1 INTRODUCTION

The Philippine Sea connects the tropical and subtropical oceans and is a key region of the Western Pacific with respect to ocean heat and mass distribution (Hu et al., 2015). The Philippine seafloor topography is divided into several basins by the ridges; it consists of the Philippine Basin, the Shikoku Basin, and the West Mariana Basin (Fig. 1). The Kyushu-Palau Ridge is the boundary between the Eastern Philippine Sea and Western Philippine Sea, and the Izu-Ogasawara, Mariana, and Yap Ridges constitute the Eastern boundary of the Philippine deep sea (Fig. 1). Deep water cannot be formed locally because the Philippine Sea is located at low latitudes (Stommel and Arons, 1960). The Philippine deep water comes from the Western Pacific through several channels in the deep layers (Kubota and Ono, 1992).

Potential temperature (θ) interfaces are used to distinguish the boundaries between deep water masses in the Philippine deep sea (Johnson and Toole, 1993; Kawabe and Fujio, 2010), which are vertically divided into two layers, i.e., the lower and upper deep layers. The Lower Circumpolar Deep Water (LCDW) flows from the Yap-Mariana Junction into the lower deep layer of the Philippine deep sea with θ≤1.2 ℃ (Mantyla and Reid, 1983). Above the LCDW, i.e., in the upper deep layer of the Philippine deep sea, both the Upper Circumpolar Deep Water (UCDW) and North Pacific Deep Water (NPDW) have a similar temperature characteristic with a range of 1.2≤θ≤2.0 ℃ (Johnson and Toole, 1993). The most significant difference between UCDW and NPDW is that the latter is characterized by decreased dissolved oxygen and increased dissolved silicate (Reid, 1997; Kawabe et al., 2009). The UCDW flows into the southern Philippine deep sea via the Eastern Mariana and Caroline Basins (Kawabe, 1993; Kawabe et al., 2003; Kawabe and Fujio, 2010). The NPDW forms in the Northeast Pacific Basin (Talley and Joyce, 1992), where the LCDW accumulates dissolved silicate by upwelling and diffusion processes and is subsequently transformed into NPDW (Talley and Joyce, 1992; Johnson et al., 2006).

Dissolved silicate is a key nutrient element in the ocean and its cycle is associated with the atmospheric carbon dioxide (Pondaven et al., 2000). There is small amount of dissolved silicate consumption in Pacific deep waters; hence, dissolved silicate is used as a non-conservative tracer for identifying NPDW pathway (Talley and Joyce, 1992). NPDW flows into the Philippine Sea through the caves of the Izu-Ogasawara and Mariana Ridges (Talley and Joyce, 1992; Kawabe, 1993; Reid, 1997; Kaneko et al., 2001) and returns to the Pacific Ocean via anticyclonic circulation. Consequently, two maximum dissolved silicate cores were found along 137°E transect (Kaneko et al., 1998).

The spatial variation of deep water characteristics in the Philippine Sea has not been thoroughly investigated (Huang et al., 2018; Ma et al., 2019; Tian et al., 2021). Earlier studies have shown that LCDW at the Challenger Deep exhibits significant seasonal variability (Huang et al., 2018) and that the potential temperature fluctuates with a period of 30–90 days (Ma et al., 2019). Wang et al. (2020, 2021) have found that the structure of deep layer circulation at the Yap-Mariana Junction exists seasonal variability. It was found that the Philippine deep sea water below 2000-m depth has been warming by (0.33±0.19)×10^-3 ℃/a over the past 30 years (Tian et al., 2021). However, the interannual variation in water mass characteristics in the upper deep layer is still unknown.

Based on repeated observations along the aforementioned 137°E transect, we investigated the interannual dissolved silicate variability in the upper deep layer of the Eastern Philippine Sea. In the following parts, we present the data In Section 2, present the distribution of the climatological dissolved oxygen and silicate in Section 3, display the interannual dissolved silicate variability in Section 4, discuss the underlying mechanism of the interannual dissolved silicate variability in Section 5, and show conclusions in Section 6.

2 DATA

The World Ocean Database (WOD18) is a collection of observations used in this study, including Ocean Station Data (OSD), conductivity-temperature-depth (CTD), and Mechanical Bathy Thermograph (MBT) data (Boyer et al., 2018). Several historical observations, including potential temperature, dissolved oxygen, and silicate measurements, have been collected along the 137°E transect from WOD18. Because the difference in dissolved silicate between the NPDW and UCDW is more significant than that in dissolved oxygen along the 137°E transect (Fig. 2), we only considered dissolved silicate to investigating the interannual water-mass variability in the upper deep layer of the Philippine Sea. Considering that most observations were conducted in winter, only winter observations was used in this study to avoid the influence of seasonal variability. Thus, we obtained the historical dissolved silicate distribution along the 137°E transect (i.e., 10°N–30°N) for 1995, 1996, 1997, 2000, 2001, 2002, 2004, 2005, 2006, and 2007.

The Climatological temperature, dissolved oxygen, and silicate were derived from the World Ocean Atlas 2018 (WOA18) data (Garcia et al., 2019). The gridded potential temperature product has a 1×1/4° horizontal resolution and 102 vertical levels. The dissolved oxygen and silicate products have a horizontal resolution of 1×1° (Boyer et al., 2019). There are 36 levels below the depth of 2 000 m with vertical resolution of 100 m (https://www.ncei.noaa. gov/access/world-ocean-atlas-2018/).

To investigate the possible mechanism of the interannual dissolved silicate variability, we used monthly velocity and potential temperature data from the Ocean Reanalysis System 4 product (hereinafter ORAS4). ORAS4 has a 1° horizontal resolution and an approximately 290-m vertical resolution below 2 000-m depth, covering the 1958–2017 period (Mogensen et al., 2012; Balmaseda et al., 2013). ORAS4 data can be download from the website of https://www.cen.uni-hamburg.de/icdc/data/ocean/easy- init-ocean/ecmwf-ocean-reanalysis-system-4-oras4.html.

3 DISTRIBUTIONS OF CLIMATOLOGICAL DISSOLVED SILICATE ALONG THE 137°E TRANSECT

Figure 1 shows the WOA18 climatological annual mean results of dissolved oxygen and silicate in the upper deep layer (i.e., vertically averaged between the 1.2- and 2.0-℃ isotherms) of the Philippine Sea. Water with relatively increased dissolved oxygen comes from the Eastern Mariana and Western Caroline Basins in the south of Philippine Sea. In the Eastern Philippine Sea, there is a meridional band with high dissolved oxygen concentration, and this band extends to the north to almost 24°N (Fig. 1). On the west side of the Kyushu Palau Ridge (between 18°N–20°N, 130°E–136°E), the dissolved oxygen concentration is higher than that in the Eastern Philippine Sea. Given that NPDW with low dissolved oxygen flows westward into the Eastern Philippine Sea from the Western Pacific at the south of 15°N and north of 25°N, the dissolved oxygen in the middle of the Western Philippine Sea is higher than both sides (Kaneko et al., 1998, 2001). Dissolved oxygen in the Shikoku Basin is less than that in the western Marina Basin. In the western Philippine Sea, dissolved oxygen is relatively high in the south, however, the concentration reduced between 12°N and 16°N along the western boundary. Dissolved silicate decreases from east to west (Fig. 1), indicating that NPDW with increased dissolved silicate enters the Eastern Philippine Sea across the Izu-Ogasawara Ridge. The increased dissolved silicate in the western boundary shown in Fig. 1 may be attributed to LCDW upwelling and NPDW production (Kawabe and Fujio, 2010).

The distribution of dissolved oxygen exhibits less horizontal difference than that of dissolved silicate in the upper deep layer along 137°E transect (Fig. 2a). There are two maximum dissolved silicate cores in the upper deep layer (Fig. 2b): the southern dissolved silicate core (i.e., dissolved silicate concentration ≥144 μmol/g) lies between 2 500 and 3 000 dbar, at approximately 15°N, while the northern core (i.e., dissolved silicate concentration ≥145 μmol/g) lies north of 30°N; The northern core has a higher dissolved silicate concentration than the southern one, while the depth of the northern core is shallower than that of the southern core.

Owing to topographical obstructions, there are two currents, carrying the NPDW with high dissolved silicate concentration and flowing into the Eastern Philippine Sea (Kaneko et al., 1998, 2001). The southern current flows from the Western Pacific through the Yap-Mariana Junction (i.e., south of 15°N), and the northern current passes through the gap between the Izu and Ogasawara Ridges (i.e., north of 25°N). Thus, two cores with increased dissolved silicate were found along the 137°E transect.

4 INTERANNUAL DISSOLVED SILICATE VARIABILITY IN THE EASTERN PHILIPPINE DEEP SEA

The observed dissolved silicate distribution along the 137°E transect in the winters of the selected 10 years (i.e., 1995, 1996, 1997, 2000, 2001, 2002, 2004, 2005, 2006, and 2007; Fig. 3) revealed the variation of the two significant maximum concentrations area, one south and one north of the 137°E transect. The southern core in winter is located at approximately 15°N and approximately 3 000 dbar, while the northern core is located north of 25°N and at approximately 2 000 dbar. In 1995, 1996, 2005, 2006, and 2007 (i.e., years with the increased dissolved silicate), the concentrations at the two dissolved silicate cores were relatively high. In 1995, both the southern and northern maximum dissolved silicate cores had a concentration of 144 μmol/g. The southern (northern) core was located between 14°N and 21°N (north of 29°N) at 2 600–3 200 (2 000–2 200) dbar in 1995. In 1996, the southern (northern) core was located between 13°N and 18°N (north of 27°N) at 2 400–3 000 (2 000–2 600) dbar. In 2005, the southern (northern) core was located between 13°N and 17°N (at 30°N) at 2 800–3 200 (2 000–2 200) dbar. In 2006, the northern maximum dissolved silicate concentration was larger than the southern one (i.e., 143 versus 142 μmol/g, respectively), and the southern (northern) core was located between 12° and 18°N (between 23°N and 30°N) at 2 800–3 200 (2 000–3 000) dbar. In 2007, the southern and northern maximum dissolved silicate cores had similar concentrations with the value of 145 μmol/g, while the southern (northern) core was located between 14° and 16°N (north of 29°N) at 2 800–3 200 (2 000–2 200) dbar. In 1995 and 2007, the spatial range with dissolved silicate concentration more than 142 μmol/g was relatively larger (i.e., between 10°N and 30°N along the 137°E transect) than those in the other three years (i.e., 1996, 2005, and 2006).

In 1997, 2000, 2001, 2002, and 2004 (i.e., years with the decreased dissolved silicate), the concentrations at the two dissolved silicate cores were smaller than 141 μmol/g. In 1997, 2000, and 2004, the concentration at the southern dissolved silicate core was equal to that at the northern core (i.e., 138, 140, and 140 μmol/g, respectively); however, in 2001 and 2002, the southern maximum concentration was higher than the northern one (i.e., > 140 versus < 140 μmol/g and 135 versus 134 μmol/g, respectively). The analysis of historical dissolved silicate distributions revealed the interannual variabilities in the depth of the cores, the maximum concentrations, and the ranges of relatively high concentrations. It is shown that there is an inconsistent variability between the southern and northern cores. We hypothesize that the variation of circulation in the upper deep layer might result in the interannual variability of the dissolved silicate (Kaneko et al., 1998, 2001; Siedler et al., 2004; Zhai and Gu, 2020).

To better understand the dynamics of the interannual variability in dissolved silicate, we performed composite and water-mass analyses. Deep dissolved silicate composites were performed based on the increased (i.e., in 1995, 1996, 2005, 2006, and 2007) and decreased (i.e., in 1997, 2000, 2001, 2002, and 2004) dissolved silicate. The 10-year-average of dissolved silicate shows that the southern (northern) maximum concentration was 141 (142) μmol/g at approximately 15°N (north of 29°N) and approximately 3 000 (2 100) dbar (Fig. 4a). Figure 4b–c shows the positive and negative dissolved silicate composite phases. The positive composite phase showed that both the southern and northern dissolved silicate cores had a concentration value of 143 μmol/g, and the southern (northern) core was located at approximately 15°N (north of 27°N) and at 2 800–3 200 (2 000–2 400) dbar. In the negative composite phase, both the southern and northern dissolved silicate cores presented a concentration value of 138 μmol/g, and the southern (northern) core was located at approximately 15°N (north of 29°N) and 2 600–3 400 (2 000–2 300) dbar.

UCDW and NPDW occupy the upper deep layer in the Eastern Philippine Sea. We performed water-mass analyses to further investigate water-mass differences among the selected 10 years. Figure 5a–b shows the potential temperature-salinity (T-S) and potential temperature-dissolved silicate (T-silicate) characteristics of the positive and negative composites. The T-S scatter diagrams were typically concentrated along the same potential density interface, and there were no significant potential temperature and salinity differences between the positive and negative years (Fig. 5a). The T-silicate characteristics revealed a significant difference between positive and negative abnormal composites, where the T-silicate scatter diagrams were typically more dispersed (Fig. 5b). The dissolved silicate range was 138.8–143.8 μmol/g (131.1–137.9 μmol/g) in the positive (negative) abnormal composite when the potential temperature was between 1.2 and 2 ℃. The mean dissolved silicate concentration of the positive composite was 4.25 μmol/g higher than that in the negative composite. Because there was a significant difference in dissolved silicate between UCDW and NPDW, it was a NPDW indicator for distinguishing these two water masses. According to the interannual variability of the dissolved silicate, there is a competitive relationship between UCDW and NPDW in the upper deep layer (i.e., the UCDW intrusion weakens when more NPDW flows into the Eastern Philippine Sea). In other words, the interannual variability of the dissolved silicate is controlled by the variation of the NPDW flowing into the Eastern Philippine deep sea.

5 DISCUSSION

Given that increased dissolved silicate in the Eastern Philippine Sea is transported by the westward current from Western Pacific, it is necessary to analyze the relationship between zonal velocity and dissolved silicate on an interannual timescale. Kaneko et al. (2001) have estimated that NPDW enters the eastern Philippine Sea south of approximately 15°N and flows out at approximately 16°N–24°N, whereas Zhai and Gu (2020) have indicated that NPDW enters the Philippine Sea through the Yap-Mariana Junction and finally transforms into an eastward current at approximately 20°N–21°N after reaching the western boundary. The possible reason that results in the difference might be that the circulation in upper deep layer in the Western Pacific presents significant interannual variability, which affects in the intensity and pathway of NPDW intrusion.

Due to limited availability of observations, we utilized the ORAS4 reanalysis data to discuss the possible mechanism driving the interannual variability of dissolved silicate. Prior to utilization of ORAS4 data, data quality was evaluated. Because the potential temperature distribution is important for deep circulation, we compared the structure along the 137°E transect derived from ORAS4 and from WOD18 data. The vertical distribution of the 10-year-averaged potential temperature along the 137°E transect from ORAS4 was similar to that from WOD18 (Fig. 6a & c). Both datasets indicated that the 2.0-℃ isotherm was relatively stable and slightly deepens southward, while the 1.2-℃ isotherm deepened rapidly at the north of 20°N and south of 15°N. The 1.2-℃ isotherm variability is determined by the path of the LCDW, which flows into the Eastern Philippine Sea and shifts westward at approximately 18°N (Uehara and Taira, 1990; Siedler et al., 2004). Figure 6b & d gives the temperature differences between positive and negative phases from ORAS4 and WOD18, respectively. Though there were some differences between WOD18 and ORAS4, they had similar structures that the positive/negative potential temperature anomaly was found above/in the upper deep layer. This indicated that the ORAS4 could capture the main signal of the interannual variability in potential temperature, as well as in the deep circulation in the Eastern Philippine Sea.

The most prominent feature of the water mass distribution in the upper deep layer along the 137°E transect is the existence of two dissolved silicate cores (Fig. 4), corresponding to the two westward currents, which originate from the Western Pacific and transport NPDW to the Eastern Philippine Sea (Kaneko et al., 1998, 2001). Thus, the interannual variability of the dissolved silicate should be closely related to the variability of the westward currents. The zonal velocity in winters corresponding to the observed years of the dissolved silicate was used here. Figure 7 shows the climatological zonal velocity in winter and the compositely zonal velocity anomalies in positive and negative phases in the upper deep layer along 137°E. We divided the transect into four regions, namely, A, B, C, and D, referring to the direction of the flow. Regions B and D corresponded to the NPDW pathways, while the eastward current in region C might be a return current of NPDW in region B (Kaneko et al., 2001; Zhai and Gu, 2020). The climatological westward currents in regions B and D along 137°E transect were approximately consistent with the distribution of the dissolved silicate cores (Figs. 4a & 7a). The negative (positive) zonal velocity anomaly in regions B and D, corresponding to the positive (negative) phase, indicated that the stronger (weaker) westward current transported more (less) NPDW into the Eastern Philippine Sea and resulted in the large (small) dissolved silicate concentration. This explained the high dissolved silicate concentration in the years of positive phase along 137°E transect. Therefore, we conclude that the interannual variability of the dissolved silicate is closely related to the interannual variability of the zonal velocity.

Kawabe et al. (2009) found that the UCDW outflow was confined to the southern Ogasawara Plateau in 2004, but extended to its northern part in 2005. Thus, in addition to considering the NPDW effect, UCDW should also contribute to the interannual variability of dissolved silicate. In addition, the upwelled LCDW will mix with the upper deep layer water masses, and then change the dissolved silicate content (Kawabe et al., 2010). Therefore, the interannual variability of LCDW is also a potential factor that influences the dissolved silicate variability. Further studies regarding the UCDW and LCDW distributing to the interannual variability of dissolved silicate are needed in the future.

Previous studies have shown that the interannual variability of the upper water masses and circulation are sensitive to El Niñ o-SouthernOscillation (ENSO) (Guan, 1986; Saiki, 1987; Hanawa and Hoshino, 1988). Thus, the circulation structure, as well as the dissolved silicate in the upper deep layer, might also be influenced by ENSO cycle. Due to the limited availability of observations in the dissolved silicate, it is difficult to reveal the relationship between ENSO and the interannual variabilities of the dissolved silicate in the upper deep layer of the Eastern Philippine Sea. More continuous observations in the upper deep layer are needed in the future.

6 CONCLUSION

Based on repeat hydrographic observations and the ORAS4 reanalysis data along a 137°E transect, we investigated the interannual variability of dissolved silicate in the upper deep layer in the Eastern Philippine Sea. Two maximum dissolved silicate cores appear at ~15°N and north of 30°N in the result of the climatological annual mean (WOA18), the dissolved silicate concentrations in the southern and northern cores are larger than 144 μmol/g and 145 μmol/g, respectively. The dissolved silicate concentration was high in the winters of 1995, 1996, 2005, 2006, and 2007 and low in the winters of 1997, 2000, 2001, 2002, and 2004. The interannual variability of the dissolved silicate was reflected in both the locations and concentrations of the northern and southern dissolved silicate cores. Through conducting the potential temperature-silicate water mass analyses, we found that the dissolved silicate concentration had pronounced differences among these years, which reached to almost 4.25 μmol/g between the positive and negative phases.

The distribution of the climatological westward velocity is consistent with the locations of the dissolved silicate cores in the upper deep layer along 137°E. Composite analysis indicated that the interannual variability of the zonal current modulated the variability of the dissolved silicate in the upper deep layer of the Eastern Philippine Sea. In years of the positive/negative phase, the westward current in the ranges of 15°N–20°N and of the north of 27°N strengthened/weakened, which transported more/ less NPDW into the Eastern Philippine Sea, thus eading to a large/small value of the dissolved silicate concentration.

7 DATA AVAILABILITY STATEMENT

The datasets analyzed during the current study are available from the corresponding authors upon reasonable request.

8 ACKNOWLEDGMENT

We are grateful to all the people working for the cruise observations along the 137°E transect in the Philippine Sea.