DATA
Spatiotemporal variations of seawater δ18O and δD in the Western North Pacific marginal seas near Japan
2024 Volume 58 Issue 3 Pages 94-108
Details
2024 Volume 58 Issue 3 Pages 94-108
The stable oxygen isotope (δ18O) values of biogenic calcium carbonates (CaCO3) present in hard tissues of marine organisms vary with the δ18O of the surrounding seawater (δ18Osw) and water temperature. Consequently, the ambient water temperature when marine organisms existed can be estimated using δ18O of CaCO3 with δ18Osw. Thus, the aims of this study were to reveal the spatiotemporal variations in δ18Osw and their correlations with salinity in the East Asian marginal seas. We collected seawater samples (n = 2222) at 1394 stations around Japan, primarily from the surface layer of the East China Sea (ECS) and Sea of Japan (SOJ), from 2015 to 2021. We analyzed δ18Osw and the stable hydrogen isotope of seawater (δDsw) along with the measured water temperature and salinity. The δ18Osw and δDsw values ranged from –3.48‰ to +0.45‰, and –21.5‰ to +2.3‰, respectively. In our full data, both δ18Osw and δDsw had positive linear relationships with salinity as follows: δ18Osw = 0.235 × salinity – 7.94 (r2 = 0.85) and δDsw = 1.56 × salinity – 52.9 (r2 = 0.85). Furthermore, the relationship between δ18Osw and δDsw for full data as follows: δDsw = 6.44 × δ18Osw – 0.18 (r2 = 0.96). These relationships varied across seasons, areas (the ECS or SOJ), and water depths. In particular, δ18Osw and δDsw of less-saline water were different in the ECS and SOJ. These fine-scale, wide-range, and high-precision δ18Osw and δDsw datasets can contribute to paleoceanography, environmental analysis, oceanography, and fisheries science.
Elemental isotope ratios in organisms provide crucial information regarding the cycles and origins of elements, environments in which organisms have been present, and food web structures. Historically, stable hydrogen and oxygen isotope ratios (δD and δ18O, respectively) have been used as indices of water cycle (Lloyd, 1966). During the evaporation of water, light isotope water (such as H216O) evaporates more than heavy isotope water (such as H218O and DH16O), leading to elevated δ18O and δD in the remaining seawater. Therefore, strong relationships between salinity and δ18O of seawater (δ18Osw), and salinity and δD of seawater (δDsw) are usually observed in the ocean. However, δ18Osw and δDsw variations are more complex than those of salinity because different freshwater δ18O and δD supplies from river and precipitation (Conroy et al., 2017; Kiran Kumar et al., 2018; LeGrande and Schmidt, 2006).
The δ18O of biogenic calcium carbonates (CaCO3) present in hard tissues of marine organisms, such as fish otoliths, foraminiferal shells, and coral skeletons, are correlated with the temperature at the time of CaCO3 formation. Thus, they are used for reconstructing the temperature of the past environment (Bemis et al., 1998; Devereux, 1967; Thompson et al., 2022; Weber and Woodhead, 1970) as well as the ratios of alkali earth metals such as Mg/Ca and Sr/Ca (Inoue et al., 2007). Furthermore, the δ18O of biogenic CaCO3 is vital for determining the paleoclimate after Emiliani (1955), in which the temperature-depended isotopic fractionation factor between CaCO3 and water is set. Recently, migration processes of many fishes have been estimated using the δ18O of fish otoliths, such as Australian salmon (Arripis trutta) (Kalish, 1991), bluefin tunas (Thunnus spp.) (Hane et al., 2020; Kitagawa et al., 2013; Rooker et al., 2008), sardines (Sardinops spp.) (Sakamoto et al., 2018; Sakamoto et al., 2017; Sakamoto et al., 2022; Sakamoto et al., 2020), juveniles of anchovy (Engraulis japonicus) and sardine (Nishida et al., 2020), chub mackerel (Scomber japonicus) (Higuchi et al., 2019), and Japanese jack mackerel (Trachurus japonicus) (Enomoto et al., 2022; Muto et al., 2022). In every case, data of δ18Osw covering the spatiotemporal variations are necessary to estimate the experienced temperature precisely. However, as shown in previous studies (Kiran Kumar et al., 2018; LeGrande and Schmidt, 2006; Schmidt, 1999), δ18Osw variations are more complex than those of salinity, and the relationship between salinity and δ18Osw is regionally specific. Thus, direct measurement of the δ18Osw, not the estimation from the δ18Osw-salinity relationship, are recommended for creating an isoscape of δ18Osw.
The East China Sea (ECS) and Sea of Japan (SOJ) are marginal seas in the Western North Pacific (WNP). The continental shelf exists widely in the ECS, with the deep Okinawa Trough on the eastern side (Fig. 1a). Along the southern edge of the continental shelf, the Kuroshio Current, i.e., the western boundary current of the North Pacific, flows from southwest to northeast. The Tsushima Warm Current (TWC) originates in the ECS and flows into the southern SOJ. The Kuroshio-origin, Changjiang-origin, and Taiwan Warm Current-origin waters comprise the TWC water (Guo et al., 2006; Isobe, 1999). The Changjiang-origin freshwater substantially impacts these seas (Senjyu et al., 2008) as well as the precipitation on these seas and Japan Archipelago and the Korean Peninsula (Kosugi et al., 2021). Japan Sea Proper Water (JSPW), whose temperature and salinity are spatiotemporally homogenous, is present below the middle layer of the SOJ (Sudo, 1986).
Maps of sampling locations. (a) Total sampling points, and those in (b) 2015, (c) 2016, (d) 2017, (e) 2018, (f) 2019, and (g) 2020 and 2021. Different shapes denote different subareas (cross, closed circle, and plus indicate the Sea of Japan (SOJ), East China Sea (ECS), and Western North Pacific (WNP), respectively). The red lines in (a) denote the boundaries of subareas. In (g), the sampling points were limited to three stations surrounding the dotted circle in 2020, and the samples from other stations were collected in 2021.
The distinct δ18Osw–salinity relationships in the ECS and SOJ (Horikawa et al., 2015; Kodaira et al., 2016) have already been studied, with slopes and intercepts of 0.27 and –9.1‰, respectively, in the TWC region (Kodaira et al., 2016) and 0.23 and –7.74‰, respectively, in the ECS (Horikawa et al., 2015). However, the reason for this difference remains unclear. It has been argued that the lower intercept in the SOJ is attributed to the contributions from local rivers (Kodaira et al., 2016), but the seasonality and sampling bias (i.e., small numbers of samples) could not be rejected because the samples only collected in limited months and the number of samples was ≤74 in these studies (Horikawa et al., 2015; Kodaira et al., 2016). The spatially different equations of these studies (Horikawa et al., 2015; Kodaira et al., 2016) can cause difficulties in calculating δ18Osw using salinity in this area. Vertically, both δ18Osw and δDsw are high in the surface layer (δ18Osw and δDsw: +0.5 and +4.5‰, respectively), but low in the middle and deep layers (δ18Osw and δDsw: 0.0 and –0.4‰, respectively, below 755 m depth) of the Okinawa Trough in the ECS (Cruz Salmeron et al., 2022), but the values were not compared to salinity in Cruz Salmeron et al. (2022). Therefore, the aims of this study were to evaluate the spatiotemporal variations in δ18O and δD values and create isoscapes of δ18Osw and δDsw in the ECS and SOJ based on big datasets (n = 2222). Consequently, we investigated the δ18Osw and δDsw values along with the relationships between different environmental parameters in the ECS and SOJ.
Water samples (n = 2222) were collected at 1394 stations along Japan from July 2015 to March 2021 (Fig. 1). Seawater samples were collected primarily from the surface layer (<11 m depth; n = 1464) at a depth of ~5 m using a bucket, Niskin bottle, or pump located at the bottom of a ship. Samples below a depth of 100 m were collected vertically from 67 stations. The vertical sampling layers varied among different cruises (year, season, and area), but the vertical samples were primarily collected below 100 m at the following seven depths (named standard depths): 100, 125, 150, 200, 300, 400, and 500 m. Seawater for on-land analysis of δ18Osw and δDsw was collected and kept in 50 ml polyethylene bottles (I-boy, AS ONE, Japan). The δ18Osw and δDsw were stable in the polyethylene bottle for at least 2 years.
The temperature and salinity of the seawater collected from the bottom of a ship were determined using a thermosalinometer (SBE45, Sea-Bird Electronics Inc., WA) and those of a Niskin bottle were using a conductivity–temperature–depth (CTD) sensor (SBE 9plus and SBE 19, Sea-Bird Electronics, WA, or AAQ-RINKO, JFE-Advantech, Japan). The temperature of the samples collected using a bucket were determined using a calibrated thermometer onboard ship, and those of salinity were collected in glass bottles and determined with a laboratory conductivity salinometer (Autosal8400B, Guildline Instruments Ltd., Ontario, Canada). The salinometer was standardized using the International Association for the Physical Science of the Oceans (IAPSO) standard seawater.
The observation area was divided into three subareas: the ECS, SOJ, and WNP. The boundary of the ECS and SOJ was set at 129°30'E in the eastern part of the Tsushima Strait (Fig. 1a). The boundary of the ECS and WNP was set at 130°30'E (Fig. 1a). Some of the stations located south of Okinawa Island are outside Kuroshio, which is the latitudinal border of the ECS and WNP; however, in this study, such stations were treated as the ECS. The sampling seasons were divided into four seasons: winter (January–March), spring (April–June), summer (July–September), and autumn (October–December). The sampling frequencies for the three areas and seasons are summarized in Table 1. In the WNP, samples were obtained only during summer. The sampling frequencies varied interannually, i.e., n = 90, 208, 214, 481, 366, 8, and 96 in 2015, 2016, 2017, 2018, 2019, 2020, and 2021, respectively (Fig. 1b–g).
Summary of the sampling frequencies
| Winter (Jan–Mar) | Spring (Apr–Jun) | Summer (Jul–Sep) | Autumn (Oct–Dec) | Total | |
|---|---|---|---|---|---|
| ECS | 382 (218) | 508 (360) | 167 (105) | 48 (29) | 1105 (712) |
| SOJ | 62 (62) | 269(257) | 703 (354) | 49 (49) | 1083 (723) |
| WNP | 0 | 0 | 34 (30) | 0 | 34 (30) |
| Total | 444 (280) | 777 (617) | 904 (489) | 97 (78) | 2222 (1464) |
The numbers in the blanket are the sampling frequencies at the surface (≤11 m depth). The ECS, SOJ, and WNP represent the East China Sea, Sea of Japan, and Western North Pacific, respectively.
Temperature is a basic parameter used to understand the environment. Temperature variations ranged from 1.2–29.3°C in the surface layer (<11 m depth), and spatiotemporal variations were significant.
Measurements of δ18Osw and δDswThe δ18Osw and δDsw were analyzed using a cavity ring-down spectroscopy (L2130-i; Picarro Inc., Santa Clara, CA, USA) at National Institute of Technology (Ibaraki College) and Kyoto University. All samples were subsampled using a 2.5 mL polypropylene syringe and filtered through a 0.45 μm polytetrafluoroethylene (PTFE) filter. Then, 2 mL of them was added to autosampler glass vials for further analysis. The δ18O provided from cavity ring-down spectroscopy is comparable to isotope ratio mass spectrometry (Walker et al., 2016), and we had checked the difference between the two methods using Ishimura et al. (2008) was smaller than the detection limit of isotope ratio mass spectrometry method, 0.2‰ (Fig. 2).
Comparison of δ18Osw values between cavity ring-down spectroscopy (CRDS) and isotope ratio mass spectrometry (IRMS). The black solid line indicates the line whose slope and intercept are 1 and 0, respectively, the dotted lines are those whose slope is 1 and intercept is ±0.2. The blue line with shadow is the regression line. The δ18Osw of seawater samples (n = 32) using the CO2-equilibrium method with the continuous-flow IRMS (Ishimura et al., 2008) was <0.2‰ different from those using the CRDS.
A batch analysis was composed with 15 field samples, 5 laboratory standards and 3 calibration standards, and this batch was analyzed in 48 hours. The five laboratory standards, which were desalted seawater (δ18O: –0.132‰ and δD: +0.155‰), comprised one laboratory standard unused in the previous batch analysis and those used in the last four batch analyses. Analysis of these standards verified long-term analytical precision and stability among the batches. The three standards used for calibration are commercially distributed (Shoko-Science Co., Ltd., Yokohama, Japan), with isotopic values of δ18O = –0.086‰ and δD = –0.27‰, δ18O = –10.82‰ and δD = –70.8‰, and δ18O = –14.51‰ and δD = –106.3‰, respectively, in the Vienna Standard Mean Ocean Water (VSMOW) scale.
All samples were analyzed in the high-precision mode as per the manufacturer’s setting. Ten injections per vial were analyzed in one set, and raw data from the first five injections were rejected to avoid memory effects because of the previously analyzed sample. The standard deviation of these five injections was always kept <0.1‰.
The analytical sequence in the batch comprised pre-analysis, analysis of standards and samples, standard analysis for three-point calibration, and post-analysis. In pre-analysis, five sets (50 injections) of laboratory standards from previous batches were used to reduce the memory effect. In the analysis of standards and samples, the unused laboratory standard was analyzed three times. If a minor isotopic drift was observed in one batch, a correction was performed in all analytical results of the batch. In the post-analysis, three types of commercial standards were analyzed and used for three-point calibration. Subsequently, 10 sets of laboratory standards were also analyzed. The salt liner and septum of the sample injection port were swapped for each batch.
The isotopic values of the water samples were reported relative to VSMOW. When the seawater sample was measured repeatedly, the external analytical precision was better than ±0.03 and ±0.2‰ for δ18O and δD, respectively (n = 15), which was treated as the detection limits of this study. We have also confirmed that this accuracy is maintained in our daily analysis.
All statistical analyses were conducted using the R software (R Core Team, 2023). A linear regression model was constructed for each relationship, and the equations were calculated using the least-squares method. To identify the effects of season, area, and sampling depth, multi-linear regression models considering these parameters were compared with a single linear regression model using analysis of variance (ANOVA). The threshold of the p-value was set as 0.05. Additionally, the Akaike information criterion (AIC) values for the multi-linear regression models were verified.
Salinity, δ18Osw, and δDsw ranged from 22.671 to 34.9448, –3.48 to +0.45‰, and –21.54 to +2.32‰, respectively. In the surface (sampling depth ≤11 m) of the central ECS, less-saline water (salinity <32) was observed from spring to autumn (Fig. 3a), whereas salinity <30 was observed during summer. In the SOJ, salinity <33 was widely observed, but the horizontal variations in surface salinity were more stable than those in the ECS (Fig. 3a). The horizontal distributions of δ18Osw and δDsw were correlated with salinity (Fig. 3b–c). Lower δ18Osw (<–1‰) and δDsw (–5‰) were observed in the central ECS during summer. In the SOJ, δ18Osw and δDsw were low during summer. Salinity, δ18Osw, and δDsw were higher in WNP than the SOJ and ECS.
Horizontal and seasonal distributions of (a) salinity, (b) δ18Osw, and (c) δDsw in the surface layer (≤11 m depth). The dots and colors represent the sampling points and 1° × 1° median values of the target parameters, respectively. The interannual variations are not considered.
At the 67 stations, the vertical distributions of salinity, δ18Osw, and δDsw below 100 m depth were similar to each other but were slightly different below 200 m depth (Fig. 4). In the ECS at 300–500 m depth, salinity slightly decreased from 34.4 to 34.3, δ18Osw decreased from +0.08 ± 0.07‰ (mean ± standard deviation [sd]) at 300 m depth to –0.83 ± 0.19‰ at 500 m, and δDsw decreased from +0.25 ± 0.51‰ at 300 m depth to –0.07 ± 0.05‰ at 500 m (Fig. 4). This decreasing trend was the same for all parameters. However, the lowest salinity was observed at 100 m depth, whereas the lowest δ18Osw and δDsw values were observed at 500 m depth. In the SOJ, samples below 300 m depth were collected only during summer, but both δ18Osw and δDsw were homogenous at 300, 400, and 500 m (mean ± sd of δ18Osw: –0.04 ± 0.03, –0.06 ± 0.02, and –0.04 ± 0.03‰ at 300, 400, and 500 m depth, respectively; those of δDsw: –0.74 ± 0.15, –0.88 ± 0.07, and –0.81 ± 0.11‰ at 300, 400, and 500 m depth, respectively). The values were almost the same at 1000 m depth in the SOJ (mean ± sd of δ18Osw and δDsw: –0.05 ± 0.02 and –0.91 ± 0.08‰, respectively).
Vertical profiles of (a, b) salinity, (c, d) δ18Osw, and (e, f) δDsw observed at the 67 stations below 100 m depth. Box plots show the mean values (thick vertical lines within boxes), standard deviations (boxes), and 95% confidence interval ranges (horizontal bars). The small dots represent the raw values, and their colors indicate seasons. During winter and autumn in the Sea of Japan (SOJ) and Western North Pacific (WNP), no samples are collected below 100 m depth vertically.
Salinity and δ18Osw had a significantly strong and positive relationship in all the datasets (r2 = 0.85; p < 10–15; Fig. 5a). The equation calculated using the least-squares method is as follows:
Relationships between δ18Osw and salinity in (a) the total and (b) below 100 m depth datasets. In (a), the blue line indicates the regression line (Eq. 1), and the red box represents the data range shown in (b). In (b), the solid line represents the regression line shown in (a), the colored and gray dots represent values below and above 100 m depth, respectively, and the color and shape indicate the sampling depth and area, respectively.
The numeric after plus-minus sign (±) in the equations denotes the standard error (se).
The δ18Osw–salinity relationships below 200 m depth were different than those of the surface in the ECS and SOJ (Fig. 5b). In the ECS, δ18Osw decreased with depth, whereas salinity remained constant at ~34.3. Furthermore, δ18Osw at 500 m depth was clearly lower than that at the surface considering its salinity. In the SOJ, δ18Osw below 200 m depth was always lower than that estimated using Eq. 1 (Fig. 5b).
Based on the multi-linear regression model approach, the seasons, areas, and sampling layers (≤200 or >200 m depth) significantly impacted the δ18Osw–salinity relationship (ANOVA, p < 0.001). Since the WNP and >200 m depth samples were limited, the equations were reconstructed using only the surface (≤200 m) values without considering the WNP values (Table 2). First, the equations without area, i.e., the slope (0.237 ± 0.002) and intercept (–7.99 ± 0.07‰) of the linear model, were reconstructed without considering the seasonality. These values were within the standard errors in all the datasets (Eq. 1). Considering the seasonality, the slope was gentle and the intercept was the highest during autumn (0.207 ± 0.011 and –6.99 ± 0.36‰, respectively), whereas the slope was steep and the intercept was the lowest during spring (0.265 ± 0.004 and –8.93 ± 0.13‰, respectively, Table 2). The coefficient of determination (r2) was lower (0.292) in winter than in other seasons (Table 2).
List of coefficients and their standard errors, intercepts and their standard errors, sampling frequencies (n), and the coefficient of determination (r2) of the salinity–δ18Osw relationship above 200 m depth
| Slope | Intercept | n | r2 | |
|---|---|---|---|---|
| Excluded WNP | 0.237 ± 0.002 | –7.99 ± 0.07 | 2101 | 0.868 |
| winter | 0.235 ± 0.018 | –7.91 ± 0.61 | 427 | 0.292 |
| spring | 0.265 ± 0.004 | –8.93 ± 0.13 | 758 | 0.865 |
| summer | 0.213 ± 0.003 | –7.20 ± 0.09 | 821 | 0.874 |
| autumn | 0.207 ± 0.011 | –6.99 ± 0.36 | 95 | 0.802 |
| The East China Sea | 0.222 ± 0.002 | –7.46 ± 0.08 | 1063 | 0.901 |
| winter | 0.286 ± 0.035 | –9.71 ± 1.23 | 365 | 0.153 |
| spring | 0.238 ± 0.005 | –8.00 ± 0.16 | 489 | 0.840 |
| summer | 0.213 ± 0.003 | –7.18 ± 0.11 | 163 | 0.961 |
| autumn | 0.233 ± 0.007 | –7.77 ± 0.24 | 46 | 0.961 |
| The Sea of Japan | 0.260 ± 0.004 | –8.79 ± 0.12 | 1038 | 0.836 |
| winter | 0.230 ± 0.038 | –7.74 ± 1.29 | 62 | 0.384 |
| spring | 0.308 ± 0.005 | –10.40 ± 0.19 | 269 | 0.923 |
| summer | 0.219 ± 0.005 | –7.42 ± 0.16 | 658 | 0.767 |
| autumn | 0.269 ± 0.021 | –9.13 ± 0.71 | 49 | 0.778 |
The coefficient and intercept are calculated using the least squares method.
In the ECS, the slope (0.222 ± 0.002) and intercept (–7.46 ± 0.08‰) were slightly gentler and lower than those in the SOJ (slope: 0.260 ± 0.004 and intercept: –8.79 ± 0.12‰; Table 2 and Fig. 6). In the ECS, the slope (0.286 ± 0.035) and intercept were the steepest and the lowest (–9.71 ± 1.23‰), respectively, during winter, whereas the slope and intercept were the gentlest (0.213 ± 0.003) and highest (–7.18 ± 0.11‰), respectively, during summer (Table 2). In the SOJ, the slope (0.219 ± 0.005) and intercept (–7.42 ± 0.16‰) were the gentlest and highest, respectively, during summer, whereas the slope (0.308 ± 0.005) and intercept (–10.4 ± 0.19‰) were the steepest and lowest, respectively, during spring (Table 2). During winter, r2 was low in both the ECS and SOJ and was dependent on a narrow range of salinity.
Seasonality in the relationships between δ18Osw and salinity in the (a–d) East China Sea (ECS) and (e–h) Sea of Japan (SOJ). The blue, black, and dotted lines are the regression lines of specific areas and seasons, specific areas (no seasonality), and other areas (i.e., the ECS one in the SOJ panels and the ECS one in the SOJ panels), respectively. The gray points indicate the value of the target area in different seasons.
The δ18Osw values of less-saline water were different in the SOJ and ECS (Fig. 6). The δ18Osw values of less-saline water (<31) were lower in the SOJ than the ECS in different observation seasons (SOJ: spring and ECS: summer). A lower δ18Osw trend was observed in the SOJ than the ECS during summer and autumn (Fig. 6).
Relationships between salinity and δDswSalinity and δDsw had significantly strong relationships like δ18Osw–salinity (r2 = 0.85; p < 10–15; Fig. 7a). The equation based on the least squares method with standard errors is as follows:
Relationships between δDsw and salinity in (a) the total and (b) below 100 m depth datasets. In (a), the blue line indicates the regression line, and the red box represents the data range shown in (b). In (b), the solid line represents the regression line shown in (a), the colored and gray dots represent values below and above 100 m depth, respectively, and the color and shape indicate the sampling depth and area, respectively.
Focusing on the vertical profile, the δDsw–salinity relationship was similar to the δ18Osw–salinity relationship: the relationship below 200 m depth was different from that at the surface in the SOJ and ECS (Fig. 7b). The δDsw below 200 m depth was at the same level in the ECS and SOJ, whereas the salinity of the ECS was high.
The linear model approach showed that the δDsw–salinity relationship varied among different areas and seasons. When the data was limited to ≤200 m depth in the SOJ and ECS, the slope (1.57 ± 0.01) and intercept (–53.2 ± 0.4‰) were within the 95% confidence intervals of Eq. 2 (Table 3). The slope of the regression line was the steepest (1.93 ± 0.07) and gentlest (1.33 ± 0.02) during winter and summer, respectively, whereas the slope was steeper in the SOJ (1.68 ± 0.02) than the ECS (1.48 ± 0.01). The intercept corresponded to the following slope values: the lowest (–45.6 ± 0.6‰) and highest (–65.5 ± 2.6‰) during summer and winter, respectively. The δDsw–salinity relationship indicated that the slope and intercept were gentler and higher, respectively, in the ECS than the SOJ. In both the ECS and SOJ, the slope was the lowest in summer among the four seasons. The steepest slopes were observed during different seasons in the ECS (winter) and SOJ (spring) (Fig. 8 and Table 3).
List of coefficients and their standard errors, intercepts and their standard errors, sampling frequencies (n), and the coefficient of determination (r2) of the salinity–δDsw relationship above 200 m depth
| Slope | Intercept | n | r2 | |
|---|---|---|---|---|
| Excluded WNP | 1.57 ± 0.01 | –53.2 ± 0.4 | 2101 | 0.889 |
| winter | 1.93 ± 0.07 | –65.5 ± 2.5 | 427 | 0.618 |
| spring | 1.74 ± 0.02 | –58.9 ± 0.7 | 758 | 0.915 |
| summer | 1.33 ± 0.02 | –45.2 ± 0.6 | 821 | 0.863 |
| autumn | 1.53 ± 0.04 | –51.9 ± 1.4 | 95 | 0.937 |
| The East China Sea | 1.48 ± 0.01 | –50.0 ± 0.4 | 1063 | 0.943 |
| winter | 1.89 ± 0.14 | –63.9 ± 4.7 | 365 | 0.351 |
| spring | 1.65 ± 0.02 | –55.6 ± 0.8 | 489 | 0.915 |
| summer | 1.31 ± 0.02 | –44.4 ± 0.6 | 163 | 0.971 |
| autumn | 1.63 ± 0.03 | –54.9 ± 1.0 | 46 | 0.985 |
| The Sea of Japan | 1.68 ± 0.02 | –57.1 ± 0.8 | 1038 | 0.839 |
| winter | 1.57 ± 0.24 | –53.3 ± 8.1 | 62 | 0.426 |
| spring | 1.90 ± 0.03 | –64.3 ± 1.1 | 269 | 0.929 |
| summer | 1.43 ± 0.03 | –48.7 ± 1.0 | 658 | 0.763 |
| autumn | 1.74 ± 0.08 | –59.1 ± 2.7 | 49 | 0.909 |
The coefficient and intercept are calculated using the least squares method.
Seasonality in the relationships between δDsw and salinity in the (a–d) East China Sea (ECS) and (e–h) Sea of Japan (SOJ). The blue, black, and dotted lines are the regression lines of specific areas and seasons, specific areas (no seasonality), and other areas (i.e., the ECS one in the SOJ panels and the ECS one in the SOJ panels), respectively. The gray points indicate the value of the target area in different seasons.
Because the variation patterns of δ18Osw and δDsw respect to salinity were similar (Figs. 5 and 7), the relationship between δDsw and δ18Osw was strongly correlated (r2 > 0.963; Fig. 9a). The equation is as follows:
Relationships between δDsw and δ18Osw in (a) the total and (b) below 100 m depth datasets. In (a), the blue line indicates the regression line, and the red box represents the data range shown in (b). In (b), the colored and gray dots represent values below and above 100 m depth, respectively, and the color and shape indicate the sampling depth and area, respectively.
This relationship did not differ with depth, even though the samples were collected below 500 m depth, as plotted in Fig. 9b. The spatial difference in the relationship was unclear below 100 m depth in the ECS and SOJ (Fig. 9b).
The seasonal variations in the δDsw–δ18Osw relationships at ≤200 m depth showed significant differences (ANOVA, p < 0.001) in the ECS (Table 4 and Fig. 10). The slope was gentler (3.62 ± 0.12) during winter in the ECS than the other seasons in the ECS and SOJ (5.64–6.86). Although the coefficient remained constant during autumn in the ECS, the intercept was low (–0.690 ± 0.030). Typically, δDsw was always lower than that estimated from the regression line estimated from the data included the other seasons (Fig. 10d). In the SOJ, seasonality was not clear in the multi-linear regression model. It was selected based on the AIC but was not significant (p > 0.05).
List of coefficients and their standard errors, intercepts and their standard errors, sampling frequencies (n), and the coefficient of determination (r2) of the δDsw–δ18Osw relationship above 200 m depth
| Slope | Intercept | n | r2 | |
|---|---|---|---|---|
| Excluded WNP | 6.43 ± 0.03 | –0.176 ± 0.007 | 2101 | 0.962 |
| winter | 4.72 ± 0.15 | 0.304 ± 0.033 | 427 | 0.697 |
| spring | 6.28 ± 0.04 | –0.104 ± 0.011 | 758 | 0.964 |
| summer | 6.17 ± 0.04 | –0.281 ± 0.010 | 821 | 0.964 |
| autumn | 6.62 ± 0.17 | –0.433 ± 0.038 | 95 | 0.941 |
| The East China Sea | 6.40 ± 0.04 | –0.102 ± 0.011 | 1063 | 0.961 |
| winter | 3.62 ± 0.13 | 0.601 ± 0.029 | 365 | 0.694 |
| spring | 6.46 ± 0.07 | –0.116 ± 0.016 | 489 | 0.949 |
| summer | 6.09 ± 0.05 | –0.175 ± 0.022 | 163 | 0.988 |
| autumn | 6.86 ± 0.12 | –0.694 ± 0.034 | 46 | 0.987 |
| The Sea of Japan | 6.33 ± 0.04 | –0.248 ± 0.008 | 1038 | 0.962 |
| winter | 5.81 ± 0.38 | –0.185 ± 0.058 | 62 | 0.801 |
| spring | 6.10 ± 0.05 | –0.097 ± 0.014 | 269 | 0.984 |
| summer | 6.35 ± 0.06 | –0.301 ± 0.011 | 658 | 0.942 |
| autumn | 5.64 ± 0.29 | –0.227 ± 0.037 | 49 | 0.892 |
The coefficient and intercept are calculated using the least squares method.
Seasonality in the relationships between δDsw and δ18Osw in the (a–d) East China Sea (ECS) and (e–h) Sea of Japan (SOJ). The blue, black, and dotted lines are the regression lines of specific areas and seasons, specific areas (no seasonality), and other areas (i.e., the ECS one in the SOJ panels and the ECS one in the SOJ panels), respectively. The gray points indicate the value of the target area in different seasons.
The residuals in Eqs. 1 (δ18Osw–salinity relationship), 2 (δDsw–salinity relationship), and 3 (δDsw–δ18Osw relationship) were calculated. The horizontal distributions of residuals in the δ18Osw–salinity relationship showed negative residuals (observed δ18Osw was lower than δ18Osw estimated using Eq. 1) in the SOJ, particularly during summer (Fig. 11a). Positive residuals were typically observed during summer and autumn in the ECS, whereas during winter and spring, the difference in the area of residuals was not clear. The residuals in Eq. 1 significantly decreased with latitude during summer and autumn (p < 0.001), but not during winter and spring (p > 0.3).
Horizontal and seasonal distributions of residuals in the (a) δ18Osw–salinity (Eq. 2), (b) δDsw–salinity (Eq. 3), and (c) δDsw–δ18Osw relationships above 200 m depth. The dots and colors represent the sampling points and 1° × 1° median values of the target residuals, respectively.
A trend observed for the residuals in the δDsw–salinity relationship was similar to δ18Osw–salinity relationship (Eq. 2; Fig. 11b), but in the δDsw–salinity relationship, the positive residuals were not clear during autumn in the ECS. Significant negative relationships (p < 0.001) were observed between the residuals in Eq. 2 and latitude during summer and autumn. In addition, a significant negative relationship was observed during winter (p < 0.001), with low residuals at latitudes >40°N. The residuals in Eq. 3, δDsw–δ18Osw relationship, differed from those in the other two (Fig. 11c). Positive residuals were observed during summer, whereas negative residuals were observed during autumn. Significant negative relationships between the residuals in Eq. 3 and latitude were observed from winter to summer (p < 0.001); however, a significant positive relationship was observed during autumn.
The interannual variations in the residuals among Eqs. 1–3 were compared after adjusting for spatiotemporal variations as follows. The seasonal 1° × 1° median values were calculated (shown in Fig. 11) and then removed from the residuals. Values for 2020 and 2021 were omitted because of sampling limitations. A significant interannual difference was observed based on ANOVA (p < 0.03; Fig. 12). Most interannual median values were within the detection limit (±0.03 and ±0.2‰ for δ18O and δD, respectively), but those of δ18Osw– and δDsw–salinity during the summer and autumn and summer, respectively, of 2019 were lower than the detection limits. The interannual median values of δ18Osw–salinity were higher than 0.03‰ during the winter of 2018.
Interannual variations in residuals among the (a) δ18Osw–salinity (Eq. 1), (b) δDsw–salinity (Eq. 2), and (c) δDsw–δ18Osw (Eq. 3) relationships above 200 m depth after considering the spatial variations. The dashed lines denote the analytical precision (δ18Osw: 0.02‰ and δDsw: 0.1‰). Box plots show the mean values (thick horizontal lines within boxes), standard deviations (boxes), 95% confidence intervals (vertical bars), and outliers (closed circles). The values for 2020 and 2021 are omitted.
Some fish species, such as the Japanese sardine (S. melanostictus) and yellowtail (Seriola quinqueradiata), migrate from the ECS to the SOJ and vice versa (Furuichi et al., 2020; Yamamoto et al., 2007), and the other fish species such as the seabass (Lateolabrax japonicus) conduct local migration (Fuji et al., 2014). Our comprehensive data in the ECS to the SOJ could provide information on the variations of δ18Osw and δDsw to the otolith isotope-based study of fish migration not only at the basin scale but also at the local scale. Our δ18Osw–salinity relationship (Eq. 2; slope: 0.235 ± 0.002 and intercept: –7.94 ± 0.07‰) was close to those of Kodaira et al. (2016) at the Tsushima Strait (slope: 0.22 ± 0.02 and intercept: –7.3 ± 0.8‰) and Horikawa et al. (2015) (slope:0.23 and intercept: –7.74‰). However, it was different from that of Kodaira et al. (2016) (slope: 0.27 ± 0.02 and intercept: –9.1 ± 0.8‰) in the TWC region. Kodaira et al. (2016) observed a limited number of months (May, September, and October). Our results for spring and autumn in the SOJ were similar to those of Kodaira et al. (2016). Therefore, we considered that the equation of Kodaira et al. (2016) reflected seasonality in the SOJ but the summer and winter relationships were different. Below 200 m depth, the low δ18Osw and δDsw in the ECS were consistent with the results of Cruz Salmeron et al. (2022).
In the ECS and SOJ, Changjiang is the primary freshwater source (Beardsley et al., 1985; Liu et al., 2009). The Changjiang discharge exhibits seasonality and is the highest and lowest during June and January, respectively (Senjyu et al., 2006). The seasonality of the Chengjiang discharge is linked to the surface salinity of the ECS and SOJ (Senjyu et al., 2006). The intercepts of the δ18Osw–salinity and δDsw–salinity relationships in this study were –7.74 ± 0.07 and –52.0 ± 0.4‰, respectively (Table 3). The intercept of δ18Osw–salinity in the Changjiang was similar to those of previous studies, ranging from –8.8 to –7.1‰ (Zhang et al., 1990) and –6.1 to –11.6‰ (Deng et al., 2016). The δDsw–salinity values were also consistent with that of a previous study (Deng et al., 2016), ranging from –37.7 to –81.2‰. These results suggest that the Changjiang is affected by the water cycle in the ECS and SOJ. In general, δ18O and δD of rainwater vary widely in the Japanese Islands but overlap with those of Changjiang (Ichiyanagi and Tanoue, 2016), suggesting that the primary source of freshwater in the seas around Japan cannot be identified. However, in the less-saline water observed during spring in the SOJ, the δ18Osw and δDsw were different than those observed during summer in the ECS (Figs. 6 and 8). Less saline water was observed during spring in the SOJ near the local Japanese rivers in Toyama Bay; therefore, the effect of local Japanese rivers cannot be ignored in the water cycle of the SOJ. Local variations in δ18Osw–salinity and δDsw–salinity are observed around the Hokkaido area (Kubota et al., 2022).
δ18O and δD of the Chengjiang freshwater exhibit seasonality (Wang et al., 2019), with the δ18O and δD values of –7.9 ± 0.58 and –50.4 ± 5.81‰, respectively, during the dry season (winter) and –10.9 ± 0.39 and –76.9 ± 2.95‰, respectively, during the wet season (summer). This variation is slightly different near the river mouth (Li et al., 2020) because the contribution from two lakes changes the δ18O values at the river month. However, in general, δ18O is lower during the dry season than the wet season. The seasonality of δ18O and δD in the rainwater is similar to those of the Changjiang water, i.e., δ18O and δD are higher during winter than summer (Ichiyanagi and Tanoue, 2016; Tanoue et al., 2013; Yoshimura and Ichiyanagi, 2009). This seasonality in the Changjiang freshwater and rainwater did not agree with our seasonality results in the ECS (Tables 2 and 3), i.e., the intercepts of δ18Osw–salinity and δD–salinity were the lowest and highest during winter and summer, respectively, owing to the time lag during transportation from the river mouth to the sampling site. In the Tsushima Strait, freshwater transport peaks in July, but it is less from January to June (Isobe et al., 2002; Morimoto et al., 2012). Freshwater transport in the Tsushima Strait is controlled by the monsoon; a strong northerly monsoon inhibits the transportation of the Changjiang diluted water to the northern part during winter (Lie and Cho, 2016). In the SOJ, the surface water from the Tsushima Strait is ejected into the North Pacific or Okhotsk Sea after more than 5 months (Kodama et al., 2016). However, less-saline water was observed during summer in the ECS (Fig. 3), suggesting that the time lag was insufficient to explain this inconsistency. Another possible explanation is the interannual variations in δ18O and δD of the Changjiang freshwater. In our study, an interannual variation was observed (Fig. 12); however, the reason for its occurrence remained unclear. For example, a catastrophic flood occurred in China due to heavy rainfall during the summer of 2016 (Kundzewicz et al., 2020), but the relationships between δ18Osw–salinity and δDsw–salinity were not different from those of the other years (Fig. 12). Therefore, the interannual variations in δ18O and δD of the Changjiang freshwater need to be examined to explain this phenomenon.
A strong linear δD–δ18O relationship was observed in our study, whose slope was lower than the global meteoric (δD = 8.00 × δ18O + 10) (Craig, 1961) and ocean (δD = 7.37 × δ18O – 0.72) (Rohling, 2007) water lines. The δD–δ18O relationship during summer in the Luzon Strait is reported using the following equation: δD = 3.51 ± 0.43 × δ18O + 1.47 ± 0.11 (Wu et al., 2021), which is consistent with our results during winter in the ECS but not with those of other seasons. In the Changjiang water, δD–δ18O relationship exhibits several patterns: δD = 5.94 × δ18O – 7.29 in Wang et al. (2019), δD = 8.5 × δ18O + 15.5 in Li et al. (2020), and δD = 8.22 × δ18O + 14.27 or δD = 7.99 × δ18O + 14.27 in Deng et al. (2016). These results differ from our results (Table 3), and thus the salinity–δDsw and salinity–δ18Osw corresponded that the freshwater origin was Changjiang in our study area, but the δDsw–δ18Osw relationship did not agree. However, δ18O and δD of Changjiang freshwater are not stable (Deng et al., 2016; Zhang et al., 1990), and thus the disagreement with the δD–δ18O relationship in Changjiang freshwater is not rejected that the Changjiang is the primary freshwater source of our study region.
We constructed the equations for calculating δ18Osw and δDsw using salinity in the surface of the ECS and SOJ. Our sampling frequencies were 10 times higher than those of the previous datasets in this area; thus, seasonal variations were also estimated. Sea surface salinity is measured not only by ships but also by satellites (Lagerloef et al., 2008) and floats. Furthermore, it is estimated using ocean general circulation models. Our equations can provide δ18Osw and δDsw in the ECS and SOJ considering the interannual variations (Fig. 12). The δ18O values of fish otoliths, foraminiferal shells, and coral skeletons reveal the temperatures of habitats after considering δ18Osw (Bemis et al., 1998; Devereux, 1967; Ishimura et al., 2012; Weber and Woodhead, 1970). Thus, our equations and datasets provide new insights into the high-precision estimation of sea surface temperature and understanding the movement of marine organisms and paleoclimate.
The raw data was in the online supplementary file.
The authors wish to thank captains, crew members and scientists who helped on the sampling and measurement at sea especially for Tsuneo Goto, Yutaka Hiroe, Misato Nakae, Naoki Iguchi, Nobuaki Nanjo, Koh Nishiuchi, Youichi Seto, Shuichi Tanaka, Michitaka Tokuyasu, Tohya Yasuda, and Goh Yasuhara, and analysis on land to Junko Ibuki. The contact author has declared that none of the authors has any competing interests. This research has been partially supported by the Japan Fisheries and Education Agency and the Fisheries Agency of Japan and KAKENHI grants (16K07831, 16H02944, 18H04921, 19K06198, 21K18653, 22H05029, 23K26978, 24H02227 and 24H00075) from the Japan Society for the Promotion of Science.
