Seasonal and interannual variations of the New Guinea Coastal Current (NGCC) and New Guinea Coastal Undercurrent (NGCUC) were investigated by examining the 5 years' data from acoustic Doppler current profiler moorings at two sites (2°S 142°E, 2.5°S 142°E) off the New Guinea coast. The NGCC flowed northwestward as is usual and intensified during the boreal summer, then weakened or even reversed direction to southeastward during the boreal winter. This seasonal change correlated to the monsoonal wind variation. However, during the 1997–1998 El Niño, the southeastward NGCC during the boreal winter was not observed, and northwestward flow was dominant throughout the year. On the other hand, the NGCUC flowed steadily northwestward all year-round and intensified during the boreal summer. During the growing phase of the El Niño, the NGCUC intensified, and its northwestward flow reached from the surface to a depth of 250 m. Comparison between the volume transport of these currents and the Sverdrup transport along 2°S in the ocean interior indicated a mean difference of 13 × 106 m3 s−1 northward. The relationship between variations of these two transports showed a negative correlation on seasonal timescales except during the El Niño. During the mature phase of the El Niño, northward Sverdrup transport was enhanced significantly, furthermore the transport of these currents was also northward. The result demonstrates a process by which anomalous water volumes can flow into the equatorial region due to an imbalance between the volume transport in the ocean interior and the western boundary.
 The western Pacific warm pool has been recognized as a key region of large-scale air-sea interaction phenomena such as El Niño/Southern Oscillation (ENSO). Besides the heat and freshwater exchanges at the sea surface, the heat and salt budgets within the upper ocean are important for a better understanding of the maintenance mechanism of the warm pool. Particularly, low-latitude western boundary currents (LLWBCs) have received much attention because they bring different water masses into the equatorial region from the subtropical gyres in both hemispheres. The LLWBCs in the Pacific consist of the Mindanao Current (MC) in the Northern Hemisphere and the New Guinea Coastal Current (NGCC), which flows in the surface layer, and the New Guinea Coastal Undercurrent (NGCUC), which flows in the subsurface thermocline layer, in the Southern Hemisphere. Lukas et al.  reviewed the observations and theories on the LLWBCs in the Pacific, including the Indonesian throughflow. The Indonesian throughflow supplies waters of the Pacific Ocean to the Indian Ocean [Godfrey and Golding, 1981; Gordon, 1986; Godfrey, 1989]. Thus the LLWBCs play an important role in the Pacific general circulation and in the heat and mass balances by connecting the subtropics to the tropics, and also the Pacific Ocean to the Indian Ocean.
 In the Northern Hemisphere, the tropical gyre consists of the North Equatorial Current (NEC) flowing westward, the MC flowing southward, and the North Equatorial Countercurrent (NECC) flowing eastward. The NEC is also the southern limb of the North Pacific subtropical gyre. Thus water subducted in the subtropical gyre at midlatitude is considered to be transported into the equatorial region by the MC. Lukas et al.  observed the vertical structure of the MC, of which speeds were over 100 cm s−1 near the surface and about 10 cm s−1 at a depth of about 700 m. In the thermocline layer, there is a low-salinity tongue of intermediate water formed in the midlatitude of the North Pacific and extending to the equator [Bingham and Lukas, 1994]. This feature suggests that the MC supplies low-salinity water to the equatorial region from the subtropics.
 On the other hand, the tropical gyre in the Southern Hemisphere extends across the equator, and therefore waters from the Southern Hemisphere easily cross the equator. These waters are a major source of the water masses of the equatorial region. Along the northern coast of New Guinea, equatorward volume transport is induced by the NGCC, even though its direction reverses seasonally [Wyrtki, 1961] and the NGCUC consistently flows equatorward. The first description of the NGCUC based on direct current measurement was reported by Lindstrom et al. . They found that this thermocline flow supplies the high-salinity waters that form in the midlatitude of the South Pacific to the equator and Northern Hemisphere, and that this flow becomes a major source of the Equatorial Undercurrent (EUC). Tsuchiya et al.  traced the EUC and NGCUC waters back to a high-salinity pool near 20°S along the northeastern coast of Australia. These high-salinity waters supplied by the NGCUC are transported eastward by the EUC and upwelled along the equator. Another high-salinity water carrier to the EUC, the New Ireland coastal undercurrent, was found by Butt and Lindstrom . The New Ireland coastal undercurrent flows equatorward along the east coast of New Ireland and is diverted to the north, and connects to the EUC along 149°E.
Butt and Lindstrom  indicated that there are three routes for LLWBCs in the region to reach the EUC: off the east coast of New Ireland, Vitiaz Strait, and St. Georges Channel. The latter two routes join southwest of Manus Island and form the NGCUC. Lindstrom et al.  measured flows and estimated volume transports for the latter two routes. They showed that the volume transport through Vitiaz Strait of 8–14 Sv is larger than that through St. Georges Channel, and the total volume transport of the NGCUC becomes 9–18 Sv. Butt and Lindstrom  showed that the volume transport of the NGCUC was about three times the transport of the New Ireland coastal undercurrent.
 The importance of the LLWBCs in the South Pacific as carriers of high-salinity waters is well recognized, but their variability is not yet clear because of insufficient observation. Observations in the past were mainly conducted using ships (i.e., hydrographic and shipboard acoustic Doppler current profiler (ADCP) measurements), and therefore temporal resolution of the current variation was limited. In addition, the path of the LLWBCs in the South Pacific is located close to the equator, so geostrophic calculation of cross-equatorial flow using hydrographic and wind stress data tends to be noisy [Joyce, 1988]. Thus not only hydrographic observations but also long-term direct current measurements by moorings are vital to study on the variability of the LLWBCs in the South Pacific.
 In order to investigate the variability of the upper ocean current systems in the Pacific warm pool, the Japan Marine Science and Technology Center (JAMSTEC) deployed an array of subsurface moorings equipped with an upward-looking ADCP. For the investigation of the southern part of the LLWBCs, two sites located at about 2°S 142°E and 2.5°S 142°E were selected. From the point of view of the volume transport, these sites were regarded as main route of the LLWBCs in the South Pacific. Kuroda  reported a preliminary result based on the first year of data from July 1995 to July 1996 and described variations of the NGCC and NGCUC on intraseasonal to seasonal timescales. The data up through September 2000, including the period of the 1997–1998 El Niño, from the mooring at 2.5°S 142°E are available now. These data enabled us to investigate the interannual variation of the current field associated with ENSO. In this paper we demonstrate the seasonal and interannual variation of the upper ocean currents off the northern coast of New Guinea corresponding to the southern part of the LLWBCs, and discuss the relationship between the variation of volume transport of these boundary currents estimated from the moored ADCP data and that of the Sverdrup transport in the ocean interior reflecting the basin-scale sea surface wind variation.
 The ADCP mooring array consists of six moorings, EQ 138°E, EQ 142°E, EQ 147°E, EQ 156°E, 2°S 142°E, and 2.5°S 142°E, in the Pacific warm pool. The location and data period of each mooring are shown in Figure 1. Two sites at 2°S 142°E and 2.5°S 142°E, located off the northern coast of New Guinea, were chosen for the investigation of the LLWBCs in the South Pacific. Each mooring was equipped with an upward-looking RD Instruments (RDI) 150-kHz ADCP, which measured hourly velocity in 35 bins at depths ranging from 30–50 to 200–330 m and a Seabird conductivity-temperature-depth (CTD) sensor (SBE16) attached 1.5 m below the ADCP. Using the depth time series measured with the CTD sensor, the bin depths were determined, and they were then interpolated vertically to extract data at each 10 m interval. The details of this procedure were described by Kutsuwada and Inaba . The hourly data for each of the subsampled layers were low-pass filtered by a 25-hour running mean, and then the data at 0000 UT were subsampled to construct a time series of daily data. The maximum record length is about 5 years at 2.5°S 142°E. Considering the topography of the northern coast of New Guinea, the axis of coordinates for the current was rotated clockwise 20° so that the alongshore component would be parallel to the 1000 m bathymetric contour.
Figure 2 shows the 2-month running mean Southern Oscillation Index (SOI), which is the monthly average sea level pressure difference between Tahiti and Darwin. Since the period indicating relatively large negative (positive) values of the SOI represents El Niño (La Niña) phase, we divided the time series used in this study into three phases of the ENSO cycle, namely the normal, El Niño, and La Niña phases.
 We conducted hydrographic cruises off the northern coast of New Guinea twice a year in summer from July to September and in winter from January to March on the research vessel (R/V) Kaiyo [Kuroda et al., 1996; Kashino et al., 2001]. Note that seasons in this paper are referred to Northern Hemisphere. We examine also the hydrographic and current data along the 142°E line basically from 2.7°S to the equator obtained on 11 cruises from 1994 to 2000 (Table 1). Temperature and salinity were measured from the sea surface to 1000 m with CTD (SBE 911plus) and expendable CTD (Tsurumi Seiki Co. Ltd.). Currents were measured by shipboard ADCP installed on the Kaiyo. A 75-kHz ADCP (RDI VM-NB 75) was used until the KY9909 cruise, and a 38-kHz ADCP (RDI OS-II 38) was used starting on the KY0006. Measured velocity profiles were corrected by Global Positioning System data. Horizontal velocity distributions along the cruise track were averaged over a horizontal scale of 0.5° by 0.5°. In the vertical direction, the data were averaged at the layers bounded at depths of 30, 50, 75, 125, 175, 225, 275, 325, 375, 425, 475 and 525 m.
Table 1. List of R/V Kaiyo Cruises From December 1994 to October 2000a
Date at 2.5°S, 142°E
Seasons in this table are defined: the period of January–March is winter, the period of July–September is summer, and the period of October–December is fall.
19 Dec 1994 to 11 Jan 1995
8 Jan 1995
30 June 1995 to 26 July 1995
10 July 1995
24 Jan 1996 to 26 Feb 1996
16 Feb 1996
7 July 1996 to 5 Aug 1996
13 July 1996
26 Jan 1997 to 1 March 1997
20 Feb 1997
3 Aug 1997 to 29 Aug 1997
21 Aug 1997
3 Jan 1998 to 1 Feb 1998
26 Jan 1998
15 Aug 1998 to 11 Sep 1998
5 Sep 1998
26 Jan 1999 to 1 March 1999
5 Feb 1999
20 Oct 1999 to 23 Nov 1999
15 Nov 1999
20 Aug 2000 to 1 Oct 2000
16 Sep 2000
 To investigate the relationship between the current variation off the northern coast of New Guinea and the basin-scale surface wind variation, the following data sets were used: the 10-day averaged satellite scatterometer wind/wind stress data sets, which are the product of Japanese Ocean Flux data sets with Use of Remote sensing Observations (J-OFURO) [Kubota et al., 2002] and the weekly optimum interpolated sea surface temperature (SST) data set [Reynolds and Smith, 1994]. The wind/wind stress and SST data sets have a spatial resolution of 1° by 1°.
 The wind/wind stress data sets were also used to calculate the curl and Sverdrup transports. The Sverdrup balance is given by
where β is the meridional derivative of the Coriolis parameter, V is the Sverdrup transport that indicates meridional volume transport in the ocean interior [Sverdrup, 1947], τ is the wind stress, and ρ is the mean density of the ocean (1.0 × 103 kg m3).
3. Current Variation Measured With ADCP Moorings
3.1. Current Variation Off the Northern Coast of New Guinea
 Profiles of the means and standard deviations for the alongshore and onshore-offshore current components at 2°S 142°E and 2.5°S 142°E were calculated over the entire mooring period (Figure 3). The means of the alongshore component at both sites are northwestward and faster with increasing depth. The velocities below 100 m at 2.5°S 142°E are faster than that at 2°S 142°E, and the maximum (greater than 50 cm s−1) velocity is shown around 200 m at 2.5°S 142°E. The standard deviations at both sites indicate a maximum value near the sea surface. In contrast to the alongshore component, the onshore-offshore component at both sites is small (less than 7 cm s−1) in all the observed layers. Therefore we turn our attention to the alongshore component.
 The temporal variation of the alongshore current component at both sites showed that the NGCC in the surface layer reversed its direction seasonally, and the NGCUC in the subsurface layer consistently flowed northwestward (Figure 4).
 Concerning the NGCC, the direction was northwestward during summer, but southeastward during winter. This variation is consistent with that of the alongshore wind component at 2°S 142°E. To estimate a critical depth, which is affected by monsoonal wind for the surface flowing NGCC, correlation coefficients between local alongshore wind and current at the each depth were calculated (Figure 5). This profile indicates a maximum at the shallowest depth. These coefficients decrease with increasing depth and become nonsignificant around 100 m. This result implies that local wind forcing is strongly related to variation of the NGCC.
 Northwesterly alongshore winds drive offshore surface Ekman flows, thereby inducing upwelling near the coast of New Guinea and subsurface onshore flows. Consistent with the upwelling, a southeastward coastal surface jet was generated associated with the shoaled pycnocline near the coast. In the case of the southeasterly alongshore winds, the opposite result occurred. The strong surface southeastward current in winter is considered as this coastal surface jet forced by monsoonal wind, of which generation mechanism was studied on the current off the west coast of Somalia by McCreary and Kundu . They formulated the structure of the surface coastal jet as the surface velocity increased with sine function and decreased exponentially, toward offshore. The observational result that the amplitude of the southeastward surface current during winter at 2°S 142°E was larger than that at 2.5°S 142°E (Figure 4) does not contradict their coastal jet structure characterized by an offshore velocity core.
 Concerning the NGCUC, a northwestward current core was generally found from 150 to 250 m and obviously intensified in summer (Figure 4).
 The most significant features in the interannual timescale are the disappearance of the southeastward NGCC and the predominance of the NGCUC with increased amplitude and shoaled velocity core during the summer and winter seasons (July 1997–February 1998) of the 1997–1998 El Niño.
3.2. Relationship Between the NGCUC and the EUC
 The hypothesis of Tsuchiya , that the source waters of the EUC come from the Coral Sea and are carried by the currents along the northern coast of New Guinea, was tested by Lindstrom et al. . These researchers revealed the existence of the NGCUC, which carries the high-salinity waters, and suggested a close relation between the NGCUC and EUC, using hydrographic and current meter measurements. Our mooring array contains four sites on the equator, and thus it enables us to compare variations of the NGCUC with those of the EUC. Figure 6 shows time-depth diagrams of the zonal current at four sites on the equator (138°E, 142°E, 147°E, and 156°E). In the upper layer, from the surface to about 150 m, the westward South Equatorial Current (SEC) was predominant and the temporal occurrence of a strong eastward current found at all the sites, indicating equatorial jets induced by westerly wind bursts. The generation and propagation processes of these jets were studied by Kutsuwada and McPhaden  based on the same moored ADCP data focusing on the onset for the 1997–1998 El Niño. They reported that these jets were forced by anomalous intraseasonal wind stress in the western Pacific and then propagated into the eastern Pacific as first mode baroclinic equatorial Kelvin waves.
 The EUC was found flowing consistently eastward below the SEC at the three sites other than that at 138°E. The site at 138°E is located close enough to the northern coast of New Guinea to be downstream of the NGCUC. Because of this geographical location, we suggest that the current variation showing a rather complex vertical structure in the subsurface at 138°E may be much affected by the NGCUC itself showing westward component, or the retroflexed flow, which the NGCUC turns to the equator after crossing the equator, showing eastward component. Thus our interests are concentrated on the other three sites on the equator.
 The seasonal variation of the EUC that intensified in summer and weakened in winter was consistent among the three sites. This seasonal variation of the EUC was also consistent with that of the NGCUC; both the EUC and NGCUC intensified in summer and weakened in winter (Figure 4). Figure 7 shows low-pass filtered time series of the zonal current at the depth of 220 m corresponding to the EUC core depth at the two sites on the equator (EQ 142°E and EQ 147°E), and the northwestward current at the depth of 220 m corresponding to the NGCUC core depth at the two sites off the northern coast of New Guinea (2°S 142°E and 2.5°S 142°E). In these time series, high correlation between the eastward flowing EUC and the northwestward flowing NGCUC was apparent, especially in the seasonal timescale.
Figure 6 also indicates the interannual variation of the EUC showing shoaling and remarkable enhancement in the summer of 1998, corresponding to the ending phase of the 1997–1998 El Niño. This interannual feature of the EUC is consistent with a result from the JAMSTEC high-resolutions global ocean circulation model [Ishida et al., 1999], which indicated that the NGCUC turned to the east on the equator without overshooting the equator and directly flowed into the EUC during the ending phase of 1997–1998 El Niño. The tendency of this feature was also recognized during 1982–1983 El Niño in their model.
4. Shipboard ADCP and CTD Measurements
4.1. Current, Temperature, and Salinity Structures
 As described in section 2, 11 cruises for shipboard ADCP and CTD measurements along 142°E have been conducted basically in winter and summer from 1994 to 2000. Using current and CTD data from eight of the cruises, except a cruise in fall (November 1999) and two cruises at the period of the El Niño phase (the summer in August 1997 and the winter in January 1998), we derived the mean structure of current, temperature, and salinity field in winter and summer.
 The mean zonal current structure along 142°E in winter (Figure 8) shows that the zonal current at the southern part of this section is basically westward, except uppermost layer. A westward current core with speed exceeding 40 cm s−1 was found from 200 to 300 m at 2.7°S. The westward current at the core depth was extended northward to 1.5°S. The position almost coincided with that of high-salinity region, and thus the NGCUC conveys the South Pacific high-salinity waters to the equator. The boundary between the surface eastward current and the subsurface westward current almost coincides with the 29°C isotherm and the 34.5 psu isohaline. This suggests that the water with high temperature and low salinity in the westernmost Pacific tends to be advected eastward during winter.
 During summer (Figure 9) the westward current off New Guinea was intensified, and the surface eastward current disappeared. A westward current core exceeding 60 cm s−1 was observed from 200 to 300 m at 2.5°S. The northern end of the westward current at the core depth was extended northward to 1°S. The position coincided with that of the salinity maximum. In the surface layer, low-salinity water under 34.5 was confined to uppermost layer in the south of 2.5°S and close to the coast.
 During the growing phase of the El Niño (in August 1997), structures of zonal current, temperature, and salinity were dramatically changed (Figure 10). The most significant feature in the current section was an upward and northward shift of the westward current core exceeding 60 cm s−1 observed from 80 to 180 m at 2°S with a maximum speed of 100 cm s−1. Associated with this change of current structure, the subsurface salinity maximum was also shifted upward and northward.
 During the mature phase of the El Niño (in January 1998), the region occupied by the westward current core exceeding 60 cm s−1 was relatively small (Figure 11) compared to the growing phase. The position of the current core was still raised but the northward extension was reduced. It is worth noting that the high-salinity core (>35.5 psu) at 150 m extended northward to 0.6°S but the westward NGCUC extended only to 1.5°S. This is evidence that the NGCUC turned directly into the EUC without mixing with the North Pacific low-salinity water as we have mentioned in section 3.2.
 The two moorings sites located at 2°S and 2.5°S are appropriate to monitor variations of the NGCC and NGCUC because they captured well the steadily flowing NGCUC core and surface reversed NGCC both confined to the New Guinea coast. We should pay attention, however, to analyze the current meter data during the El Niño period because the NGCUC core had moved northward, although the moorings still captured the enhancement of the NGCUC.
4.2. Volume Transport From Shipboard ADCP
 As an index of the volume transport of the northwestward flowing LLWBCs off the New Guinea coast, volume transports from 40 to 500 m across 142°E were calculated by integrating zonal velocity component in the 11 cruises' shipboard ADCP sections. In order to take account of transport change due to the local wind forcing, namely seasonal reversals of the NGCC, the integration was carried out for both zonal directions above 100 m, whereas for only westward current below 100 m to exclude the transport of the EUC. The difficulty in estimating the volume transport is in the definition of the current width. To distinguish the transport of the NGCC and NGCUC from that of the SEC, the integration range of the transport was defined subjectively as from 2.7°S to 1°S by the authors referring the averaged current sections (Figures 8 and 9).
 Note that if there was no observation the south of 2.5°S, the volume transport from 2.5°S to 2.7°S was extrapolated by a linear regression of the volume transports between 2.3°S–2.5°S and 2.5°S–2.7°S for the other sections. The correlation coefficient was 0.94. The value was calculated from the six sections, which covered to 2.7°S.
 Time series of the volume transport from 40 to 500 m across 142°E was shown in Figure 12 and Table 2. Gouriou and Toole  estimated the mean transport south of 1°S down to 400 m along 142°E at about 23.8 Sv by direct measurements. Their value is somewhat biased toward summer because most of their cruises were carried out in summer. The mean value of 26.9 Sv derived from our four non-El Niño cruises conducted in summer is consistent with their value. In general, the volume transports were smaller in winter than in summer. Volume transports in summer of 1995 and 1997 showed remarkably large values.
Table 2. Zonal Volume Transports Across the 142°E Section Obtained From Shipboard ADCP Data of 11 Cruises Between 1°S and 2.5°S in the Layer of 40–500 ma
Date (Boreal Season)
The negative sign indicates westward. Units are in Sv = 106 m3 s−1.
Jan 1995 (winter)
July 1995 (summer)
Feb 1996 (winter)
July 1996 (summer)
Feb 1997 (winter)
Aug 1997 (summer)
Jan 1998 (winter)
Sep 1998 (summer)
Feb 1999 (winter)
Nov 1999 (fall)
Sep 2000 (summer)
5. Relationship Between the Sverdrup Transport and LLWBCs in the South Pacific
5.1. Recharge Oscillator
 Investigating warm water accumulation in the equatorial region is key for a better understanding of the ENSO cycle [Wyrtki, 1985; Cane et al., 1986]. Jin [1997a, 1997b] proposed a “recharge oscillator” mechanism, which represents a crucial contribution of the warm water volume to the onset of El Niño and La Niña. In their theory, easterly winds weakening during El Niño cause a poleward discharge of upper layer warm water in the equatorial region, which results in thermocline shoaling. As a result, the SST in the equatorial region becomes low, which causes the termination of El Niño. The reversed process occurs during La Niña. The key process in that theory is the recharge or discharge of upper layer warm water in the equatorial region, which is produced by the Sverdrup transport due to an anomalous wind stress curl pattern. This recharge oscillator was confirmed by observations; however, the contribution of the western boundary current was not clear because the data were insufficient [Meinen and McPhaden, 2000, 2001]. Thus we examine first Sverdrup transport in the ocean interior, which reflects basin-scale wind variation, and second the transport of the NGCC and NGCUC as LLWBCs, and discuss their relationship.
5.2. Basin-Scale Wind Variation
 Before treating the Sverdrup transport, we investigate basin-scale sea surface wind stress variation, which induces variation of Sverdrup transport. Figure 13 shows the mean wind stress vector with mean SST and mean wind stress curl field during winter and summer calculated using 5 years' monthly data from 1995 to 2000 excluding the El Niño period from May 1997 to April 1998. This figure shows the difference in the distribution of the sea surface wind and SST in winter and summer clearly. Strong easterly winds are located along 8°N in winter and along 8°S in summer. These locations correspond to eastward extensions of higher SSTs. The seasonal reversal of monsoonal winds along the northern coast of New Guinea is also apparent. As a result of this sea surface wind pattern, a positive wind stress curl is dominant in winter, whereas a negative wind stress curl is dominant in summer.
 Since SST distribution changed dramatically during the El Niño, the distribution of wind stress was also altered (Figure 14). During the growing phase of the 1997–1998 El Niño, the high SST region over 29°C was found in the central equatorial Pacific. As a result, winds tend toward the high SST region, and the northward component in the South Pacific was particularly enhanced. Concerning wind stress curl, positive areas were found between 8°N and 4°N, whereas the maximum negative value was found around 6°S 170°E. The SST distribution during the mature phase of the 1997–1998 El Niño indicates a zonally uniform structure. According to this SST distribution, southward wind stress was dominant in the equatorial region. As a result of this wind stress pattern, areas of positive wind stress curl were dominant around the equator with a zonally broad structure. Due to the drastic change of SST distribution, the wind stress pattern during the El Niño greatly differed from that of non-El Niño years. Particularly, the predominance of the meridional wind stress component produced anomalous wind stress curl in the equatorial Pacific.
5.3. Sverdrup Transport in the Ocean Interior
 The mean and standard deviation of the Sverdrup transport calculated from wind stress curl from 1995 to 2000 are shown in Figure 15 with a latitude-time diagram of the Sverdrup transport along the western boundary, which is integrated from the eastern boundary. The contour lines of the mean Sverdrup transport show a northward transport maximum at the western boundary near 8°N and southward transport maxima at the western boundary near 20°N and at the center of the South Pacific near 3°S. The large standard deviation of the Sverdrup transport at the western boundary near 5°N corresponds to the seasonal migration of the ITCZ. The Sverdrup transport in the ocean interior integrated from the eastern boundary to the western boundary tends to be southward during summer and northward during winter. This suggests that the water volume in the ocean interior of the equatorial region is built up from the Northern Hemisphere during summer and from the Southern Hemisphere during winter. Concerning the interannual timescale, the Northern Hemisphere water penetrated the Southern Hemisphere largely during the growing phase of the El Niño, and the water around the equator then flushed out quickly northward during the mature phase of the El Niño.
5.4. Time Series of Estimated Volume Transport From ADCP Mooring Data
 The variation of the volume transport obtained from shipboard ADCP (Figure 12) seems to be consistent with the time series of alongshore currents observed with the ADCP moorings (Figure 4). Therefore to derive time series of volume transport from the single point ADCP mooring data, we calculated correlation coefficients by linear regression between transports from shipboard ADCP and vertically averaged velocities from the mooring at 2.5°S (Figure 16). The correlation coefficient was remarkably high (r = 0.79) between the vertically averaged velocities from 30–50 to 250–280 m. The correlation coefficient became higher (r = 0.93) if the point for winter 1998, corresponding to the mature phase of the El Niño, is eliminated.
 It is apparent that the large westward vertically averaged velocities derived from the mooring data correspond well to larger westward volume transports. There is an isolated datum obtained in January 1998, however, in this relationship, which is for the mature phase of the El Niño. If we estimate the volume transport from vertically averaged velocity based upon this relationship for this period of the mature phase of the El Niño, the estimated westward volume transport will be higher. The zonal current section in this period (Figure 11) shows that the westward velocity core speed increased, but the area of westward current remarkably reduced and it resulted in a small transport as same as the transports in another winters (Figure 12).
 We compare the time series of this estimated volume transport with that of the Sverdrup transport in the ocean interior for a better understanding of the water volume budget in the equatorial region. The time series of the estimated volume transport is shown along with Sverdrup transport integrated from the eastern boundary to 143°E along 2°S in Figure 17a (for convenience, transport direction of northwestward and southeastward along the coast were expressed as northward and southward, respectively). Note that this estimation contains errors caused by insufficient vertical integration above 30–50 m and below 200–330 m and by discrepancy of regression particularly during the mature phase of the El Niño.
 The mean value of estimated volume transport for the whole period was 21.4 Sv northward, whereas that of Sverdrup transport was southward at 8.4 Sv. As a result, there was a mean difference of 13.0 Sv northward.
 The time series of anomalies of both transports are shown in Figure 17b. The amplitude of the seasonal variation of the estimated volume transport, except for the mature phase of the El Niño, was somewhat smaller than that of Sverdrup transport. Both transports show a negative correlation except for winter and during the mature phase of the El Niño. Because of the dominance of a semiannual signal from 1995 to 1997, an out of phase relationship was found in winter. This may produce water volume inflow into the equatorial region on seasonal timescale. The semiannual signal was not present in the Sverdrup transport variation (Figure 18a). The semiannual signal in the alongshore current has a maximum near 90 m at 2.5°S 142°E (Figure 18b), whereas it was not present at 2°S 142°E (Figures not shown). This suggests that the semiannual signal in the estimated volume transport might be influenced by local phenomena at 2.5°S 142°E.
 During the growing phase of the El Niño, the southward Sverdrup transport in the ocean interior along 2°S was most dominant, whereas the northward estimated volume transport also showed their greatest value at that time. It suggests that the southward transport at the southern end of the equatorial band tended to be compensated by the northward transport of the LLWBCs.
 On the other hand, during the mature phase, the reversed northward Sverdrup transport of the ocean interior was remarkably large, whereas the northward estimated volume transport was decreased but never reversed, even if we choose the transport value (i.e., 12 Sv, northward) from shipboard ADCP section data for that period. As a result, the water volume inflow into the equatorial region must have increased rapidly during this phase. This suggestion is consistent with the model result examined by Zebiak  that waters were concentrated from the southern tropics into the equatorial Pacific not only in the ocean interior but also at the western boundary during the mature phase of the El Niño.
6. Summary and Conclusion
 The seasonal and interannual variability of the LLWBCs in the South Pacific was investigated using 5 years' ADCP mooring data obtained off the northern coast of New Guinea. The currents flowed basically equatorward along the coast. The current in the upper 100 m, however, was affected by monsoonal wind forcing and changed direction seasonally: it flowed southeastward during winter, whereas northwestward during summer. On the other hand, the current in the thermocline also showed seasonal variation, namely the northward current was weakened during winter and intensified during summer. Concerning the interannual variation, the northward current in the subsurface layer was intensified and shoaled during the growing phase of the 1997–1998 El Niño. Furthermore, the seasonal reversal of the surface current did not appear during the El Niño because of weak southeasterly monsoonal winds.
 There is a large excess northward transport of 13 Sv between the mean estimated volume transport off the northern coast of New Guinea and the mean Sverdrup transport in the ocean interior, while the amplitudes of seasonal variation of the two transports did not indicate such large differences in general. On the seasonal timescale, the estimated volume transport indicated northward anomaly in summer and southward anomaly in winter, although, northward anomaly appeared in winter from 1995 to 1997 because of the dominance of a semiannual signal. On the interannual timescale the northward anomaly in the estimated volume transport was maximum during the growing phase of the El Niño, whereas the southward anomaly in the mature phase of the El Niño was remarkably suppressed. Because both the volume transport off the northern coast of New Guinea and the Sverdrup transport indicated northward anomaly during the mature phase of the El Niño, water volume transport into the equatorial region from Southern Hemisphere was dominant.
 In the recharge oscillator theory [Jin, 1997a, 1997b], the variation of the SST field in the tropical Pacific, namely the migration of the Pacific warm pool, controls the wind field pattern, and the wind field pattern then affects the variation of the Sverdrup transport in the ocean interior. In this paper we showed that the variation of Sverdrup transport may affect the volume transport by the LLWBCs that is not taken into consideration in the recharge oscillator theory. In future, the relationship between variation of the volume transports by the LLWBCs in both the hemispheres and that of the warm water volume change in the equatorial region should be investigated. For that purpose, observation of the LLWBCs in the Northern Hemisphere, namely the MC, will be vital because the amplitude of the Sverdrup transport variation in the ocean interior along 5°N is much larger than that along 2°S (Figure 15c).
 Meanwhile it was suggested that there was another factor controlling the volume transport off the northern coast of New Guinea because of a large bias between the mean estimated volume transport and the mean Sverdrup transport. This means that the currents off the northern coast of New Guinea are not induced only by wind variation in the equatorial Pacific. Johns et al.  studied the variability of the North Brazil Current and showed a large bias between the transport induced by wind stress curl and that observed by moorings at the western boundary. They pointed out that a thermohaline component became one of the factors in that large bias. A thermohaline component may also be a key factor in the Pacific. Moreover, the low-latitude western boundary in the equatorial South Pacific corresponds to the path of the circulation around the Australian continent. Therefore investigations taking into account the Indonesian throughflow are needed in future study.
 The seasonal variation of subsurface salinity and its effect on dynamic height in the warm pool region was investigated by Ueki et al. . They also showed the depth of the subsurface salinity core, which has a large variability, was coincident with the depth in which the water masses from the Southern Hemisphere exist. It suggests that the variation of the volume transport by the LLWBCs might be the primary factor controlling salinity variation in the subsurface layer of the downstream region. Therefore understanding of the volume transport of the LLWBCs in the South Pacific is required from the point of view of salinity budget in the western Pacific warm pool. In the warm pool region, the Triangle Trans Ocean buoy Network (TRITON) buoys, which observe subsurface salinity, have been deployed since 1998, and will thus enable us in the future to investigate these water mass changes in conjunction with LLWBC variations.
 The deployment and recovery of ADCP mooring buoys was carried out by R/V Kaiyo and R/V Mirai. We would like to thank the captains and crews of these cruise vessels. The mooring operations were achieved by technicians of Marine Works Japan Ltd. and Nippon Marine Enterprises Ltd. We also express thanks to the members of the TOCS project: Kei Muneyama, Kentaro Ando, and Kunio Yoneyama who conducted these cruises. We are grateful to Kentaro Ando and Akio Ishida for their valuable comments. Many constructive comments from the editor and reviewers were helpful for the improvement of this paper. The TOCS project is supported by the Ministry of Education, Culture, Sports, Science and Technology, Japan.