A different zonal current system along the meridional transect of East Java (EJ; AEJ-BEJ-CEJ) can clearly be seen in Fig. 6a–f. On average, for the period 1950 through 2013, zonal climatological current at AEJ (nearshore area) generally flows eastward from the sea surface to 100 m depth (Fig. 6a and d) and reaches a maximum value of about 0.16 m s-1. It is suggested that the average zonal current at this point is mainly attributed to SJC, and it shows seasonal variations. During the SE monsoon (JJA), the strength of climatological eastward SJC at this point in upper 10 m depth reduces (Fig. 6d). Meanwhile, during the NW monsoon (DJF), the current in the upper 10 m (Fig. 6d) flows more eastward in response to the prevailing northwesterly winds (Fig. 7). In general, the mean eastward current at AEJ, during DJF was attributed to local winds. Interestingly, during this monsoon period (DJF), the eastward current at AEJ, particularly that at depths beneath 100 m, strengthens and occurs up to ∼ 800 m. Other physical processes may account for the enhanced eastward current at this point. The SJC and SJUC, which are seasonally varying currents and predominantly eastward, are defined as the surface current in the upper 150 m and the subsurface current beneath 150 m down to 1000 m, respectively (Iskandar et al., 2006). The eastward-flowing SJC and SJUC are intensified, coinciding with the arrival of a seasonal downwelling Kelvin wave along the southern coast of Java (e.g., Sprintall et al., 1999, 2000; Iskandar et al., 2006). Downwelling Kelvin waves originating in the equatorial Indian Ocean during the transitional monsoons propagate along the coasts of western Sumatra and southern Java with phase speeds ranging from 1.5 to 2.9 m s-1 (e.g., Sprintall et al., 2000; Syamsudin et al., 2004; Iskandar et al., 2005). These phase speeds indicate that the downwelling Kelvin waves will arrive at AEJ in 21–41 d. In this case, downwelling Kelvin waves generated during the monsoon transition period in November may arrive at AEJ in December–January. Therefore, in addition to the local eastward winds, the downwelling Kelvin waves may also contribute to strengthen the eastward currents at AEJ during the NW monsoon, including those at depths beneath 100 m.

Meanwhile, the average current at BEJ (the transitional zone) is westward. It is suggested that the mean westward current at the point BEJ is more dominated by the ITF (shown by black lines in Fig. 6b and e). Based on observations of the exit passages (Lombok Strait, Timor Passage, and total ITF along exit passages), ITF in JJA is stronger than that in DJF (e.g., Sprintall et al., 2009). In this study, however, it is found that westward current at the point BEJ at 100 m depth is stronger during DJF than JJA. This phase changing (delay) of the ITF seasonality from JJA to DJF at this point is also found in the Ombai Strait as documented by Sprintall et al. (2009, their Table 3; 2010, their Fig. 3). Moreover, Sprintall et al. (2010) found cores of subsurface maximum ITF during DJF extending from 100–250 m (100–800 m) depth at the northern (southern) part of the strait. In the present study, this seasonal feature of the subsurface maximum ITF is also found at BEJ, in which the corresponding westward current at this point reaches its maximum values at ∼ 100 m depth and the maximum westward current is stronger during DJF than JJA (Fig. 6b and e). Hence, we suggest that the primary driver for zonal westward current at BEJ is the ITF coming from the southern Ombai Strait. To confirm the above relation, we have calculated the correlation between zonal westward current at a depth of ∼ 100 m at point BEJ and that representing subsurface (∼ 200 m) maximum ITF in the southern Ombai Strait (Sprintall et al., 2010). The correlation coefficient between the zonal westward current at ∼ 100 m at the BEJ and that of the southern Ombai Strait is 0.58, with a 95 % significance level of approximately ± 0.33. This study shows that the zonal westward current at 100 m depth at BEJ has a strong correlation with the subsurface (∼ 200 m) maximum ITF in the southern Ombai Strait, confirming that the ITF flowing from the Ombai Strait is the primary driver for zonal westward current at BEJ.

In the offshore region of the study area, zonal current at CEJ (Fig. 6c and f) flows westward throughout the year and has average velocity around 0.20 m s-1 in the upper 100 m. Under such characteristics, we propose that the westward current at this point is the SEC in the southeast Indian Ocean, which joins the ITF flowing out from the Lombok and Ombai straits and Timor Passage. The HYCOM westward current at this point is stronger during JJA than DJF, which is associated with seasonal characteristics of the ITF in Lombok Strait and Timor Passage and of the total ITF through the Lombok and Ombai straits and Timor Passage (Potemra, 1999; Sprintall et al., 2009). The westward current at CEJ (Fig. 6c and f) reaches its maximum value of about 0.31 m s-1.

Figure 6g–l show the vertical structure of the zonal current along the meridional transect of West Java (WJ; AWJ-BWJ-CWJ). Similar to AEJ, mean zonal current at AWJ (nearshore region) is attributed to the SJC, which generally flows eastward in the upper 100 m depth (Fig. 6g and j) and reaches a maximum value of about 0.12 m s-1. Our simulation shows that during the monsoon transitions (MAM and SON), SJC is eastward and intensified by the propagation of coastal Kelvin waves associated with the Wyrtki Jet in the equatorial Indian Ocean, which is forced by the local equatorial zonal winds during both monsoons. These waves propagate along the Sumatran and Javan coasts (i.e., Sprintall et al., 2000; Druskha et al., 2010; Iskandar et al., 2009) and some portions propagate northward into the Lombok and Makassar Straits (Susanto et al., 2000, 2012; Pujiana et al., 2013), whereas the remaining parts continue eastward (Syamsuddin et al., 2004). Furthermore, the present study shows that the eastward current during SON is stronger than that during MAM, which is consistent with mooring observation in the Makassar Strait (Susanto et al., 2012; their Fig. 3). The stronger eastward current during SON was supposed to be attributed to the faster and more intense climatological Wyrtki Jet during SON rather than during MAM (Knox, 1976; McPhaden, 1982; Han et al., 1999; Qiu et al., 2009; McPhaden et al., 2015; Figs. 1d and 2e of Duan et al., 2016) and also associated with stronger wind forcing over the eastern equatorial Indian Ocean during the SON period than the MAM period (figure not shown), which in turn causes the jet.

Moreover, it can be seen that during the NW monsoon the eastward current at AWJ (Fig. 6g and j) is weaker than that at AEJ (Fig. 6a and d). The weaker current at AWJ may exist as a consequence of the weaker mean NW monsoon at this point compared with that at AEJ (Fig. 7). Interestingly, at a depth of 100 m, there is a maximum westward current at AWJ during DJF with velocity of about 0.1 m s-1 (Fig. 6g and j). Here, we suggest that ITF is the cause of the westward current at 100 m at AWJ during the DJF. In regard to the ITF, Fig. 3 of Sprintall et al. (2010) shows cores of subsurface maximum ITF extending from 100 to 250 m depth in the northern part of the Ombai Strait and from 100 to 800 m depth at the southern part of the strait during DJF. Meanwhile, the influence of ITF on the zonal current at AEJ at 100 m is weaker as a consequence of the stronger NW monsoon at AEJ compared with that at AWJ (Fig. 7), and thus the current instead flows eastward at AEJ during DJF (Fig. 6a and d).

To further investigate which one is more influential out of the ITF and the NW monsoon in terms of forcing the zonal current at the AWJ and AEJ at 100 m depth, we have carried out correlations between the zonal current at both points (each at a depth of ∼ 100 m) and both the NW zonal wind and the zonal current representing subsurface (∼ 200 m) maximum ITF in the southern Ombai Strait (Table 1). Here, the ITF in the southern part of the Ombai Strait was chosen for carrying out the correlations because the ITF mainly flows through the southern part of the passage (Sprintall et al., 2010). It was observed that the subsurface maximum ITF during DJF exists at a depth of about 200 m in both the northern and southern parts of the Ombai Strait and that it is stronger during DJF than JJA in both parts of the strait (Fig. 3 of Sprintall et al., 2010). In this study, the DJF zonal currents in the period of 2004 through 2006 in the southern Ombai Strait derived from the INSTANT program (http://www.marine.csiro.au/~cow074/instantdata.htm, last access: 17 May 2020) were used for the correlation analysis.

It is found that during DJF the zonal current at AWJ at 100 m shows high correlation with the subsurface (∼ 200 m) maximum ITF in the southern Ombai Strait, whereas its correlation with the NW zonal wind is weak (Table 1). Moreover, although during DJF the correlations between the zonal current at AEJ at 100 m and both the NW zonal wind and the subsurface (∼ 200 m) maximum ITF in the southern Ombai Strait are below the significance level, the NW zonal wind is more influential regarding force variation of zonal current at AEJ at 100 m than the ITF. Hence, during DJF we suggest that the westward current simulated at AWJ at 100 m is ITF related, whereas that at AEJ is relatively NW zonal wind related. As already discussed, in addition to the local eastward winds during DJF, it is suggested that the arrival of downwelling Kelvin waves in December–January at AEJ may contribute to a net eastward current across the water column, which in turn reduces the influence of ITF at this point.

In the transition region, the mean current at BWJ is westward and is more dominated by the ITF (denoted by black lines in Fig. 6h and k). Similar to BEJ, the seasonal feature of the subsurface maximum ITF is also found at BWJ, where the corresponding westward current at this point reaches its maximum value at ∼ 100 m depth and is stronger during DJF than JJA (Fig. 6h and k). In this study, it is also found that the zonal westward current at 100 m depth at BWJ has a strong correlation with the subsurface (∼ 200 m) maximum ITF in the southern Ombai Strait, with a correlation coefficient of about 0.77 and a 95 % significance level of approximately ± 0.33, corroborating that the ITF flowing from the Ombai Strait is the main driver for zonal westward current at this point.

Furthermore, like CEJ, characteristics of persistent westward currents exist in the offshore region (CWJ), attributed to the SEC, and the westward current has a mean velocity of around 0.22–0.33 m s-1 in the upper 100 m (Fig. 6i and l). The simulated westward current at CWJ shows seasonal variations and reaches its maximum value at about 0.48 m s-1.

Vertical structures of zonal current along the meridional transect of Sumatra (SM; ASM-BSM-CSM) are shown in Fig. 6m–r. Similar to AEJ and AWJ, mean zonal current at ASM (nearshore region) is eastward, attributed to SJC, and associated with the Kelvin wave propagation. However, due to ASM located in front of western Sumatra (Fig. 2) and oriented in the northwest–southeast direction, the meridional component of velocity at this point is also dominant (Figs. 1a and 2). Therefore, zonal currents at ASM are relatively weak compared to those at AWJ and AEJ, which are located in front of southern Java and oriented in the west–east direction. For example, during SON, the eastward current reaches its maximum velocity of about 0.05 m s-1 at ASM (cyan lines in Fig. 6m and p), whereas it is about 0.23 m s-1 (at AWJ; Fig. 6g and j) and 0.20 m s-1 (at AEJ; Fig. 6a and d) at ∼ 30–50 m depth.

Furthermore, results of this study show that a maximum value of the eastward current at ASM, AWJ, and AEJ is found at a certain depth (at ∼ 30–50 m depth), and this strengthening of eastward flows is supposed to be attributed to a baroclinic Kelvin wave. The baroclinic Kelvin wave propagating vertically and horizontally along its waveguide can exert the most energy at a certain depth (Drushka et al., 2010; Pujiana et al., 2013; Iskandar et al., 2014). According to laboratory experiment observations conducted by Codiga et al. (1999) and Hallock et al. (2009), Kelvin waves can be trapped in a slope and propagate along an isobath. This phenomenon is known as a slope-trapped baroclinic Kelvin wave. Moreover, Kelvin waves that propagate along continental slope with strong stratification can cause strong current velocity. Codiga et al. (1999) also found that this slope Kelvin wave is formed after encountering a canyon-like bathymetry. Meanwhile, Pujiana et al. (2013) showed that Kelvin wave propagation from Lombok Strait to Makassar Strait, across the Sunda continental slope, is along isobaths at depths greater than 50 m. In this present study, the eastward current along the Transect A has a maximum current velocity at ∼ 30–50 m depth. Therefore, it is suggested that this maximum eastward current at ∼ 30–50 m depth is associated with a slope-trapped Kelvin wave that propagates at that depth along the southern coasts of Sumatra and Java.

In the transition region, the characteristics of the average zonal current (the climatological current field) at BSM (Fig. 6n and q) are different from those at BWJ (Fig. 6h and k) and BEJ (Fig. 6b and e). The average current at BSM is eastward, while at points BWJ and BEJ it is westward. During NW and transitional periods of the monsoon, zonal current at BSM flows eastward and reaches its maximum velocity of about 0.12 m s-1 at a depth of 40 m within the period of SON (Fig. 6q). Meanwhile, during the SE monsoon, the zonal current at this point flows westward. In contrast to the mean zonal currents in the nearshore region (ASM), it seems that the average zonal current field at BSM is not attributed to SJC. The reason is the BSM location, which is far from the coasts of Mentawai Islands and Enggano Island off the western coast of Sumatra (430 km away). This distance is more than Rossby radius of deformation at this latitude (∼ 90 km). Thereby, Kelvin waves, which affect the SJC variations, do not exist at this point. We suggest that the current variability at BSM is influenced by tropical current systems in the Indian Ocean, such as the Equatorial Counter Current (ECC), Southwest Monsoon Current (SWMC), and Wyrtki Jet. Here, we displayed seasonal averaged surface currents over 64 years (1950–2013) and schematics of the tropical current systems in the Indian Ocean as supporting evidence (Fig. 8).

Figure 8 shows that BSM is located in an area that is affected by the ECC, SWMC, and Wyrtki Jet. It can be seen in Fig. 8a that during DJF surface currents along the equatorial Indian Ocean are dominated by the westward North Equatorial Current (NEC) and the eastward ECC. Meanwhile, during JJA (Fig. 8c) the NEC disappears and the ECC becomes absorbed into the SWMC, which dominantly flows eastward in the northern Indian Ocean (Tomczak and Godfrey, 1994). In addition, during the transitional periods (MAM and SON) the jet is generated and causes a strengthening of eastward flows along the equatorial Indian Ocean (Fig. 8b and d). This explains why the climatological current at BSM flows eastward and reaches its maximum velocity during SON and MAM. These currents (the ECC, SWMC, and Wyrtki Jet) flow eastward before they turn and some part of their flow feed into the SEC in the southern Indian Ocean.

Current characteristics in the offshore region (CSM) generally show similarities with those at CWJ and CEJ, as shown in Fig. 6o and r. The current at CSM is attributed to the SEC and flows westward year round, with a mean velocity around 0.18–0.3 m s-1 in the upper 100 m. In addition, the strength of westward current at CSM varies seasonally and reaches its maximum value of about 0.42 m s-1 during SON (Fig. 6r).

Figure 10a–b show the vertical structure and temporal variability of the EOF1 (58 % of total variance) at AWJ, respectively. In order to see the temporal variation of the EOF1 more clearly in Fig. 10b, we have also provided data for the last 8-year period of the EOF first temporal mode (Fig. 10c, as an example). Current velocity variability relative to the mean flow can be obtained by multiplying the vertical mode structure (Fig. 10a) with the temporal variability (Fig. 10b).

Intraseasonal, semiannual, and annual variabilities of the EOF first temporal mode at AWJ as a result of the EEMD analysis are displayed in Fig. 10d–f, where their power spectra (Fig 10, left column) show the maximum energy for 3-month, 6-month, and 12-month periods, respectively. At this point, the highest power spectrum occurs at semiannual variability (Fig. 10e). In this figure (right), the semiannual variability of the EOF first temporal mode at AWJ clearly shows the presence of an eastward anomaly of the zonal current during the MAM and SON, which may be enhanced by downwelling Kelvin waves associated with the Wyrtki Jet in the equatorial Indian Ocean. Meanwhile, the anomaly of the zonal current at AWJ is westward during JJA in response to the prevailing southeasterly local winds during the SE monsoon. On the other hand, during DJF the anomaly of the zonal current at AWJ is not associated with the prevailing northwesterly local winds during the NW monsoon, in which the current anomaly is westward during this monsoon (Fig. 10e). As already discussed in Sect. 3.2 (Table 1 and Fig. 7), this may be attributed to the ITF, which has more influence on variation of zonal current at AWJ during DJF than the NW local wind.

Similar to AWJ, the first mode of EOF vertical structure and its temporal variability (64 % of total variance) at BSM show seasonal pattern (Fig. 11a–c). It is also found that signal on a 6-month (semiannual) period is quite dominant at BSM (Fig. 11e). In order to see the seasonal variation more clearly, we have provided a probability distribution function of the EOF1 of zonal currents for the NW, SE, and transition seasons at BSM at a depth of ∼ 40 m (Fig. 14). The 40 m depth was selected as an example because the most obvious seasonal variation of currents is present at this depth. It is found that variation of zonal current at BSM is dominantly eastward during DJF (Fig. 14a) and that this eastward current is enhanced during MAM and SON (Fig. 14b and d), which may be attributed to the tropical current systems in the Indian Ocean (ECC, SWMC, and Wyrtki Jet). Meanwhile, during JJA (Fig. 14c) the dominance of eastward current reduces, and the current tends to be dominantly westward. Furthermore, Fig. 12a–c show the first mode of EOF vertical structure and its temporal variability (72 % of total variance) at CEJ. In general, the anomaly of the zonal current at CEJ is westward, which is supposed to be associated with the meeting of SEC driven by trade winds and the ITF at this region. The EEMD analysis of the EOF1 of zonal current at CEJ also shows intraseasonal–interannual variabilities (Fig. 12d–g), where it is found that interannual timescale dominates the zonal current variation at CEJ (0.017 power per year).

To obtain a better understanding of the zonal current characteristics at AWJ, BSM, and CEJ, we have summarized the maximum energy density of zonal currents at intraseasonal, semiannual, annual, and interannual timescales that exists at each point based on power spectrum calculation in Figs. 10–12 (Table 3). It is shown that the zonal currents at AWJ have peak energies that are consecutively dominated by semiannual, intraseasonal, and annual signals, while the interannual signal is weaker than them at this point. Furthermore, although semiannual and intraseasonal signals are dominant at BSM, there is pronounced interannual variation of the zonal current at this point. In contrast, the zonal current variability at CEJ is dominated by interannual signal.

Furthermore, based on the power spectrum calculation shown in Fig. 12 (Table 3), it is found that intraseasonal variability of the SEC (zonal current at CEJ) is also prominent (∼ 0.012 power per year) in addition to the interannual signal (∼ 0.017 power per year). Meanwhile, based on sea level anomaly data in the period of October 1992 to the end of 1998 (about 6 years), Feng and Wijffels (2002) suggested that the strongest intraseasonal variability in the SETIO occurs in the SEC during the July–September season with baroclinic instability seeming to be the leading cause. On the other hand, in this study we found that the strongest intraseasonal variability occurs in the SJC (zonal current at AWJ). This different result seems to be due to differences in the length of data used in this study (64 years) and that in Feng and Wijffels (2002) (6 years). In addition, in this study we analyzed intraseasonal variability from the signal of the EOF first temporal mode of zonal currents (accounting for 58 %, 64 %, and 72 % of total variance at AWJ, BSM, and CEJ, respectively), whereas Feng and Wijffels (2002) analyzed the intraseasonal variation from standard deviation of the 6-year sea level anomaly data based on the 100 d high-pass filtered altimeter data during the four seasons (January–March, April–June, July–September, and October–December). Moreover, some of the differences may also be due to the fact that altimeter data do not resolve coastal processes well. However, further study is required to address this issue.

In this study, it is found that the most energetic zonal current variations of EOF1 at AWJ, BSM, and CEJ exist at ∼ 30 m depth (Figs. 10a, 11a, and 12a). To exclusively investigate the ocean currents at an interannual timescale, lagged correlation analyses have been applied between the zonal currents at a depth of about 30 m at points AWJ, BSM, and CEJ and each of the climatic indices (e.g., ONI and DMI), as shown in Table 4. The ONI and DMI indices from 1950 to 2013 used in this study are shown in Fig. 13.

The analysis of the lagged correlation indicates that the currents at BSM and CEJ show positive correlations with the ONI, namely r(18)=0.24 and r(4)=0.27, respectively, with the 95 % significance level approximately ± 0.07, indicating that an El Niño (La Niña) event is favorable for an eastward (westward) currents at these points (Figs. 11g and 12g) and showing that ITF transport is lower (higher) during El Niño (La Niña) events (Fieux et al., 1996; Meyers, 1996; Gordon and Susanto, 1999; Ffield et al., 2000; Susanto et al., 2001; Susanto and Gordon, 2005; Susanto et al., 2012; Liu et al., 2015; Susanto and Song, 2015; Zhang et al., 2016). ENSO seems to have a strongest influence on the zonal current variability at CEJ (Table 4), which is located close to the exits of the ITF. The ENSO signals penetrate into the SETIO mainly through the equatorial Pacific and coastal ocean Indonesian waveguides (Wijffels and Meyers, 2004; Zhang et al., 2016). Meanwhile, the present study shows that the correlation between the zonal current at AWJ and ONI is weak and below the significance level.

Furthermore, negative correlation is found between IOD and zonal currents at AWJ [DMI-U: r(9)=-0.09], BSM [DMI-U: r(1)=-0.28], and CEJ [DMI-U: r(11)=-0.13]. The correlation analysis indicates that IOD is most influential in forcing the interannual variation of the zonal currents at BSM, with the IOD leading the zonal currents by 1 month. The influence of interannual phenomena at BSM, such as IOD, is stronger and relatively instantaneous compared to that at points CEJ and AWJ. This may be due to the location of BSM, which is close to the center of the eastern pole of the IOD (5∘ S, 100∘ E; Saji et al., 1999). In contrast to ONI, there are IOD signals at AWJ, although the IOD signals at this point are weak compared to BSM and CEJ (Table 4). This indicates that some of the IOD signals are coastally trapped.

Table 5 lists extreme and neutral years and their concurrent events through 1950–2013. To further investigate interannual variation of zonal current, we summarized presence of major climate modes (ENSO and/or IOD) and the corresponding current anomalies at the points of BSM and CEJ (Table 6) based on the lagged correlation analyses in Table 4, the interannual variations of zonal current (Figs. 11g and 12g), and the ONI and DMI (Fig. 13). In Table 4, the ONI-U and DMI-U correlations are independent of IOD and of ENSO, respectively. Meanwhile, the current anomalies, which are attributed to the presence of major climate modes (ENSO and/or IOD) shown in the Table 6, could be forced by ENSO, IOD, or their combined effect. In this study, the amounts of the contribution values of ENSO and IOD or their combined effect on the current anomalies shown in the Table 6 are still unknown. Further studies are thus required to more quantitatively determine the contribution values of each of climate mode on zonal current variations in the study area as well as their possible teleconnection.

In addition to the lagged correlation analysis (Table 4), partial correlation analysis was also conducted since the IOD tends to co-occur with ENSO. Table 7 shows the partial correlation coefficients between zonal currents at 30 m on an interannual timescale for both ONI and DMI. As for ONI, the currents revealed significant positive correlations at CEJ during all monsoon seasons. This positive correlation suggests that El Niño (La Niña) events caused an eastward (westward) anomaly of currents at this point. Meanwhile, the partial correlation between the currents and the DMI showed significant negative correlation at BSM, in which it occurred only during the SE monsoon (JJA), as shown in Table 7. This negative correlation indicates that an eastward (westward) anomaly of the currents was induced by negative (positive) IOD. The results of the partial correlation analysis confirm and complement the previous findings in Table 4 that ENSO mainly contributed to the zonal current variability at CEJ in DJF, MAM, JJA, and SON, whereas the IOD had a significant influence on the variability of the current at BSM and only in JJA. In this present study, however, determining the causes of the influence of IOD on the current variability at BSM only in JJA is still a work in progress. Further research is necessary to explain the dynamical links of this matter. Additionally, the last mode (Figs. 10h, 11h, and 12h) represents long-term variation and trends, which may be associated with long-term internal variability within the Indian Ocean or remote forcing from the Pacific Ocean, as will be discussed in detail in a future paper.

To confirm possible influences of wind forcings on dominant variations of zonal current at AWJ, BSM, and CEJ, we have calculated the correlation between them. In this study, it is found that the zonal currents at AWJ (close to the shore) have peak energy over a semiannual period (0.140 power per year; Table 3). The semiannual variations of the zonal current at AWJ show the presence of an eastward anomaly of the zonal current during MAM and SON, which may be associated with Kelvin waves forced by winds over the equatorial Indian Ocean (Wyrtki, 1973; Quadfasel and Cresswell, 1992; Sprintall et al., 1999, 2000, 2010). Furthermore, we have calculated the correlation between zonal currents in the upper layer (30 m) at AWJ and zonal winds for the semiannual signals extracted using the EEMD method (Fig. 15). The 30 m upper-layer flows at AWJ show a strong positive correlation with the zonal winds over the equatorial Indian Ocean, with the winds leading the current by approximately 1 month. The positive correlation indicates that the flows are to the east when the winds blow from the west to the east and vice versa for the easterly wind. The 1-month lag between the flows at AWJ and the zonal winds in the equatorial Indian Ocean is in agreement with the expected arrival time of Kelvin waves at this point, suggesting that it is of about 18–35 d, with phase speeds ranging from 1.5 to 2.9 m s-1 (e.g., Sprintall et al., 2000; Syamsudin et al., 2004; Iskandar et al., 2005). Interestingly, there is also a weaker positive correlation between the 30 m upper-layer flows at AWJ at a lag of about 1 month and zonal trade winds in the western equatorial Pacific Ocean (WEPO) at a semiannual timescale, indicating that a strengthening (weakening) of easterly trade winds over the WEPO is favorable for anomalous westward (eastward) currents at AWJ. The strengthening of easterly trade winds over the WEPO will increase the sea level in the northern waters of West Papua and New Guinea, enhancing the Pacific-to-Indian pressure gradient across the Indonesian seas and forcing strengthened ITF transport. Since the currents at AWJ are strongly correlated to the ITF (Table 1), it is suggested that this possible dynamic could result in anomalous westward currents at AWJ and vice versa for the weakening winds over the WEPO.

Semiannual (0.135 power per year) signal of current variations is also dominant at BSM, but it is weaker than that at AWJ. In addition, there is pronounced interannual (0.012 power per year) variation of the zonal current at BSM (Table 3 and Fig. 11g), in which IOD is most influential in forcing interannual variation of currents at this point (at 30 m), as shown in Table 4. Like at AWJ, we also look for the relationships between the upper-layer flow (30 m) at BSM and the zonal winds but for the interannual signal obtained using the EEMD method (Fig. 16). At an interannual timescale, the 30 m upper-layer flows at BSM show a strong positive correlation with the zonal winds over the eastern tropical Indian Ocean, in which the response of the flows to the zonal winds are relatively instantaneously at a lag of about 1 month (Fig. 16). The location of the zonal winds affecting interannual variations in the upper-layer flows at BSM is in accord with the eastern pole region of IOD (10–0∘ S, 90–110∘ E; Saji et al., 1999).

Furthermore, as already explained, the zonal current variability at CEJ (close to the exits of the ITF) is dominated by an interannual (0.017 power per year) signal where the influence of ENSO is strongest at this point at depth of 30 m (Table 4). To enhance our understanding of the possible relationship of zonal currents at CEJ to wind forcings at an interannual timescale, we have also calculated the correlation between the upper-layer flow (30 m) at CEJ and the zonal winds, particularly in the Pacific Ocean. Like at BSM, the interannual signals of both flows and winds are extracted using the EEMD method. At an interannual timescale, the flows at CEJ at 30 m show a significant positive correlation with the local winds and the remote winds over the equatorial Pacific Ocean, in which the response of the flows to the zonal winds are about 4 to 6 months. In addition, we also found that the 4-month lag signal is stronger than the signals with the 5 to 6 months of lag. Figure 17 shows a correlation map between the Pacific winds and the currents at CEJ in the case of a 4-month lag.

A previous study conducted by Wijffels and Meyers (2004) showed that the variability in the ITF region was associated with Kelvin and Rossby waves originating in the Indian and Pacific Oceans, respectively. They have revealed the pathways for equatorial Pacific wind energy traveling down the Papuan–Australian shelf break and radiating westward-propagating Rossby Waves into the Banda Sea and southeastern Indian Ocean (their Fig. 20). Hence, there is a contribution from the westward-propagating Rossby waves to the ITF variability inside the Indonesian seas or at the ITF exit regions (the Ombai and Lombok straits and the Timor Passage), which lead into the SETIO and the western coast of Sumatra and southern coast of Java. Our simulation (Fig. 2) clearly shows that ITF flowing from the exit passages of Indonesian seas (Lombok, Ombai, and Timor passages) feeds into the SETIO region. Moreover, Wijffels and Meyers (2004) have computed the remotely driven Pacific Rossby wave speeds as a function of latitude. The phase speeds have been compared with the theoretical Rossby wave speeds based on the atlas of Chelton et al. (1998). In this study, we have estimated the travel time of the westward-propagating Rossby waves excited by the wind anomalies in the central and western Pacific to the SETIO, especially at point CEJ, based on the pathways for the Pacific signals introduced by Wijffels and Meyers (2004). In general, it was found that the equatorial Pacific signals around 130∘ W took approximately 3.01 months to arrive at CEJ based on the mean phase speed of about 0.2 cm s-1 taken from Wijffels and Meyers (2004). This travel time estimation was within the range of the 4-month lags between the flows at CEJ and the Pacific winds derived from the lagged correlation analysis in Fig. 17.