Otolith biogeochemistry reveals possible impacts of extreme climate events on population connectivity of a highly migratory fish, Japanese Spanish mackerel Scomberomorus niphonius

Introduction

Climate change is precipitating large changes to a range of ecological conditions, leading to directional shifts in species' distributions, richness, abundance, demography and phenology (Chen et al. 2023; García Molinos et al. 2015; Poloczanska et al. 2013, 2016). These collective impacts of climate change are likely to alter population connectivity, the exchange of larvae, juveniles or adults across a species range, which is the vital component of metapopulation and fishery ecology (Kool et al. 2013; Wasserman et al. 2012). Due to their reliance on seasonal resource availability in habitats that can be far apart, highly migratory species appear to be particularly susceptible to these climate change-induced consequences, which may make adaptation challenging (Robinson et al. 2009). Previous studies exploring the potential effects of climate change on population connectivity focused mainly on the pelagic larval duration, highlighting the importance of ocean fronts, currents and circulations (Bharti et al. 2022; Carson et al. 2010; Munday et al. 2009; Wolanski et al. 2021; Woodson et al. 2012). Nevertheless, evaluating those effects is more challenging for highly migratory fish, as explicit assessment of species migration potential and understanding connectivity patterns at a large spatial scale are necessary.

In marine environments, instead of reacting to changes in long-term mean conditions, responses to extreme climatic events may be more overarching (Frolicher and Laufkotter 2018; Wernberg et al. 2012). The biggest source and control of yearly fluctuations in the climate is the El Niño Southern Oscillation (ENSO) (Cai et al. 2014; Santoso et al. 2017; Wang et al. 1999). Though characterized by recurring (2 to 7-year) oscillations between a warming and a cooling phase in tropical Pacific sea surface temperatures (SST), it is a "global pattern of anomalies" (Cane 1986) that has strong impacts on global and regional marine ecosystems. The China Seas have shown multiple physical and biochemical responses to El Niño events, including increased SST (Ma et al. 2019), lower sea levels (Wang et al. 2018), abnormal currents (Li 2016), weakened monsoons (Zhou et al. 2007), and southward-shifted rain bands (Zhang et al. 2017). The abundance and distribution of marine organisms were significantly impacted by this environmental variation, for example, phytoplankton showed significant decreased correlation with elevated SST (Liu et al. 2019) whereas some small pelagic fish increased substantially in biomass (Ma et al. 2019). By impacting physiological processes or disrupting biotic interactions (e.g., predator–prey interactions), ENSO events are likely to cause great changes in population connectivity of migratory fish in China Seas.

Japanese Spanish mackerel, Scomberomorus niphonius, is widely distributed in the temperate waters of the western North Pacific, supporting an economically valuable commercial fishery in China, Japan and Korea (Horikawa et al. 2001; Qiu and Ye 1996; Shoji and Tanaka 2005). S. niphonius undertakes long, seasonal migrations and has been observed to move into coastal waters of China Seas from spring to summer to breed and spawn, and move back to deeper waters from fall to winter (Fig. 1). The traditional acknowledgement claimed that there are two populations along the coast of China: (1) the Yellow Sea and Bohai Sea population; and (2) the East China Sea population (Horikawa et al. 2001). However, most recent studies applying advanced technologies namely mitochondrial DNA analysis (Shui et al. 2009), otolith phenotypic analysis (Zhang et al. 2016) and otolith chemistry analysis (Pan et al. 2020b) provided evidence for the existence of a metapopulation and large-scale connectivity between the Yellow Sea and the East China Sea. Since the 1990s, the distribution of S. niphonius showed a northward expansion with increasing water temperature (Fujiwara et al. 2013; Yang et al. 2022). Furthermore, the spawning grounds of S. niphonius would move northward when SST was high (Shui et al. 2009). Considering that the 2015–2016 El Niño has spread its impacts to the ecosystem of China Seas (Yin et al. 2021), the climate-induced range shifts and migration alteration probably influence the population connectivity of S. niphonius.

Figure  1.  Map indicating sampling regions (Qingdao (QD), Lvsi (LS), Xiangshan (XS) and Fuzhou (FZ)) of S. niphonius (Japanese Spanish mackerel). The four sampling regions are from three main spawning areas: northern Yellow Sea (NYS), southern Yellow Sea (SYS) and East China Sea (ECS) which are the natal origins indicated by Pan et al. (2020b)

DownLoad: Full-Size Img PowerPoint

Despite this breadth of background, the complex life histories of marine migratory fish make the investigations of their responses to extreme climate events difficult and rare. Otolith biogeochemistry offers a potentially efficient and cost-effective means to evaluate the impacts of ENSO events on population connectivity (Pan et al. 2020a; Reis-Santos et al. 2023; Walther 2019). As they continuously accrete, trace and minor metals from the local environment are permanently incorporated into the crystalline matrix of the otolith. Otolith increments are unlikely to be subject to resorption due to the metabolic inertness (Powles et al. 2006). Although otolith microchemistry can represent a combination of local ambient chemistry and individual physiology, the resulting elemental composition may produce a distinctive chronological 'signature' that could be used as a natural tag (Campana and Thorrold 2001; Elsdon and Gillanders 2002) to distinguish location and infer ontogenetic change. This approach has been applied successfully to identify the population connectivity and natal origin in the highly migratory species, such as Chilean jack mackerel (Trachurus murphyi), bigeye tuna (Thunnus obesus) and yellowfin tuna (Thunnus albacares) (Ashford et al. 2011; Rooker et al. 2016). In regions where there is a paucity of long-term data, fish otoliths have been used increasingly to evaluate the climatic effects on fish populations (Lee and Punt 2018; Reis-Santos et al. 2021; Wu et al. 2023). However, studies using otolith biogeochemistry to evaluate the impacts of extreme climate events on population connectivity are still limited.

In this study, we assessed the population connectivity of S. niphonius collected from the main spawning grounds throughout its distribution in the Yellow Sea and East China Sea. To mitigate the impact of fish/otolith size on the concentration of micro-chemical elements, all samples used in our study were 1-year-old spawning adults, carefully selected to ensure a relatively uniform fork length. By analyzing the otolith biogeochemistry signatures of samples from 2016 to 2018, during which 2015–2016 was a strong ENSO year, we aimed to evaluate the temporal variations in population connectivity and determine the relationship with the ENSO events. First, we compared the chemistry of the whole life to examine empirically how otolith chemistry varies spatially and temporally. Then, we focused on the larval and adult stages, using random forest classification and clustering to identify the natal origins and to determine if the source-sink mechanism differed greatly for the ENSO years. Our study will help identify potential changes in population connectivity of highly migratory species in response to extreme climate events, and highlight the further application of otolith biogeochemistry to explore the climate-induced impacts on fish populations.

Materials and methods

Samples collection

S. niphonius were collected from four main spawning grounds along the coast of the Yellow Sea and the East China Sea: Qingdao (QD), Lvsi (LS), Xiangshan (XS) and Fuzhou (FZ) (Fig. 1), which cover nearly all known spawning areas across this region (Zhang et al. 2016). Based on the former study, they could be treated from three natal origins: the East China Sea, the southern Yellow Sea and the northern Yellow Sea (Pan et al. 2020b). Samples in this study were collected in April and May from 2016 to 2018. All fish were sampled randomly from catches by commercial vessels operating gill net (mesh size 90–110 mm). All specimens were refrigerated immediately and transported to the laboratory. The laboratory procedures followed those of Pan et al. (2020b). A total of 220 1-year old individuals in spawning or spent condition, confirmed respectively by otolith year ring and gonad status, were selected for elemental analysis. S. niphonius could become sexually mature at one year old (Qiu and Ye 1996). Therefore, the life history of these adults covered one complete migration cycle. The Oceanic Niño Index (ONI) is NOAA's primary index for tracking the oceanic part of ENSO, calculated by the rolling 3-month average temperature anomaly—difference from average—in the surface waters of the east-central tropical Pacific Ocean, near the International Dateline. Index values of + 0.5 or higher indicate El Niño. Values of -0.5 or lower indicate La Niña. Based on ONI, samples collected in 2016 have experienced the process of the 2015–2016 El Niño event (Fig. 2). Detailed information of samples for otolith chemistry analysis is summarized in Table 1. It should be noted that the samples from 2016 were different from that of our former study (Pan et al. 2020b). The former experimental analysis work was finished in 2017. All the samples in this study were analyzed in the same single batch in 2021 (including preserved 2016 samples) to make the results more precise and convincing.

Figure  2.  Schematic illustration of the periods of elemental analysis (gray shadow) in this study during different stages of El Niño events. The Oceanic Niño Index (ONIs) were derived using the data from NOAA climate prediction center (http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml)

DownLoad: Full-Size Img PowerPoint

Table  1.  Detailed information of S. niphonius (Japanese Spanish mackerel) for otolith chemistry analysis

Sampling sea area	Sampling region	Capture time	Number	Fork length (mm)	Mean age (days)	Sex (M/F)
NYS	Qingdao	May 2016	15	450 ± 42	344	7/8
	Qingdao	May 2017	20	443 ± 23	332	10/10
	Qingdao	May 2018	20	449 ± 19	324	10/10
SYS	Lvsi	May 2016	15	440 ± 21	335	7/8
	Lvsi	May 2017	20	452 ± 16	352	9/11
	Lvsi	May 2018	20	441 ± 36	340	11/9
ECS	Xiangshan	April 2016	15	455 ± 35	352	9/6
	Xiangshan	April 2017	20	445 ± 23	338	12/8
	Xiangshan	April 2018	20	439 ± 39	336	10/10
	Fuzhou	April 2016	15	448 ± 23	342	8/7
	Fuzhou	April 2017	20	450 ± 24	346	10/10
	Fuzhou	April 2018	20	461 ± 33	350	13/7
Mean values ± standard deviation. The four sampling regions are from three main spawning areas: northern Yellow Sea (NYS), southern Yellow Sea (SYS) and East China Sea (ECS) which are the natal origins indicated by Pan et al. (2020b)

 | Show Table

DownLoad: CSV

Otolith preparation and elemental analysis

In the laboratory, one whole sagitta from each pair was taken at random for each fish and cleaned with deionized water. After air drying, the otoliths were embedded in epoxy resin (Epofix; Struers, Copenhagen, Denmark), which were later sectioned to approximately 400 μm in thickness in the transverse section-plane to incorporate the core, along the proximal–distal axis with an Isomet^® 1000 precision linear saw (Buehler, Coventry, UK). Each otolith section was mounted on a glass microscope slide and grounded by hand with 500, 1200, 2000, and 4000 grit paper (Struers) (Wu et al. 2023; Xu and Chan 2002). The sections were then polished using aluminum micropolisher (0.3 μm) with an automated polishing wheel (Roto Pol-35; Struers, Copenhagen, Denmark) to expose the core. Thereafter, polished otoliths were sonicated in an ultrasonic bath for 5 min, and rinsed with Milli-Q water (Millipore, Molsheim, France). After decontamination, all samples were oven-dried at 38 ℃ overnight for chemical analyses.

Laser ablation inductively coupled plasma mass spectrometry (LAICP-MS, Laser–New Wave UP213, ICP-MS–Agilent 7500ce) was used to obtain time-related element: Ca profiles. Laser ablation was programmed at a wave length of 213 nm, high voltage of 10 kV, pulse rate of 10 Hz. Ablation spots of 40 μm diameter were located at 80 μm intervals (from center to center) along the longest transect from the otolith core to the dorsal-distal edge (energy density of 9.29 J/cm², dwell time of 5 s and wash-out time of 5 s, dwell depth: 50 μm). This axis was chosen for easy interpretation, which has been explained in Pan et al. (2020b). Sample gasses were extracted from the chamber through a smoothing manifold facilitated by a helium and argon stream (He gas flow 800 mL min⁻¹). Two standard samples NIST612 and MACS-3 (National Institute of Standards and Technology, USA; United States Geological Survey, USA) were analyzed at the beginning and end of the sampling sessions for every five samples. The analysis involved a 100 s background count at the beginning and at the end to determine the actual limits of detection (LODs), and relative standard deviations (% RSD) based on replicated measurements of the standard sample were calculated to reflect the level of precision achieved for each element. All results were expressed as concentration ratios of elements to calcium (Amano et al. 2013). Otoliths were analyzed for the presence of eight elements (⁷Li, ²⁴Mg, ⁴⁴Ca, ⁵⁵Mn, ⁵⁶Fe, ⁵⁹Co, ⁸⁸Sr, ¹³⁸Ba). However, we only focused on Li: Ca, Mg: Ca, Sr: Ca and Ba: Ca in the present study, as they were most informative to reveal the migration pattern and population connectivity of S. niphonius (Pan et al. 2020a, b).

Otolith microstructure

Digital photographs were taken of these otolith sections after LA-ICPMS analysis (Nikon SMZ 1000) and the distance of each laser ablation pit relative to the otolith core was expressed as distance from their center to the first pit center (0 μm). The average of all the data spots along the ablation line was used as the elemental signature for the whole life of S. niphonius (Fig. 3). The laser pits were then each assigned to a life-history stage based on otolith increment analyses. It has been proven that the immature fish mix to a large extent when they feed and overwinter in the extensive offshore waters (Pan et al. 2020b). In this study, we only focused on the larval (core) and adult (edge) stages. For each life stage, the spots falling within that area identified in the otolith were averaged (Fig. 3). Typically, the core zone of the otolith was visible, and was formed in the first 9–10 days for Scomberomorus fishes (Lewis and Mackie 2002). For the edge data, the last two spots at the edge of each otolith were deemed representative of the adult phase. The last two spots represent a period of the life history of 20–25 days, which should cover the whole stage from their gonads maturing to atrophying (Qiu and Ye 1996). These spots were used to estimate average elemental concentrations representative of the sampling locations and the likely most recent environment inhabited by each fish.

Figure  3.  S. niphonius (Japanese Spanish mackerel) otoliths section showing the line of ablation spots analyzed from core to edge. The analysis line is representing the entire life history (whole life), the core zone is representing the early larval phase (Larval) and the edge zone is representing the adult phase (Adult). The figure was adapted from Pan et al. 2020b

DownLoad: Full-Size Img PowerPoint

Data analysis

Average elemental concentration was calculated as element-to-Ca ratios for the whole life, larval and adult stages, respectively. Since data could not be transformed to satisfy both normality and homoscedasticity assumptions, Kruskal–Wallis tests were carried out to determine whether the mean elemental ratios for the whole life differed among sampling locations and sampling years. A Dunn's multiple comparison test was used to identify pairwise differences in element ratios between spawning grounds and sampling years using the dunn.test package (v1.3.3) in R (R Core Team 2015). The population connectivity was estimated by the ability of the otolith chemical signature of the otolith core and edge to discriminate among spawning grounds according to Random forest (RF) classification (Breiman 2001). The main advantage of this method is that it makes no assumptions on variable distributions or on linear relations between variables (Mercier et al. 2011; Stekhoven and Bühlmann 2012) and accommodates continuous as well as ordinal or categorical variables. Therefore, Li: Ca, Mg: Ca, Sr: Ca and Ba: Ca in our analysis, whether they met parametric tests' assumptions or not, were used as predictors. In RF, each tree is grown using 2/3 of the available data, the remaining 1/3 of the data not used to grow a particular tree being used to assess the classification accuracy of each tree. RF classification error is then calculated as an aggregate error from all trees.

Clustering analysis was performed on the larval (natal) region of the otolith to gain insights into the number of sources of adults and the contribution rate of each identified source to the four sampling locations in the present study to interpret source–sink dynamics. The clustering method developed by Shi and Horvath (2006) was applied, using RF (a supervised learning method) in an unsupervised way, hereafter called RF clustering. RF clustering is a two-step process. In the first step, unsupervised RF was applied to the larval stage data set. Capture location was ignored in this process. Instead, a synthetic dataset was created through random sampling from the product of empirical marginal distributions of the different elements (Shi and Horvath 2006). Then, RF was used to separate observed from synthetic data (used as the class factor) and to produce a similarity matrix, defined by the frequency at which two individuals end up in the same terminal node of the trees (Breiman 2001). In the second step, the similarity matrix was transformed into a dissimilarity matrix (dissimilarity = √(1 − similarity)) and used as input for partitioning around medoid (PAM) clustering (Kaufman and Rousseeuw 2009). The appropriate number of clusters was determined using minimum average silhouette width (Kaufman and Rousseeuw 2009).

Results

Spatial and temporal variability in otolith elemental signatures

Otoliths were successfully analyzed from 220 age-1 spawned S. niphonius (Table 1). A total of 4908 detected spots were obtained with a relatively narrow range of 19 to 21 in one otolith. Of all the analyses measured, limits of detection (LODs) (μmol/mol) for Li: Ca (0.817), Mg: Ca (1.970), Sr: Ca (0.852) and Ba: Ca (0.0191) were all well below the detected concentrations in otoliths. The analytical accuracy of the standard across all samples was high for all elements with relative standard deviation (%RSD) ranging from 1.70 (Li) to 5.73 (Ba).

A summary of elemental ratios (Li: Ca, Mg: Ca, Sr: Ca and Ba: Ca) for the whole life is provided in Fig. 1. Overall differences in mean elemental ratios among locations were significant for Mg: Ca and Ba: Ca in all three sampling years (P < 0.001, Table 2). Post-hoc Dunn's test comparisons indicated that values for Mg: Ca were higher in LS than in other locations (P < 0.05) in all 3 years. For Ba: Ca, the elemental ratios were higher in FZ than in other locations (Table 3, Fig. 4). Also, it is significant for overall differences in mean elemental ratios among sampling years for Mg: Ca and Ba: Ca in all four locations (P < 0.001). Mg: Ca and Ba: Ca were both significantly higher in 2016 than in the other 2 years for almost all locations (Table 3, Fig. 4).

Table  2.  Kruskal–Wallis results of elemental profiles across S. niphonius (Japanese Spanish mackerel) otoliths whole life chemistry collected at four locations

Collection year	Among locations: H (df) = p
	Li: Ca	Mg: Ca	Sr: Ca	Ba: Ca
2016	4.88 (3) = 0.18	13.50 (3) < 0.001	1.08 (3) = 0.56	16.49 (3) < 0.001
2017	6.85 (3) = 0.08	12.11 (3) < 0.001	5.40 (3) = 0.14	13.00 (3) < 0.001
2018	10.48 (3) = 0.01	11.74 (3) < 0.001	6.42 (3) = 0.09	13.23 (3) < 0.001
Bracketed numbers indicate degrees of freedom

 | Show Table

DownLoad: CSV

Table  3.  Kruskal–Wallis results of elemental profiles across S. niphonius (Japanese Spanish mackerel) otoliths whole life chemistry derived at three collection years

Location	Among collection years: H (df) = p
	Li: Ca	Mg: Ca	Sr: Ca	Ba: Ca
FZ	11.50 (2) < 0.001	26.63 (2) < 0.001	4.00 (2) = 0.26	19.69 (2) < 0.001
XS	10.69 (2) < 0.001	21.61 (2) < 0.001	10.48 (2) = 0.01	14.31 (2) < 0.001
LS	1.49 (2) = 0.47	29.72 (2) < 0.001	15.10 (2) < 0.001	18.31 (2) < 0.001
QD	0.61 (2) = 0.89	25.45 (2) < 0.001	15.49 (2) < 0.001	17.44 (2) < 0.001
Bracketed numbers indicate degrees of freedom FZ Fuzhou, XS Xiangshan, LS Lvsi, QD Qingdao

 | Show Table

DownLoad: CSV

Figure  4.  Mean elemental ratios (μmol/mol) and standard deviations for S. niphonius (Japanese Spanish mackerel) otoliths whole life chemistry from each of four sample locations and sampling years. Significant differences between spawning grounds based on Dunn's multiple comparison test are indicated by different capital letters (P < 0.05). Different letter forms represented spatial difference at different sampling years, bold letters for 2016, hollow letters for 2017 and underlined letters for 2018. Significant differences (P < 0.05) among sampling years are indicated by different lowercase letters. Insignificant signals are not depicted in this figure. FZ Fuzhou, XS Xiangshan, LS Lvsi, QD Qingdao

DownLoad: Full-Size Img PowerPoint

Although not showing significant overall differences among locations for Li: Ca and Sr: Ca, there is an increasing spatial variation of elemental ratios from 2016 to 2018 (Table 2). In 2018, Li (H = 10.48, 3 d.f, P = 0.01) differed most among locations (Fig. 4). The post-hoc Dunn's test indicated that values for Li were lower in LS than in other locations (P < 0.05). Even in 2018, Sr (H = 6.42, 3 d.f, P = 0.09) showed limited spatial variations among sampling locations. For temporal variations, Li: Ca was significantly different in FZ and XS (P < 0.001; Table 3), both showing lower level in 2016 than in 2017 and 2018 (Dunn's test, P < 0.05; Fig. 4). Sr: Ca was significantly different in LS and QD (P < 0.001; Table 3), both showing lower level in 2016 than the other two years (Dunn's test, P < 0.05; Fig. 4).

Multi-elemental classification

Overall spatial differences were reflected in the assignment to individual grounds and the classification accuracy varied from 2016 to 2018. The random forest (RF) classification in both larval and adult stages of all sampling years were driven mostly by Li: Ca and Ba: Ca, which explained the greatest changes in classification accuracy in order of importance (accounting for over 15% decrease in accuracy when excluded, respectively). The other elements Mg: Ca and Sr: Ca had a smaller impact on classification accuracy, ranging from 3% to 4%. For the larval stage, individuals from different spawning grounds were difficult to discriminate in 2016. The overall classification accuracy from the RF classification approach was relatively low, ranging from 13.3% to 63.4% (overall 36.7%; Tables 4, 5). In 2017, the classification accuracy increased significantly (over 10%) to 49.2%. The overall highest RF classification accuracy was found in 2018, which was 53.3%. In 2017 and 2018, the RF classification success for the East China Sea was relatively high, i.e., 77.5% and 85% (Tables 4, 5). In contrast, QD contributed the lowest RF classification success in all sampling years (no more than 30%; Tables 4, 5), with large parts discriminated as from LS.

Table  4.  Percentage of adult individuals assigned to their original spawning areas based on a supervised RF clustering of larval and adult stages otolith chemistry

Adult original area	Predict region
	2015/2016				2016/2017				2017/2018
	ECS	SYS	NYS		ECS	SYS	NYS		ECS	SYS	NYS
ECS	76.7	13.3	10.0		90.0	7.5	2.5		85.0	10.0	5.0
SYS	33.3	46.7	20.0		15.0	45.0	40.0		35.0	50.0	15.0
NYS	40.0	26.7	33.3		24.5	35.0	40.0		35.0	20.0	45.0
Overall accuracy	52.2				58.3				60.0
Numbers in bold refer to individual assignment to capture spawning areas ECS East China Sea, SYS southern China Sea, NYS northern China Sea

 | Show Table

DownLoad: CSV

Table  5.  Percentage of larval individuals assigned to their original spawning areas based on a supervised RF clustering of larval and adult stages otolith chemistry

Larval original area	Predict region
	2015/2016				2016/2017				2017/2018
	ECS	SYS	NYS		ECS	SYS	NYS		ECS	SYS	NYS
ECS	63.4	16.6	20.0		77.5	15.0	7.5		85.0	10.0	5.0
SYS	46.7	33.3	20.0		25.0	45.0	30.0		35.0	45.0	20.0
NYS	60.0	26.7	13.3		40.0	35.0	25.0		35.0	35.0	30.0
Overall accuracy	36.7				49.2				53.3
Numbers in bold refer to individual assignment to capture spawning areas ECS East China Sea, SYS southern China Sea, NYS northern China Sea

 | Show Table

DownLoad: CSV

The classification accuracy for the adult stage was lower in 2016, ranging from 33.3% to 76.7% (overall 52.2%; Table 4). Compared with the larval stage, the overall classification accuracy for the adult stage in 2017 and 2018 was only slightly increased to 58.3% and 60%. In 2017, the East China Sea contributed the most RF classification success among sampling locations and years (90%, Tables 4, 5). Although the northern Yellow Sea still contributed the lowest RF classification success, the overall classification accuracy was over 50% from 2016 to 2018. Furthermore, most of the misidentified classification in the northern Yellow Sea was from the southern Yellow Sea. Thus, 35%–40% samples were misidentified from each other in 2017 and 2018 (Tables 4, 5).

Natal origins

RF clustering identified three clusters of chemically distinct larval elemental signatures (Fig. 5). The contribution of three identified clusters for different sampling locations showed some variations from 2016 to 2018. There was some temporal difference for the elemental ratios of the three clusters, for example, the values of Mg: Ca were higher in 2016 for all clusters (Table 6). Despite the temporal difference, the identified clusters from 2016 to 2018 were characterized by more contrasting spatial difference in otolith chemistry signatures. Cluster 1 was characterized by the highest value of Li: Ca (7.27–7.89 μmol/mol) and Sr: Ca (2461–2699 μmol/mol) among all clusters with medium Mg: Ca (300.56–400.68 μmol/mol) and relatively low Ba: Ca (2.99–3.68 μmol/mol). Cluster 2 was characterized by higher Mg: Ca (447.52–546.64 μmol/mol) with low level of Li: Ca (3.78–4.31 μmol/mol), Sr: Ca (2135.13–2483.02 μmol/mol) and Ba: Ca (2.99–3.33 μmol/mol). Cluster 3 was characterized by the highest value of Ba: Ca (4.96–5.76 μmol/mol) with the lowest value of Li: Ca (3.49–3.84 μmol/mol), Mg: Ca (254.32–363.17 μmol/mol) and Sr: Ca (2009.47–2303.17 μmol/mol) among all clusters.

Figure  5.  Chart of the sampling locations and pie charts representing the contribution of the three larval sources identified through Random Forest clustering. FZ Fuzhou, XS Xiangshan, LS Lvsi, QD Qingdao. Cluster 1 points to the East China Sea (ECS), Cluster 2 points to the northern Yellow Sea (NYS) and Cluster 3 points to the southern Yellow Sea (SYS)

DownLoad: Full-Size Img PowerPoint

Table  6.  Mean elemental signature of the three larval clusters, northern Yellow Sea (NYS), southern Yellow Sea (SYS) and East China Sea (ECS), identified by RF clustering

Cluster	2015/2016					2016/2017					2017/2018
	Li	Mg	Sr	Ba		Li	Mg	Sr	Ba		Li	Mg	Sr	Ba
1	7.27	400.68	2461.63	3.68		7.66	336.58	2631.44	3.31		7.89	300.56	2699.94	2.99
ESC	± 2.53	± 137.49	± 250.45	± 0.69		± 2.37	± 90.68	± 221.10	± 1.09		± 3.11	± 120.00	± 121.10	± 1.13
2	3.78	546.64	2135.13	3.33		3.95	464.52	2336.12	2.99		4.31	447.52	2483.02	3.02
NYS	± 1.45	± 190.94	± 245.64	± 1.22		± 0.68	± 144.78	± 200.01	± 0.98		± 1.22	± 104.66	± 115.61	± 0.99
3	3.49	363.17	2009.47	5.76		3.84	254.32	2106.31	5.07		3.79	298.32	2303.17	4.96
SYS	± 0.99	± 170.48	± 188.55	± 01.56		± 0.77	± 101.64	± 186.55	± 1.36		± 1.06	± 91.14	± 105.38	± 1.52
Mean values ± standard deviation (μmol/mol)

 | Show Table

DownLoad: CSV

In 2016, all three clusters could be found in all sampling locations. Cluster 1 was the main contributor to FZ, XS and LS but the part it contributed decreased from south to north. Cluster 2 contributed most in QD, lowest in LS and equally in FZ and XS. Contrary to cluster 1, the part cluster 3 contributed increased from south to north but it contributed most in LS instead of QD. QD reflected roughly equal contributions from all three clusters. In 2017, cluster 1 still contributed most to FZ and XS, but the contribution to LS and QD decreased significantly. LS was dominated by cluster 3 whereas cluster 1 and cluster 2 contributed equally in LS. For QD, it was almost dominated by cluster 2 and cluster 3 whereas cluster 1 made significantly less contribution. In 2018, FZ and XS were similar to those in 2017 and were dominated by cluster 1 with negligible contribution of other clusters. Cluster 2 and Cluster 3 were the main contributors to QD and LS, respectively. The contribution of cluster 1 to LS and QD were limited.

Discussion

This is the first study using otolith chemistry to evaluate the impacts of ENSO events on population connectivity of a highly migratory fish in China Seas. The extensive and consecutive sampling of S. niphonius throughout its range confirmed that the spatial variability of otolith chemistry signature among different spawning grounds tended to be less significant in the El Niño year. Clustering of near core chemistry pointed to three sources, meanwhile the source–sink pattern showed significant temporal variability: a large-scale connectivity between the East China Sea (ECS) and the Yellow Sea (YS) happened in the El Niño year while the local mackerel assemblages in different spawning areas seemed to be more self-sustaining afterwards. Population connectivity is an adaptive ability mitigating against local mortality events affecting population resilience and persistence (Hastings and Botsford 2006; Prince and Hilborn 1998). Understanding the climatic effects on population connectivity is crucial for any conservation or fisheries management. Moreover, the development of otolith chemistry techniques provided much more potential for working as a climatic index to further evaluate climate-induced individual and population level fluctuations.

ENSO drives variations in element pattern

Ontogeny influenced the element pattern of S. niphonius, leading to significant differences of elemental concentrations among different life stages (Pan et al. 2020a, b). ENSO is the source of year-to-year variations in global climate and its impacts continued all the year round through winter (Cai et al. 2014; Santoso et al. 2017; Wang et al. 1999). Thus, we used the mean elemental concentration of the whole otolith (covered a span of 1 year) to extract the spatial and temporal variations in element pattern. From 2016 to 2018, strong temporal variations for Mg: Ca and Ba: Ca were found on all the spawning grounds along the Chinese coast, which were probably related to changing environmental conditions driven by ENSO. The process could be completed in two ways: directly influenced by the shifts in extrinsic factors such as ambient chemistry, temperature and salinity, or indirectly through influencing intrinsic factors, i.e., metabolism or growth (Elsdon and Gillanders 2002; Limburg et al. 2018; Sturrock et al. 2015). Mg: Ca is regulated by temperature and food sources simultaneously (Grammer et al. 2017; Sturrock et al. 2012). During the El Niño period from 2015 to 2016, increased SST was found at a wide range of the China Seas (Ma et al. 2019; Yin et al. 2021). In addition, Tian et al. (2023) found that there was an increase in horse mackerel and anchovy during this period, which were the main prey species for S. niphonius (Sui et al. 2021). Probably, those factors explained the elevated Mg pattern in the El Niño year.

Positive relationships between otolith Ba: Ca and elemental concentration in ambient water are well documented. These underpin the use of otolith Ba: Ca to reconstruct environmental life histories (Izzo et al. 2018). The Ba: Ca level was highest in the El Niño year and a decreased pattern from 2016 to 2018 was found in FZ and XS. This general pattern could be understood easily as ENSO promotes increased rainfall, floods and outflow from inland areas (Reis-Santos et al. 2021). According to the established trend, Ba: Ca increased along with the lower salinity waters (Albuquerque et al. 2012). The coastal region's salinity variation range is constrained and predominantly well within typical marine values (between 30 and 35). In the China Seas, ENSO also caused southward moving rain-bands and would remain influencing the rainfall to the summer of the following year (Park et al. 2015; Xu and Chan 2002; Zhang et al. 2017), explaining the decreasing trend in the ECS.

Lithium is another element which is most likely to be reflective of the physico-chemical environment (Sturrock et al. 2012; Grammer et al. 2017). Otolith Li has been proven to be strongly correlated (positive) with chlorophyll a levels in the environment, and could be a potential indicator of productivity within the marine environment (Grammer et al. 2017). It has been reported that El Niño could induce an abrupt decline in primary and secondary production to many ecosystems (Ohman et al. 2017). In the ECS, this process could be achieved by influencing the variability of upwelling. As the sea surface wind was much weaker in the ECS during the El Niño year, the coastal upwelling was corresponding weak (Hong et al. 2011). Therefore, the weak pattern of Zhejiang and Fujian upwellings was expected, leading to less supplementation by nutrient-rich deep water, further contributing to the decreased Li level in FZ and XS in 2016. Unlike Li: Ca, the decreased Sr level in 2016 was found in LS and QD. Under wider salinity variations, Sr: Ca correlates closely with salinity variations (Campana 1999; Gillanders 2005; Walther and Limburg 2012). In a marine system, physiological processes may outweigh the waterborne sources in explaining otolith Sr (Grammer et al. 2017; Sturrock et al. 2015). The decreased growth to climatic influence of ENSO has been found for snapper (Martino et al. 2019) and dusky grouper (Reis-Santos et al. 2021). It is possible that the decreased Sr pattern in our study was also correlated with the ENSO effects on S. niphonius growth.

ENSO drives variations in population connectivity of S. niphonius

The existence of the large-scale spatial connectivity (> 1000 km) in the spawning population of S. niphonius along the coast of China has been indicated, with a strong spatial homogeneity in elemental composition (Pan et al. 2020b). However, based on the inter-annual variability on the RF classification and clustering that we found, this spatial connectivity was strengthened by the ENSO events in 2016. Although showing consistent overlap in otolith elemental composition across locations from 2016 to 2018, especially obvious in Sr: Ca and Li: Ca (Table 2, Fig. 4), the classification accuracy in the larval and adult stages was both higher to varying degrees in 2017 and 2018. For the larval stage, significantly less individuals could be successfully assigned to their spawning areas in 2016 (36.7% vs 49.2% and 53.3%; Table 5). In addition, there was increased contribution rates of ECS natal origins to YS (Fig. 5). T Either, this suggested that a strong El Niño could enhance the connectivity of S. niphonius spawning assemblages between ECS and YS or the homogenization of the environmental baseline of elemental concentrations by the ENSO event in 2016.

Here, the former deduction seemed more possible based on the following reasoned arguments. First, we admitted that through teleconnection, El Niño events might bring changes on physical and biochemical conditions in the China Seas (Park et al. 2015; Xu and Chan 2002; Zhang et al. 2017), which impressed corresponding chemical signatures on S. niphonius otolith (as we have discussed in the first part of discussion). The inter-annual variability on element pattern was also presented in the identified natal clusters (Table 6). However, the spatial variability of the otolith edge elemental signatures, representing the most recent period before being caught, was not eliminated in 2016. The overall classification accuracy was over 50 percent (Tables 4 and 5). Moreover, the three natal origins identified in Pan et al. (2020b) were clearly identified from 2016 to 2018, following the same chemical characters, which indicated few chances that the discrimination was confounded by environmental conditions. Second, if the latter deduction was true, the homogenized environment in the China Seas by the ENSO events would also decrease significantly the classification accuracy in the adult stage in 2016. At least a similar decreased pattern as in the larval stage (over 10%) would be expected, which was not accorded with our results (48.3% vs 52.5% and 53.8%; Table 5).

For highly migratory fish, the strong swimming ability through life helps to maintain the ocean-scale connectivity (Longmore et al. 2014; Wright et al. 2018). The overall classification accuracy in the adult stage is over 50%, lower than similar studies conducted on other species, such as Atlantic spadefish Chaetodipterus faber (Soeth et al. 2019) or Warsaw grouper Hyporthodus nigritus (Sanchez et al. 2020). The major reason for this difference is that the population structure of S. niphonius was formed through two main processes: the connectivity between wintering to spawning grounds and the migration of spawning adults in the spawning seasons (Horikawa et al. 2001; Pan et al. 2020b; Shoji and Tanaka 2005). Adults that first arrive at southern locations for breeding may move northward, leading to spawning mackerel groups in the northern area being composed of individuals from multiple natal origins. When solely looking at the East China Sea natal origin, the classification accuracy was over 80% in all sampling years (Table 4).

Clusters 1, 2 and 3 were referred to the natal origins of East China Sea (ECS), northern Yellow Sea (NYS) and southern Yellow Sea (SYS) (Pan et al. 2020b). The spawning components of FZ and XS were well connected in all three sampling years showing that ENSO created no biogeographic barriers in the open ESC (Fig. 5). In 2016, some portions of spawning components in ECS were originated from YS whereas it tended to be self-sustaining in 2017 and 2018. The spawning period for S. niphonius in the north is later than that in the south (Shui et al. 2009). Therefore, the spawners which were hatched in YS but spawned in ECS probably learned the migratory behavior of the ECS adult group when they mixing together in the wintering stage (McQuinn 1997; Petitgas et al. 2010). Although showing similar decreasing trend for the portion of ECS originated spawners in YS, the contribution of ECS spawning components to YS was probably caused by the northward moving adults (Pan et al. 2020b). The partially mixed pattern between NYS and SYS was consistent from 2016 to 2018 with no major shifts. The effects of ENSO on population connectivity are realized in two main ways. Firstly, ENSO affects the extent of mixing in winter, which results in a greater number of young fish originating from the Yellow Sea (YS) were adopted by the East China Sea (ECS) spawning groups. Secondly, it influences the northward movement of adult fish from the ECS, leading to more spawners reaching the YS in spring.

Climate alters population connectivity through influencing population functional traits such as reproductive timing (Carson et al. 2010), larval dispersion (Bharti et al. 2022; Lopera et al. 2020) and distribution range (Inoue and Berg 2017; Wasserman et al. 2012). The mean annual SST in El Niño years was markedly higher than that in ENSO-neutral years (Ma et al. 2019; Yin et al. 2021). The mean annual SST in 2016 was higher than in 2017 and 2018 in both YS and ECS (Supplementary Figs. S1 and S2). It has been indicated that the increased SST would cause S. niphonius adults to reach sexual maturation earlier, with the distribution and spawning grounds moving northward (Shui et al. 2009; Yang et al. 2022). Therefore, spawning adults of ECS may well arrive on their original spawning grounds in advance in 2016, and spawned in batches as they moved northward in the spawning season (Horikawa et al. 2001; Shoji and Tanaka 2005), leading to high contribution rates to YS. Under the El Niño event, a shrunken ecosystem size, increased energetic efficiency and less organized ecosystem was found in the Yellow Sea (Yin et al. 2021). Given the situation that increased prey species for S. niphonius occur in the China seas (Ma et al. 2019), as a top predator, enough trophic supplement combined with increased energetic efficiency lead to stronger swimming ability. During the feeding period, an extended feeding ground range for young fish in YS could be expected, which increased the occurrence rate to encounter the adult group from ECS.

Shortcomings and prospects

In line with the growing interest in accruing knowledge on the trends of population responses to local and large-scale environmental conditions and climatic anomalies, we provided the potential evidence that population connectivity in a highly migratory fish in the Northwest Pacific is linked to the ENSO events. However, speculation was based on indirect evidence from otolith chemistry results. Interpreting otolith elemental chemistry is challenging because of the environmental variability, the complex process by which fish absorb these elements and incorporate them into the otolith and the way elements bind to the otolith's carbonate matrix, which is not yet fully understood. (Izzo et al. 2018; Reis-Santos et al. 2023; Walther 2019). A thorough survey of the environmental elemental baseline of the studied area in depth and long-term persistent archival tissues collection will greatly improve the application on resolving the fish population connectivity and movement pathways.

Whereas this study has successfully integrated temperature data to explore the association between migration of S. niphonius and the El Niño event, we recognize the absence of specific salinity and food abundance data as a limitation. Salinity and food availability are crucial environmental parameters that significantly influence marine life, potentially affecting growth rates, migration patterns, and the bioavailability of trace elements. The lack of these data restricts our ability to fully understand the complex interactions between these environmental factors and otolith composition. For example, variations in salinity could alter the solubility and hence the uptake of trace elements, whereas food abundance may influence the metabolic rate and overall health of the fish, thereby affecting otolith development and composition (Izzo et al. 2018).

Meanwhile, the Yangtze and Yellow Rivers significantly influence the seawater chemistry of the East China Sea and the Yellow Sea, impacting the trace element composition of fish otoliths. These rivers transport vast amounts of freshwater, nutrients and trace metals into marine ecosystems, altering salinity, nutrient levels, and the bioavailability of trace elements. Such changes may be reflected in the otoliths of fish, as these ear stones incorporate trace elements from their environment during growth. Seasonal and interannual variations in river discharge, affected by monsoonal patterns and anthropogenic activities may lead, therefore, to observable changes in otolith composition. This suggests that otolith analysis could serve as an indirect method for assessing the impact of riverine inputs on marine environments.

This gap highlights a critical avenue for future research. Subsequent studies could aim to collect comprehensive environmental data, including salinity levels and food abundance, alongside temperature and trace element concentrations. We should also focus on quantifying how these inputs affect seawater and otolith trace element concentrations, providing insights into environmental history and the adaptation of marine life to changing conditions. Such data would enable a more holistic understanding of the environmental determinants of otolith composition and provide insights into how these factors interplay to shape the life history strategies of marine species.

Climate may affect marine fish populations through many different pathways (Otterson et al. 2010). Our study demonstrated that extreme climate events, such as ENSO, may result in profound changes in the extent, pattern and connectivity of populations of highly migratory fish and discussed the possible mechanisms of how it was realized. In the future, ENSO events may be more frequent, persistent and intense with global warming (Cai et al. 2015). A good understanding of its effects on population connectivity would allow us to prepare for adjustments on adaptive management in advance.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s42995-024-00229-x.

Acknowledgements

The authors are grateful for all scientific staff and crew for their assistance with sample collection and experiment implementation. This work was partially supported by the National Natural Science Foundation of China (NSFC) (Grant No. 41930534). The senior author's study at Stony Brook University is supported by China Scholarship Council.

Author contributions

XDP reviewed the literature and wrote the manuscript. YC, TJ, JY and YJT improved and corrected the manuscript. The above authors approved this manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Declarations

Conflict of interest

The authors verify that there is no incompatibility of interest. Yongjun Tian is a member of the Editorial Board, but he was not involved in the journal's review of, or decision related to this manuscript.

Animal and human rights statement

This article does not contain any experiments with human participants conducted by the authors. All the experiments were conducted following standard procedures.