Analysis of wind data over the past 40 years and results from a high-resolution general circulation model has revealed the existence of a previously undescribed circulation that connects the tropical Pacific with Indian Ocean. As a direct response to the Pacific wind, water of the Pacific origin enters the South China Sea through Luzon Strait, and from there part of the water continues southward into the Java Sea and returns to the Pacific through Makassar Strait. This circulation contains a strong signal of El Niño and Southern Oscillation and appears to have a notable impact on the Indonesian Throughflow heat transport.
 Tropical Pacific water weaves through the complicated passages of the Indonesian archipelago to the Indian Ocean. This throughflow, referred to as Indonesian Throughflow (ITF), is the only place in the global ocean where warm tropical water flows from one ocean into another, forming a choke-point in the global distribution of heat in the climate system [Godfrey, 1996]. As an element of the global, meridional overturning circulation [Gordon, 1986], the ITF is influential in regulating the sea surface temperature (SST) pattern of the Indian Ocean. Without the constant supply of warm ITF water from the Pacific, the west coast of Australia would be as dry and cold as the west coast of South America [Wajsowicz and Schneider, 2001].
 The ITF has been shown to be part of the circulation around Australia-New Guinea, forced by the South Pacific wind [Godfrey, 1996]. Applying the “island rule” [Godfrey, 1989] to Australia-New Guinea, based on the recent ECMWF (European Center for Medium Range Weather Forecasting) 40 year Re-Analysis (ERA-40) wind, yields a mean transport estimate of 14.6 Sv (1 Sv = 106 m3 s−1). Despite some quantitative differences for different wind products, the mean ITF estimated by the “island rule” is well consistent with earlier observations [Godfrey, 1996; Meyers et al., 1995]. With the baroclinic adjustment over the sills in the Indonesian archipelago, the “island rule” can also be used to describe variations of the ITF on interannual time scale or longer [Wajsowicz, 1994]. The ITF estimated by the “island rule”, based on the ERA-40 wind, has a good correspondence with El Niño and Southern Oscillation (ENSO). In almost all cases during the period from 1958 to 2002, the ITF estimated by the “island rule” tends to be weaker during El Niño years and stronger during La Niña years (Figure 1). A composite El Niño event has a mean ITF of 12.8 Sv in November/December, about 3.7 Sv below its long-term average at this season. Although no such long-term observations are available, the expendable bathythermograph (XBT) measurements clearly show a minimum ITF in the 1986/87 and 1991/92 El Niño events and a maximum ITF in the 1988/89 La Niña event [Meyers, 1996].
 Among a number of passages in the Indonesian archipelago, the Makassar Strait is the primary pathway of the ITF. Recent mooring observations showed that the meridional velocity profile in Makassar Strait has a strong vertical shear, with the surface flow in the opposite direction to the subsurface flow during the boreal winter [Gordon et al., 1999, 2003; Susanto and Gordon, 2005]. Gordon et al.  attributed the seasonal variation of the Makassar Strait surface flow to the regional winds of the Java Sea. They argued that during the boreal winter, the southeastward monsoon wind drives buoyant, low-salinity Java Sea surface water into the Southern Makassar Strait, creating a northward pressure gradient in the surface layer of the Strait that inhibits the warm surface water from the Pacific from flowing southward into the Indian Ocean. The monsoon wind reversal eliminates the obstructing pressure gradient, and as a consequence the surface flow in Makassar Strait turns southward in the boreal summer.
 With what discussed above for the seasonal variation, one might expect an eastward zonal wind anomaly in the Java Sea during an El Niño event, when the ITF approaches its minimum strength in Makassar Strait (Figure 1a). This, however, doesn't seem to be the case in reality. Analysis of the ERA-40 wind indicates that the Java Sea zonal wind anomaly is highly correlated (r = 0.87) with the Southern Oscillation Index (SOI; Figure 1a), and during the mature phase (November/December) of El Niño, the regional zonal wind anomaly is actually westward (Figure 2a). The westward zonal wind anomaly cannot explain the weakening ITF (Figure 1a), in particular in the surface layer of Makassar Strait, as will be shown in the following sections.
2. A Hypothesis
 Here we hypothesize that the ITF in Makassar Strait is a consequence of interplay between two circulations. One is around Australia-New Guinea in the thermocline, and the other around Philippine-Borneo near the surface, both forced by the large-scale wind in the Pacific (Figure 3). The circulation around Philippine-Borneo provides an important pathway, through which the impact of Pacific variability can be conveyed into the Indian Ocean.
3. Results From a High-Resolution GCM
 Results from a high-resolution general circulation model lend support for our hypothesis. The model used for this study is based on Modular Ocean Model version 2, and has a horizontal resolution of degree and 55 levels in the vertical [Ishida et al., 1998]. The model is zonally global, with its southern and northern boundaries closed at 75°S and 75°N, respectively, and has a realistic coastline and bottom topography derived from the National Geophysical Data Center's Earth Topography Five Minute (ETOPO5) dataset.
 The model was forced with the Hellerman and Rosenstein  annual mean wind stress for the first 2 years and with their monthly wind stress afterward for 18 years. A highly scale-selective biharmonic operator was used for horizontal turbulent mixing, with a coefficient of −1 × 1019 cm4 sec−1 for both momentum and tracers, and the Pacanowski and Philander  formulation was used for vertical mixing. SST was relaxed to Reynolds sea surface temperature and sea surface salinity to Levitus  monthly climatology. Then, the model was further integrated for 17 years with 3-day averaged ECMWF wind stress from January 1982 to December 1998.
 During the period of integration, the model ITF is well in phase with what estimated by the “island rule” (Figure 1b). Its correlation with SOI reaches 0.65, despite the concurrence of positive Indian Ocean Dipole (IOD) events in 1982/83 and 1997/98 [Saji et al., 1999]. During these positive IOD events, the negative sea level anomalies along the Indonesian coast may have enhanced the pressure head that drives the ITF [Wyrtki, 1987] and thus reduced the effect of El Niño [Meyers, 1996].
 The ITF is mostly confined in the depth-range of the thermocline. In Makassar Strait, the mean flow at 100 m is southward with a typical speed of 30 cm s−1 (Figure 4a). The flow is much reduced in the surface layer (Figure 4b), showing a northward “relative” surface flow (relative to 100 m) consistent with observations [Gordon et al., 2003, their Figure 2]. The surface flow in Makassar Strait can be traced to the Luzon Strait (Figure 4b), indicative of a Pacific origin. The intrusion of the Pacific water through Luzon Strait, often referred to as the Luzon Strait Transport (LST), has been extensively studied in the past decades [Wyrtki, 1961; Metzger and Hurlbert, 1996; Qu, 2000]. Here, we note that the LST is part of the circulation around Philippine-Borneo (Figure 3), and can be estimated explicitly from the integral of the long-path component of wind stress along the closed path extending from the southern tip of Borneo to the South American coast and from the North American coast to the northern tip of Luzon [Godfrey, 1989]. Based on the ERA-40 wind, the ‘island rule’ yields a mean LST estimate of 5.8 Sv. With friction associated with the narrow and shallow passages in the South China Sea, the actual value of LST is much less, mostly to a range of 2–4 Sv [Metzger and Hurlbert, 1996; Qu et al., 2000]. The annual mean LST from the model is 2.4 Sv. As circulation around islands tends to flow in the westernmost channel [Wajsowicz, 1993], the outflow through Karimata Strait is, on the average, about two times larger than that through Mindoro Strait [Qu et al., 2004], suggesting that Karimata Strait is the primary but not the only pathway for the Pacific influence.
 On the interannual time scales, the LST estimated by the “island rule” is out of phase with SOI (Figure 1a). Its correlation with model LST reaches 0.69 (Figure 1b), implying that the large-scale wind in the Pacific is the primary forcing for this variability. Variability in the Pacific may first affect the bifurcation of the North Equatorial Current (NEC) [Qu and Lukas, 2003], shifting it northward during El Niño years and southward during La Niña years [Kim et al., 2004]. The northward (southward) shift of the NEC bifurcation further results in a weaker (stronger) Kuroshio east of Philippines, providing a favorable (unfavorable) condition for Pacific waters to enter the South China Sea through Luzon Strait [Yaremchuk and Qu, 2004]. Thus, both the LST and KST tend to be stronger during El Niño years and weaker during La Niña years (Figure 1b) [Qu et al., 2004].
 As part of the circulation around Philippine-Borneo, the surface (0–100 m) ITF correlates (r = 0.76) well with LST in the model. During El Niño years and in particular when not concurred with positive IOD, an anomalous southward flow is seen to extend from Luzon Strait to the Java Sea (Figure 2b). The anomalous surface flow turns northward in Makassar Strait, inhibiting the surface Pacific water from flowing southward. The situation tends to be reversed during La Niña years, except in 1988/89 when the KST anomaly was weakly positive, showing additional evidence that the impact of Pacific variability is not always conveyed through Karimata Strait (Figure 1b).
 The variation in the surface ITF is of particular importance to the inter-ocean heat flux. In the model, the transport-weighted mean temperature of the ITF is 15.9°C, indicative of a mean heat flux of 9.5 × 1014 W from the Pacific to the Indian Ocean if a reference temperature of 0°C is selected, in a reasonable agreement with earlier observations [Ffield et al., 2000]. During the period from January 1982 to December 1998, the ITF heat transport in the model contains a strong ENSO signal, getting smaller during El Niño years and larger during La Niña years. As the Makassar Strait volume transport is mostly surface trapped [Gordon et al., 1999], the variability of the ITF heat flux defined by the root mean square is 4.0 × 1014 W in the upper 100 m, accounting for 89% of the total variability (Figure 5), most (∼99%) of which is due to the variability in volume transport. During the mature phase of El Niño, when the circulation around Philippines-Borneo reaches its maximum strength, the total ITF heat transport is reduced by as much as 1/3 of its long-term average.
 Many observational and model studies have suggested that the SST pattern of the Indian Ocean is strongly influenced by the ITF heat transport. Here we show that the circulation around Philippines-Borneo is an important pathway for the Pacific influence. It is this circulation that alters the meridional velocity profile in Makassar Strait, which otherwise would be constant from the sea surface to the thermocline, thus contributing significantly to the interannual variation of surface ITF heat transport. We note that the influence may also come from the Indian Ocean, in particular during the positive IOD events. The integration of the model used for this study is not long enough to provide a statistically significant description of the Indian Ocean influence. Further experiments with atmospheric and oceanic models will be needed to understand the full implications of the Pacific pathway through the South China Sea, as well as the role of Indian Ocean variability in modulating the ITF heat transport.
 This research was supported by the National Aeronautics and Space Administration through grant NAG5-12756. TQ was also supported by Japan Agency for Marine-Earth Science and Technology through its sponsorship of the International Pacific Research Center (IPRC), GM by the CSIRO Wealth from Oceans Program, and DW by the National Science Foundation of China through grants 40136010 and 40406006. The authors are grateful to E. J. Lindstrom and T. Yamagata for useful communications on the topic and to two anonymous reviewers for thoughtful comments on the manuscript. School of Ocean and Earth Science and Technology contribution number 6690, and IPRC contribution number IPRC-357.