The Fukushima Daiichi Nuclear Power Plant (FDNPP) accident on March 11, 2011 released large quantities of artificial radionuclides into the ocean (Aoyama et al., 2012; Buesseler et al., 2012; Honda et al., 2012; Nakano and Povinec, 2012; Kaeriyama, 2017). Although there is considerable uncertainty, the estimated total ¹³⁷Cs inventory derived from the direct discharge of the contaminated water from Fukushima was approximately 3.0−4.0 PBq (Kaeriyama et al., 2016). The half life of ¹³⁷Cs is about 30 years. Moreover, additional tritium which half life is 12.3 years is planned to be released into the ocean from FDNPP in 2023. The FDNPP is located between the subtropical and subpolar gyres (Yoshida et al., 2015). The Kuroshio and Oyashio Currents, which are the strongest currents of both gyres, respectively, can actively disperse the radioactive materials, which can subsequently pollute fishery products in the surrounding seas (Yoshida and Kanda, 2012; Wada et al., 2013; Nakata and Sugisaki, 2015; Kaeriyama et al., 2016). In particular, trans-Pacific migratory fish, such as the Pacific bluefin tuna (Thunnus orientalis) are vulnerable to radioactive materials (Fujioka et al., 2018). It is, therefore, crucial to investigate the path of dispersion, and concentration of the radioactive material to reduce their potential risks.

After the FDNPP accident, many observational and numerical modeling studies were undertaken to determine the path and concentration of the radioactive material. Honda et al. (2012) observed ¹³⁴Cs and ¹³⁷Cs in the sea surface layer near the FDNPP one month after the accident and found concentrations in zooplankton and suspended matter that were two orders of magnitude higher than those recorded before the accident. The half life of ¹³⁴Cs is about 2.1 years. Moreover, Buesseler et al. (2012) tracked ¹³⁴Cs and ¹³⁷Cs released from FDNPP in seawater 30–600 km offshore and found a high correlation between near-shore eddies and radioactive material distribution. The radioactive material was distributed widely in the surface layer along the Kuroshio Current and its extension (Behrens et al., 2012; Buesseler et al., 2012; Honda et al., 2012). Meanwhile, Kaeriyama et al. (2013) assessed the Cs concentration by collecting seawater in wide areas of the Western and Central Pacific Ocean in July 2011, October 2011, and July 2012, and tracked the path of surface dispersion from FDNPP using a simple particle tracking numerical model. Similarly, Zhao et al. (2021) simulated the path and concentration of tritium released from FDNPP under various future release duration scenarios over an extended period using a numerical model. They proposed that the simulated tritium in the surface commonly moved eastward over time regardless of the considered scenario. Collectively, the findings of these observational and numerical studies suggest that radioactive Cs from FDNPP, in 2011, rapidly spread eastward via surface dispersion.

The Kuroshio current and its extension are major deliverers of radioactive materials and serve as the southern boundary of their surface movement (Buesseler et al., 2012; Honda et al., 2012). However, a different distribution pattern was observed for radioactive materials in the subsurface. More specifically, Kaeriyama et al. (2014) observed a ¹³⁴Cs concentration peak in the subsurface (100–500 m) south of the main axis of the Kuroshio extension and suggested that FDNPP-derived radioactive Cs was included in Subtropical Mode Water (STMW) owing to winter mixing. Moreover, Kumamoto et al. (2014) observed high concentration of ¹³⁴Cs in the subsurface layer along 149°E in February 2012. Meanwhile, Kaeriyama et al. (2016) observed meridional transects of the vertical distribution of ¹³⁷Cs along 147°E and 155°E from October to November 2013, 19 months after the accident, reporting its peak concentration in the subsurface, which corresponded to the depth of STMW and Central Mode Water (CMW).

Although previous studies suggest that STMW and CMW are potential deliverers of radioactive materials, observations were limited in time and space, thus impeding elucidation of the precise path taken and travel time of radioactive material in the subsurface. Numerical model studies may prove useful for tracing the path of radioactive material in the subsurface. In fact, Zhao et al. (2021) suggested that simulated tritiums, primarily distributed at 400–600 m depth, is correlated with mode waters. However, the associated coarse model resolution prevented resolution of the different paths taken by the two mode waters.

Due to various model uncertainties, modeling studies often fail to accurately simulate the subsurface path of radioactive material (Nakano and Povinec, 2012; Zhao et al., 2021; Garraffo et al., 2016; Zulema et al., 2016). Moreover, it is challenging to simulate the realistic formation and subduction of mode waters in the study area. Therefore, the contribution of mode waters to the dispersion of radioactive materials in the subsurface remains unclear.

This study seeks to clarify the subduction of the radioactive substances released directly from FDNPP with mode waters, as well as their paths taken and travel time within the subsurface. To this end, a high resolution three-dimensional numerical model is employed. The model reliability is assessed by comparing the observed and simulated characteristics of mode waters, including temperature, salinity, and location depth.

The radioactive material released from FDNPP spread quickly eastward by the Kuroshio extension after the accident (Buesseler et al., 2012; Honda et al., 2012). Figure 6 displays the monthly surface ¹³⁷Cs concentration from the model results in April, August, December 2011 and annually from February from 2012 to 2020. In the early stage, most of the released ¹³⁷Cs moved eastward while some of the ¹³⁷Cs moved southeastward by the effect of meandering Kuroshio currents ( Figure 6A ). After that, the high concentrated ¹³⁷Cs core moved eastward ( Figures 6B, C ) and was then distributed widely in the west, which was the STMW formation region in February 2012 ( Figure 6D ). It was located primarily within the central area where CMW formed in February 2013 ( Figure 6E ). Some ¹³⁷Cs in the sub-arctic gyre region moved south near the FDNPP in 2014 ( Figure 6F ). This recirculation by the sub-arctic gyre increased the ¹³⁷Cs concentration off the FDNPP after 2015. Most of the ¹³⁷Cs moved further east and reached the west coast of North America in 2015 ( Figure 6G ). This simulation result was consistent with results of previous studies (Behrens et al., 2012; Zhao et al., 2021), which simulated the arrival of a high ¹³⁷Cs concentration to the western coast of North America after 5 years of the accident.

The ¹³⁷Cs off the western coast of North America moved further north and south by the currents ( Figures 6H-J ). The northward moving ¹³⁷Cs entered the sub-arctic gyre region. A portion of the southward moving ¹³⁷Cs moved to the west by the Equatorial Current in 2019 and 2020 ( Figures 6K, L ).

Notably, A ¹³⁷Cs concentration > 1 Bq/m³ was not detected to the southwest of the Kuroshio extension from 2012 to 2020, as the Kuroshio current and its extension acted as a southern boundary for the dispersion of radioactive material along the surface. Similarly, the model result did not show a high concentration of ¹³⁷Cs to the south of the Kuroshio extension, which was consistent with previous studies (Kaeriyama et al., 2013; Yoshida et al., 2015; Aoyama et al., 2016; Periáñez et al., 2019; Wu et al., 2019; Garraffo et al., 2016). However, our model result does not account for ¹³⁷Cs and ¹³⁴Cs in the east of Taiwan Island observed by Men et al. (2015) in May–June 2012. Moreover, the ¹³⁷Cs and ¹³⁴Cs detected east of Taiwan in 2012 may have resulted from atmospheric deposition.

Observations indicated that high concentrations of ¹³⁷Cs appeared in the subsurface at an approximate depth of 400 m south of 35°N, whereas the ¹³⁷Cs concentration in the surface was not significant ( Figure 5 ). This suggests that the ¹³⁷Cs dispersion in the subsurface traveled along a different path than that in the surface, which is consistent with suggestions made in previous studies (Kumamoto et al., 2014; Kaeriyama et al., 2016; Zhao et al., 2021; Garraffo et al., 2016). Two vertical sections were selected to examine the temporal change in the dispersion of ¹³⁷Cs by the STMW and CMW.

The vertical section along line L1, which crosses the STMW formation area, showed subsurface dispersion over time ( Figure 7 ). The subducted ¹³⁷Cs in March/April 2011 ( Figure 7A ) remains in the subsurface in August and December 2011 ( Figures 7B, C ) and connected to the newly subducted ¹³⁷Cs in February 2012 ( Figure 7D ). The surface ¹³⁷Cs subducted to the subsurface along the density of 25.0–25.9 kg/m³ corresponding to the STMW. In 2013, the ¹³⁷Cs concentration decreased in the surface layer, while it dispersed widely in the subsurface ( Figure 7E ). All ¹³⁷Cs became isolated from the surface after 2014 and moved south over time ( Figures 7F-I ). In 2018, the ¹³⁷Cs core moved to 10°N at approximately 200 m ( Figure 7J ). The concentration of the southern tip at a depth of 200 m increased in 2020 ( Figure 7L ).

However, the subduction of ¹³⁷Cs along the line L2, which crosses the CMW formation area, does not occur before December 2011 due to time delay of ¹³⁷Cs arrival in the surface ( Figures 8A-C ). After that, the vertical section along line L2 also showed the subduction of ¹³⁷Cs over time, similar to line L1 ( Figures 8D-L ). The ¹³⁷Cs in the surface layer along line L2 presented a longer retention duration than that near line L1. The ¹³⁷Cs in the subsurface along line L2 was isolated from the surface after February 2016 and moved southward over time. The ¹³⁷Cs concentration along line L2 was higher and its distribution was wider than that along line L1. Most ¹³⁷Cs moved along the 25–26.3 kg/m³ density layer, corresponding with the STMW and CMW layers.

The ¹³⁷Cs concentration along line L2 was high in STMW and CMW, which is consistent with the observation of a high ¹³⁷Cs concentration in multiple density layers corresponding to mode waters along the 152°W vertical section in the North Pacific (Macdonald et al., 2020). The ¹³⁷Cs concentration from the model was also high in both STMW and CMW along line L2. Wu et al. (2019) observed a high Pu isotopic ratio in the tropical region at 10–20°N in winter in 2014 and suggested a possibility that it was dispersed by CMW. The CMW, formed in the Central North Pacific, moves east by the North Pacific current, then turns clockwise to the south and west in the subsurface layer (Oka and Suga, 2005; Kumamoto et al., 2014).

Previous observations have reported the subduction of radioactive Cs due to mode water formations but did not show the detail dispersion path and its travel time by mode waters in the subsurface (Kaeriyama et al., 2014; Kumamoto et al., 2014; Kaeriyama et al., 2016). Zhao et al. (2021) suggested that simulated radioactive tritium could be dispersed by mode waters; however, they did not resolve the separate dispersion by two mode waters. Our model result suggests a three-dimensional dispersion path of radioactive material by STMW and CMW.

The long term evolution of the ¹³⁷Cs distribution by STMW and CMW in the subsurface was examined over the whole simulation period. Figure 9 shows the monthly mean horizontal ¹³⁷Cs concentrations on 25.4 and 26.3 kg/m³ density layers, corresponding to STMW and CMW. Most ¹³⁷Cs moving with the STMW dispersed east, however, some also moved southward after the accident, which is consistent with the observation of Kumamoto et al. (2014). The southward movement of ¹³⁷Cs by STMW continued until 2014 ( Figures 9A, B ). The eastward movement of ¹³⁷Cs by STMW turned southward in the Central North Pacific in 2014 ( Figures 9A-C ). It then moved westward in 2017 and reached the east of Taiwan in 2020 ( Figure 9D ). The ¹³⁷Cs moved by the CMW followed a clockwise path as in STMW. However, its path was larger, with a longer radius than that in STMW. The ¹³⁷Cs arrived east of the Philippines in 2020 ( Figure 9D ). The horizontal distribution of each year from 2012 to 2020 is shown in Supplementary Figure 8 .

Vertical sections of ¹³⁷Cs distribution across the overlaid region for the two mode waters displayed their different depths. The ¹³⁷Cs in the STMW of the meridional section along 180° was located at a depth of 200–300 m at approximately 30°N, however that in CMW was distributed from the surface to a depth of approximately 500 m in February 2014 ( Figure 9E ). The descending structure of the ¹³⁷Cs suggested CMW formation between 35°N and 40°N. The ¹³⁷Cs moved further south by STMW and CMW in February 2020 ( Figure 9F ). The mid-depth of ¹³⁷Cs moving by the STMW and CMW was approximately 200 and 400 m, respectively. The zonal section along 25°N showed that the moving depth by the STMW was shallower than that by the CMW in February 2017. The depths of both mode waters became deeper in the west ( Figure 9G ). This spatial distribution of ¹³⁷Cs persisted from February 2017 to 2020 ( Figure 9H ).

Figure 10 shows the time series of the spatial mean concentrations of ¹³⁷Cs in each mode water from the model. Tracers located south of 40°N were considered. Both mean concentrations decreased due to diffusion and decay over time. The mean concentration of ¹³⁷Cs in CMW was higher than that in STMW for the entire period, and both concentrations in the STMW and CMW showed weak seasonal changes with high concentrations in winter. The increased concentration of ¹³⁷Cs in winter may have resulted from the supply owing to vertical mixing. Considering the half-life of ¹³⁷Cs is as long as 30 years, the main cause of the decrease in the concentration is likely diffusion.

A schematic diagram showing the dispersion of ¹³⁷Cs released into the ocean by STMW and CMW over long periods of time has been proposed based on the model result ( Figure 11 ). Some ¹³⁷Cs moving via STMW moved south after the accident, however most moved clockwise at a relatively shallow depth (200–300 m). The ¹³⁷Cs moving via CMW also exhibited clockwise movement; however, its path had a larger and deeper radius than STMW (300–500 m). In contrast, the real depth of CMW in the Central North Pacific was likely shallower considering the depth difference between the model and observation, as indicated in Figure 4 .

Considering that tritium is to be additionally released into the ocean at the FDNPP planned for 2023, the subsurface dispersion of ¹³⁷Cs that leaked into the ocean in 2011, as described in this study, is expected to provide important insights regarding the path and effect of tritium after 2023. Because the discharging point is consistent with this experiment, it is expected that the discharging tritium from 2023 will give insight into understanding the dispersion in the subsurface, which is unclear. In addition, Pacific blue fin tuna (Thunnus orientalis) spend long periods of time in both mode water layers, whereas skipjack tuna (Katsuwonus pelamis) passes STMW during northward migration (Fujioka et al., 2018; Kiyofuji et al., 2019). Hence, the fishery products may face a risk of pollution; therefore, continuous tracking and observation of radioactive material by each mode water are necessary.

In this study, the subsurface dispersion of radioactive materials released into the ocean from the FDNPP accident were simulated by a three dimensional numerical model from March 11, 2011, to December 31, 2020. The model performance was validated by comparisons with various observed parameters, including water temperature, salinity, and distribution of radioactive materials at the ocean surface and subsurface.

The model results revealed that the subsurface dispersion path of the radioactive material differed from that in the surface. The ¹³⁷Cs spread by two mode waters in the subsurface, whereas the surface ¹³⁷Cs was driven by the strong Kuroshio and its extension, which also acted as a southern boundary for ¹³⁷Cs dispersion.

Previous observational studies suggested the possibility of the subsurface dispersion of radioactive Cs by mode water. However, the dispersion path and travel time of radioactive materials in the subsurface have not been known due to the spatiotemporal limitation of observational data. Our numerical model result enabled us to decipher the path and travel time of ¹³⁷Cs by STMW and CMW. ¹³⁷Cs subducted to the subsurface by STMW in the western region bifurcated into two paths; one moved southward while the other moved clockwise. ¹³⁷Cs subducted by CMW in the Central North Pacific was located beneath the STMW. The ¹³⁷Cs in the CMW moved clockwise along an outer and deeper path than that in the STMW. The final destination for the ¹³⁷Cs moving with STMW and CMW was east of Taiwan and the Philippines nine years after the accident, respectively.

The ¹³⁷Cs concentration in STMW and CMW primarily decreased due to diffusion over time; however, it increased slightly during each winter, when mode waters were supplied by vertical mixing. The mean concentration of ¹³⁷Cs in CMW was higher than that in STMW; in addition, the concentration in CMW decreased faster than that in STMW. Because the area and volume of CMW is larger than those in the STMW, the decrease of mean concentration in CMW by diffusion might be faster.

Our model will have important implications for estimating the potential impact of Japan’s plan to discharge contaminated water into the subsurface ecosystem. Therefore, special attention is needed to the effect on Pacific bluefin tunas (Thunnus orientalis) or skipjack tunas (Katsuwonus pelamis), which spends extended durations in STMW and CMW.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Conceptualization, S-TL and Y-KC; Methodology, JJ and S-TL; Writing Draft, S-TL; Review and Editing, Y-JT, GK and Y-KC; Visualization, S-TL; Supervision, Y-KC. All authors contributed to the article and approved the submitted version.