1.    Marine fish dispersal during early life stages is a critical process that affects population connectivity and dynamics, yet individual-level field observations remain difficult. This limits our ability to identify the factors controlling the dispersal process, including the active swimming of larvae and pelagic juveniles.

Here, we propose a method to reproduce ontogenetic vertical and horizontal movements of fish on continental shelves. The approach combined high-resolution profiles of oxygen stable isotope values (δ¹⁸O) from otoliths, hydrodynamic simulations, and a hidden Markov model. The shelf environment was compressed into a representative cross-section, in which fish movements were predicted by analysing temperature and salinity distributions in relation to otolith δ¹⁸O.

Applying the method to Japanese jack mackerel in the East China Sea shelf revealed general ontogenetic vertical migration and variable cross-shelf movements. Twelve fish collected from the southern to northern shelves originated mainly from the southern outer shelf, which is influenced by the northward-flowing Kuroshio current. Fast growers in the first 30 days moved to the inner shelf earlier and stayed south, while others remained in the outer shelf were transported to the northern shelf.

These findings indicate that horizontal movements, potentially influenced by active swimming, play an important role in the retention and dispersal of fish that spawn at shelf edges. This novel method can greatly advance our understanding of marine fish dispersal and improve conservation efforts.

Juvenile T. japonicus were collected during a bottom trawl survey in the East China Sea conducted by the Japan Fisheries and Education Agency from May to June 2014. We analysed geographic differences in transport and migration by selecting otoliths from 12 juveniles at four stations with relatively large catches: Sts. S and C on the southern shelf and Sts. N-1 and N-2 on the northern shelf. Station St. N-2, near N-1, served to confirm regional movement differences.

The extracted otoliths (sagittae) were cleaned, air-dried, and embedded in epoxy resin. To facilitate the observation of daily increments and subsequent micromilling, the otoliths were ground with sandpaper to expose the core, polished with an alumina polishing suspension (BAIKOWSKI International Corporation), and coated with nail polish. The daily increment widths were quantified from the core to edge using an otolith measurement system (RATOC System Engineering Co., Ltd.). As an indicator of the early life growth rate, the SL at 30 dph was estimated from the otolith radius at 30 dph, based on the reported linear relationship between the otolith and body sizes (Xie et al., 2005):

SL₃₀ = 2.65 + 0.0425* OR₃₀

where SL₃₀ and OR₃₀ are the SL and otolith radius at 30 dph, respectively.

Before micromilling, the polished nail surface was removed using acetone. Otolith portions were formed over an average of 5 days (18 days at the core, 7 days at the edge, and 4 days for the remaining sections on average). These portions were sequentially milled from the core to edge using a high-precision milling system (Geomill 326, Izumo-web). The milling depth was set to 50 μm from the surface. Following each milling, the extracted otolith powder was collected in a small reaction tube for isotope analysis, and the otoliths were cleaned with an air duster to prevent cross-contamination. The δ¹⁸O was measured using a microscale stable isotope analytical system (MICAL3c with IsoPrime 100, Ishimura et al., 2004; 2008). The otolith powders reacted with 104% phosphoric acid at 25 °C, resulting in the release of CO₂ gas, which was purified and introduced into the mass spectrometer. The δ¹⁸O values were expressed in δ-notation relative to the Vienna Pee Dee Belemnite (VPDB) reference standard, reported as ‰ values, with an analytical precision better than ± 0.10 ‰. These procedures yielded otolith δ¹⁸O profiles with corresponding calendar date ranges for each value.