The overall TLs of Japanese glass eels in Yilan, Taiwan, ranged from 52 to 62 mm, forming a normal distribution with an average length of 56.9 mm (Fig. 2). However, the mean TLs of each batch ranged from 54.6 to 59.9 mm. Based on otolith measurements, it has been suggested that the total length of the Japanese glass eel is highly correlated with age23,38,39. In this study, the estimated age of each arrival wave in Yilan was found to positively correlate with the mean TL (Fig. 3), supporting the relationship between age and TL of the glass eel. The variated mean TL of each arrival batch in Taiwan suggested that each wave possessed a different mean transport time. Variations in mean ages between arrival waves may reflect differentiations in the current speed during larval transport from the spawning site to Taiwan. Alternatively, these batches of glass eels may spawn at different sites in the range of 12–16° N during the new moon period, and then drift along different routes on the NEC and Kuroshio to the Yilan offshore area. Han et al. (2016) found that each batch of glass eels caught offshore in December of three consecutive year classes at Yilan, had a narrow range of ages between years. This implies that most individuals in the arrival peaks spawn in the same month and thus belong to the same cohort. Furthermore, tracer simulation also supported that each batch of glass eels takes approximately 160–170 days to reach the Luzon Strait from their spawning site (Fig. 6). All of this data suggests that the potential mixing of glass eels between monthly spawning cohorts was low, at least in Taiwan.
One of the main dispersal routes of Japanese eel larvae is from Taiwan to Japan by the Kuroshio stream. It is estimated that the time needed for the larval eels to travel from Taiwan to Kyushu, Japan (approximately 1,000 km) is 25.3 ± 6.4 days14. The mean velocity of the Kuroshio along the Pacific coast of Japan is approximately 0.7–1.4 m/s40,41; thus, it takes 8–16 days for the eel larvae to be transported from western to eastern Japan (approximately 1,000 km). This means that the recruitment time lag between Taiwan and Japan is expected to be close to one month, and the mean TLs of the glass eels in Japan should be greater than those in Taiwan because of their increased mean ages. The mean TLs of the glass eels in Japan ranged between 57.9 and 60.1 mm42, which is greater than that of the glass eel in Taiwan (56.9 mm). In addition, the recruitment of the Japanese glass eel usually started in November in Taiwan and December in Japan, and the arrival peak of the Japanese glass eel was December/January in Taiwan and January/February in Japan (Fig. 7), coinciding with a lag of approximately one month. Since the estimated age of the glass eel was positively correlated with its TL (Fig. 3), the mean TLs of the glass eels in Japan are expected to be larger than those of glass eels in Taiwan. According to an eel trader, the Japanese glass eels from Chiba, Japan, which is the easternmost part of the Kuroshio and the farthest from the spawning site, usually have the largest TLs.
Japanese eels spawn in restricted areas (12–16° N and 141–142° E) during a specific time span (May through September)14,39. The larvae are passively transported via the NEC to the east Philippines for 3-4 months and then enter the Kuroshio and move toward their habitats in East Asia. For the glass eels that arrived in Taiwan, the dispersal distance was longer in the NEC (approximately 2,500 km) than in the Kuroshio (closer to 1,000 km), and the mean NEC velocity was generally low compared to that of the Kuroshio (Fig. 5). However, the monthly velocity variations of the NEC were low compared with those of the Kuroshio (Fig. 5). A previous study has also demonstrated that recruitment dynamics and distribution of Japanese glass eels might chang with global warming, due to changes in the velocity, structure, and location of the NEC, NEC bifurcation, and Kuroshio with warming climate43. This suggests that the transport time the glass eels need to reach the Luzon Strait from the spawning site is significantly impacted by both the NEC and the Kuroshio. Eel larvae that spawn between May and June (early recruits) require a longer transport time (167–174 days) than those spawning between August and September (late recruits, 160 days) (Fig. 6). Therefore, a greater percentage of early recruits may metamorphose in the southern areas of East Asia due to their slower dispersal. By contrast, more late recruit eel larvae may metamorphose in the northern areas of East Asia due to their faster transport. Indeed, the relative abundance of glass eels near Taiwan relative to Japan was significantly higher in earlier recruitment months (November–December), and significantly lower during later recruitment months (February–March) (Fig. 8). Seasonal variations in the location of the NEC bifurcation can be entertained as one of the most important factors, since it is located at its southernmost latitude in May, and moves northward after that, reaching its northernmost point in September. In other words, the distance between the NEC bifurcation and Kuroshio might be wider when the NEC bifurcation is in the southern region of its range, thereby causing a longer duration of drift among the earlier recruitment cohort. In March, Japanese glass eels were scarce in Taiwan; however, they were abundant in Japan in April. This supports the concept that late recruits, which were transported with faster oceanic current velocity, would bypass Taiwan and metamorphose mainly downstream of the Kuroshio and then continue on to Japan.
Although the mean TLs of Japanese glass eels in Taiwan decreased monthly with significance, the mean TLs of each batch ranged between 54.6 and 59.9 mm with high variation. Multiple biotic/abiotic factors must be involved in this TL variation. In a previous study, it was found that ENSO events had a substantial impact on the TL of glass eels in Taiwan23. The TLs of Japanese glass eels were significantly greater during El Niño years and less during La Niña years. The northward shifts in the NEC bifurcation and the southward shift of the salinity front during the El Niño years might have led to larvae encountering slower currents and broadening the distance between the spawning site (south of the salinity front) and the NEC bifurcation13,44,45, thereby extending the journey time required for the larvae to enter the Kuroshio from their spawning grounds23. Since ENSO events occur periodically, the effect of ENSO on the transport time of eel larvae in each month would be offset over the long term.
It has been suggested that local eddy currents along the transport route may trap some leptocephali and result in a small amount of mixing between monthly cohorts33. Chang et al. (2018) indicated that v-larvae may be able to remain in eddies passively (physical trapping) due to mesoscale eddy nonlinearity, and/or actively (biological attraction) due to rich food supplies in those eddies. In addition, otolith analysis indicated that Japanese eels may have a recruitment route through the mesoscale eddies to the east of Taiwan, in addition to the direct transfer route from the NEC to the Kuroshio42. However, based on the present study, the transport of each eel recruitment batch was generally stable in Taiwan, suggesting that the simultaneously spawning eel cohort usually forms a group and moves together. The eel larvae transported by eddies, if present, may mainly merge downstream of the Kuroshio, thus reducing the mixing degree between monthly cohorts in Taiwan. In either situation, direct observational fish larvae data from the open ocean in mesoscale eddies is necessary to evaluate their true degree of contribution to eel larvae dispersal.