Chinese Journal of Oceanology and Limnology   2017, Vol. 35 issue(6): 1287-1302     PDF       
http://dx.doi.org/10.1007/s00343-017-6141-2
Institute of Oceanology, Chinese Academy of Sciences
0

Article Information

ZHOU Wenzheng(周文正), YU Fei(于非), NAN Feng(南峰)
Water exchange through the Kerama Gap estimated with a 25-year Pacific HYbrid Coordinate Ocean Model
Chinese Journal of Oceanology and Limnology, 35(6): 1287-1302
http://dx.doi.org/10.1007/s00343-017-6141-2

Article History

Received May. 8, 2016
accepted in principle Sep. 26, 2016
accepted for publication Nov. 9, 2016
Water exchange through the Kerama Gap estimated with a 25-year Pacific HYbrid Coordinate Ocean Model
ZHOU Wenzheng(周文正)1,2, YU Fei(于非)1, NAN Feng(南峰)1        
1 Institute of Oceanology, Chinese Academy of Sciences, Qingdao 266071, China;
2 University of Chinese Academy of Sciences, Beijing 100049, China
ABSTRACT: Variations in water exchange through the Kerama Gap (between Okinawa Island and Miyakojima Island) from 1979 to 2003 were estimated with the 0.08·Pacific HYbrid Coordinate Ocean Model (HYCOM). The model results show that the mean transport through the Kerama Gap (KGT) from the Pacific Ocean to the East China Sea (ECS) was 2.1 Sv, which agrees well with the observed mean KGT (2.0 Sv) for 2009-2010. Over the time period examined, the monthly KGT varied from -10.9 Sv to 15.8 Sv and had a standard deviation of ±5.0 Sv. The water mainly enters the ECS via the subsurface layer (300-500 m) along the northeastern slope of the Kerama Gap and mainly flows out of the ECS into the southwest of the Kerama Gap. The seasonal and interannual variations of the KGT and the Kuroshio upstream transport were negatively correlated. The Kuroshio upstream transport was largest in summer and smallest in autumn while the KGT was smallest in summer (1.02 Sv) and largest in spring (2.94 Sv) and autumn (2.44 Sv). The seasonal and interannual variations in the Kuroshio downstream (across the PN-line) transport differed significantly from the Kuroshio upstream transport but corresponded well with the KGT and the sum of the transport through the Kerama Gap and the Kuroshio upstream, which indicates that information about variation in the KGT is important for determining variation in the Kuroshio transport along the PN-line.
Key words: Kerama Gap     Kuroshio     Ryukyu Current     HYbrid Coordinate Ocean Model (HYCOM)     water exchange    
1 INTRODUCTION

Water exchange often occurs between two basins that are interconnected by a shallow or a deep sill (Song, 2006). The Ryukyu Island chain acts as a barrier to deep water exchange between the Northwestern Pacific (NP) and the East China Sea (ECS) (Jin et al., 2010). The Kerama Gap, which extends over about 250 km in the center of the Ryukyu Island chain between Miyakojima Island and Okinawa Island (Fig. 1) (Nakamura et al., 2013) and has a sill depth of more than 1 050 m, is the only deep channel in the chain (Sibuet et al., 1995). In recent decades, water exchange through the Kerama Gap has attracted much attention from oceanographers (Andres et al., 2008b; Nakamura et al., 2013; Na et al., 2014; Yu et al., 2015).

Figure 1 Maps of the East China Sea (ECS) and the Northwestern Pacific (NP) with 1 000 m interval bathymetry The thin black line between the Miyakojima and Okinawa represents the Kerama Gap. Red (or blue) lines represent the trajectory of the drifters through the Kerama Gap. The solid circles represent the start position of the drifters.

The upper circulation around the Kerama Gap is dominated by strong and persistent subtropical current systems, namely the Kuroshio and the Ryukyu Currents (Zhu et al., 2003; Centurioni et al., 2004). The Kuroshio enters the ECS mainly through the East Taiwan Channel (ETC) and flows out through the Tokara Strait, while the Ryukyu Currents flow northeastward along the eastern slope of the Ryukyu Islands (Nakamura et al., 2013). The exchange of upper-layer water between the ECS and the NP is mainly through the Kerama Gap, except at the entrance from the ETC where the sill depth is 775 m and at the exit to the Tokara Strait where the sill depth is 690 m (Yu et al., 2015); deep water exchange only occurs through the Kerama Gap. Soeyanto et al. (2014) pointed out that the volumes transported by the Kuroshio upstream and downstream transects were spatially consistent, but there was a significant discrepancy in the volume transported between the Kuroshio upstream and downstream. Andres et al. (2008a) pointed out that the Kerama Gap is a bottleneck for interactions between the Ryukyu Currents and the ECS-Kuroshio, and, even though the bulk of the Kuroshio enters the ECS through the ETC, the Kerama Gap may still be an important communication conduit with the ECS. They also suggested that the effect of mesoscale eddies arriving at Okinawa from the interior of the Pacific Ocean may drive changes in the Kuroshio transport across the PN-line (KNT) through the Kerama Gap (Fig. 2). Zheng et al. (2008) found that the flow from the Ryukyu Currents into the ECS through the Kerama Gap enhanced the KNT and that there was a significant discrepancy between the volumes transported by the Kuroshio upstream and the Kuroshio downstream. Jin et al. (2010) found that a shift in the Kuroshio axis determined the water exchange through the Kerama Gap, and that the water flowed from the ECS (NP) to the NP (ECS) when the Kuroshio axis shifted to (away from) the Kerama Gap. Part of the KNT is supplemented by the volume transported through the Kerama Gap (KGT), and the different water masses from the open ocean will modify the properties of the Kuroshio water mass (Oka and Kawabe, 1998).

Figure 2 Mean surface currents from 1992 to 2002 as shown by the Pacific HYCOM model The five red sections represent the Kerama Gap, Kuroshio upstream section (PM-line), Kuroshio downstream section (PN-line), continental shelf section (MN-line) and ETC. Sections 1 and 2 represent the section through the Kerama Gap monitored by the World Ocean Database (WOD).

To date, there have been few studies of the KGT (Andres et al., 2008b; Na et al., 2014). Yuan (1995) used current meter moorings and estimated that a volume of 2.4 Sv entered the ECS from the KGT between November 1991 and September 1992. Morinaga et al. (1998) also used current meter measurements and estimated that the KGT transport from the NP into the ECS was 7.2 Sv during the summer of 1992. Na et al. (2014) measured the annual mean KGT with an array of current and pressurerecording inverted echo sounders (CPIES) and current meter moorings from June 2009 to June 2011, and found that the 2-year mean KGT from the NP into the ECS was about 2.0 Sv. The many numerical models that have been used to estimate the KGT have all shown that the KGT transport is from the NP to the ECS. Using a 1/6° (~18 km) model, You and Yoon (2004) reported that the KGT was 5.6 Sv. Guo et al. (2006), by running a 1/18° (~6 km) nested ocean model from September 1991 to December 1998, found that the inflow through the Kerama Gap was 0.49 Sv; they also found that the Kuroshio onshore fluxes across the shelf break of the ECS were largest in autumn and smallest in summer. Soeyanto et al. (2014), with a data-assimilative ocean model (JCOPE2), estimated that the KGT from 1993 to 2012 was 0.18 Sv. Their model comprised two submodels connected by a one-way nesting method. The inner model had a horizontal resolution of 1/12° (~9 km) with 46 vertical layers and the outer model had a horizontal resolution of 1/4° (~27.5 km) with 21 vertical layers. Yu et al. (2015) studied the seasonal variations in the KGT from 1993 to 2012 using the 1/12.5° global HYCOM reanalysis and estimated that the KGT was 2.03 Sv. They also pointed out that the HYCOM reanalysis showed bottom-trapped inflow and that the maximum velocity occurred near the depth of the sill, which is not consistent with the observations.

Understanding the variations in water exchange through the Kerama Gap will contribute new insights into variability in the ECS-Kuroshio (Na et al., 2014). In this study, we investigated seasonal and interannual variations in the KGT and discussed the relationship between the KGT and the ECS-Kuroshio transport using the Pacific HYCOM model. We found that, in some ways, the Pacific HYCOM model had more advantages than the global HYCOM reanalysis, as discussed further in Section 2.3. In this paper, we describe the data, the numerical model, and model verification in Section 2; present the results in Section 3; discuss seasonal and interannual variability of the KGT in Section 4, and report our main findings in Section 5.

2 MATERIAL AND METHOD 2.1 Satellite-tracked drifting buoys

Satellite-tracked drifting buoys (drifters) collect direct velocity measurements worldwide from the sea surface layer at a nominal depth of 15 m as part of the Global Drifter Program. The raw data of each drifter is quality controlled and interpolated to 6-h intervals (Centurioni et al., 2004). Most drifters enter the ECSKuroshio through the ETC and drift northeastward along the Kuroshio principal axis, but some drift out of the Kuroshio through the Kerama Gap. Thirtythree drifters passed through the Kerama Gap from 1979 to 2015 (Fig. 1). Sixteen of them (red line) drifted from the ECS to the NP mainly through the southern part of the Kerama Gap and the other 17 (blue line) moved from the NP to the ECS, mainly through the northern part of the Kerama Gap. Out of the 16 drifters that crossed the Kerama Gap into the NP, 14 drifted from the Kuroshio upstream and the other two drifted from the Kuroshio downstream (across the PN-line). Almost all the drifters that moved from the NP to the ECS through the Kerama Gap travelled northeastward along the mean Kuroshio path, with most transport in autumn. The drifters' tracks suggest that water exchange through the Kerama Gap may lead to variations in the volumes transported between the Kuroshio upstream and the Kuroshio downstream.

2.2 Numerical model

In this study we used the Pacific HYbrid Coordinate Ocean Model (HYCOM PACa0.08) with an equatorial horizontal resolution of 0.08° (2 294×1 362 grid points, spaced at an average of 6.5 km) that covered the North and Equatorial Pacific regions from 98°E to 78°W and from 20°S to 66°N. The model has 33 hybrid vertical coordinate layers and the bathymetry is derived from a quality controlled ETOPO 2.5 dataset. Surface forcing of the model was done with 6-hourly gridded data from the European Centre for Medium-Range Weather Forecast (ECMWF) reanalysis product (ERA-15 forcing), including wind stress, wind speed, heat flux and precipitation (Kelly et al., 2007). The model outputs included monthly sea surface height (SSH), temperature, salinity, and currents for the period from 1979 to 2003. There is no data assimilation in this model. A more detailed description of the Pacific HYCOM model is provided by Kelly et al. (2007).

2.3 Model verification

Kelly et al. (2007) pointed out that the high resolution HYCOM model was much better than lowresolution models for reflecting the characteristics of the Kuroshio region and gave results that were closer to the observations. They also reported some bias between high resolution models without assimilation and the assimilated observations but considered they were still useful for studying ocean dynamics and for predictions of seasonal and interannual variabilities. To verify the reliability of the Pacific HYCOM, we compared the current velocity, temperature, and salinity distributions with the observed data.

2.3.1 Current velocity in the Kerama Gap

The Kuroshio entered the ECS mainly through the ETC and flowed out of the ECS through the Tokara Strait, as shown in Fig. 2. Although the Kuroshio is a persistently strong western boundary current, there was significant discrepancy in the seasonal variability between the upstream and downstream. The KNT comprises the Kuroshio transport of the PM-line (KMT), the KGT, and the transport of the continental shelf (MN-line) (MNT). There was a strong correlation between seasonal variability in the KMT and transport of the ETC (correlation coefficient of 0.96), and a weak correlation between the KMT and the KNT (correlation coefficient of 0.33) (Table 2). The Kerama Gap, between the PM-line and the PN-line, plays an important role in water exchange between the ECS and the NP, as shown by the drifter tracks in Fig. 1. The mean MNT was only about 1.0 Sv and varied from -0.8 to 2.8 Sv, while the mean KGT was 2.1 Sv and varied from -10.9 Sv to 15.8 Sv (Table 1), which indicates that the KGT may have more influence on the temporal variability of the KNT than the MNT. This is discussed further in Section 4.

Table 2 Correlation coefficients (month/season/year) among all the sections
Table 1 Statistics of transport volumes (Sv) through all the sections

The vertical structure of the velocity along the Kerama Gap transect from the global HYCOM reanalysis (Fig. 3a), Pacific HYCOM (Fig. 3b), and observations of Na et al. (2014) (Fig. 3c) are compared in Fig. 3. This figure shows that there is better agreement between the current structure and the observations with the Pacific HYCOM than with the global HYCOM reanalysis. The global HYCOM seems to have a bottom-trapped inflow, with a maximum velocity close to the sill depth (Yu et al., 2015). There is a discrepancy between the Pacific HYCOM and the global HYCOM because the Modular Ocean Data Assimilation System (MODAS) database is used to project surface information downward to the water column in the global HYCOM. Cummings and Smedstad (2013) noted that MODAS was only marginally useful in areas where the historical profile data seemed inadequate to statistically represent the Ryukyu Current in the northeast of the Kerama Gap. The Pacific HYCOM adopted an improved methodology, the Improved Synthetic Ocean Profile (ISOP), that adjusts the model forecast density field so that it agrees with the observations.

Figure 3 Vertical distribution of velocity (m/s) along the Kerama Gap transect a. global HYCOM reanalysis; b. Pacific HYCOM; c. linear interpolation of the observations (Na et al., 2014) (The figure is from Yu et al., 2015).

The Pacific HYCOM has more advantages than the global HYCOM reanalysis. For example, the Pacific HYCOM agrees that there is a thin vertical layer near the bottom with intensified inflow across the Kerama Gap, as suggested by Nakamura et al. (2013). The Pacific HYCOM indicates that 65% of the KGT is in the upper 500 m, which is closer to the observations (60%) (Na et al., 2014), whereas the global HYCOM reanalysis indicates that 61% of the KGT is in the upper 750 m (Yu et al., 2015). Although the current structures of the two models are significantly different, Yu et al. (2015) argued that the mean inflow into the ECS from both the Pacific HYCOM (2.1 Sv) and the global HYCOM (2.03 Sv) agreed well with the observations (2.0 Sv) and did not seem to be impacted by the difference in the flow structure in the deep layer. However, the seasonal variability of the KGT derived from the two models is different.

2.3.2 Temperature and salinity

We chose two observed sections (section 1 and section 2 in Fig. 2) from the World Ocean Database (WOD) (Boyer et al., 2013) to verify the model results. Figures 4 and 5 show that the temperature and salinity distributions from the Pacific HYCOM agreed well with the observations. Figure 4a and 4b show that the temperature of section 1 decreased vertically from the surface (23℃) to the bottom (5℃). Figure 4c and 4d show that the salinity of section 1 was more than 34.9 at depths of between 150 and 200 m and less than 34.4 at depths of between 600 and 700 m. Figure 5a and 5b show that the temperature of section 2 decreased vertically from the surface (23℃) to the bottom (3℃). Figure 5c and 5d show that the salinity was more than 34.8 between 150 and 250 m deep and less than 34.2 between 600 and 700 m deep. Figure 5 also shows that the properties of the Okinawa Trough water mass and the Pacific water were similar, which indicates that the Pacific water can influence the water mass properties of the Okinawa Trough through the Kerama Gap.

Figure 4 Vertical distributions of the temperature (units: ℃) and salinity along section 1 from the Pacific HYCOM (a, c) and WOD (b, d) The HYCOM data is for February 1988 and the observed WOD is from 22 February 1988 to 23 February 1988.
Figure 5 Vertical distributions of the temperature (units: ℃) and salinity along section 2 from the Pacific HYCOM (a, c) and WOD (b, d) The HYCOM data is from January 1999 and the observed WOD is from January 23 1999 to January 24 1999.
3 RESULT 3.1 Seasonal variability in the vertical distributions of velocity

There were significant seasonal variations in the vertical distributions of the velocity in the upper and deep layers of Kerama Gap. The drifter tracks (Fig. 1) illustrate that surface water of the Kerama Gap flows into and out of the ECS throughout the year and reaches a maximum in the autumn, when it drifts from the NP into the ECS. Figure 6 further verifies that the maximum and minimum surface water volumes flowed into the ECS from the Kerama Gap in autumn and summer, respectively. Variation in the Kuroshio is important for determining the water exchange through the Kerama Gap (Jin et al., 2010). The distance between the Kuroshio Current axis and the Kerama Gap can determine the effect of the Kuroshio on the Kerama Gap. When the transport volume of the Kuroshio reaches a maximum, the flow amplitude of the Kuroshio will be at its widest and the current axis will be closest to the Kerama Gap, which will cause water to flow out of the ECS. In contrast, when the volume transport of the Kuroshio is at a minimum, the flow amplitude of the Kuroshio is narrowest and the current axis is farthest away from the Kerama Gap; when this happens, the water will flow into the ECS and increase the transport volume of the Kuroshio. Zheng et al. (2008) suggested that inflow from the Ryukyu Currents through the Kerama Gap could enhance the volume transported downstream by the Kuroshio.

Figure 6 Average vertical distributions of the velocity (m/s) of the Kerama Gap transect for the (a) spring, (b) summer, (c) autumn, and (d) winter between 1979 and 2003 Positive values indicate the inflow into the ECS from the NP.

Compared with the surface velocity, the subsurface velocity (300–500 m) is much stronger from the NP to the ECS and persists throughout almost the whole year with a minimum in summer and a maximum in spring. The Kuroshio upstream was strongest (Table 1), and the Kuroshio Current axis was closest to the Kerama Gap, in summer, which may cause the water to flow out of the ECS. The maximum and minimum deep water exchange (from the NP to the ECS) through the Kerama Gap were in summer and winter, respectively. The distributions of the surface, subsurface, and deep-layer horizontal velocities will be discussed in next section.

3.2 Horizontal velocity fields around the Kerama Gap

Although the mean current vectors measured with current meters in the upper (400–500 m), middle (600–700 m), and deep levels (950–1 050 m) in the Kerama Gap have been studied by Na et al. (2014) and Yu et al. (2015), they were only able to study two or three observation points because of a lack of observed data. To date there have been no studies of the seasonal horizontal velocity distributions in the upper and deep layers in the Kerama Gap and the surrounding region, even though they are needed to provide an understanding of the dynamics of the Kerama Gap overflow and the seasonal variations in the water exchange through the Kerama Gap. In particular, while the control of the Kuroshio on surface flow, and mesoscale eddies have both been studied (Andres et al., 2008b; Jin et al., 2010; Nakamura et al., 2013), there is little information about subsurface and deep-layer flows surrounding the Kerama Gap. In this study therefore, we examined the seasonal horizontal velocity distributions at the surface (10 m), subsurface (300 m), and in the deep layers (1 000 m) in the Kerama Gap and in the surrounding region using the Pacific HYCOM.

Figure 7 shows the seasonal variability of the surface horizontal velocity fields surrounding the Kerama Gap for the period from 1979 to 2003 modelled by Pacific HYCOM. In spring, the Kerama Gap overflow (from the ECS to the NP) was mainly in the southwest of the gap and the inflow (from the NP to the ECS) was mainly in the northeast of the gap. The overflow was intensified along Miyakojima. The inflow grew weaker in summer, intensified in autumn throughout the entire Kerama Gap, and then became weaker again in winter. Surface horizontal velocity fields show that surface water mainly flowed out of the ECS into the southern part of the Kerama Gap in summer and mainly flowed into the ECS into the northern part of the Kerama Gap in autumn, which is consistent with the drifter tracks (Fig. 1).

Figure 7 Average (a) spring, (b) summer, (c) autumn, and (d) winter horizontal velocity fields around the Kerama Gap over the period from 1979 to 2003 for the surface layer (10 m) from Pacific HYCOM

Figure 8 shows the seasonal variability of the horizontal velocity fields around the Kerama Gap for the period from 1979 to 2003 at a depth of 300 m. We chose this depth because the velocity at 300 m provides a good representation of the core subsurface inflow from the NP to the ECS through the Kerama Gap. Nagano et al. (2007) indicated that the maximum velocity of the Ryukyu Currents was quite variable between 300 and 800 dbar deep. Thoppil et al. (2016) thought that the subsurface velocity of the Ryukyu Currents was at a maximum between 800 and 1 000 m and that there was a transient shallow velocity core at about 300 m followed by anticyclonic eddies. Nakamura et al. (2013) showed that the intermediate water of the NP intensified along the northeast slope of the Kerama Gap into the ECS and could be regarded as a persistent flow throughout the year. The results of this study show that the subsurface water of the NP bifurcated from the Ryukyu Current System and flowed into the ECS along the northeastern slope of the Kerama Gap throughout the year, with a maximum in spring and a minimum in summer.

Figure 8 Same as Fig. 7 but for the subsurface layer (300 m)

The Kerama Gap plays an important role in deep water exchange between the NP and the ECS but there is little information about the deep flow pattern in the Kerama Gap and even the direction of the mean flow has not yet been firmly established (Andres et al., 2008a). Thoppil et al. (2016) found that there were significant variations in the Ryukyu Currents with maximum currents in winter and spring and minimum currents in summer. Nakamura et al. (2008) pointed out that the deep countercurrent beneath the Kuroshio in the northern Okinawa Trough was approximately bounded by the 1 000 m isobaths and was much stronger during the winter-spring period than in the summer-autumn period. The horizontal velocity of the deep layer (1 000 m) surrounding the Kerama Gap (Fig. 9) suggests that the Kuroshio had vanished and that the deep countercurrents beneath the Kuroshio were more stable along the continental shelf of the northern Okinawa Trough during the winter-spring period than during the summer-autumn period. The Kerama Gap deep outflow was also significant during the winter-spring period and the inflow was noticeable during the summer-autumn period, which indicates that deep water exchange through the Kerama Gap may be dominated by the deep countercurrent beneath the Kuroshio. We can therefore conclude that the Kuroshio and the mesoscale eddies result in upper water exchange through the Kerama Gap (Andres et al., 2008b; Jin et al., 2010; Na et al., 2014), but the deep countercurrent beneath the Kuroshio determines the deep water exchange of the Kerama Gap. Besides, the upper water entered the ECS with a maximum velocity of 10 cm/s in spring while the deep water flowed into the ECS with a maximum velocity of 5 cm/s in summer.

Figure 9 Same as Fig. 7 but for the deep layer (1 000 m)
4 DISCUSSION 4.1 Seasonal variation in the KGT

Although the mean KGT was only 2.1 Sv and only contributed about 6% of the mean Kuroshio transport, the fact that it varied widely (from -10.9 Sv to 15.8 Sv with a standard deviation of 5.0 Sv) means that it may have had a significant impact on the seasonal variability of the ECS-Kuroshio transport. The large variation in the KGT may be caused by the bifurcation of the Kuroshio and Ryukyu Currents and the high frequency of mesoscale eddies. Yu et al. (2015) calculated the standard deviation of the KGT and the KNT and concluded that the yearly transport standard deviation of the KNT was highly correlated with the standard deviation of the KGT (correlation coefficient of 0.64). Although they reported that there was a relationship between the KNT and the KGT, they did not discuss the relationship between the KGT and the KMT and the continental shelf transport (MNT). In this study, we calculated the ETC transport, KMT, KNT, KGT, and MNT (Table 1) for each month from 1979 to 2003 and discussed the relationships between them.

The most significant feature of the monthly variability is that the KNT was positively correlated with the sum of KGT and the KMT, with a correlation coefficient of 0.87 (Fig. 10a); the KGT was negatively correlated with the KMT with a correlation coefficient of -0.66 (Fig. 10b). Nakamura et al. (2008) reported that a small portion of the Kuroshio flow separated from the western slope in the southern Okinawa Trough and moved toward the Kerama Gap and that its greatest portion flowed along the eastern slope of the northern Okinawa Trough. When the Kuroshio goes toward (away from) the Kerama Gap, the water will flow from the ECS (NP) to the NP (ECS), thereby causing a decrease (increase) in the KGT, which is agreement with the findings of Jin et al. (2010). The PM-line is between the ETC and the PN-line, and they are all part of the Kuroshio. It is interesting to find that the monthly variability of the KMT was strongly correlated with the transport of the ETC (correlation coefficient of 0.93), but not so strongly correlated with the KNT (correlation coefficient of 0.27) (Table 2). This shows that there was little variability in the Kuroshio upstream transport (between the ETC and the PM-line) but that there was significant variability in the Kuroshio midstream transport (between the PMline and the PNline), which is consistent with the results reported in other studies (Lin et al., 2004; Zheng et al., 2008; Soeyanto et al., 2014). The location of the Kerama Gap, between PM-line and the PN-line, may be an important control on the variation between the KMT and the KNT. The monthly variability in the KNT was positively correlated with the KGT and MNT, with correlation coefficients of 0.45 and 0.43, respectively, but was weakly correlated with the KMT, with a correlation coefficient of 0.27. Besides, the monthly variability of KNT is more positively correlated with sum of KMT and KGT (correlation coefficients of 0.87) than the sum of KMT and MNT (correlation coefficients of 0.33) (Table 2). We can conclude that the variation in the KNT was determined by the variability in the KGT, KMT, and MNT, but that the KGT was perhaps the most important factor.

Figure 10 Comparisons of the monthly volume transports from 1979 to 2003 from the Pacific HYCOM a. PN-line (blue) and the sum of transports through the Kerama Gap and the PM-line (green); b. Kerama Gap (blue) and the PM-line (green). Positive volume transport through the Kerama Gap is in the direction from the NP to the ECS.

The most significant feature of the seasonal variability is that it was consistent with the monthly variability. The KNT was positively correlated with the sum of KGT and KMT (correlation coefficient of 0.90) and the KGT was negatively correlated with the KMT (correlation coefficient of -0.63) (Fig. 11). The positive correlation between the seasonal variability in the KNT and the MNT was stronger than the correlations with either KGT or KMT (Table 2); the standard deviation and the range of the KGT (5.0 Sv, 26.7 Sv) were bigger than those of the KMT (4.5 Sv, 23.8 Sv) and the MNT (0.6 Sv, 3.6 Sv) (Table 1), which also demonstrates that KGT may have more control on the seasonal variation in the KNT than either the KMT or MNT.

Figure 11 Comparisons of the seasonal cycle of volume transports through the Kerama Gap (green), PN-line (red), PM-line (black) and total transports through the Kerama Gap and the PM-line (blue) from 1979 to 2003 from the Pacific HYCOM model

The 25-year mean KGT from the NP into the ECS derived from the Pacific HYCOM was about 2.1 Sv, and had a standard deviation of ±5.0 Sv. The temporal variability of the total transport (~250 km) was strongly correlated (correlation coefficient of 0.97) with the deep gap (~50 km) transport (1.93 Sv). Figure 12 shows that the 25-year monthly mean KGT was largest in April (3.83 Sv) and smallest in July (0.31 Sv), which is not consistent with the findings of Yu et al. (2015), who reported that the KGT was largest in October (3.04 Sv) and smallest in November (0.54 Sv). The discrepancy between our results and those of Yu et al. (2015) may reflect the bottom velocity of the model. As they pointed out, in the global HYCOM reanalysis the inflow is bottomtrapped with a maximum velocity occurring near the sill depth, but the current structure in the Pacific HYCOM model shows better agreement with the observations. Although the Pacific HYCOM model has no data assimilation, it is important for predicting seasonal and interannual variability (Kelly et al., 2007). Seasonal variability of the KGT was greatest in spring (2.94 Sv) and autumn (2.44 Sv) and was least in summer (1.02 Sv). The KMT, with a maximum in summer (34.4 Sv) and a minimum in autumn (32.5 Sv) (Table 1), was negatively correlated with the KGT but was consistent with the MNT. The KNT, with a maximum in spring (37.5 Sv) and a minimum in winter (34.7 Sv), was positively correlated with the sum of KGT and KMT. This also illustrates that the discrepancy in the seasonal variations between the KNT and the KMT is related to variations in the KGT.

Figure 12 Monthly mean transports of the Kerama Gap, PN-line, and PM-line, and the sum of the transports of the Kerama Gap and the PM-line from1979 to 2003 from the Pacific HYCOM model
4.2 Interannual variability

The yearly transport and its standard deviations through the Kerama Gap, PM-line, PN-line, and the sum of the transports through the Kerama Gap and the PM-line from 1979 to 2003 estimated by the Pacific HYCOM model are shown in Fig. 13. The interannual variation of the KGT was positively correlated with the KNT (correlation coefficient of 0.40) but was negatively correlated with the KMT (correlation coefficient of -0.64). The correlation coefficient between the KNT and the KMT was 0.34, confirming that the interannual variability of the KNT also corresponded more closely with variability of the KGT than of the KMT. Yu et al. (2015) estimated that the two time series of the yearly standard deviation between the KGT and the KNT were highly correlated, with a correlation coefficient of 0.64 between 1992 and 2012, which is less significant than the correlation reported in this study; the correlation coefficient between the KNT and the total of KGT and the KMT was high (0.89). The yearly mean MNT ((1.0±0.6) Sv) was not useful for estimating the contribution to the KNT. The interannual variability of the KNT was mainly determined by the sum of the KGT and KMT (correlation coefficient of 0.86).

Figure 13 Yearly mean transport (a) and its standard deviation (b) in the Kerama Gap (green), PN-line (blue), PM-line (black), and total transports through the Kerama Gap and the PM-line (red) from 1979 and 2003 derived from the Pacific HYCOM model
5 CONCLUSION

The Kerama Gap plays an important role in water exchange between the ECS and the NP and makes a significant contribution to the variability between the KMT and the KNT. We investigated variations in water exchange through the Kerama Gap from 1979 to 2003 with a 0.08° Pacific HYCOM model. The 25-year mean KGT from the NP into the ECS was (2.1±5.0) Sv. The monthly KGT varied widely and reached a maximum of 15.8 Sv (March 1999) and a minimum of -10.9 Sv (June 1995). The mean transport of the deep gap (~50 km) was (1.93±4.3) Sv and varied from -9.9 Sv to 13.9 Sv. The monthly mean KGT was greatest in April (3.83 Sv) and smallest in July (0.31 Sv). There was significant seasonal variation in the KGT, with maximum values in spring (2.94 Sv) and autumn (2.44 Sv) and a minimum in summer (1.02 Sv). The KMT reached a maximum in summer (34.4 Sv) and a minimum in autumn (32.5 Sv).

Water flows into the ECS mainly from the northern part of the Kerama Gap and flows out of the ECS from the southern part of the Kerama Gap in the surface layer. The subsurface currents are the most important factor in water exchange through the Kerama Gap. They persistently flow into the ECS along the northeastern slope of the Kerama Gap throughout the year, with a maximum in spring and a minimum in summer. The deep water exchange through the Kerama Gap is determined by the deep countercurrents beneath the Kuroshio and is dominated by strong inflows during the summer-autumn period and an outflow during the winter-spring period.

The seasonal variability of the KNT and the sum of KGT and KMT were positively correlated, while the KGT and the KNT were negatively correlated. The KMT was almost consistent with the Kuroshio transport (across the ETC), but differed significantly from the KNT, indicating that the KGT plays an important role in the variation of the volume transported by the PM-line and the PN-line. The correlation between the KNT and the KGT was larger than the correlation between the KNT and the KMT, which also demonstrates that the variations in the KNT are determined by the variability of the KGT rather than the KMT. The interannual variability was almost consistent with the seasonal variations. We found that both the Kerama Gap and the Kuroshio transports varied on seasonal and interannual time scales, which will be useful for future observations (Qu and Song, 2009). The water mass properties from the NP to the ECS-Kuroshio through the Kerama Gap may be modified. Future studies will focus on (1) tracing the intermediate and bottom water from the Pacific Ocean to the Okinawa Trough through the Kerama Gap, with reference to Chen. (2005), and (2) the effect of the Kuroshio and mesoscale eddies on water exchange via the Kerama Gap.

6 ACKNOWLEDGEMENT

The satellite-tracked drifting buoy data are available from the Global Drifter Program (GDP), with support from their website (ftp://ftp.aoml.noaa.gov/pub/phod/buoydata/). The World Ocean Database (WOD) is available from the website (http://www.nodc.noaa.gov/OC5/SELECT/dbsearch/dbsearch. html). The HYCOM data are available for download from the Global Ocean Data Assimilation Experiment (GODAE) from the website (ftp://ftp.hycom.org/datasets/PACa0.08/).

References
Andres M, Park J H, Wimbush M, Zhu X H, Chang K I, Ichikawa H, 2008a. Study of the Kuroshio/Ryukyu current system based on satellite-altimeter and in situ measurements. Journal of Oceanography, 64(6): 937–950. Doi: 10.1007/s10872-008-0077-2
Andres M, Wimbush M, Park J H, Chang K I, Lim B H, Watts D R, Ichikawa H, Teague W J, 2008b. Observations of Kuroshio flow variations in the East China Sea. Journal of Geophysical Research:Oceans, 113(C5): 15.
Boyer T P, Antonov J I, Baranova O K, Garcia H E, Grodsky A, Johnson D R, Locarnini R A, Mishonov A V, O'Brien T D, Seidov D, Smolyar I V, Zweng M M, 2013. World Ocean Database 2013. Data Science Journal, 90(49): 153–173.
Centurioni L R, Niiler P P, Lee D K, 2004. Observations of inflow of Philippine Sea surface water into the South China Sea through the Luzon Strait. Journal of Physical Oceanography, 34(1): 113–121. Doi: 10.1175/1520-0485(2004)034<0113:OOIOPS>2.0.CO;2
Chen C T A, 2005. Tracing tropical and intermediate waters from the South China Sea to the Okinawa trough and beyond. Journal of Geophysical Research:Oceans, 110(C5): C05012.
Cummings J A, Smedstad O M. 2013. Variational data assimilation for the global ocean. In: Park S K, Xu L eds. Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications (Vol. Ⅱ). Springer, Berlin Heidelberg. p. 303-343.
Guo X Y, Miyazawa Y, Yamagata T, 2006. The Kuroshio onshore intrusion along the shelf break of the East China Sea:the origin of the Tsushima warm current. Journal of Physical Oceanography, 36(12): 2205–2231. Doi: 10.1175/JPO2976.1
Jin B G, Wang G H, Liu Y G, Zhang R, 2010. Interaction between the East China Sea Kuroshio and the Ryukyu current as revealed by the self-organizing map. Journal of Geophysical Research:Oceans, 115(C12): 168–172.
Kelly K A, Thompson L, Cheng W, Metzger E J, 2007. Evaluation of HYCOM in the Kuroshio Extension region using new metrics. Journal of Geophysical Research:Oceans, 112(C1): 18.
Lin X P, Wu D X, Lan J, 2004. The intrusion and influences of intraseasonal long Rossby waves in the East China Sea. Journal of Hydrodynamics, 16(5): 621–632.
Morinaga K, Nakagawa N, Osamu K, Guo B. 1998. Flow pattern of the Kuroshio west of the main Okinawa Island. In: Proceedings of Japan-China Joint Symposium on Cooperative Study of Subtropical Circulation System. Seikai National Fisheries Research Institute, Nagasaki, Japan. p. 203-210.
Na H, Wimbush M, Park J H, Nakamura H, Nishina A, 2014. Observations of flow variability through the Kerama Gap between the East China Sea and the Northwestern Pacific. Journal of Geophysical Research:Oceans, 119(2): 689–703. Doi: 10.1002/2013JC008899
Nagano A, Ichikawa H, Miura T, Ichikawa K, Konda M, Yoshikawa Y, Obama K, Murakami K, 2007. Current system east of the Ryukyu Islands. Journal of Geophysical Research:Oceans, 112(C6): 137–154.
Nakamura H, Nishina A, Ichikawa H, Nonaka M, Sasaki H, 2008. Deep countercurrent beneath the Kuroshio in the Okinawa Trough. Journal of Geophysical Research:Oceans, 113(C6): 1394–1404.
Nakamura H, Nishina A, Liu Z J, Tanaka F, Wimbush M, Park J H, 2013. Intermediate and deep water formation in the Okinawa Trough. Journal of Geophysical Research:Oceans, 118(12): 6881–6893. Doi: 10.1002/2013JC009326
Oka E, Kawabe M, 1998. Characteristics of variations of water properties and density structure around the Kuroshio in the East China Sea. Journal of Oceanography, 54(6): 605–617. Doi: 10.1007/BF02823281
Qu T D, Song Y T, 2009. Mindoro Strait and Sibutu Passage transports estimated from satellite data. Geophysical Research Letters, 36(9): 269–277.
Sibuet J C, Hsu S K, Shyu C T, Liu C S. 1995. Structural and kinematic evolutions of the Okinawa Trough backarc basin. In: Taylor B ed. Backarc Basins. Springer, New York. p. 343-379.
Soeyanto E, Guo X Y, Ono J, Miyazawa Y, 2014. Interannual variations of Kuroshio transport in the East China Sea and its relation to the Pacific Decadal Oscillation and mesoscale eddies. Journal of Geophysical Research:Oceans, 119(6): 3595–3616. Doi: 10.1002/2013JC009529
Song Y T, 2006. Estimation of interbasin transport using ocean bottom pressure:theory and model for Asian marginal seas. Journal of Geophysical Research:Oceans, 111(C11): 36–45.
Thoppil P G, Metzger E J, Hurlburt H E, Smedstad O M, Ichikawa H, 2016. The current system east of the Ryukyu Islands as revealed by a global ocean reanalysis. Progress in Oceanography, 141: 239–258. Doi: 10.1016/j.pocean.2015.12.013
You S H, Yoon J H, 2004. Modeling of the Ryukyu current along the Pacific side of the Ryukyu islands. Pacific Oceanography, 2: 44–51.
Yu Z T, Metzger E J, Thoppil P, Hurlburt H E, Zamudio L, Smedstad O M, Na H, Nakamura H, Park J H, 2015. Seasonal cycle of volume transport through Kerama Gap revealed by a 20-year global HYbrid Coordinate Ocean Model reanalysis. Ocean Modelling, 96: 203–213. Doi: 10.1016/j.ocemod.2015.10.012
Yuan Y, 1995. The western boundary current east of the Ryukyu Islands. Oceanographic Literature Review, 42(11): 929–930.
Zheng X T, Liu Q Y, Hu H B, Miyazawa Y, Jia Y L, 2008. The study of the temporal and spatial characteristics of western boundary current East of Ruykyu submarne ridge and the transport of Kuroshio in East China Sea. Acta Oceanologica Sinica, 30(1): 1–9.
Zhu X H, Han I S, Park J H, Ichikawa H, Murakami K, Kaneko A, Ostrovskii A, 2003. The northeastward current southeast of Okinawa Island observed during November 2000 to August 2001. Geophysical Research Letters, 30(2): 61–79.