The Northwest Pacific, from the subarctic to the subtropical region, interacts with the Sea of Okhotsk, Sea of Japan, East China Sea, South China Sea, and the Bering Sea.

The Kuroshio Current and Oyashio Current are the most important currents in the Northwest Pacific ( Figure 1 ). The Kuroshio originates from the western boundary current to the east of the Philippine Islands, and is characterized by high temperature, high salinity oligotrophic water (Nitani, 1972). It flows northward along Taiwan Island, passes through the East China Sea continental shelf, then flows into the Pacific Ocean through the Tokara Strait, and turns eastward at about 35°N. The Oyashio is formed by mixing of the East Kamchatka Current and Okhotsk water, which is characterized by low temperature and low salinity water, rich in oxygen and nutrients (Rogachev et al., 2000). As the Oyashio front intrudes southward, it interacts with warm and saline Kuroshio waters to form North Pacific Intermediate Water (NPIW). This region is defined as the Mixed Water Region (MWR) (stations P1-1~P1-4; Talley et al., 1995; Yasuda et al., 1996; Yasuda, 2004; Qiu and Chen, 2011; Hu et al., 2015).

At stations P1-7~P1-28 ( Figure 1 ), North Pacific Tropical Water (NPTW) is the dominant North Pacific subsurface water mass, and is characterized by a salinity maximum (S>34.60, Suga et al., 2000). NPIW is formed in the MWR and flows cyclonically in the North Pacific. It is usually defined by a potential density of 26.6-27.5 kg/m³ and features a vertical salinity minimum in the subtropical Pacific (Yasuda et al., 2001; Nishioka et al., 2007; Kim et al., 2017).

Seawater samples shallower than 2000 m were collected along a longitudinal transect P1 (13°N ~ 40°N, 150°E) during a cruise on the R/V Dongfanghong 3 (October 31st – December 1st, 2019) in the Northwest Pacific. The stations are shown in Figure 1 . Stations P1-1~P1-4 are located in the MWR and stations P1-7~P1-28 are located in the subtropical region.

Seawater was sampled from 12 L Niskin-X bottles and filtered immediately through 0.45 μm membrane filters (polyether sulfone) into pre-cleaned low-density polyethylene (LDPE) bottles (500 mL). The filtering process was conducted in a clean bench onboard. Samples were then acidified to pH≤ 2 using 6 M ultra-pure hydrochloric acid (Optima grade, Fisher Chemical), and stored in double bags.

Water temperature and salinity profiles were measured using conductivity-temperature-depth sensors (CTD, Sea-Bird 911 plus). The dissolved oxygen (DO) concentration was determined via Winkler titration (Bryan et al., 1976). AOU was calculated by the equation below,

where, DO^s and DO represent the saturated oxygen concentration at the given pressure/depth and the observed oxygen concentration, respectively. Chlorophyll was measured using a HITACHI F-4700 fluorometer (Parsons et al., 1984; Li et al., 2022). Nitrate and nitrite concentrations were measured using an autoanalyzer (Seal analytical AA3) (Li et al., 2022).

Seawater samples were pre-concentrated with NOBIAS PA1 resin in a Class 1000 clean room and the REE concentrations were determined by inductively coupled plasma mass spectrometry (ICP-MS; Takata et al., 2009; Persson et al., 2011; Hatje et al., 2014; Liu et al., 2022). The results were corrected using a Lu (lutecium) internal standard. The recoveries of the internal standard were greater than 90%. The concentration of REEs in blank samples was within 3% of that measured in surface seawater, which usually has the lowest concentration in the ocean. Relative standard deviations (RSDs) of REEs were determined on replicate measurements of surface (n=4) and 4000 m deep (n=11) seawater, which were both less than 5%. The sample measurement errors are reported in Supplementary Table S1 .

The Ce anomaly (Ce/Ce^*) was calculated as (Zhang and Nozaki, 1998; Lacan and Jeandel, 2001):

where (Ce) _N , (La) _N ,and (Pr) _N represent the Post Archean Australian Shale (PAAS) (Taylor and McLennan, 1985) normalized Ce, La, and Pr, respectively. The RSD of Ce/Ce^* in Pacific deep water (4000 m, n=11) is 3.5%. By convention, an anomaly value >1 (or <1) is referred to as a positive (or negative) anomaly.

The La/Yb ratio was calculated as (Zhang et al., 2008):

where (Yb) _N represents the PAAS normalized Yb. The RSD of the La/Yb ratio in Pacific deep water is 2.4% (4000 m, n=11).

The location of station P1-15 (26°N, 150°E) in this study is close to station TPS 24 271-1 (24.29°N, 150.45°E) from previous trans-Pacific sections (Piepgras and Jacobsen, 1992). The REE concentrations were comparable between the two stations (1000 m and 2000 m) with RSDs <10 % except for Ce ( Supplementary Table S2 ). Considering that Ce is susceptible to contamination and has a relationship with latitude (see section 4.2), the Ce concentration RSDs of 18 % and 19 % are acceptable. The intercalibration reported here basically follows GEOTRACES protocols (https://geotracesold.sedoo.fr/Cookbook.pdf).

To evaluate the quasi-conservative behavior of REEs from NPIW (26.6~27.5 kg/m³) in the MWR (stations P1-1~P1-4), we used a mixing model to determine the Oyashio and Kuroshio mixing ratios at three isopycnal surfaces: 26.6, 27.0 and, 27.5 kg/m³ (Du et al., 2013; Wu et al., 2015; Li et al., 2021). The equations are as follows:

where, f_OC and f_KC represent the fractions of the Oyashio and the Kuroshio waters, respectively, and the θ_OC (θ_KC ), S_OC (S_KC ),and Nd _OC (Nd _KC ) terms are the potential temperatures (θ), salinities, and Nd concentrations of Oyashio (Kuroshio) end-members, respectively. We selected stations located in regions before the end-member currents enter the MWR since water mass properties may change after formation during transport to the study area. End-member characteristics of the Oyashio and the Kuroshio and their uncertainties are listed in Table 1 . The subscript “ _M ” represents measured values. Calculated results were fitted by least-squares optimization. Non-conservative Nd (ΔNd) is defined as the difference between the observed concentrations and calculated values due to water mass mixing; ΔNd values greater than 0 or less than 0 represent additions or removals from our defined endmember region to the study region, respectively.

The potential temperature-salinity diagram ( Figure 2 ) shows a potential temperature range from 1.7 to 29.5°C. At low latitudes, the surface water (≤5 m) has higher potential temperatures than that at high latitudes ( Figure 3A ). Potential temperature in surface water at stations P1-1~P1-4 (MWR, 37°N-40°N) ranges from 14.5 to 22.3°C. At stations P1-7~P1-28 (subtropical region, 13°N-34°N), the potential temperature in surface water ranges from 22.3 to 29.4°C. Vertically, the potential temperature decreases rapidly to <5°C from the surface to 1000 m and then remains constant below 1000 m ( Figure 3A ).

Salinity ranges from 33.39 to 35.12. In the upper water column (≤300 m), stations P1-1 ~P1-4 have low salinities, 33.39 to 34.55, which is mainly influenced by the Oyashio. In contrast, the other stations have high salinities (34.12 to 35.12), where the NPTW forms ( Figure 3B ). Below 300 m, the salinities at stations P1-1 and P1-2 gradually increase with depth. However, the salinities of the other stations reach their minimum values (<34.3) between 300 m and 1000 m, which indicates the core area of NPIW (26.6-27.5 kg/m³, Figures 3B , 4A ). The salinities at all stations are constant at 34.20-34.63 in deeper water (1000 m-2000 m).

Detailed REE concentration data is shown in Supplementary Table S1 . Similar distributions were observed for Nd and the sum of LREEs (Light rare earth elements ; La , Ce, Pr, Nd, Sm, and Eu; except for Ce, which has significant redox-related behavior that other LREEs do not; R²=0.98), and Yb and the sum of HREEs (R²=0.98) ( Supplementary Figures S1 , S2 ), Therefore, we use Nd as a representative element for LREEs and Yb as a representative element for the HREEs (e.g., Zheng et al., 2016; de Baar et al., 2018) in this study.

The concentrations of Nd and Yb for all stations along the P1 transect are shown in Figures 3D, E . Concentrations of Nd and Yb ranged from 2.62 ~ 26.66 pmol/kg and 0.83 ~ 10.02 pmol/kg, respectively. In general, REE concentrations in the MWR were higher than those in the subtropical region. In the MWR (37-40°N, stations P1-1~P1-4), the water samples from less than 300 m depth (<26.6 kg/m³) were characterized by high REE concentrations (Nd=13.02±2.91 pmol/kg, Yb=3.60±1.03 pmol/kg), which may indicate significant influence by the Oyashio with low salinity and high REEs. Below 300 m, the REEs increased with depth ( Figures 3D, E , 4B, C ). At 2000 m, the concentrations of Nd and Yb were 22.70±5.39 pmol/kg and 9.68±2.91 pmol/kg, respectively. In the subtropical region (13-34°N, stations P1-7~P1-28), the lowest REEs concentrations were found in surface and subsurface water (≤105 m, Nd=4.45±1.20 pmol/kg, Yb=1.18±0.26 pmol/kg), which may be due to particle scavenging (Stichel et al., 2015; Fröllje et al., 2016). Below 105 m, the REE concentrations increased with depth ( Figures 3D, E , 4B, C ). At 2000 m, the concentrations of Nd and Yb reached up to 21.44±2.92 pmol/kg and 9.04±0.68 pmol/kg, respectively.

Due to its significant redox behavior, Ce concentration distributions differed from other REEs ( Figure 3F ). The highest Ce concentration (6.29±1.94 pmol/kg) was found at a water depth less than 300 m in the MWR. The lowest Ce concentration was found in water samples from 500 m to 2000 m depths at stations P1-19~P1-28 (1.77±0.22 pmol/kg). Overall, at the same depth, Ce concentrations were higher at high latitudes than at low latitudes (except for ~1500 m at stations P1-7~P1-9). At ~1500 m of stations P1-7~P1-9, Ce concentrations (5.49 ±1.12 pmol/kg, Figure 3F ) were slightly higher than the surrounding seawater and the oxygen was the lowest (46.69±1.50 µmol/kg, Figure 3C ) as well.

REE concentrations were strongly positively correlated with AOU below the depth of the chlorophyll maximum (DCM) layer ( Figure 5A , the DCM in each station is shown in Supplementary Table S1 ), suggesting that the remineralization of sinking organic matter contributes to the increase of REEs (except for Ce). This is consistent with findings in the central and intermediate waters of the West Pacific by Behrens et al, (2018a, b). To systematically understand the influence of remineralization on REEs, we summarize the relationship between Nd and AOU in the subsurface and intermediate water (σ_θ< 27.5 kg/m³, depth >DCM), where the remineralization processes are usually pronounced, based on our observations and data from the literature (Behrens et al., 2018a; Behrens et al., 2018b). The results show that Nd increased rapidly as AOU increased in subsurface water (σ_θ< 26.6 kg/m³, depth >DCM) with the relationship yielding about a 4-fold higher slope (0.15±0.06) in the MWR (stations P1-1~P1-4, 37~40°N) vs. south of the MWR (0.04±0.003) (-15~34°N; Figure 5A ). However, Nd remained almost constant with increasing AOU in the intermediate water (26.6-27.5 kg/m³) of the MWR (stations P1-1~P1-4, 37~40°N; Figure 5A ). Therefore, in the MWR remineralization processes mainly influenced the REEs in subsurface water, rather than in intermediate water. This may indicate the rapid decomposition of organic matter in the subsurface water, with few organic particles sinking to intermediate water depths in the MWR. Similarly, Nd is positively correlated with nitrate + nitrite in subsurface and intermediate water (σ_θ<27.5 kg/m³, depth >DCM) to the south of the MWR as well as in subsurface water (σ_θ< 26.6 kg/m³, depth >DCM) of the MWR ( Figure 5B ). This also suggests that the remineralization of sinking organic particles plays a critical role in the production of REEs. We also estimated the addition or removal of Nd in the intermediate water of the MWR (stations P1-1~P1-4, 26.6~27.5 kg/m³) by using the isopycnal mixing model (section 2.4). The results show that, at 26.6 kg/m³, the proportion of Nd added to the observed Nd (ΔNd/Nd) was 21±16%. At 27.5 kg/m³, there was almost no addition or removal of Nd (ΔNd/Nd=-6±12%), which is consistent with the above conclusion that remineralization contributes negligible Nd to the intermediate water of the MWR. Overall, our results suggest that the influence of remineralization on REEs varies regionally.

Remineralization processes are related to the ambient environmental context, such as microbial community structure, nutrient supply and exogenous labile organic matter inputs (Church et al., 2000; Carlson et al., 2004; Mills et al., 2008; Carlson et al., 2009; Li et al., 2021). In the MWR, the intrusion front of the Oyashio mainly exists at σ_θ < 26.7 kg/m³ (Zhu et al., 2019), which is almost consistent with the depth (σ_θ < 26.6 kg/m³) where REEs appear to be significantly released by remineralization. The intrusion of the Oyashio into the Kuroshio, with distinct environmental contexts (e.g., temperature, salinity, nutrient supply, and organic matter inputs), may enhance the remineralization in the MWR subsurface water (<26.6 kg/m³). Enhanced remineralization due to strong mixing is also found in the Luzon Strait near the Kuroshio intrusion (Li et al., 2021). In addition, the regional differences in remineralization might be associated with the plankton community structure. For instance, in the MWR, the phytoplankton are dominated by haptophytes, while in the oligotrophic subtropical region, dinoflagellates are dominant (Suzuki et al., 1997; Lin et al., 2020; Wang et al., 2021; Wang et al., 2022). We suggest that organic detritus from different organisms may have distinct ratios of consumed oxygen and released REEs during the remineralization process. This supposition requires more evidence, such as the ratio of REE concentrations to organic matter in haptophytes vs dinoflagellates. Thus, the water masses mixing and/or the plankton community structure could be responsible for the regional differences of the influence of remineralization on REEs.

Overall, our results emphasize regional differences in the release efficiency of REEs by remineralization. For other trace elements, like iron, the efficiency of organic remineralization is speculated to be one of the reasons for the difference in dissolved iron concentrations between the western and eastern subarctic Pacific (Nishioka et al., 2013). Our results may provide evidence for regional differences in trace elements caused by organic matter remineralization.

Figure 6 shows the distribution of LREEs (La, Ce, Pr) and La/Yb ratios in surface water (<10 m) in the Northwest Pacific. REE (La, Ce, Pr) concentrations near the margin are higher than those farther away. We suggest that the high REEs in the surface water of the MWR and Oyashio likely originate from the Kuril Islands (Morton et al., 2019). In addition, high La/Yb ratios (>0.35) ( Figures 6E and 7A ) were observed in northern stations (stations TPS 47 39-1 and SEEDS-II; Piepgras and Jacobsen, 1992; Hara et al., 2009), which reflects the effect of lithogenic material dissolution (Zhang et al., 2008; Pearce et al., 2013). The La/Yb ratio could be the Aleutian Islands signal caught up in the East Kamchatka Current since the elevated La/Yb ratios (0.51±0.51, Supplementary Figure S3A ) have been observed in the Aleutian Islands. Furthermore, the high Nd isotopic composition (-2.2 at station CM-S-3 and -2.0 at station TPS 47 39-1) of surface water in the Oyashio also verifies the input of REEs from islands (the Kuril and Aleutian Islands), as volcanic islands usually have more radiogenic Nd (Piepgras and Jacobsen, 1988; Amakawa et al., 2004a, Amakawa et al., 2004b; Fuhr et al., 2021). Therefore, islands are important lithogenic sources of REEs to surface waters of the Northwest Pacific subarctic region and the MWR.

In addition, the Sea of Okhotsk is an important source of materials to the Oyashio and NPIW in the Northwest Pacific, such as Fe (Nishioka et al., 2013), Zn (Kim et al., 2017), Cd (Yang et al., 2018) and anthropogenic CO₂ (Yasuda, 2004). Based on the fact that the Sea of Okhotsk receives a large number of trace elements from rivers and sediments, we believe that Okhotsk seawater may also be an important source of REEs to the Oyashio (Shulkin and Bogdanova, 2003; Nishioka et al., 2007; Nishioka et al., 2013; Nishioka et al., 2014; Kim et al., 2015).

As discussed above, the REEs in the Oyashio Current are sourced from the islands and the Sea of Okhotsk and then supplied to the NPIW in the subtropical region. In the study area, REE concentrations (except for Ce) are mainly influenced by remineralization processes during NPIW transport (see above section 4.1), and their concentration changes vertically, increasing with depth ( Figures 3D, E , 4B, C ). However, unlike other trivalent REEs, dissolved Ce is easily removed by particles in seawater. This is because Ce (III) is easily oxidized to the tetravalent state (Ce(IV)) in seawater and subsequently precipitated as CeO₂ or Ce(OH)₄, resulting in the depletion of dissolved Ce and negative Ce anomalies in seawater (Alibo and Nozaki, 1999; Tazoe et al., 2011). Following Morton et al. (2019), in the subarctic region, the Ce concentration was high but Ce/Ce^* showed an extremely negative anomaly (<0.2) in surface water ( Figures 6C, F ), which suggests Ce inputs from the shelf as well as preferential removal of Ce relative to La and Pr by Mn-oxides (Morton et al., 2019). In this study area (transect P1), Ce concentrations were higher at high latitudes than at low latitudes (except for ~1500 m depth at stations P1-7 ~P1-9, Figures 3F ): that is, the Ce concentration decreased with distance from the shelf, due to the rapid oxidation and removal of Ce from seawater. At ~1500 m depth at stations P1-7~P1-9, Ce concentrations were slightly higher than in the surrounding seawater ( Figure 3F ), and the oxygen concentration was the lowest ( Figure 3C ). This might be explained by fewer oxide particles in the water column or the effect of lateral transport from a low oxygen region, and needs further research. In addition, a strong particulate negative Ce anomaly was observed in the subarctic region (station 2, 155°E, 44°N) by Morton et al. (2019), which may rule out the possibility of supply from an atmospheric source because aerosols generally have no Ce negative anomaly (Morton et al., 2019).

The Oyashio Current acquires the imprints of high La/Yb ratios from the islands through the East Kamchatka Current. As shown in Figure 7A , the highest La/Yb ratios (>0.35) were observed at stations TPS 47 39-1 (Piepgras and Jacobsen, 1992) and SEEDS-II (Hara et al., 2009) (< 100 m) in the East Kamchatka Current. In the low salinity water (<34.25, mainly the Oyashio and NPIW), the high value signature is gradually diluted as the water mass flows southward. It should be noted that because of the lack of REE data in the Sea of Okhotsk, we cannot determine whether the water masses from the Sea of Okhotsk also provided elevated La/Yb ratios to the Oyashio and NPIW. At all stations (≤2000 m), La/Yb is significantly negatively correlated with salinity (R²=0.63, P<0.01) ( Figure 7B ). This indicates that the La/Yb ratios are mainly affected by water mass mixing even though it is generally accepted that the particulate process should cause a fractionation of LREEs and HREEs due to their different particle adsorption affinities (Byrne and Kim, 1990; Sholkovitz et al., 1994). Similarly, the Oyashio may also have acquired the signals of Pr/Yb and Nd/Yb ratios from nearby islands (Aleutian and Kamchatka Islands; Figures 8A, B ). The islands’ Pr/Yb and Nd/Yb signals are more significant than the La/Yb signal ( Supplementary Figures S3B, C ). NPIW has shown elevated ratios of Pr/Yb and Nd/Yb ( Figures 8C, D ); however, there is no correlation between Pr/Yb (or Nd/Yb) and salinity (R² values were 0.06 and 0.01, respectively). This indicates that La/Yb more faithfully retains the original signature of an end-member than Pr/Yb and Nd/Yb in seawater. This may be related to the fact that La is the only lanthanide element that lacks a 4f electron, resulting in a solution stability that differs from other LREEs in seawater. Otherwise, in seawater, the longer residence time of La compared to other LREEs (Li, 1991; Alibo and Nozaki, 1999; Nozaki, 2001) might also contribute to the water masses tracing values of La/Yb. In summary, the La/Yb ratio provides a useful proxy for studying the flow paths of the Oyashio and NPIW in the North Pacific, and the NPIW is critical for transporting substances that affect marine ecosystems and climate, such as Fe and CO₂ (Yasuda, 2004; Nishioka et al., 2007; Nishioka et al., 2013; Nishioka et al., 2014; Nishioka and Obata, 2017).

REEs ratios have also been used for tracing Southern Ocean water masses (Zhang and Nozaki, 1996; Osborne et al., 2015; Behrens et al., 2018a). For instance, the Ho/Dy ratio may be used to differentiate intermediate/deep water in the Southwest Pacific from Antarctic Bottom Water (Zhang and Nozaki, 1996). The low Dy/Er and high HREE/LREE ratios formed in the East Pacific Rise hydrothermal plume trace the Upper Circumpolar Deep Water (Osborne et al., 2015; Behrens et al., 2018a). These examples indicate that the ratios of REEs in the oceans are valuable water mass tracers and should be widely focused on in future studies.

Taken together, lithogenic material input from islands (including the Kuril Islands and the Aleutian Islands) and the Sea of Okhotsk are the important sources of REEs and other trace elements (Fe, Zn, Cd, etc.) to the Oyashio current. It is also an important source for NPIW due to the supply of Oyashio water. The La/Yb ratio is mainly affected by water mass mixing and therefore is a very good tool for tracing the Oyashio and NPIW.

The marginal seas (including the Sea of Okhotsk, Sea of Japan, East China Sea, South China Sea, and the Bering Sea) may be important REE sources to the Northwest Pacific. We estimated the net Nd fluxes ( F N d n e t ) between the marginal seas and Northwest Pacific based on the following equation:

Where, F w i n f l o w (or F w o u t f o l w ) represents the water flux from the marginal sea to the Northwest Pacific (or from the Northwest Pacific to the marginal sea). Nd ^m (or Nd ^p ) represents the end-member values of Nd concentration in the marginal seas (or the Northwest Pacific). The water fluxes and Nd concentrations are shown in Table 2 . The net Nd flux from the Bering Sea was not estimated due to a lack of REE data.

A portion of the East Kamchatka Current flows into the Sea of Okhotsk through the Northern Kuril Islands Strait (mainly in the Kruzenshtern Strait) and then flows out to the Pacific Ocean through the South Kuril Islands Strait (mainly through the Bussol’ Strait, water depth ~2300 m) ( Figure 1 ; Hill et al., 2003; Shu et al., 2021). Because the water flow flux into the Sea of Okhotsk through the Kruzenshtern Strait is thought to be nearly equal to the outflow flux from the Bussol’ Strait (Ohshima et al., 2010), we assume that both the inflow and outflow water fluxes are 8.2~8.8 Sv (Katsumata, 2004) (excluding the water exchange caused by tidal processes). Due to the lack of a vertical profile of REEs in the Sea of Okhotsk, the Nd concentration of Okhotsk surface water (Amakawa et al., 2004a) was used to represent the average Nd concentration of the full water column. The net input flux of Nd from the Sea of Okhotsk to the Pacific Ocean was estimated to be 29~32 tNd/y ( Table 2 and Figure 9 ), which is comparable to the input from the Yangtze and Korean rivers combined (20~24 and 7 tNd/y, respectively; Behrens et al., 2018b; Che et al., 2022).

The Tsugaru Strait is the only pathway directly connecting the Sea of Japan and the North Pacific, at a depth of <200 m ( Figure 1 ; Ito et al., 2003). The Tsugaru Warm Current originates from the Sea of Japan and flows into the North Pacific through the Tsugaru Strait with a water flux of 1.1~2.1Sv (Ito et al., 2003). Using above water flux and the weighted average value of Nd in the Sea of Japan (<200 m, Seo and Kim, 2020), we calculated the net Nd input flux to be 159~302 tNd/y ( Table 2 and Figure 9 ). This Nd input flux from the Sea of Japan is of the same order of magnitude as the atmospheric and riverine contributions of Nd to the global oceans (400-630 tNd/y and 260-660 tNd/y, respectively ; Greaves et al., 1994; Tachikawa et al., 2003; Arsouze et al., 2009), which are considered important sources of REEs in the oceans (Lacan and Jeandel, 2001; Flierdt et al., 2004; Arsouze et al., 2009).

In the East China Sea, the channels in the Ryukyu Island chain are important links to the Northwest Pacific. The East Taiwan Channel and the Tokara Strait are the main pathways connecting the East China Sea and the Northwest Pacific (Johns et al., 2001; Isobe et al., 2002; Teague et al., 2002; Takikawa et al., 2005; Isobe, 2008). Therefore, we only focus on the fluxes across these two straits. The results show that the net Nd flux from the East China Sea is 142~616 tNd/y, which is relatively large and similar to or greater than the input from the Sea of Japan (159~302 tNd/y, Table 2 and Figure 9 ). On the other hand, the net Nd input to the Pacific Ocean from the East China Sea shelf at depths of <250 m was estimated to be 113 t/y using Nd isotopes and a box model (Lacan and Jeandel, 2005), which is higher than our result for depths of <200 m (4~47 tNd/y, Table 2 ). This indicates that our result might be an underestimate, probably due to the fact that we selected a lower average Nd endmember concentration in the Northwest Pacific for our estimation than that used by Lacan and Jeandel (2005).

The water exchange between the Northwest Pacific and the South China Sea occurs mainly through the Luzon Strait. In the Luzon Strait, the surface and deep waters flow from the Pacific to the South China Sea. At intermediate depths, the flow direction is reversed (Chao et al., 1996; Gan et al., 2006; Qu et al., 2006). The Nd fluxes between the Northwest Pacific and the South China Sea at different water layer depths are shown in Table 2 . The calculated Nd fluxes for 0-500 m, 500-1500 m, and 1500 m-bottom are -287~-137, 125~557, and -400~-236 t/y (negative values represent input from the Northwest Pacific to the South China Sea), respectively. The net Nd input fluxes for the full water column are -298~34 t/y ( Table 2 and Figure 9 ). This indicates that the South China Sea is a sink of Nd from the Northwest Pacific at 0-500 m and 1500 m-bottom, and a source of Nd in the 500-1500 m water layer. Overall, the South China Sea may be either a net source or a net sink of Nd to the Northwest Pacific depending on the selection of the endmembers in the present estimation. For instance, when the estimated water fluxes at 0-500 m, 500-1500 m, and 1500m-bottom are -9.0, 5.0, and -2.0 Sv, respectively (Tian et al., 2006), the South China Sea is a source of REEs to the Northwest Pacific with net Nd flux of 34 t/y. When the water fluxes in 0-500 m, 500-1500 m, and 1500 m-bottom are estimated to be -5.1, 2.8, and -3.8 Sv, respectively (Xu and Oey, 2014), the South China Sea is a sink of REEs in the Northwest Pacific with net Nd flux of -298 t/y. It should be noted that there are large uncertainties in the estimated net fluxes of Nd. In particular, the complex water structure (Chen and Wang, 1998; Wang et al., 2010; Zhang et al., 2015) and seasonal variations of water exchange in the Luzon Strait (Xu and Oey, 2014 and references therein) lead to a wide range of water fluxes ( Table 2 ). In addition, there may be seasonal variations in Nd concentrations, which introduce uncertainties in Nd net flux estimates.

In summary, the Sea of Okhotsk, the Sea of Japan and the East China Sea are Nd sources to the Northwest Pacific, with a relatively low net flux of Nd from the Sea of Okhotsk (29~32 t/y) but high fluxes from the Sea of Japan (159~302 t/y) and the East China Sea (142~616 t/y). In contrast, the Nd exchange flux between the South China Sea and the Pacific is subject to water exchange occurring at different water layer depths in the Luzon Strait. The South China Sea is a sink of Nd from the Northwest Pacific at the water layers of 0-500 m (-287~-137 t/y) and 1500m-bottom (-400~-236 t/y), and a source of Nd at 500-1500 m (125~557 t/y), with a net Nd flux range of -298~34 t/y.