During the observation period, seasonal thermocline, halocline, and subsurface oxygen maximum gradually formed in the upper mixed layer, separating the Subtropical Mode Water (STMW) from the sea surface (Figs. 3a–c). Although a small semidiurnal perturbation of the internal tidal wave appeared for all pressure levels on a nearly half a day cycle, the vertical pressure level and vertical gradient of the main thermocline were relatively stable. However, in the main thermocline at levels of 400–900 dbar, small perturbations were observed from late-April to mid-May. Because this study focused on the perturbations equivalent to small water mass parcels, our analysis highlighted the details in the small water parcels for a 9-day period from April 28 to May 6, 2014. During this period, the Seaglider was positioned around 31.7–32.8°N, 144.5–144.7°E, which was the southwestern side of an anticyclonic eddy and far west (over 250 km) of a cyclonic eddy detached from the KE (Fig. 1a). At the same time, the northwestward current velocity estimated by the Seaglider gradually increased and eventually reached around 40 cm s− 1 (Fig. 1b).
During the highlighted period, small water parcels with temperature, salinity, and DO anomalies were observed below the winter mixed layer depth (MLD) along 26.0–27.0 σθ isopycnal surfaces (Figs. 4a–c), which almost corresponded to the density ranges of the North Pacific Central Mode Water (NPCMW; 26.2–26.4 σθ; Nakamura 1996; Suga et al., 1997) and the North Pacific Intermediate Water (NPIW; ~26.8 σθ; Talley et al., 1995). In the layer below the MLD, many anomalies in temperature, salinity, and DO were detected with values of < − 0.5 °C, − 0.05 PSS-78, and > 15 µmol kg− 1, respectively. The water parcel sizes varied significantly. The horizontal width ranged from a few kilometers to several ten kilometers, while the vertical thickness was mostly 0.01–0.5 σθ. Low (high) temperature anomalies corresponded with low (high) salinity anomalies.
The high (low) DO anomalies did not always correspond with the low (high) temperature and salinity anomalies. Hence, the water parcel characteristics demonstrate that subducted water has complicated water mass properties. For example, water parcels observed around 26.7–26.9 σθ at a distance of 830–850 km along the x-axis have properties of low/high temperature and salinity anomalies, but only low DO anomalies. On the other hand, water parcels around 26.4–26.9 σθ at a distance of 790–820 km have low/high temperature, salinity, and high/low DO anomalies.
Regarding planetary potential vorticity (f ρ−1 dρ dz− 1, hereafter PV), the significance of the relationship between low PV and low temperature/salinity and high DO anomalies was unclear below the main thermocline. Thus, the water parcels originate from the north of the KE where high PV water is dominant (not shown here). Detecting a wide potential density range suggests that small water parcels may come from a broader area on the northern part of the KE, including levels of > 26.8 σθ surface, which are not in contact with the winter cooling atmosphere (e.g., Talley et al.,1995; Yasuda et al., 1996).
To clarify whether anomaly bias exists, Figs. 5a–c plot the appearance frequencies of temperature, salinity, and DO anomalies divided into three levels of upper (26.2–26.4 σθ), middle (26.45–26.6 σθ), and lower (26.7–26.8 σθ) layers, respectively. The frequency was counted from the interpolated data on the isopycnal surfaces every 0.01 kg m− 3 for all profiles. The characteristics of the water parcels are likely biased with high DO, low temperature, and low salinity concentrations in the middle and lower levels. By contrast, such biases are not obvious in the upper layer. Especially in the middle layer, negative-ward/positive-ward salinity/DO anomalies exist over a broad range (− 0.38 to 0.15 PSS-78 / −21 to 68 µmol kg− 1), which also suggests that water parcels are subducted and transported from a broader range north of the KE. Additionally, the size of each water parcel varies; the water parcels with a minimum 5–10 km horizontal width and 30–100 dbar vertical thickness frequently appeared throughout the observation period. The detected spatial size based on the Seaglider’s results means the Seaglider observation has a large potential to accurately measure the size of the water parcels even below the main thermocline using a fine vertical sampling rate.
Figure 6 shows a θ–S diagram of the Seaglider data throughout the whole observation period along with the Argo float data in the western North Pacific Ocean observed in 25–45°N, 140–160°E for January 1 – Aprirl 30, 2014 for comparison. Based on the θ–S relationship from the Argo float data, a wide range of water mass characteristics were observed (saltier/fresher side of the figure is the southern/northern part of the KE). The water mass characteristics obtained from the Seaglider were broadly detected and within the range of those from the Argo float data. Because the Seaglider collected data south of the KE in the subtropical region, the subtropical water mass characteristics were dominant and corresponded to the saltier side of the θ–S diagram (orange dots). To represent the water parcels with anomalous low salinity or high DO water, salinity and DO anomalies over three times of the standard deviations (3 σ) estimated from those average values through the observed period were highlighted. The characteristics of some water parcels (colored with over 3 σ of low salinity and/or high DO concentration variability) observed in 26.0–26.9 σθ were elongated toward the fresher side of the diagram (green dots), indicating that the water parcels may originate from the northern part of the KE (close to plots colored with green from Argo data). The lower salinity of the water parcels around 3–9 °C is significant (< 34.0 PSS-78), which is the level in the lower layer of the main thermocline.
Figure 7 shows the σθ–S diagram of the Seaglider data using the same color scheme as in Fig. 6. The water parcels along 26.2, 26.5, and 26.7 σθ displayed low salinity characteristics, corresponding to minimum salinity values of 33.8, 33.6, and 33.5, respectively. The low salinity properties extended along the isopycnal surfaces, indicating that the small water parcels occasionally intrude at arbitral isopycnal surfaces. Figure 8 shows the σθ–DO diagram of the Seaglider data using the same color scheme as in Figs. 6 and 7. In 26.2–26.7 σθ isopycnal surfaces, the DO concentration was basically the same within 150–200 µmol kg− 1, while a high DO water over 200 µmol kg− 1 was detected within the isopycnal surfaces, which had a maximum of 250 µmol kg− 1. However, a high DO water does not correspond well with the low salinity water shown in Figs. 7 and 8. In addition, the high DO value below 26.7 σθ was unclear, indicating that higher DO water tends to be distributed in the upper layer relative to the lower salinity water in the lower layer. Figures 7 and 8 indicate that the σθ level appearing in low salinity water differs slightly from that of high DO water, suggesting that the origin and characteristic of subducted water parcels at the sea surface north of the KE differ from each other.
Based on the in situ DO concentration, temperature, and salinity values, the apparent oxygen utility (AOU) and saturated oxygen concentration were calculated. Hence, the relative residence time of water parcels was estimated since subduction from the sea surface saturated the seawater with oxygen by contacting the atmosphere. Figure 9 shows σθ–AOU/Cs (a ratio of AOU and saturated oxygen concentration, Cs) diagram from the Seaglider data. Note that small water parcels with larger negative salinity anomalies and the positive DO anomalies were located below the euphotic zone (here we assume above 200 m). Here, a low AOU/Cs ratio means nearly saturated oxygen water, which occurs closer to the surface, while a high value of the ratio indicates far away from the sea surface. In other words, the ratio of water parcels positioned on the left side of the diagram indicates that less time has passed since the water parcels were on the sea surface. Assuming the water parcels move with time along a σθ surface after subduction, the relative age of a water parcel to the surrounding water mass can be estimated by comparing the AOU/Cs ratios. At a few σθ surfaces where water parcels with over 3 σ of DO and/or salinity anomalies were detected, AOU/Cs ratios of 0.18, 0.18, and 0.25 on 26.2–26.5 (26.7) σθ were observed, while those in the surrounding water were 0.3–0.4, 0.35–0.45, and 0.4–0.5, respectively. If the water parcels are advected with only some diffusive processes after subduction, then the water parcels with high DO and low salinity should be about 30–60% younger than the surrounding water on the same σθ, assuming that the water is unaffected by other processes to control the DO concentration. However, it is impossible to estimate the absolute age of a water parcel. Here, we discuss the age of water parcels as well as possibilities to detect younger water parcels using the high-resolution numerical model OFES.