LIU Chunlin ,ZHANG Chi ,TIAN Yongjun ,WANG LiangmingLIN Longshan,LI Yuan,and WATANABE Yoshiro

1) Key Laboratory of Mariculture, Ministry of Education,Ocean University of China,Qingdao 266003,China

2) Frontiers Science Center for Deep Ocean Multispheres and Earth System (FDOMES), Ocean University of China,Qingdao 266100,China

3) Laboratory for Marine Fisheries Science and Food Production Processes, Pilot National Laboratory for Marine Science and Technology, Qingdao 266237, China

4) Third Institute of Oceanography,Ministry of Natural Resources, Xiamen 361000,China

5) Atmosphere and Ocean Research Institute,The University of Tokyo,Kashiwa 277-8564,Japan

Abstract After decades of low year classes,the stock of Japanese sardine (Sardinops melanostictus) has begun to recover since the mid-2000s.The hatch dates and otolith growth rates of age-0 juvenile sardine,which were collected in the subarctic Oyashio waters in autumn 2018,were determined from an otolith microstructure analysis.The sardines were hatched from late January to late April,while mostly in February and March.The otolith growth rate increased continuously up to 60 d after hatching and thereafter decreased.The revealed growth rate in a crucial growth period is faster than that reported for juvenile sardines collected in the 1990s,which is coincided with the recent recovery trend of the sardine stock.Two groups with different hatch dates,growth histories,and migration routes were identified using unsupervised random forest clustering analysis.They were considered inshore and offshore migration individuals in accordance with recent researches.In the offshore group,a high proportion of sardine juveniles hatched late and grew faster in the Kuroshio– Oyashio transitional waters,a finding consistent with the hypothesis of growth-rate-dependent recruitment.This finding on the population composition and growth rate of juvenile sardine in the Oyashio waters can be a basis for an improved prediction of their survival and provides us with valuable information on the recruitment processes of this stock during the period of stock recovery.

Key words Sardinops melanostictus;otolith microstructure;growth history;unsupervised random forest clustering;recruitment

1 Introduction

The Japanese sardine (Sardinops melanostictus) is one of the major commercial small pelagic fish species in the western North Pacific,with annual catches once contributing more than 4 million tons,accounting for one-third of Japan’s total catches and 1/20 of the global capture production (Yasudaet al.,1999).However,the sardine population off the Pacific coast of Japan has undergone dramatic fluctuations over multi-decadal timescales (Yasudaet al.,1999).Such variation is suggested to be driven by the annual variation in recruitment abundance,given that recruitment depends on the growth and survival rates during early life history,especially the juvenile stage after metamorphosis (Watanabeet al.,1995).Sea temperature and food density along the age-0 sardines’ migration routes are considered the main factors that affect survival and determine recruitment abundance (Takasukaet al.,2007;Nishikawaet al.,2013).Low temperatures,as well as high plankton biomass generated by intensive vertical mixing provide a favorable environment for sardine growth,which is critical for desirable survival rates and eventual abundant recruitment (Takahashiet al.,2008;Nishikawaet al.,2011,2013;Nishikawa,2019).Favorable environmental conditions in recent years,together with fisheries management based on improved knowledge (Ishidaet al.,2009;Yatsu,2019),have led to the general recovery of sardine stock since the mid-2000s after the dramatic decline in the 1990s.As a result,the catch in 2018 was 451000 t,exceeding 10 times of the catch in 2008 (27000 t).The current data indicate a trend of continuous increase in the overall stock of Japanese sardine (Japan Fisheries Research and Education Agency,2020).

Despite a series of studies,the effects of specific envinmental factors on sardine survival have not been entirely specified.One of the main reasons is the difficulty of accurately defining the environmental history of sardine during northward migration from the spawning ground in the Kuroshio waters to the feeding ground in the Oyashio waters across the nursery ground in the Kuroshio– Oyashio Transition Zone (KOTZ) (Fig.1).Sakamotoet al.(2018)provided a potential solution by the interdisciplinary combination of high-resolution otolith stable oxygen isotope ratio (δ18O) analyses and numerical simulations.Their results indicated that the migration routes of sardines arriving in feeding grounds in the Oyashio waters differed during high-stock and low-stock periods.Furthermore,under conditions of increased biomass,sardine populations can migrate farther offshore and then divide into several groups taking several different migration routes in the inshore and offshore waters (Sakamotoet al.,2018),a conclusion that supported the inference proposed by Noto and Yasuda (1999).Based on backtracking by otolith daily age,Sakamotoet al.(2018) further suggested that the sardines start moving northward departing the Kuroshio Extension from mid-May to early June,that is,at ages 41– 70 d (Takahashiet al.,2008).The growth during this late larval and early juvenile stages within KOTZ is sensitive to environmental factors and is accordingly crucial for the sardines’ survival to their eventual recruitment success (Takahashiet al.,2008;Takahashiet al.,2009;Sakamotoet al.,2017).Therefore,the different environmental conditions experienced by the sardine groups taking different northward migration routes should result in various growth patterns;consequently,the survival rate and contribution of each migration group to age-0 year class size can differ because of habitat-specific influences.The proportion that each migration group occupies in the nursery ground is essential information needed to fully understand the population dynamics and improve stock assessments.In addition,the population dynamics of sardine in the Oyashio feeding ground are closely related to the abundance of age-1 recruits (Wada,1988);however,little ecological information is available for this region given the difficulty of sample collection in the offshore waters.Thus,more data are required to understand the key processes in the early life history and recruitment phase of Japanese sardine,especially during the period of stock recovery.

Otoliths are calcium carbonate crystals formed in the inner ear of fishes,and otolith microstructure analysis has become an essential tool in various fields of fish biology (Campana and Thorrold,2001).The growth trajectory recorded in otolith microstructure enables the tracking of the early life history and separation of groups from a mixed stock(Brophy and Danilowicz,2002;Clausenet al.,2007).Here,we aimed to reconstruct the early life growth history of juveniles of Japanese sardine in the Oyashio waters during a period of stock recovery.This information will improve our understanding of the recent population dynamics of the Japanese sardine.We used the hatch date distribution and growth histories during critical pre-recruit periods,which were recorded by daily increments in otoliths,to identify potential juvenile groups with different growth patterns.New information on the population composition can benefit the estimation of sardine survival among groups and ultimately provide a precise management tool for this important resource.

2 Materials and Methods

2.1 Sample Collection

Juveniles of Japanese sardine were collected from a light falling-net survey in offshore waters of the Oyashio Current in the western North Pacific (41˚– 43˚N,151˚– 160˚E)(Fig.1).The survey was conducted on 4– 15 September 2018.Individuals were randomly obtained from each survey sample,preserved by freezing (-20℃),and thawed prior to the analysis.The standard length (SL) and total length of each juvenile were measured to an accuracy of 1 mm.

Fig.1 Schematic of the current structures (dotted line with arrow,Oyashio Current;dashed line with arrow,Kuroshio Current and Kuroshio Extension) and important locations (i.e.,spawning,nursery,and feeding grounds) in the early life of Japanese sardine on the Pacific side of Japan.The dark-gray and light-gray regions in the lower left corner indicate spawning grounds in the low-stock and high-stock periods,respectively.The approximate area of KOTZ is demarcated by the gray dashed lines;the survey area in 2018 is marked by an ellipse.

2.2 Otolith Measurement

Sagittal otoliths were dissected out,cleaned with pure water,and air-dried,and the right one was embedded in epoxy resin.After hardening,the otoliths were ground with sandpaper (240– 4000 grit) on the sagittal plane and finally polished with alumina polishing suspension (0.3 µm) until the primordia and daily rings can be identified (Ohshimoet al.,1997).The otoliths were photographed with photomicrographic equipment (OLYMPUS BX53) at 400 × magnification (Fig.2).Using ImageJ software,the number of daily increments was counted twice,with a 20-day interval for cross-validation.Only otoliths from individuals whose counting error was <5% were included in the subsequent analysis.Of the 350 samples captured (SL:101– 149 mm),219(SL:101– 145 mm) were precisely aged and used for the otolith microstructure analysis.The SL distribution of the 219 sub-sampled fish approximate that of the total sample(Fig.3).

Fig.2 Otolith microstructure observed in the sagittal section from a 192-day old sardine with 123 mm SL.The white line shows the measurement axis.

Fig.3 SL distribution of (a) the total captures and (b) the subsamples used for cluster analysis.The numbers above the columns indicate the quantities of Japanese sardines in each length category.

The daily-age-at-capture of each juvenile was calculated by adding 2 to the average number of otolith daily increments because the first daily ring is formed 2– 3 d after hatching (Hayashiet al.,1989).The hatching dates were then back calculated from the sampling dates and daily ages.

To investigate the early growth history of the sardines,we measured the otolith increments from hatching to the age of 100 d,following the methods of Takahashiet al.(2008).The increment width was measured along the axis from the nucleus to the posterior margin of the otolith.The mean increment width in 10-day intervals was calculated for each individual as follows:

whereR10iandR10(i-1)are the increment radii at ages 10i(i=2,3,···,10).Given that the first daily increment corresponds to the age of 3 d,I10(i=1) was calculated exceptionally asI10=(R10-R3)/7.

2.3 Clustering Analysis

Unsupervised random forest (URF) clustering analysis was applied to detect the potential groups of sardines with different migration routes (Breiman,2001;Afanadoret al.,2016).URF clustering is an unsupervised classification approach that leads to the measurement of natural dissimilarity between observations by generating an ensemble of individual tree predictors (Seligsonet al.,2005).Classifier ensembles promote an optimal trade-off between diversity and accuracy.Random forests bootstrap the data from the samples from the training set with randomly selected feature subsets that were evaluated at each node of the decision tree,whereas the final decision is made by the decision fusion of all trees by majority voting (Kandaswamyet al.,2011).Compared with other Euclidean distance-based clustering methods,random forest dissimilarity can maximize the elimination of autocorrelation between variables.With the R package ‘randomForest’ (Liaw and Wiener,2002;Shi and Horvath,2006),I10i(i=5,6,7) characterizing the juvenile growth rate after metamorphosis and the part of the lifespan that is important for recruitment were selected as clustering factors (Takahashiet al.,2008).Hatch dates were also used because certain time inconsistencies were expected in different year classes.Prior to clustering,we determined the number of clusters using the R package ‘cluster’.

Among the identified clusters,growth differences between ages and clusters were compared by repeated-measures analysis of variance (RMANOVA) and the post hoc multiple comparison verification,because there can be correlations between differentIvalues for each fish due to the dependent repeated measurements appear in the growth rates at different life periods,which cannot be solved by analysis of variance (ANOVA) which requires an independence assumption.because of the correlation between the differentIvalues for each fish.Before this analysis,Mauchly’s test of sphericity was performed to judge whether the variances of the differences between various measurements were equal among repeatedly measured data.In addition,the appearance of differences in SLs between identified groups was defined by Student’st-test.

3 Results

3.1 Hatch Dates

For the 219 fish samples,the hatch-dates was from 24 January to 26 April,indicating that the sardine spawning lasted for about three months.The hatch dates were concentrated to February and March,accounting for 90.4% of all the individuals (Fig.4a).

3.2 Growth Rates Through Early Life Stages

The estimated 10-day mean otolith daily growth rates during the early life stages showed a dome-shaped configuration (Fig.4b).For all the 219 samples,the values ofIincreased from 3.8 µm d-1(I10) to the maxima of 11.9 µm d-1at age 50– 60 d and then decreased to 6.1 µm d-1in fishI100.

Fig.4 (a) Hatch date distribution and (b) mean 10-day-interval otolith growth rates (I) during the early life stages calculated for the total 219 samples of Japanese sardine.Error bars:95% confidence intervals.

3.3 Group Identification

Two groups of juvenile sardines were identified by URF clustering as shown in Multidimensional Scaling (MDS)(Fig.5).The group sizes weren=175 for G1 andn=44 for G2.The SLs of G2 were greater than those of G1,as determined byt-test (P<0.001) (Fig.6).

Fig.5 MDS cluster dendrograms of Japanese sardine collected in autumn 2018.

Fig.6 Identified SL distribution of the Japanese sardines in the two migration groups.

The hatch date of G1 sardines was from 24 January to 26 April,and the peak was in March (61.7%).G2 sardines finished hatching by late March (24 January to 26 March),with most members hatching in February (63.6%) (Fig.7).

Fig.7 Frequency distribution of hatch dates of Japanese sardine in groups G1 (n=175) and G2 (n=44).

The otolith 10-day growth incrementIof G1 (Fig.8)showed a similar pattern as that of whole samples (Fig.4b),with an increase from 4.0 µm d-1during the first 10 d to a maximum of 12.8 µm d-1at 51– 60 d after hatching and thereafter dropping abruptly to 5.4 µm d-1(I100).By contrast,G2 showed a relatively slower growth (Fig.8).TheIvalues took longer to reach the highest value atI80and thereafter decreased less rapidly compared with G1.The 10-day mean otolith incrementIthrough the larval stage to early juvenile stage increased with age from 2.9 µm d-1(I10) to 9.8 µm d-1(I80) and thereafter decreased to 8.6 µm d-1(I100).

Fig.8 Changes in the mean otolith growth rates (I) by 10-day intervals in the early life stage of juvenile Japanese sardine identified in two groups.Error bars:95% confidence intervals.

The Greenhouse-Geisser-corrected result of RMANOVA showed significant differences between the sardine age and groups (at 99.9% confidence interval,same as below).Not only did both factors showed statistical significance on the otolith growth rate,butIalso varied from group to group with ontogeny (P<0.001).

The post-tests showed a significant inter-group difference in every life stage (P<0.001).In the first 70 d from hatching,G1 grew faster than G2.Then,their growth rates crossed,and theIvalue of G2 became larger sinceI80.

4 Discussion

This study provided information on early growth history and hatching date distribution of Japanese sardineSardinops melanostictusin the subarctic Oyashio feeding ground in the western North Pacific in recent years.Spawning of the 2018-year class recruited to the feeding grounds in Oyashio waters had occurred mainly in February and March,which is consistent with the surveys conducted during the past high-stock periods in the 1980s (Watanabeet al.,1996,1997;Nishikawaet al.,2013).Our results are similar to the data of Furuichiet al.(2020),who reported that the hatch dates in 2018 were from 14 February to 24 March.The slightly longer hatching duration found in our research may attribute to a larger number of samples,which can interpret the spawning event in 2018 in a more reliable manner.Although Japanese sardine spawns from late winter to early spring,spawning can change temporally among different year classes.In the low-stock period of 1995–2004,spawning trends were detected in later months,with peaks in March and April (Takasukaet al.,2008;Nishikawaet al.,2013).The shift of spawning to earlier months in 2018,which was consistent with the past high-year classes,implies that the sardine stock recovered from the low numbers in the 1990s.Niinoet al.(2021) also presented an earlier hatching trend since 2013 and a faster otolith growth rate during 40– 60 d in 2018.The understanding of early spawning as a sign of stock recovery of sardine coincides with the higher otolith growth rates at age 41– 70 d,which are considered as an index of the abundance of recruits,than the rates reported in 1996– 2003 when the stock was low (Takahashiet al.,2008).This finding is consistent with the results of Furuichiet al.(2020).The results also provide robust evidence of stock recovery and depict a strong year class.

Two groups with different growth histories in 2018 were successfully identified by URF clustering.Given the dissimilar hatch dates and otolith growth trajectories,the two migration groups probably arrived in the Oyashio waters from different spawning areas and possibly along inshore(G2) and offshore (G1) migration routes (Okunishiet al.,2012).Based on previous studies,we inferred that the G1 sardines hatched near Kuroshio meanderings and then migrated along an offshore route by the Kuroshio and Kuroshio Extension Currents,whereas G2 individuals hatched far away from the Kuroshio Current but were attracted by a coastal flow and migrated northwardviaa coastal route in which transportation was less dependent on the eastward currents (Japan Fisheries Research and Education Agency,2020).The hatching of the offshore migration group G1 occurred later than that of the inshore migration group G2 and then passively transported eastward (Japan Fisheries Research and Education Agency,2020).In comparison with inshore individuals,the offshore group that followed the Kuroshio and Kuroshio Extension Currents exhibited a significantly higher growth rate at the larval stage (Itohet al.,2011).The offshore group would have experienced low sea surface temperatures before departing from the Kuroshio Extension to the Oyashio waters (Sakamotoet al.,2018);hence,the offshore group would show a higher growth rate(Takahashiet al.,2009).The inshore group included individuals that hatched early in low-latitude waters and those hatched in the same spawning ground as the offshore group but were not transported by eastward currents (Sakamotoet al.,2018).The environmental sea temperatures they experienced would have been higher than that for the other group,thus leading to a slower growth.

The contributions of the two groups to the sardine juvenile population in the Oyashio waters were different.G1 sardines grew faster than G2 sardines in the first 70 d and especially during the 41– 70-day period after hatching in which the growth processes essential for recruitment occurred,accounting for the majority (79.9%) of the Oyashio stock.Takahashiet al.(2008) detected the ‘growth-dependent recruitment’ process for Japanese sardine in KOTZ(Fig.1).As survival and recruitment are a positive function of growth rates in the juvenile stage (Takahashiet al.,2008),the abundance of survivors is a direct function of fish growth.Given the successful feeding and favorable physical conditions,the group with an overall high growth rate in a suitable environment has low vulnerability to starvation and exhibits resistance to predation,thus leading to excellent recruitment (Anderson,1988;Okunishiet al.,2012).Fast-growing and large-sized sardine larvae have a high tolerance to starvation (Zenitani,1999) with improved survival in coastal waters.Okunishiet al.(2012) also regarded low-weight Japanese sardines as abortive individuals owing to their vulnerability to survive the environment.

Our results on otolith growth were further compared with the growth rates of sardine in the low-stock period reported by Takahashiet al.(2008).The estimates ofI50andI60for G2 were similar to those calculated for the juveniles obtained in 1997 and 1998.In these years,with low-stock size,sardine was considered to have spawned in the inshore waters,have been transported,and migrated along the inshore route.In this study,G2 fish accounted for about 20% of the whole mixed stock,and the total abundance of age-0 individuals was approximately five times that recorded in 1997 and 1998.These proportions are similar to the separate estimates made by the Japan Fisheries Research and Education Agency (2020).These results highlight a correlation between the growth rate,population composition,and recruitment abundance of the sardine.Further insights into the details of these relationships would help in the comprehensive evaluations of the survival rates and recruitment of sardine in the western North Pacific.

We observed that the SLs of the G2 group were greater than those of the G1 group (Fig.6).Considering that the otolith growth rate of G2 fish surpassed that of G1 fish as juveniles older than 80 d (I80),and the mean age of G2 fishes was greater than that of G1 fishes due to earlier hatch month,this result is understandable.The SL compositions indicate that SL-at-capture alone cannot be used to discriminate G1 from G2 for management.

In conclusion,this study provided valuable information on the hatch dates and early growth history of the Japanese sardine in the Oyashio waters.Our results provide credible evidence of the current recovering condition of the sardine stock.Otolith microstructure analysis was proved to be an effective method for separating migration groups with different growth patterns and contributes to the improved description of the recruitment process during a high-stock period.Future research can focus on studying the migration routes and composition of inshore and offshore groups in detail.Long-term multi-year or inter-year field studies,combined with other technologies such as particle tracking and simulations,are required to continuously track fish movements in the KOTZ.We propose that Japanese sardines on the Pacific side of Japan should be considered as two groups for management and stock assessments because their recruitment trends will likely continue to vary.

Acknowledgements

We acknowledge the crew members of the Zhongtai Oceanic Fishery Co.for their help with the sample collection.This work was supported by the National Natural Science Foundation of China (Nos.41930534,41861134-037,and 41876177) and funding to the Third Institute of Oceanography through the National Program on Global Change and Air-Sea Interaction (No.GASI-02-PAC-YDaut).