records at Timor Passage. along the southern arc of Nusa Tengarra. Indeed, pres- The Roti record shows an agreement to winds on both sure at Bali, North Ombai, South Ombai, and Roti are the Paci®c and Indian Ocean sides. Similar to the Lom- all well correlated (r ϭ 0.75 with an 8-day lag between bok correlations (Fig. 6) there is a patch of high cor- Bali and Roti). The gap between Roti and Ashmore, relation in the equatorial Indian Ocean although it is however, is 136 km, larger than the ®rst mode Rossby over a much broader region. Unlike the Lombok Strait, radius (130 km). Consequently the pressure signal at however, the correlations at Roti are negatively corre- Ashmore is not signi®cantly correlated to the pressure lated to winds in the equatorial Paci®c Ocean at all lags. at any of the other sites, suggestive of a different pri- This is perhaps indicative of larger-scale wind variations mary forcing mechanism [e.g., remotely generated simultaneously over the Paci®c and Indian Ocean af- coastal Kelvin waves from the Paci®c (Potemra 1999]. fecting Roti, while at Lombok more local (e.g., MJO- Figure 8 shows the correlation of ECMWF zonal wind type) winds are responsible. Signi®cant correlations of
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC 1848 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
FIG. 9. Sea level variations (m) from observed pressure gauges (solid lines) is compared to sea level from the numerical model (dashed line) over the time period of the observations. The upper panels are for both sides of the Lombok Strait, the center panels are for the Ombai Strait and the lower panels are for both sides of the Timor Passage. winds with Ashmore pressure (not shown) are limited occurs mainly in two layers (more than 95% of the by the shorter time series, but Potemra (2001) showed variance from EOF analysis). how low-frequency changes in sea level along the north- This is in agreement with recent in situ current meter west Australian shelf (using a series of tide gauges) were measurements at Ombai Strait (Molcard et al. 2001) and not correlated to local forcing but were due to remove at Timor Passage (Molcard et al. 1996). ADCP transects waves generated in the equatorial Paci®c. taken during the present SPGA program also indicate a two vertical mode structure to the ¯ow (Hautala et al. 3. Observations and model analysis 2001). Since the POCM does not accurately resolve the three Results from a numerical model are used to assist in straits measured by the SPGA, a more simple model the analysis of the observed pressure data. One dif®culty 1 1 was run with a higher horizontal resolution ( ⁄Њ by ⁄Њ). in modeling the through¯ow is that it requires a large- A2 -layer reduced-gravity model (henceforth referred scale domain (one that includes Paci®c and Indian ba- ½ sins) and at the same time high horizontal resolution in to as UHLM) similar to the model described in Potemra the Indonesian region to resolve the narrow straits. The (1999) was run using 3-day ECMWF wind stress as the Parallel Ocean Climate Model (POCM) (Semtner and only external forcing (there are no thermodynamics in Chervin 1992) is a global model with relatively high the UHLM). The two-mode vertical structure derived vertical resolution (20 levels), but with an average of 2⁄Њ from the POCM (and the few in situ observations) was horizontal resolution it does not accurately resolve a used to determine the initial vertical structure of the few key straits, for example, Lombok Strait (closed) or UHLM; the initial layer thicknesses were chosen to be Torres Strait (too wide). Analysis of the POCM results 100 m for the upper layer and 400 m for the second does however show that ¯ow through the out¯ow straits layer. The model was integrated from rest for 20 years
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC JUNE 2002 POTEMRA ET AL. 1849
FIG. 10. UHLM sea level variations are given for 15 years of model integration. The results have been low-pass ®ltered using a 1-yr running mean. The upper panels are for Lombok Strait, the center panels for Ombai Strait, and the lower panels for Timor Passage. The right panels show the variance preserving power spectra at each site.
using a climatological mean 1985±99 ECMWF wind pendent bathymetry. Nonetheless, the good agreement stress then forced with 3-day winds for 1985±99. between UHLM and observed sea level is remarkable. The model does accurately reproduce the observed sea The model also does well in simulating observed sea level variations. Figure 9 shows a comparison of model level at Ombai Strait and Timor Passage (Fig. 9). The sea level variations with those based on the observed correlation of UHLM sea level to observed sea level is pressure variations (converted to sea level by ЈϭЈ/ 0.90 for North Ombai, 0.89 for South Ombai, 0.84 for g). In the layer parameterization of the model, sea level Roti, 0.77 for the ®rst half-record at Ashmore, and 0.86 variations are approximated by (gЈ/g)hЈ, where gЈ is the for the second-half record at Ashmore. reduced gravity and hЈ is the upper-layer thickness with The UHLM results were ®rst used to put the SPGA the long-term mean removed. This assumes that sea level period into a longer-term perspective. Figure 10 shows variations are linked to thermocline variations, and bathy- the UHLM results and the power spectral density of the metric effects are necessarily neglected. 15-yr records. The long-term spectra (of 3-day model The UHLM resolves the Lombok Strait at its most output) show less power in the higher frequencies (pe- narrow point with two grid points. The resulting model riods less than semiannual) than in the 3-yr spectra (see sea level is well correlated to the observed sea level Fig. 5). Consistent with the hypothesis outlined earlier, variations at Bali and Lombok (r ϭ 0.80 and 0.67, re- sea level at Timor Passage has more low frequency en- spectively; see Fig. 9) even though the UHLM does not ergy (stronger Paci®c in¯uence) than Lombok and Om- include any thermodynamic effects nor have depth-de- bai Straits, which have more annual and semiannual
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC 1850 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
energy (Indian Ocean in¯uence). There is also more V0toЈϭz0 ␣11Јϩ␣ 22Ј (5) annual and semiannual energy in sea level on the west- with ern and northern sides of the straits (Bali, North Ombai, Roti) than the eastern and southern sides (Lombok, z0 South Ombai, Ashmore). ␣iiϭ c ⌬x (z) dz. (6) The model sea level variations (3-day) were low-pass ͵0 ®ltered with a 1-yr running mean to highlight the low- To demonstrate with an example, Eq. (5) can be re- frequency variations (Fig. 10). In all three straits, large written for the case of ¯ow in the strait that is in geo- amplitude ¯uctuations (10 cm) occur during the SPGA strophic balance: period (1995±99). Prior to this period, low frequency VЈϭ(␣ ϩ ␣ )Јϩ␣ (ЈϪЈ). (7) variations are reduced. In the model, therefore, inter- 0toz0 121221 annual variations during the late 1990s are two to three The ®rst term represents changes in transport due to times larger than during the years prior. changes in sea level at only one side of the strait (e.g., Finally, the UHLM results were used to estimate from large-scale coastal waves) and the second term is transport variations from the observed sea level varia- equivalent to a cross-strait geostrophic balance; that is, tions. The approach taken was to ®nd a relationship g (ЈϪЈ) gz VЈϭ21 ⌬x⌬z ϭ 0(ЈϪЈ). (8) between sea level variability (at each side of a strait) 0toz0 f ⌬xf21 and through-strait transport using the model sea level and transport ®elds. This relationship was then applied In this case, ␣ 2 can be compared to gz0/ f. to the observed sea level variations. This assumes that At this point, there has been no assumption about the the relationship of sea level to transport is accurate in change in velocity with depth. The few observations of the model. the depth-dependent ¯ow (e.g., Molcard et al. 1996,
Surface to depth (z0) integrated transport variations, 2001; Hautala et al. 2001), as well as model results indicate an exponential decay of velocity with depth. V 0toЈ z0, are given by The relationship derived above can be applied to an zx0 2 example of exponential decay of velocity with depth by VЈϭ Ј(x, z) dx dz, (1) 0toz0 ͵͵ setting (z) ϭ eϪz. In this case, Eq. (6) becomes 0 x1 ⌬x where Ј(x, z) represents velocity variations in the strait ␣ ϭ c (1 Ϫ eϪz0). (9) ii bounded by x1 and x 2. Velocity variations averaged across the strait [strait width at depth z given by L(z)], To determine the constants ␣1,2, sea level from the Ј(z), are de®ned by model (at each side of a particular strait) was regressed to transport through the strait (also from the model). 1 x2 Ј(z) ϭ Ј(x, z) dx. (2) Regressing sea level at each side (not the sea level dif- L(z) ͵ ference) provides an additional check on the geostrophic x1 approximation. This calculation was done for each of Using a change in mean cross-strait velocity with the three straits as described below. depth of (z), where (z) ϭ 1 at the surface (z ϭ 0), Eq. (1) can be rewritten using Eq. (2): a. Lombok Strait zz00 VЈϭ ЈЈ(z)L(z) dz ϭ (0)L(0)(z) dz To examine the relationship of sea level to through- 0toz0 ͵͵ 00 strait transport in the model, a linear regression was used to best ®t the UHLM sea level variations at Bali and z0 ϭ Ј(0)⌬x (z) dz (3) Lombok (separately) to the model upper-layer transport ͵ variations (the upper 100 m, approximately): 0 V , (10) where ⌬x is the distance across the strait [i.e., L(0)]. layer1Јϭ␣ 1 BALЈϩ␣ 2 LOMЈ
Assuming a linear relationship between sea level var- whereV layer1Ј is the variation in transport (in Sv) inte- iations at the two sides of a strait,1,2Ј , and surface ve- grated over the model upper layer and Ј are the var- locity variations (averaged across the strait), Ј(0), of the iations in sea level (in m) at each side of the strait (Bali: form Ј(0) ϭ c1 12Ј ϩ c2Ј, Eq. (3) may be rewritten as BAL, Lombok: LOM) weighted by ␣1 and ␣ 2. For the
z 3-yr period coinciding with the observations, ␣1 ϭ 10.7 0 and 13.2. The magnitude of the two coef®cients VЈϭ(c Јϩc Ј)⌬x (z) dz. (4) ␣ 2 ϭϪ 0toz0 11 22 ͵ are nearly equal, suggesting that ¯ow is consistent with 0 a cross-strait geostrophic balance (the narrow gap, 36 The variations in transport through the strait are there- km, at Lombok Strait is much smaller than the ®rst mode fore given as a function of sea level variations at either Rossby radius of 130 km, so it is unlikely that Kelvin side of the strait waves will affect only one side of the strait).
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC JUNE 2002 POTEMRA ET AL. 1851
Equation (10) gives an estimate of transport variations V layer1Јϭ␣ 1 NOMЈϩ␣ 2 SOMЈ . (11) that is well correlated (r ϭ 0.79) to depth-integrated In the case of the Ombai Strait, the best ®t of UHLM model transport variations through Lombok Strait. For sea level variations to UHLM upper layer transport var- the entire UHLM integration (1985±99), the coef®cients iations occurs for ␣1 ϭ 27.9 and ␣ 2 ϭϪ29.0. This ␣1 ϭ 10.8 and ␣ 2 ϭϪ14.1 are similar to those found parameterization of transport has a correlation of 0.90 for the SPGA period, and the correlation of transport to the transport directly integrated from the model re- given by Eq. (10) to depth-integrated transport increases sults. The application of Eq. (11) to the observed sea slightly to 0.85. The depth-integrated UHLM transport level variations is given in Fig. 11. The best ®t of Eq. variations are not well correlated to sea level at Bali (11) using the entire model integration time period from (0.58) or Lombok (0.20), contrary to the hypothesis that 1985 to 1999 gives similar coef®cients, ␣ ϭ 24.9 and transport through the strait can be determined from sea 1 ␣ 2 ϭϪ25.9, and the correlation of estimated to direct level variations at only one side. transport increases slightly to 0.93. Equation (10), with ␣1,2 obtained from the UHLM For both the entire 15-yr model run and over the 3- over the SPGA time period, and observed sea level var- yr period of observations, the relationship of model sea iations from the SPGA array were used to estimate upper level difference across Ombai Strait to transport through ocean transport and the result compared to the estimate the strait is nearly a linear constant. Unlike Lombok based on model sea level (Fig. 11). The low frequency Strait, the constant in this case is about 30 Sv mϪ1. trends are in good agreement, with larger than normal Using Eq. (9) as described above (assuming an expo- transport towards the Indian Ocean during May±Octo- nential decay of velocity with depth), this gives a decay ber. The model has some higher frequency variations factor () of 0.01 mϪ1 at Ombai Strait. This implies that that are driven by sea level on the Bali side (Fig. 9; e.g. velocity decays to zero slower (at a deeper depth) at in mid-1996) that are not in the observations. In general Ombai Strait than at Lombok Strait. For ϭ 0.01, the the model transport variations are larger in magnitude velocity at 100 m is about 35% of its surface value than what is inferred from the observations, probably assuming an exponential decay with depth. At Lombok, due to the ®nite resolution of the model. In the model the surface velocity was 3% of its surface value at 100 the Bali and Lombok sites are only separated by three m, while at Ombai Strait surface velocity does not decay grid points. Sea level is therefore highly correlated be- to 3% of its surface value until about 400 m. tween the two in the model results, and the difference It is dif®cult to verify that the decay of velocity with in sea level between the two sides of the channel is depth is slower at Ombai than at Lombok. However the lower than the observations. The coef®cients in Eq. (1), observations of (Molcard et al. 2001) show ¯ow in Om- then, are probably too large. The important point, how- bai Strait to decay to 5% of the surface value at about ever, is that the two coef®cients (for Bali and Lombok) 1200 m. Individual ADCP records, taken across the are nearly the same, indicating that geostrophy holds straits (as part of the SPGA program) are not conclusive, for the strait. but are consistent; Hautala et al. (2001) show there is The change in velocity with depth can be inferred still signi®cant ¯ow toward the Indian Ocean at 250 m. from Eq. (10), given that transport variations are ap- The mean vertical pro®le of velocity at Ombai Strait in proximately equal to a constant times variations in the POCM does show an agreement with the UHLM cross-strait sea level difference. At Lombok Strait the estimate at decays to zero at about 400 m. constant is approximately 11.0 Sv mϪ1. If an exponential decay of velocity with depth is assumed, for example, c. Timor Passage Eq. (9) can be solved for the decay constant . For Lombok Strait is approximately 0.04 mϪ1. This rep- A linear regression of sea level variations at either resents a velocity at 100 m that is about 3% of the side of Timor Passage was made to transport through surface value. The observed ADCP velocity at Lombok the passage: was found to decay to zero at about 200 m in March V layer1Јϭ␣ 1 ROTЈϩ␣ 2 ASHЈ . (12) 1997 and 1998 (Hautala et al. 2001). In December 1995, when the model ¯ow is to the north (see Fig. 9), the For the UHLM results over the observation period (in tidally corrected ADCP velocity observations also found this case just the second half of the Ashmore record), 30.2 and 33.6. Again, the close magnitude northward ¯ow from the surface to 300 m, ranging from ␣1 ϭ ␣ 2 ϭϪ of the two coef®cients suggests that the geostrophic re- 60 cm sϪ1 near the surface to 20 cm sϪ1 at 200 m (Hautala et al. 2001). lationship holds. In the UHLM case, the sea level based approximation of transport has a correlation of 0.92 to transport (over b. Ombai Strait the 3-yr model record). For the entire model record (15 years), ␣1 ϭ 35.9 and ␣ 2 ϭϪ39.7, and the correlation The UHLM sea level variations were ®t to the model to integrated transport is 0.96. Application of the model integrated transport at Ombai Strait similar to the pro- derived transport estimate [Eq. (12)] to the observed cedure described for the Lombok Strait: pressure signal is shown in Fig. 11.
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC 1852 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
FIG. 11. Through-strait transport variations were estimated from the observed pressure variations (solid line) and the UHLM sea level (thin dotted line). The depth-integrated model transport is given with the dashed line. The upper panel is for Lombok Strait, the center panel is for Ombai Strait and the lower panel is for Timor Passage.
4. Discussion forcing mechanisms described above could lead to vary- ing ¯ow at two different levels. Generally speaking, The SGPA measurements of bottom pressure provide higher sea level on the Lombok side is indicative of a relatively long term, high frequency record of vari- higher southward through-strait geostrophic transport ability at the straits connecting the Indonesian seas to and is accompanied by warmer, fresher water in the the Indian Ocean. One striking feature of the dataset is the intraseasonal variability at the out¯ow passages sur- strait. Higher sea level on the Bali side, indicative of veyed that is most dominant in the west (Lombok Strait) lower southward through-strait geostrophic transport, is and decreases to the east (Timor Passage). This vari- accompanied by higher temperatures and at most times ability appears to be forced on the Indian Ocean side is unrelated to salinity. Analysis of the SPGA temper- of archipelago in response to MJO wind activity. It is ature and salinity data will be presented in a future work. speculated that low-frequency pressure variations (at pe- It also seems that wind forcing in the Indian Ocean riods longer than 6 months), mainly on the eastern side is responsible for high-frequency sea level variability at of Lombok Strait (at Lombok), respond more to Paci®c Ombai Strait and to a lesser degree at the northern side forcing and high frequency variations, mainly on the of Timor Passage (at Roti). A combination of coastal western side (at Bali), respond to Indian Ocean forcing. Kelvin waves forced by equatorial Indian Ocean winds, There are also differences in observed temperature MJO activity, and local alongshore winds account for and salinity variations at the two sides of Lombok Strait much of the variation in pressure at Lombok Strait but (not presented here) leading to the possibility that the less so at the more eastern Ombai Strait, which has more
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC JUNE 2002 POTEMRA ET AL. 1853 of a Paci®c in¯uence. The southernmost strait, Timor m, to capture the two modes derived from the model Passage, seems primarily controlled by Paci®c wind var- results. If these observations con®rm the model results, iability. Paci®c-forced variability at Ombai Strait is ap- then long term monitoring of transport could be done parent in a higher vertical mode. Model variations seem using just sea level, or even wind stress, if the link to con®rm this, but more detailed model experiments between winds and sea level can be determined more are needed to speci®cally address these hypotheses. accurately. In addition, a large part of the variations of The longer-term, high resolution UHLM results show sea level and therefore transport seems to occur in short, that the observational period of 1996 through mid-1999 10 to 50 day bursts. Future monitoring would need to is a time of increased interannual and intraseasonal var- include measurements at these higher frequencies. iability and reduced seasonal cycle. The observations and model results are dominated by intraseasonal var- Acknowledgments. The success of the Shallow Pres- iability during this period. For example, 50% of the sure Gauge Array (SPGA) program that provided the variance in transport through Timor Passage from the data for this work was due in large part to the efforts 20-yr POCM results is in intraseasonal variations, 48% of Nan Bray, who, among other things, not only initiated in the seasonal cycle (annual and semiannual harmonic) the program but also served as chief scientist on the and the remainder in lower frequencies. Whereas in ®rst two cruises. We also gratefully acknowledge the 1995±98 (SPGA period), 50% of the total variance is commanders and crew of the R/V Baruna Jaya I and again in intraseasonal variability, but only 30% is in the R/V Baruna Jaya IV for their capable and enthusiastic seasonal cycle and 20% in low frequency variations. In support of the project. We also thank our colleagues at the POCM Ombai Strait there is not much change in BPPT Indonesia, particularly Dr. Indroyono Soesilo and the distribution of variance during the different time Mr. Basri Gani, and Dr. Gani Ilahude at LIPI for their periods, possibly due to the relatively larger contribution organizational efforts and warm collegiality. The ®eld of low frequency Paci®c forcing. The possibility of support, technical expertise, and scienti®c insight of changes in through-strait transport during ENSO years, Werner Morawitz, Thomas Moore, Paul Harvey, and speci®cally a change in the ratio of transport carried in Jackson Chong were an important part of the program. each strait remains to be determined. Similar questions Financial support was provided by the National Science remain for the in¯ow straits on northern side of the Foundation (OCE-9819511, OCE-9818670, OCE- Indonesian seas, where POCM results (not shown here) 9505595, OCE-9415897, and OCE-9529641) and the indicate the through¯ow is split between Makassar Strait Of®ce of Naval Research (N00014-94-1-0617). Finally, and ¯ow to the east of Sulawesi (approximately 4 Sv we thank Gary Meyers and an anonymous reviewer and 3 Sv, respectively). whose comments greatly improved the manuscript. The high resolution, reduced gravity model (UHLM) was also used to derive a relationship between sea level and depth-integrated transport. At each of the three REFERENCES straits, the best ®t of sea level at either side of the strait to transport through the strait occurs when a constant Arief, D., and S. P. Murray, 1996: Low-frequency ¯uctuations in the is multiplied by the sea level difference across the strait, Indonesian through¯ow through Lombok Strait. J. Geophys. Res., 101, 12 455±12 464. consistent with the geostrophic approximation. Chong, J. C., J. Sprintall, W. L. M. Morawitz, S. Hautala, N. A. Bray, Sea level on the two sides of each strait is correlated, and W. Pandoe, 2000: Transport through the out¯ow straits of suggesting variations from one side could be used to Indonesia. Geophys. Res. Lett., 27, 125±128. estimate transport as in Murray and Arief (1988). The Cresswell, G., A. Frische, J. Peterson, and D. Quadfasel, 1993: Cir- culation in the Timor Sea. J. Geophys. Res., 98, 14 379±14 389. highest correlation of sea level on two sides of a strait Godfrey, J. S., 1989: A Sverdrup model of the depth-integrated ¯ow occurs at Ombai: 0.90 between North Ombai and South for the world ocean allowing for island circulations. Geophys. Ombai. In the UHLM, transport based on one side of Astrophys. Fluid Dyn., 45, 89±112. the strait has a correlation of only 0.47 to model trans- Han, W., J. P. McCreary, D. L. T. Anderson, and A. J. Mariano, 1999: port, while using the cross-strait pressure difference Dynamics of the eastward surface jets in the equatorial Indian Ocean. J. Phys. Oceanogr., 29, 2191±2209. yields a estimate with a correlation of 0.93 to the in- Hautala, S. L., J. Sprintall, J. T. Potemra, A. G. Ilahude, J. C. Chong, tegrated model transport. It is true, however, that at most W. Pandoe, and N. Bray, 2001: Velocity structure and transport times a high cross-strait pressure difference occurs when of the Indonesian through¯ow in the major straits restricting ¯ow pressure on both sides is high (see Fig. 4). into the Indian Ocean. J. Geophys. Res., 106, 19 527±19 546. Lukas, R., T. Yamagata, and J. P. McCreary, 1996: Paci®c Ocean low The detailed investigation of observed pressure var- latitude western boundary currents and the Indonesian through- iations, in conjunction with numerical models may shed ¯ow. J. Geophys. Res., 101, 12 209±12 216. some light for any planned observational work in the Madden, R. A., and P. R. Julian, 1972: Description of global-scale region. It seems clear from the models (and past in situ circulation cells in the tropics with at 40±50 day period. J. Atmos. work) that the vertical structure of the ¯ow in the straits Sci., 29, 1109±1123. Meyers, G., R. J. Bailey, and A. P. Worby, 1995: Volume transport is complex, and it would be of great bene®t to have in of Indonesian through¯ow. Deep-Sea Res., 42A, 1163±1174. situ measurements of ¯ow in at least four depth ranges, Molcard, R., M. Fieux, J. C. Swallow, A. G. Ilahude, and J. Ban- including near surface, 100 m, 200 m, and below 400 jarnahor, 1994: Low frequency variability of the currents in In-
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC 1854 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32
donesian channels (Savu-Roti and Roti-Ashmore Reef). Deep- from a global eddy-resolving model. J. Geophys. Res., 97, 5493± Sea Res., 41, 1643±1661. 5550. ÐÐ, ÐÐ, and A. G. Ilahude, 1996: The Indo-Paci®c through¯ow Smith, W. H. F., and D. T. Sandwell, 1997: Global sea¯oor topography in the Timor Passage. J. Geophys. Res., 101, 12 411±12 420. from satellite altimetry and ship depth soundings. Science, 277, ÐÐ, ÐÐ, and F. Syamsudin, 2001: The through¯ow within Ombai 1956±1962. Strait. Deep-Sea Res., 48, 1237±1253. Sprintall, J., J. Chong, F. Syamsudin, W. Morawitz, S. Hautala, N. Murray, S. P., and D. Arief, 1988: Through¯ow into the Indian Ocean Bray, and S. Wijffels, 1999: Dynamics of the South Java Current through the Lombok Strait, January 1985-January 1986. Nature, in the Indo-Australian Basin. Geophys. Res. Lett., 26, 2493± 333, 444±447. 2496. Potemra, J. T., 1999: Seasonal variations of the Paci®c to Indian ÐÐ, A. L. Gordon, R. Murtugudde, and R. D. Susanto, 2000: A Ocean through¯ow. J. Phys. Oceanogr., 29, 2930±2944. semi-annual Indian Ocean forced Kelvin wave observed in the ÐÐ, 2001: The potential role of equatorial Paci®c winds on southern Indonesian seas in May 1997. J. Geophys. Res., 105, 17 217± tropical Indian Ocean Rossby waves. J. Geophys. Res., 106, 17 230. 2407±2422. Webster, P. J., A. M. Moore, J. Loschnigg, and R. Leben, 1999: Qiu, B., M. Mao, and Y. Kashino, 1999: Intraseasonal variability in Coupled ocean-atmosphere dynamics in the Indian Ocean during the Indo-Paci®c through¯ow and the regions surrounding the 1997-98. Nature, 401, 356±360. Indonesian seas. J. Phys. Oceanogr., 29, 1599±1618. Wyrtki, K., 1987: Indonesian through¯ow and the associated pressure Quadfasel, D., and G. R. Cresswell, 1992: A note on the seasonal gradient. J. Geophys. Res., 92, 12 941±12 946. variability of the South Java Current. J. Geophys. Res., 97, 3685± Yu, L., and M. M. Rienecker, 1999: Mechanisms for the Indian Ocean 3688. warming during the 1997-98 El NinÄo. Geophys. Res. Lett., 26, Semtner, A. J., and R. M. Chervin, 1992: Ocean general circulation 735±738.
Unauthenticated | Downloaded 09/28/21 06:11 PM UTC