Abyssal Circulation in the Philippine Sea

ZHAI Fangguo, and GU Yanzhen

College of Oceanic and Atmospheric Sciences, Ocean University of China, Qingdao 266100, China

Abstract The abyssal circulation in the Philippine Sea (PS) is investigated, with outputs from the Simple Ocean Data Assimilation version 2.2.4 (SODA224). The deep-water currents in SODA224 are carefully evaluated, with sparse in situ observations in the North Pacific Ocean. In the upper deep layer (2000–3000 m) of the PS, a strong westward current, which originates from the Northeast Pacific Basin and enters the PS through the Yap-Mariana Junction, exists along 11˚–14˚N. This strong westward current bifurcates into two western boundary currents off the Philippines. The northward-flowing current flows out of the PS around 20˚–21˚N, whereas the southward-flowing current transports deep water from the northern hemisphere to the southern hemisphere. In the lower deep layer (3000 – 4500 m), the inflow water first flows northward to the east of the Western Mariana Basin and then turns westward at approximately 18˚N. The inflow water mainly enters the Philippine Basin (PB), with a small part turning southward to constitute a weak cyclonic circulation. The water entering the PB mainly merges into a strong southward western boundary current in the southern PB. In the bottom layer (below 4500 m), both the northeast and northwest PB show single cyclonic gyres, whereas the south PB shows a single anticyclonic gyre. Moreover, comparisons with the observations indicate the possible existence of a cyclonic sense of circulation over the Philippine Trench. The current study provides the implications for future observations, which are needed to further investigate the temporospatial variations of the abyssal circulation in the PS on multiple scales.

Key words Philippine Sea; deep ocean circulation; Simple Ocean Data Assimilation version 2.2.4; Yap-Mariana Junction; mean structure

1 Introduction

The Philippine Sea (PS) is located in the far western Pacific Ocean in the northern hemisphere (Fig.1). The PS is separated from the other part of the North Pacific Ocean in the east by the Izu-Ogasawara-Mariana-Yap Ridges and from the adjacent China Seas in the west by the Ryukyu Islands, Taiwan Island, and Philippines. In the upper layer,the North Equatorial Current (NEC) flows westward and bifurcates in the Philippine coast into the northward- flowing Kuroshio Current and southward- flowing Mindanao Current (Hu et al., 2015). At the subsurface, the eastwardflowing North Equatorial Undercurrent, southward- flowing Luzon Undercurrent, and northward- flowing Mindanao Undercurrent exist (Qiu et al., 2015). These currents have considerable influence on the regional and global circulations and climate variations, and thus have been extensively investigated with observations, numerical simulations, and data assimilations (e.g., Hu et al., 2015; Qiu et al., 2015; references therein).

However, in the deep layers below approximately 2000 m, little is known about the temporospatial variations of the abyssal circulation in the PS mainly because of the lack of direct in situ observations (Zhai et al., 2014). In the literature, only sparse observations were conducted in different time periods (Yoshioka et al., 1988; Uehara and Taira, 1990; Chaen et al., 1993; Kaneko et al., 1998, 2001;Nagano et al., 2013, 2018; Zhai et al., 2014; Wang et al.,2017). By analyzing full-depth high- resolution hydrographic measurements in the summer of 1994 along the 137˚E meridian, Kaneko et al. (1998) determined that the deep water flowing into the Western Mariana Basin (WMB)from the North Pacific Ocean first flows northward and then spreads to the Shikoku Basin (ShB) and Philippine Basin (PB). By applying an inverse model to the World Ocean Circulation Experiment hydrographic observations,Kaneko et al. (2001) estimated the circulation of the deep and bottom waters of the PS. They indicated that in the deep layers (approximately 1500–3500 m), the North Pacific Deep Water (NPDW) first enters the PS in regions south of approximately 15˚N and then flows out at approximately 16˚–24˚N. In the bottom layer, inflow water exists at the southern end of the PB, resulting in a northward western boundary current in the PB. This finding is nearly consistent with the observations of Reid (1981) but different from the observations of Uehara and Taira (1990)and Wang et al. (2017). With a current meter moored at a depth of approximately 4000 m, which is approximately 600 m above the sea bottom in the WMB, Yoshioka et al.(1988) observed a southward current throughout the entire time period from July 1985 to July 1986 at 137˚E, 12.5˚N.On the basis of the hydrographic observations obtained in February 1987, Uehara and Taira (1990) observed the northward geostrophic current at a pressure of 1500 dbar around the same place. By contrast, Uehara and Taira (1990)observed the southward geostrophic current in the narrow western boundary region of the WMB.

Along the northern periphery of the ShB, the direct observations indicate that the westward/southwestward boundary current exists parallel to the local bottom contour in the deep layer (Nagano et al., 2013, 2018). However, this boundary current may not be a stable current with a predominant direction according to the observations of Chaen et al. (1993). In regions southeast of Okinawa and Taiwan Islands, mean currents in the deep layer were observed to flow southwestward/southward parallel to the local bottom contour (Chaen et al., 1993). In regions just east of the Mindanao Island, Wang et al. (2017)observed that the mean currents in the deep layer flow equatorward.

Moreover, these observations indicate that the deepwater currents show significant tidal oscillation and variations on the mesoscale and seasonal and interannual time scales. The tidal oscillation is related to the deeply intruded internal tides generated at the steep topographies(Nagano et al., 2013). A significant mesoscale variability is also observed in the Faroe Bank Channel overflow(Darelius et al., 2011), deep western boundary current in the Atlantic Ocean (Schott et al., 2006), and deep currents in the Pacific Ocean (Rudnick, 1997; Kawabe et al., 2003).This variability is quite possibly related to the vertically propagating Rossby waves (Yoshioka et al., 1988). Seasonal and interannual variations have been observed by Wang et al. (2017) with approximately 4 years of continuous mooring observations just east of the Mindanao Island.

However, these observations are far from enough for us to obtain a full picture of the abyssal circulation in the PS.Therefore, the current study investigates the mean structure of the abyssal circulation in the PS, with outputs from a global data assimilation, namely the Simple Ocean Data Assimilation version 2.2.4 (SODA224; Carton and Giese, 2008). The remainder of this paper is organized as follows: Section 2 briefly describes the data sets used in the current study. Section 3 extensively compares the outputs from the global data assimilation with sparse observations in the Pacific Ocean to evaluate the performance of the SODA224. The results show reasonably good consistency.Section 4 presents the abyssal circulation in the PS. Section 5 provides a brief discussion and the conclusion. The results derived in the present work need to be further confirmed in the future. However, the results may provide the implications for future observations, which are important in advancing our knowledge.

Fig.1 Map of the Pacific Ocean. The blue and black lines show the isobaths at depths of 2000 and 4000 m, respectively. PB,Philippine Basin; KPR, Kyushu-Palau Ridge; ShB, Shikoku Basin; WMB, Western Mariana Basin; EMB, Eastern Mariana Basin; IOR, Izu-Ogasawara Ridge; MR, Mariana Ridge; WCB, Western Caroline Basin; ECB, Eastern Caroline Basin;MB, Melanesian Basin; JT, Japan Trench; WIP, Wake Island Passage; HR, Hawaiian Ridge; SaB, Samoan Basin; SP, Samoan Passage; TKT, Tonga-Kermadec Trench. The pink point indicates the Yap-Mariana Junction (YMJ).

2 Data Sets

2.1 SODA224

Given the lack of in situ deep ocean observations in the PS, we adopt the ocean reanalysis data in the current study,which uses an ocean model in conjunction with data assimilation of limited ocean observations to estimate the state of oceans. On the basis of the complex topography of the PS, the SODA224 (e.g., Carton and Giese, 2008) is selected from a number of available ocean reanalysis and numerical simulation outputs. The SODA224 is based on the Parallel Ocean Program (POP) ocean model and SODA software. The ocean model adopts the POP version 2.0.1 numerics, with an average 0.25˚×0.4˚ horizontal resolution and 40 vertical levels increasing from approximately 5 m to 5375 m with 10 m spacing near the sea surface.There are 14 levels below 2000 m, ranging from 2125 m to 5375 m with an interval of approximately 250 m. In this study, we use the outputs mapped onto a uniform global 0.5˚×0.5˚ horizontal grid. The sea surface wind forcing in SODA224 is derived from the 20th Century Atmospheric Reanalysis product (Compo et al., 2011).

The reanalysis product spans a long period of approximately 140 years, i.e., from 1871 to 2010. In this study,we utilize the results obtained in the past six decades, i.e.,from 1950 to 2010, and focus on the mean abyssal circulation during this period. Given the scarcity of observations in deep-water layers, the SODA224 aims to construct a reasonably good retrospective analysis of the timeevolving state of the upper layers of the ocean (Carton and Giese, 2008). It has been extensively proven to perform well at describing the temporospatial characteristics and dynamics of deep ocean waters and its circulation(Schott et al., 2009). In Section 3, the water masses and circulation of deep layers mainly below 2000 m in the Pacific Ocean in SODA224 are extensively compared with a gridded observation product and those sparsely observed in the literature. Notably, the comparisons are reasonably good in terms of the spatial circulation pattern,which supports the finding that the SODA224 is capable of describing the abyssal circulation in the PS.

2.2 EN4

To evaluate the performance of the SODA224 and describe the deep-water masses, we also adopt a kind of objective ocean analysis in the current study, which is the version 4 of the Met Office Hadley Center ‘EN’ series of data sets (EN4; Good et al., 2013). Monthly objective analyses of potential temperature and salinity of EN4 are obtained from the global quality control data. The EN4 data set incorporates all types of observed profiles of temperature and salinity (if available). The main data source for EN4 is the World Ocean Database 2009. Moreover,three other data compilations, namely, the Arctic Synoptic Basin-wide Oceanography project, Global Temperature and Salinity Profile Program, and global Argo project, are included. The details of the data sources, quality control,bias correction, and objective analysis method adopted in the EN4 data set can be derived from the study of Good et al. (2013).

The EN4 provides global monthly objective analyses of the potential temperature and salinity on a horizontal grid of 1.0˚×1.0˚ from 1900 to the present. There are 42 vertical levels increasing from approximately 5 m to approximately 5350 m, of which 12 levels are below 2000 m with the depth interval increasing from approximately 275 m to approximately 299 m. In the current study, we utilize the objectively analyzed potential temperature and salinity during the period of 1950–2010, which are consistent with the SODA224.

3 Abyssal Circulation in the Pacific Ocean and Comparisons with the Observations

Before using the SODA224 outputs to examine the abyssal circulation in the PS, we need to evaluate its performance in describing the Pacific Ocean abyssal circulation, which is considered to have an important influence on the abyssal circulation in the PS (Uehara and Taira,1990; Kawabe and Fujio, 2010).

3.1 General Features

As an example, Figs.2a–c show the vertical sections of the climatological mean potential temperature, salinity,and density along 199.75˚E in SODA224. Meanwhile,Figs.2d–f show the vertical sections of the climatological mean potential temperature, salinity, and density along 200˚ E in the EN4 data set. Notably, the two data sets are quite consistent, showing nearly the same spatial distributions along the meridional section across the Pacific Ocean, not only in the upper layers but also in the deep layers. The two data sets are also consistent with the Levitus data set analyzed by Ishizaki (1994).

One of the prominent features in Fig.2 is the northward deepening of isothermals, isohalines, and isopycnals from 60˚S to 40˚S, which are associated with the eastward Antarctic Circumpolar Current (ACC; Whitworth and Peterson, 1985). Both the observed and assimilated salinity fields show a low-salinity core in the upper layer and a high-salinity core in the deep layer. These layers correspond to the Antarctic Intermediate Water and Circumpolar Deep Water (CDW), respectively, which are influenced by the high salinity of the North Atlantic Deep Water (Kawabe and Fujio, 2010). According to Kawabe and Fujio(2010), the high-salinity CDW in the Pacific Ocean mostly comprises Antarctic Bottom Water (AABW), which can be clearly observed in Fig.2. From south to north, the AABW, which is modulated by the ocean bathymetry and other processes, dominates below the thermocline, and its thickness generally decreases.

The cold water around 200˚E can be observed near the sea bottom, with the potential temperature of θ < 0.9℃.However, the cold water in both SODA224 and EN4 exists only south of the Line Islands Ridge, which separates the Northeast Pacific Basin (NEPB) from the Central Pacific Basin (CPB) and Southwest Pacific Basin (SWPB)(Kawabe and Fujio, 2010). This means that the cold water with the potential temperature of θ < 0.9℃ originates from the south and proceeds northward in the deep circulation.However, the cold water does not enter the NEPB, at least its northern part. This finding is consistent with previous in situ observations (e.g., Kawabe et al., 2003). These observations indicate that the cold water flows into the CPB through the Samoan Passage at approximately 10˚S but does not penetrate into the northern part, which is blocked by the sill around the Wake Island Passage (WIP).

By integrating vast historical observations, Kawabe and Fujio (2010) indicated that the overall circulation in the Pacific Ocean can be described comprehensively by the components of three layers. North of the midlatitude in the South Pacific Ocean, the three layers correspond to the potential temperature of θ < 1.2℃ at depths greater than approximately 3500 m, 1.2℃ to 2.2℃ at depths of approximately 2000–3500 m, and θ > 2.2℃ at depths less than 2000 m (Kawabe et al., 2003). The depth ranges for the three layers are quite consistent with the data assimilation in SODA224 and objective analyses in EN4 (Fig.2).

Fig.2 Comparisons of the potential temperature (℃; a and d), salinity (b and e), and density (kg m-3; c and f) along 199.75˚E in SODA224 (left panels) and 200˚E in EN4 (right panels).

Fig.3 further compares the depth of and the potential temperature and salinity on the 27.77σθ-isopycnal surface in the Pacific Ocean in SODA224 and EN4. The spatial distributions in the two data sets are quite consistent. Generally, all of the three variables vary significantly in the meridional direction but vary insignificantly in the zonal direction.From south to north,the depth of the 27.77σθisopycnal surface deepens,whereas both the potential temperature and salinity decrease, consistent with the previous observations (e.g., Kaneko et al., 1998; Wijffels et al., 1998; Kato and Kawabe, 2009). However, we also note that the 27.77σθ-isopycnal surface in the PS is deeper than those in the central and eastern Pacific Ocean at the same latitudes. This finding is also consistent with the previous observations (e.g., Uehara and Taira, 1990;Siedler et al., 2004).

Fig.3 Comparisons of the depth (m; a and d), the potential temperature (℃; b and e) and salinity (c and f) on the 27.77σθ-isopycnal surface in SODA224 (left panels) and EN4 (right panels).

Fig.4 shows the zonally averaged meridional velocities in the Pacific Ocean derived from SODA224. The most prominent feature is the existence of a single meridional circulation cell in the deep layer below the average depth of approximately 1750 m. The lower branch flows northward from the ACC region to the North Pacific Ocean in the abyssal layer below approximately 3500 – 4000 m, carrying the bottom waters of Antarctic origin. By contrast,the upper branch flows southward from the North Pacific Ocean to the ACC region below approximately 1500–2000 m, carrying the NPDW, which is transformed from the bottom waters (Kawabe and Fujio, 2010). Overall, the SODA224 reasonably reproduces the single meridional circulation cell in the deep layer, which is confirmed by the observations, theoretical studies, and model simulations presented in the literature (e.g., Ishizaki, 1994).

Fig.4 Zonally averaged meridional velocities (m s-1) in the Pacific Ocean derived from SODA224.

3.2 Abyssal Circulation

In this subsection, we examine the abyssal circulation at different depth layers in SODA224 and its comparisons with the historical observations in detail, focusing on deep currents in the CPB, NWPB, and adjacent small basins,which are considered to be important for those in the PS(e.g., Kawabe and Fujio, 2010). Following Ishizaki (1994),the abyssal layer deeper than 1750 m is divided into three sublayers, namely, the bottom layer (below 4250 m), lower deep layer (3250 – 4250 m), and upper deep layer (1750 –3250 m).

3.2.1 Bottom layer

Figs.5a and 5b present the horizontal velocity fields in the Pacific Ocean at 5125 and 4375 m, respectively. The northward-flowing current can be clearly observed in the SWPB, carrying the cold and saline Lower Circumpolar Deep Water (LCDW; Roden, 2000). As shown in Fig.5b,most of the water flows into the Samoan Basin and enters the CPB mainly through the Samoan Passage, with a minor part passing through a gap in the Robbie Ridge(Ishizaki, 1994; Rudnick, 1997). The left part of the water turns northeastward into the Penrhyn Basin and flows mainly along the eastern flank of the Manihiki Plateau(Ishizaki, 1994), eventually entering the NEPB. In the CPB,the main body of the water flows northward as a western boundary current along the east side of the Marshall- Gilbert Islands (Ishizaki, 1994; Kawabe et al., 2003). However, at shallower depths of the bottom layer, part of the water enters the Melanesian Basin (MB) through gaps in the Gilbert and Ellice Islands, eventually flowing into the Eastern Mariana Basin (EMB; Ishizaki, 1994; Kawabe et al., 2003).

Fig.5 Horizontal velocity fields (cm s-1) at (a) 5125 m and(b) 4375 m. In (a) and (b), the thin gray lines show the 5125 and 4375 m isobaths, respectively.

The northward western boundary current in the CPB bifurcates at approximately 10˚N, south of the Wake-Necker Seamounts. The eastern branch flows eastward along the northern margin of the CPB and eventually enters the NEPB predominantly through the Clarion Passage,consistent with both the numerical simulation (Ishizaki,1994) and in situ observations (Kato and Kawabe, 2009).The western branch flows northwestward between the Wake-Necker Seamounts and the Marshall Seamounts and further bifurcates into two branches around 169˚E,16˚N, south of the WIP, flowing northward and westward,respectively. The main branch flows northward into the NWPB through the WIP (Ishizaki, 1994; Kawabe et al.,2003). After entering the basin, the main branch takes an anticyclonic route (Yanagimoto and Kawabe, 2007). One part of it enters the dead end between the Hawaiian Ridge and the Wake-Necker Seamounts (Ishizaki, 1994). The other part turns northeastward and diverges there. A small body of the water flows into the NEPB through the Midway Island Passage between the Hess Rise and the Hawaiian Ridge. This finding is consistent with the numerical simulation of Ishizaki (1994) and speculations based on observations of Roden (2000) but still needs to be further confirmed with direct current observations in the future(Kawabe et al., 2003; Yanagimoto and Kawabe, 2007). A large body of the water flows along a cyclonic pathway southeast of the Shatsky Rise and enters the regions east of the Izu- Ogasawara Ridge and Japan by passing southwest of the Shatsky Rise, which is consistent with the observations (Kawabe et al., 2003; Yanagimoto and Kawabe,2007).

The westward branch south of the WIP flows into the EMB, with a minor part turning northward into the NWPB through passages to the west of the WIP in the Mid- Pacific Seamounts (Yanagimoto and Kawabe, 2007). This finding is slightly different from the results derived from the conductivity-temperature-depth (CTD)-oxygen observations of Kawabe et al. (2003), who indicated that the westward branch entirely enters the NWPB. However, the observations of Roden (2000) and Kawabe et al. (2003;their Fig.9a) indicated that the meridional geostrophic transport across the zonal section at 18˚20΄N is mostly northward, nearly consistent with the finding that this branch flows slightly toward the west by north in the EMB. In the northern EMB, the westward branch meets the water coming from the MB through gaps in the Marshall Seamounts and from the southern EMB through the gap between the Marshall Seamounts and the Magellan Seamounts. Then, the confluent water flows northwestward/northward into the NWPB through the passage between the Ogasawara Plateau and the Mid-Pacific Seamounts, which is consistent with the in situ observations(Wijffels et al., 1998; Kawabe et al., 2003; Yanagimoto and Kawabe, 2007).

The circulation just southeast of Japan shows a pattern with multiple cyclones and anticyclones (Fig.6), which can also be derived from the sparse direct current observations (e.g., Fujio et al., 2000). Then, the water flows northeastward mainly through two routes separated by the Hokkaido Rise. One route follows the Kuril Trench and has been confirmed by the observations (e.g., Kawabe and Fujio, 2010). The other route is between the Hokkaido Rise and the Shatsky Rise. These two flows first merge together in the northeast part of the NWPB and then enter the NEPB through two passages located at the northern end of the Emperor Seamounts and in the Main Gap. The water passing through the northern end of the Emperor Seamounts follows the Aleutian Trench in the NEPB(Warren and Owens, 1985; Roden, 2000; Kawabe and Fujio, 2010). The passage through the Main Gap appears in the numerical simulation of Ishizaki (1994) but not in the schematic of the deep currents of Kawabe and Fujio(2010) derived by integrating the previous observations.As shown in Fig.6b, aside from the water originating from the north of the Shatsky Rise, the water originating from the south of the Shatsky Rise also flows through the Main Gap into the NEPB. This finding has been confirmed by previous observations (e.g., Kawabe and Fujio,2010).

Overall, the bottom water of southern origin enters the NEPB along various paths and converges there to form the NPDW, which is consistent with the findings of previous studies (e.g., Ishizaki, 1994; Kawabe and Fujio,2010).

Fig.6 Same as Fig.5 but for the Northwest Pacific Basin.

3.2.2 Lower deep layer

Figs.7a and 7b show the horizontal velocity fields in the Pacific Ocean at 3875 and 3375 m, respectively. The circulation pattern in the lower deep layer is quite different from that in the bottom layer. At 3875 m, the northward-flowing water first passes through the gap in the Robbie Ridge from the SWPB into the CPB and then proceeds northward into the EMB in the North Pacific Ocean through the MB. At 3375 m, no influx from the South Pacific Ocean to the North Pacific Ocean is observed. Instead, the western boundary region is dominated by southward-flowing water from the North Pacific Ocean to the South Pacific Ocean. This finding is nearly consistent with the coarse-resolution model simulation of Ishizaki (1994).

Fig.7 Horizontal velocity fields (cm s-1) at (a) 3875 m and (b)3375 m. In (a) and (b), the thin gray lines show the 3875 and 3375 m isobaths, respectively.

The entire North Pacific Ocean, except for the PS, is dominated by an anticyclonic gyre, with currents in the western and northern margins being similar to those in the bottom layer. However, in the low-latitude regions between the equator and approximately 18˚N, the dominant flow is westward. The westward flow in the NEPB is blocked by the Line Islands and enters the CPB mainly through the Horizon Passage, Clarion Passage, and southern end of the Line Islands (Fig.7; Firing et al., 1998). At 3875 m in the CPB, the westward currents that pass through the Horizon and Clarion Passages flow northward and circulate in the large-scale anticyclonic gyre. The westward current passing around the southern end of the Line Islands is consistent with the previous direct current observations (e.g., Firing et al., 1998) and bifurcates at approximately 177˚E, 1˚S just east of the Gilbert Ridge.One branch flows northward as a western boundary current in the CPB. The other branch flows southward and eventually enters the SWPB through the Samoan Passage,consistent with the direct current observations obtained by Rudnick (1997). At 3375 m, the bifurcation point in the CPB shifts northward to approximately 10˚N. The westward current is strongest around 12˚N, consistent with the geostrophic calculation of Kato and Kawabe (2009). The deep water entering the EMB from the CPB splits into two branches in the west corner of the EMB. Most of the water flows southward as a strong western boundary current through the EMB and MB and eventually into the South Pacific Ocean, and this phenomenon has been observed by Ishizaki et al. (2012). The rest of the water enters the WMB through the YMJ (Siedler et al., 2004).

3.2.3 Upper deep layer

Fig.8 shows the horizontal velocity fields in the Pacific Ocean at 2625 m, which is representative of the upper deep layer. The circulation pattern in the upper deep layer is nearly similar to that in the lower deep layer. The entire North Pacific Ocean is dominated by a single anticyclonic gyre, with the southern boundary located at approximately 16˚–19˚N (Ishizaki, 1994). However, the circulation pattern is complex, which is also consistent with the previous observations. For example, in the subarctic Pacific Ocean, a cyclonic circulation structure exists above the Aleutian Trench, which can also be observed in the lower deep layer (Fig.7). The cyclonic circulation structure consists of two relatively strong zonal currents, i.e., one flowing westward just along the Aleutian Island Arc and the other flowing eastward above the Aleutian Rise and Aleutian Trench just to the south of the westward flow.These findings are consistent with the results derived by Warren and Owens (1985) by analyzing deep current meter records from five moorings along 175˚W and CTD sections along 165˚W, 175˚W, and 175˚E in the subpolar region of the North Pacific Ocean.

In the low-latitude regions, the most prominent feature is the existence of the broad westward-flowing water stemming from the NEPB (Wijffels et al., 1998), which mainly consists of two currents located at approximately 6˚N and 12˚N. The current located at approximately 6˚N is blocked by the northern end of the Solomon Rise and turns southward around 158˚E. By contrast, the current located at 12˚N enters the PS through the YMJ, continues flowing westward, and bifurcates off the Philippines into two western boundary currents, i.e., flowing southward and northward. The southward- flowing current is stronger than the northward-flowing current and turns eastward along the equator around 130˚E. This current is called the Circum-Philippine Sea Deep Current (CPDC), which has been discussed and compared with the observations by Ishizaki (1994). The southward western boundary current in the PS has been well observed by Wang et al. (2017)using the Aanderaa Data Instruments SeaGuard Platform mounted at 2500 m. The eastward current enters the MB through gaps in the northern Solomon Rise, joins the southward-flowing current, and crosses the equator as a western boundary current. This finding is consistent with the meridional circulation pattern illustrated in Fig.4,which shows an outflux from the North Pacific Ocean to the South Pacific Ocean.

Fig.8 Horizontal velocity fields (cm s-1) at 2625 m. The thin gray lines show the 2625 m isobaths.

4 Abyssal Circulation in the PS

In this section, we examine the climatological mean abyssal circulation deeper than 2000 m in the PS with SODA-224. The abyssal circulation is strongly influenced by the complex topography. On the basis of the characteristics of the horizontal circulation, the abyssal layer in the PS can be divided into three sublayers, which are the upper deep layer (2000–3000 m), lower deep layer (3000– 4500 m), and bottom layer (below 4500 m).

4.1 Upper Deep Layer

Figs.9a and 9b present the horizontal velocity fields in the PS at 2125 and 2625 m, respectively. The most prominent feature in this layer is the existence of the strong westward current along 11˚–14˚N, which originates from the NEPB carrying the NPDW (Wijffels et al., 1998) and enters the PS through the YMJ (Fig.8; Kaneko et al., 2001).As discussed in Subsection 3.2.3, when approaching the Philippines, this strong westward current bifurcates into two western boundary currents, i.e., flowing southward and northward. The bifurcation latitude is approximately 13˚N, similar to that of the NEC at sea surface (Hu et al.,2015). As stated previously, the southward-flowing current is part of the CPDC and eventually flows out of the North Pacific Ocean into the South Pacific Ocean (Ishizaki, 1994). The northward-flowing current separates off the coast toward the interior of the PS around 20˚–21˚N and turns into an eastward current. At 2125 m, the northward-flowing current seems stronger than the southwardflowing current, and the resultant eastward current flows out of the PS through gaps in the Ogasawara Ridge. This finding is consistent with the flow patterns derived from the observations (Wijffels et al., 1998; Kaneko et al., 2001).In the northeast PS, inflow water flows from the NWPB to the PS through the gaps between the Izu Ridge and the Ogasawara Ridge, consistent with the previous observations of Kaneko et al. (1998, 2001). In the northern ShB, a significant anticyclonic gyre that extends downward to the ocean bottom exists (see also Figs.10 and 11). The anticyclonic gyre above 2000 m is consistent with the Argo float observations. However, the anticyclonic gyre below 2500 m is different from the observations of Nagano et al.(2018), which exhibit westward/southwestward-flowing current that extends downward to the ocean bottom.

Fig.9 Horizontal velocity fields (cm s-1) at 2125 m (a) and 2625 m (b) in the PS. In (a) and (b), the thin gray lines show the 2125 and 2625 m isobaths, respectively. The red point indicates the observation site of Wang et al. (2017) at 127˚3΄E, 8˚N.

At 2625 m, the eastward-flowing current around 20˚–21˚N is largely blocked by the Ogasawara Ridge, which quite possibly results in the northward-flowing current becoming weaker than the southward-flowing current off the Philippines, as shown in Fig.9b. The strong southward current has been well observed by Wang et al. (2017)using the Aanderaa Data Instruments SeaGuard Platform mounted at 2500 m at 127˚3΄E, 8˚N (red point in Fig.9b)from December 2012 to October 2013. The velocity magnitude in SODA224 is comparable to the observations.However, Wang et al. (2017) determined that the observed mean zonal velocity is approximately 1.59 cm s-1,approximately two times larger than the mean meridional velocity, which is approximately -0.99 cm s-1. One possible reason for the discrepancy is that the observation site is located far from the strong southward current. The other possible reason for the discrepancy is that the southward current has a strong interannual variation. The exact reasons should be further explored with more observations in the future.

The significant westward current around 11˚–14˚N does not reach the ocean bottom in the PS. Instead, this current is blocked by the Palau Ridge around 134.75˚E.To illustrate this clearly, Fig.10 shows the climatological mean zonal velocity across 134.75˚E in SODA224. As shown in the figure, the westward current around 11˚–14˚N seems to extend downward to the deep layers of the strong westward NEC, which is mainly located in the upper layer and has its largest westward velocity near the sea surface. The deepest depth of the westward current is approximately 3400 m just above the Palau Ridge. Actually, multiple zonal jets with alternately reversing directions exist in this layer in the PS (Fig.9). As shown in Fig.10, these zonal jets are downward extensions of those in the surface/subsurface layers, which have been well confirmed by the observations (e.g., Qiu et al., 2015).

Fig.10 Zonal velocity (cm s-1) across 134.75˚E above (a)and below (b) 1875 m.

4.2 Lower Deep Layer

In the lower deep layer, the topography is more complex, which considerably influences the horizontal circulation (Fig.11). Pacific water still flows into the PS through the YMJ (Siedler et al., 2004), which significantly decreases with the increase in depth. The incoming water flows northward along the western flank of the Mariana Ridge in the WMB, which is consistent with the schematic LCDW flow pattern in the WMB derived by Siedler et al. (2004) from the direct current observations and by Kawabe (1993) and Kaneko et al. (1998) from the analysis of the distributions of water properties. Then, the incoming water turns westward at approximately 18˚N.This is different from that in the upper deep layer (Fig.9).Finally, the major part of the westward- flowing water around 18˚N enters the PB (Uehara and Taira, 1990; Kaneko et al., 1998) through gaps in the Kyushu-Palau Ridge.The minor part turns southward along the east flank of the Kyushu-Palau Ridge, constituting a weak cyclonic circulation in the WMB. This finding is consistent with the geostrophic current calculated by Uehara and Taira (1990)using the in situ hydrographic observations at 12˚– 13˚N,which shows the northward current in the broad regions in the east and the southward current in the western boundary regions of the WMB.

The water entering the PB flows westward around 22˚N,turns southwestward, and merges into the strong southward western boundary current in the southern PB. We note that the current southeast of Okinawa and Taiwan Islands at 3625 m flows southwestward–southward parallel to the local bottom contour, which is quite consistent with the observations of Chaen et al. (1993). The northern PS, north of approximately 22˚N, is dominated by a weak anticyclonic circulation flowing around the northern PB and ShB. By contrast, the southern PS, south of approximately 12˚–13˚N, shows a large-scale cyclonic circulation.We compare the result in SODA224 with the geostrophic calculation in the PB of Uehara and Taira (1990), which are quite consistent with each other and show broad southward meridional current across 12˚–13˚N in the western part of the basin. The significant southward western boundary current just off the Mindanao Island observed in the upper deep layer also exists in this layer. However, the main body of this current circulates in the southern PB because of the blockage of the Yap Ridge and only a small part enters the Western Caroline Basin.

Fig.11 Horizontal velocity fields (cm s-1) at 3125 m (a) and 3625 m (b) in the PS. In (a) and (b), the thin gray lines show the 3125 and 3625 m isobaths, respectively.

4.3 Bottom Layer

Figs.12a and 12b present the horizontal velocity fields in the PS at 4625 and 5125 m, respectively. As shown in the figure, no bottom water directly flows into the PS from the Pacific Ocean. The entire PB can be divided into three subbasins, namely, the south, northeast, and northwest PB (Fig.12b). The northeast and northwest PB are separated by ridges along 128˚E and separated from the south PB by ridges along 16˚N. Notably, both the northeast and northwest PB are dominated by a single cyclonic gyre, whereas the south PB is dominated by a single anticyclonic gyre.

By using a Doppler Volume Sampler and the Aanderaa Data Instruments SeaGuard Platform, Wang et al. (2017)observed that the mean current at approximately 5600 m at 127˚3΄E, 8˚N east of the Mindanao Island but west of the Philippine Trench (red point in Fig.12b) was southward during 2011–2013. It is different from that in SODA224,as illustrated in Fig.12b, which shows a northward- flowing current east of the Mindanao Island. The discrepancy can be attributed to two reasons. The first reason is the different time periods in the observations and SODA224.We note that though the mean meridional velocities in observations and SODA224 are with opposite directions,they show similar magnitudes and seasonal variations.Given the configurations of the horizontal model resolution and the depth levels, no output can be obtained from the observation site, time, and depth in SODA224. In comparison with the observations, Fig.13 presents the assimilated monthly time series of the meridional velocity from January 2001 to December 2010 at 5375 m at 127.25˚ E, 7.75˚N, which is the nearest point to the observation site of Wang et al. (2017). The annual mean assimilated meridional velocity is approximately 1.12 cm s-1, similar in magnitude to that observed by Wang et al. (2017), which is approximately -1.61 cm s-1. From 2011 to 2013, the observed abyssal current at 5600 m shows a significant seasonal variation, i.e., southward in late fall and winter,whereas northward in summer and early fall. This is also the situation in most years from 2001 to 2010 in SODA-224. In addition to the seasonal variation, both the directly observed and SODA224 assimilated abyssal meridional velocities also exhibit significant interannual variation.For example, the observed abyssal northward current in August 2013 is stronger than that in August 2011 (Wang et al., 2017). Therefore, the abyssal circulation in the PS quite possibly undergoes significant low- frequency variations on multiple time scales, which could result in mean currents with different directions in different time periods.

Fig.12 Horizontal velocity fields (cm s-1) at 4625 m (a) and 5125 m (b) in the PS. In (a) and (b), the thin gray lines show the 4625 and 5125 m isobaths, respectively. In (b), the color shading denotes the bathymetry (m) below 4000 m.

Fig.13 Monthly time series of the meridional velocity(cm s-1) at 127.25˚E, 7.75˚N in SODA224.

The second reason is the possible existence of a cyclonic sense of circulation over the Philippine Trench. The cyclonic circulation structure has also been observed above the Japan Trench (Mitsuzawa and Holloway, 1998),Mariana Trench (Huang et al., 2018), Aleutian Trench(Warren and Owens, 1985), and other trenches, as documented by Johnson (1998). Meanwhile, currents flowing along isobaths with upward slope on the right side have also been observed over other trenches (Nagano et al.,2013). According to Johnson (1998), as the trench bathymetry unambiguously results in a region of closed planetary geostrophic contours, a significant cyclonic circulation around these contours would be driven by upwelling within the trench through vortex stretching (Kawase and Straub,1991; Kawase, 1993). The cyclonic circulation over the Philippine Trench results in southward and northward currents on the western and eastern slopes of the trench,respectively. The southward current on the western slope of the Philippine Trench is consistent with the observations of Wang et al. (2017). We note that the abyssal northward current off the Mindanao Island in SODA224 is not part of the cyclonic circulation, as the Philippine Trench is not well represented in the SODA224 model.The abyssal northward current is part of the single anticyclonic gyre dominating the south PB, as shown in Fig.12.

Next, we discuss the mechanisms responsible for the cyclonic/anticyclonic gyres in the three subbasins of the entire PB. The bottom water circulation in a concave abyssal basin has been discussed in the literature, (e.g., Kawase and Straub, 1991; Kawase, 1993). The bathymetry of the concave abyssal basin with weak stratification produces a region of closed planetary potential vorticity f/H contours(Fig.14a), which approximates closed geostrophic contours. Here, f is the local Coriolis parameter and H is the thickness of the bottom layer (below 4500 m). Then, the upwelling/downwelling within the basin drives the cyclonic/anticyclonic gyre through vortex stretching/ shrinking (Kawase and Straub, 1991; Kawase, 1993). As expected in SODA224, upwelling dominates in the northeast and northwest PB, whereas downwelling dominates in the south PB (Fig.14a). Moreover, Fig.14b shows the potential density at a pressure of 4000 dbar (σ4) at 5125 m, which could provide insight into the abyssal geostrophic currents. In the entire PB, σ4has two maxima located in the northwest PB and southwest of the south PB. We note that, in the south PB, σ4is larger in the west than in the east, consistent with the observations of Uehara and Taira (1990) at 12˚N. According to the geostrophic balance, the local σ4maxima induce anticyclonic geostrophic circulation around them. Therefore, in the northeast and northwest PB, it is quite possibly that upwelling predominantly drives the cyclonic gyres. However, in the south PB, both the downwelling and horizontal pressure gradient force contribute to the formation of the anticyclonic gyre.

Fig.14 (a) Potential vorticity (×10-8 m-1 s-1) at the bottom layer in the PB. Regions with upward vertical velocity at 4625 m are indicated by gray points. (b) Potential density (σ4, kg m-3) at a pressure of 4000 dbar at 5125 m in the PB.

5 Discussion and Summary

In the current study, the abyssal circulation in the PS is examined using the outputs from the global ocean data assimilation product SODA224. To evaluate the performance of SODA224, we extensively compare the assimilated deep-water circulation in the North Pacific Ocean with sparse in situ observations. The comparisons show quite good consistency, indicating that the results from SODA224 can be used to describe the deep-water circulation in the PS.

Overall, the abyssal circulation in the PS is strongly influenced by the complex topography. On the basis of the characteristics of the horizontal circulation, the abyssal layer below 2000 m in the PS is divided into three sublayers, namely, the upper deep layer, lower deep layer,and bottom layer. The main characteristics of the circulation in the three layers are summarized as follows:

1) The most prominent feature in the upper deep layer is the existence of the strong westward current along 11˚–14˚N, which originates from the NEPB carrying the NPDW and enters the PS through the YMJ. After reaching the Philippines, the strong westward current bifurcates into two western boundary currents, i.e., flowing southward and northward. The southward-flowing current is part of the CPDC and eventually flows out of the North Pacific Ocean into the South Pacific Ocean, transporting deep water from the northern hemisphere to the southern hemisphere. The northward-flowing current separates off the coast toward the interior of the PS around 20˚–21˚N and turns into an eastward current. The significant westward current around 11˚–14˚N is blocked by the Palau Ridge around 134.75˚E below about 3400 m.

2) Pacific water still flows into the PS through the YMJ in the lower deep layer, which decreases with the increase in depth. Different from that in the upper deep layer, the incoming water first flows northward along the western flank of the Mariana Ridge in the WMB and then turns westward at approximately 18˚N. Finally, the major part of the westward-flowing water around 18˚N enters the PB,whereas the minor part turns southward along the east flank of the Kyushu-Palau Ridge, constituting a weak cyclonic circulation in the WMB. The water entering the PB flows westward around 22˚N, turns southwestward,and merges into the strong southward western boundary current in the southern PB. The northern PS, north of approximately 22˚N, is dominated by a weak anticyclonic circulation flowing around the northern PB and ShB. By contrast, the southern PS, south of approximately 12˚–13˚N, shows a large-scale cyclonic circulation. The significant southward western boundary current just off the Mindanao Island as observed in the upper deep layer also exists in this layer. However, its main body circulates in the southern PB because of the blockage of the Yap Ridge.

3) In the bottom layer, no bottom water flows directly from the Pacific Ocean into the PS. Both the northeast and northwest PB show a single cyclonic gyre, whereas the south PB shows a single anticyclonic gyre. Further analysis indicates that the cyclonic gyres in the northeast and northwest PB are quite possibly driven by the upwelling within the basins through vortex stretching, whereas the anticyclonic gyre in the south PB is induced by both the downwelling and horizontal pressure gradient force. Moreover, comparisons with the observations of Wang et al.(2017) indicate the possible existence of a cyclonic sense of circulation over the Philippine Trench.

The assimilated abyssal circulation in the PS is generally consistent with most historical sparse observations.However, differences between data assimilation and previous observations can be detected. A significant difference can be observed in the ShB, where SODA224 shows an anticyclonic gyre throughout the three layers (Fig.10),whereas some observations (e.g., Nagano et al., 2018)indicate a westward/southwestward-flowing current that extends downward to the ocean bottom below 2500 m. As discussed previously, the differences between the SODA-224 results and the observations can be attributed to three reasons. The first reason is the different time periods.Both observations and SODA224 support the finding that the abyssal circulation undergoes significant low-frequency variations, which may result in different circulation patterns during different time periods. The second reason is the horizontal resolution used in preparing SODA224,which is not fine enough to resolve the complex topography. The third reason is the possible errors contained in SODA224.

The current study provides the implications for future observations, which are needed to confirm and further investigate the spatial structure and low-frequency variations of the abyssal circulation in the PS. Moreover, longterm warming and freshening have been observed in global abyssal waters (Sloyan et al., 2013). As abyssal warming could contribute largely to the increase in the global and regional sea levels (Purkey and Johnson, 2010), one immediate question is whether there is a long-term trend in the abyssal circulation and water mass in the PS. Therefore, the temporospatial variations on multiple scales and the underlying dynamics of the abyssal circulation in the PS and their effects on the regional and global climate variations should be further investigated with fine observations and high-resolution numerical simulations in the future.

Acknowledgements

The present study is sponsored by the Aoshan Science and Technology Innovation Project (No. 2016ASKJ12),the Open Fund of the Key Laboratory of Ocean Circulation and Waves, Chinese Academy of Sciences (No. KLO CW1503), and the National Natural Science Foundation of China (Nos. 41606107, 41506008, 41776012, 41476002).We sincerely thank two anonymous reviewers and the editor for their constructive comments on the manuscript.