Recruitment is a fundamental process in population dynamics, which is highly variable in time, space and vulnerable to fishing gear (Gaillard et al., 2008). Understanding the spatial pattern of juvenile distribution is of great interest to fishery science and marine ecology, since juvenile is a critical stage of fish stocks. Therefore, reducing the fishing effort with non-selective gears in recruitment areas will help avoid recruitment overfishing (Caddy, 2000; Paradinas et al., 2015). Considering the vulnerability of recruitment and their role in population dynamics and stock size fluctuation, the protection of important habitats is critical for the conservation of marine fish and even the conservation of biodiversity (Hunt et al., 2020). A recommended management tool is establishing a network of Fisheries Restricted Areas (FRA) in regions where the target species is known to aggregate during critical stages of their life cycle (Caddy, 2009, 2000; Colloca et al., 2015; Paradinas et al., 2015; Petza et al., 2019). However, ensuring the effectiveness of such protected areas requires a comprehensive understanding of species distribution and habitat relationships. This ecological knowledge is becoming increasingly important for the sustainable exploitation of commercially important marine populations (Colloca et al., 2015; Paradinas et al., 2015).

Linking the environmental processes and ecological interactions to recruitment success may help explain the dynamics of fish populations and their productivity (Zimmermann et al., 2019). Biogeographers studied the environmental drivers of recruitment distribution and variation at the regional and other extents and identified basin scale oceanographic processes as the key drivers (Castillo-Jordán et al., 2016; Koenigstein et al., 2016; Zimmermann et al., 2019). As powerful physical forces in the seas, oceanic currents have a major influence on the distribution of water masses, nutrients and the productivity of ecosystems, thus influencing the drift, dispersal, and settlement of fish at early life stages (White et al., 2019; Bashevkin et al., 2020). The thermal environments that organisms experience strongly affect their survival rates, distribution, and abundance (Kingsolver, 2009; Isaak et al., 2015). Therefore, temperature defines the habitat of a species’ realized thermal niche (Armstrong et al., 2021). Dissolved oxygen in the aquatic habitat is an important requirement for fish growth and survival, because the energetic demands of organisms increase with temperature and activity, and must be met by an adequate supply of oxygen (Smith and Delorme, 2010; Deutsch et al., 2020). In the California Current large marine, Keller et al. (2015) indicated that catch and species richness exhibited positive relationships with near-bottom oxygen concentration. However, fish and organisms have a different tolerance for low dissolved oxygen concentrations, for instance, pelagic fish require higher levels of dissolved oxygen, while benthic organisms such as crabs and shrimp are more tolerant to the lower oxygen concentration (Brennan et al., 2016). Environmental predictors have been used to accurately predict the distribution of species in several studies (Bateman et al., 2012; Albuquerque and Beier, 2015a,b; Young and Carr, 2015). Such analyses concluded that abiotic variables could be used to predict the location of irreplaceable areas for conservation efforts (Lomolino et al., 2010). At large marine ecosystem scales, average trends in recruitment capacity in each ecosystem were found to be significantly related to environmental variables (Britten et al., 2016). Therefore, a better understanding of the environmental preference, the temporal and spatial scales at which species move could provide information on habitat use for the effective management. The East China Sea (ECS) is situated in the western Pacific Ocean. It is a semi-enclosed marginal sea spanning an area of 744,000 km². More than 50% of the harvested stocks are overexploited due to the unselective exploitation pattern (Teh et al., 2019) and opportunistic fishery behavior. The minimum mesh size of trawl fisheries had been implemented since 2004. Due to poor enforcement and reduced trawl selectivity, 43.59% of the catch in bottom trawl fisheries are low-valued feed-grade fish that dominated by juveniles of food fish (Zhang et al., 2020). Such an exploitation pattern hinders the achievement of Maximum Sustainable Yield (MSY) for fisheries, as required by the new fisheries law and total allowable catch guide policy during China’s 13th Five-Year Plan. The current annual amount of feed-grade fish in China has been estimated to be over 4 mmt, corresponding to a percentage of 34.7% of the overall catch (Zhang et al., 2020). The protection of main commercial species nurseries is increasingly viewed as a major step toward achieving more sustainable exploitation patterns.

Largehead hairtail (Trichiurus japonicus) is a flagship species in China. The annual catch in China was over 1.4 mmt in 2006 (see Chinese Fishery Statistical Yearbook, 2016), but declined to 0.92 mmt in 2019 (see Chinese Fishery Statistical Yearbook, 2019). Due to the wide distribution of spawning grounds and high fecundity, largehead hairtail catch remained high for a long time despite persistently high-intensity fishing pressure. Although largehead hairtail fisheries have shown a surprisingly high resilience to human exploitation than other fisheries, the most important commercial species is now fully exploited. Important negative impacts have also been detected, such as a high proportion of juvenile fish in the catch because of non-selective gears and fishing grounds overlap with the juvenile largehead hairtail distribution.

The Chinese government has taken a series of practical measures to address these problems. In 2010, the State Council issued and implemented the strategy and action plan for biodiversity conservation in China (2011–2030), which provided guidance for the priority areas of inland, water, ocean, and coastal biodiversity conservation. By 2019, China had established 271 protected areas that include oceans (Zhou et al., 2021). They cover 12.4 million hectares of the ocean but represent only 4.1% of China’s maritime area (Zhou et al., 2021). Because of small conservation areas, the ability to protect migratory fishes is likely hindered by their fragmentation. China holds an annual fishing moratorium lasting nearly 4 months in the Yellow Sea, the ECS, and the South China Sea and established ECS largehead hairtail national aquatic germplasm resources conservation zones (TCZ) to protect spawning and juvenile largehead hairtail stock. However, the current TCZ cannot keep up with the demand for effective resource protection. How to accurately identify and determine the priority reserve is a prerequisite for conservation activities. Accurate identification and determination of priority reserves can guide the rational allocation of resources and maximize the protection benefits (Wu et al., 2016).

Species distribution models (SDMs) have been widely used across terrestrial, freshwater, and marine realms to identify species-environment relationships and to identify priority areas, ‘hotspots’ of vulnerable species (Ovando et al., 2019; Pennino et al., 2019a; Lyons et al., 2020; Epele et al., 2021) or essential fish habitat (Colloca et al., 2015; Laman et al., 2018). A variety of methodological approaches have been developed over the last decades to generate SDMs, such as Neural Networks (Özesmi and Özesmi, 1999), Boosted Regression Trees (Elith et al., 2008; Wege et al., 2021), Maximum Entropy (Phillips et al., 2006), Generalized Linear Regression Model (Guisan et al., 2002), and Additive Regression Models (Swartzman et al., 1992; Austin, 2007). However, the statistical challenges using SDMs have increased as datasets have become more complex over time (Orúe et al., 2020; Lloret-Lloret et al., 2021). Recently, Bayesian hierarchical spatiotemporal models with the Integrated Nested Laplace Approximation (INLA) methodology has acquired an important role, since it is ideally suited for fitting complex spatiotemporal covariance structures (Rue et al., 2009; Silva et al., 2017). INLA provides accurate approximations to the posterior marginal distributions of latent Gaussian Markov Random Field (GMRF) models with considerably less computational effort (Rue and Held, 2005). INLA handles models with the additional advantage of integrating common modeling approaches and uncertainty analyses into a more general hierarchical framework. Because of these advantages, Bayesian methods are increasingly used in fisheries studies in recent years (Pennino et al., 2014, 2019a,2019b; Nikolioudakis et al., 2019; Cavieres et al., 2021; Hintzen et al., 2021).

This study aims to provide seminal information on the key marine areas of the largehead hairtail in ECS to support conservation efforts by elucidating the spatial relationship between largehead hairtail distribution and oceanography. To accomplish this, (i) species distribution models of the INLA-SPDE approach was developed using Generalized Additive Models (GAM) to define oceanographic features of key largehead hairtail habitats in the ECS and identify contributions of each predictor. (ii) abundance indices from model output to explore the spatial and temporal patterns of juvenile largehead hairtail abundance in the ECS. (iii) we discussed the results within the current conservation and draw attention to the main threats for the studied species. We argued that an improved understanding of the spatial distribution of largehead hairtail could help to determine and manage MPAs by taking into account sensitive habitats.

The red line for ecological protection marks the minimum ecological protection space designated by the Chinese government based on the requirements for protecting the integrity and connectivity of the ecosystem so as to maintain regional ecological security and sustainable development (Xu et al., 2015; Lin et al., 2016). Protecting sensitive ecosystems and important ecological functions is an important element of the red line for ecological protection. Marine fish spawning grounds and nursery grounds are important for replenishing the stock and maintaining biodiversity, while non-selective fishing often leads to massive bycatch of juveniles. Therefore, conserving nursery grounds is important for the sustainable development of fisheries resources, and the spatial characteristics of key habitats for exploited stocks need to be understood. Thus, identifying the spatio-temporal distribution of nursery grounds is a prerequisite for setting the red line of ecological protection. In this paper, nursery grounds were identified based on the persistence of the hotspots of the juveniles spatial distribution. Specifically, two spatio-temporal structure models were adopted to represent the juveniles’ occurrence and abundance, which were combined to present the hotspots, and the persistence index was used to map the spatial distribution of nursery grounds. This method of using abundance distributions to determine nursery grounds is particularly suitable when the contribution of nursery grounds to the adult stock and the ecological processes are unclear (Beck et al., 2001; Gillanders et al., 2003; Maynou et al., 2003; Tserpes et al., 2008; Garofalo et al., 2011; Colloca et al., 2015; Paradinas et al., 2015). Many researches focused on identifying the nursery grounds by developing different metrics (Colloca et al., 2009, 2015). Geostatistical techniques are often adopted in identifying the nursery grounds of fish species such as red mullet, European hake, and deep-water rose shrimp in the southern Adriatic Sea (Carlucci et al., 2009), as well as European hake nursery grounds in the central Mediterranean Sea (Colloca et al., 2009). Colloca et al. (2015) argued that using unique standardized method of spatial modeling for all species/stages/areas was unfeasible due to the large observed differences in the availability of data and causal mechanisms. Colloca et al. (2015) adopted adaptive methods for different species to overcome the issues of autocorrelation, non-linearity and large amount of zero catch. In this study, we used the ZAP models to deal with zero inflated data, that was similar to the zero inflated generalized additive models adopted by Colloca et al. (2015). In addition, alternative Bayesian spatio-temporal models were fitted using INLA-SPDE approach, and we selected the most appropriate model that captured spatial and temporal effects, which explained additional sources of uncertainty on the structure of dependence, thus was able to accurately describe the spatial distribution of nursery grounds. In this case, areas of high abundance but low probability of occurrence may not be considered important nursery habitats. Instead, only locations of high abundance and high probability of occurrence are identified as the cores of nursery grounds. In this study, the concepts of habitat suitability index and hotspots were used to identify core areas of nursery grounds (Colloca et al., 2015). Specifically, hotspots were identified by a suitability index standardized between 0 and 1, which were then classified into important nursery grounds using different thresholds. In this way, the spatial distribution of nursery grounds can be more flexibly compared under the same scale, making it possible to distinguish whether there are unique spatial processes or whether the spatial structure has changed over time. Compared to the geostatistical aggregation curve (Colloca et al., 2009; Tamdrari et al., 2010) or the Getis-Ord hotspot analysis (Colloca et al., 2015; Zacher et al., 2018; Milisenda et al., 2021), the suitability index analysis does not require complex calculation but lacks statistical interpretations and objectivity in the hotspot classification criteria. In this study, since the spatial distribution pattern of juvenile largehead hairtail was inter-annually stable with slight variations in the southern part of the ECS, the persistence index could be used for nursery grounds identification. In addition, two thresholds were used to identify important nursery grounds based on different aggregation intensities.

Due to the presence of permanent largehead hairtail nursery grounds in the study area, different thresholds were adopted in the classification of nursery grounds hotspots so that the importance of different areas can be evaluated. Three permanent nursery grounds were identified under a suitability index of 0.6, and the two larger ones were in the waters of Taizhou Islands and the eastern waters of Yushan Island, both of which had relatively high abundance and probability of occurrence. Parts of the two nursery grounds were inside the TCZ, which was established for the exclusive protection of largehead hairtail, especially the spawning stock and juveniles. The TCZ is rich in feedstuff organisms, and trawling is prohibited from April to September, which provides a stable and high-quality ecosystem for juvenile largehead hairtail and improves their survival rate. However, parts of the two nursery grounds were still not within the TCZ. The smallest nursery ground was in the southern waters of the Beijishan Islands and Nanjishan Islands. Although the abundance of the smallest nursery ground was lower than those of the two northern nursery grounds, the location of the core area is stable, and its abundance was even higher in specific year. Since 2017, the area has been managed under the local government’s spawning ground protection policy and is protected from any fishing operations other than longline fishing from April to August each year. The permanent nursery grounds were extremely small and outside the TCZ under a suitability index of 0.7, which may be related to the life history stage of the population. Since the TCZ was set up to protect the spawning stock and juveniles, the juveniles could be transported to the shallow nearshore waters and feeding around (Wu, 1991). The juvenile largehead hairtail in our study were mainly immature individuals under 20 g (123 mm). They have migratory ability and move across TCZ for seeking feedstuff around the spawning grounds, which is also a common behavior of many fish species.

Functional response curves are important for understanding the optimal environmental adaptability and tolerance of a certain species as they provide information on how species respond to environmental gradients. The range of environmental features for potential juvenile largehead hairtail habitats was obtained based on the findings in this study. The occurrence and abundance distribution of juvenile largehead hairtail were closely related to water depth and hydrological conditions. The abundance distribution showed optimum areas within the depth range from 60 to 100 m with a turning point at 80 m. Abundance was positively correlated with water depth in the range of 40–80 m and negatively correlated with water depth above 80 m. Beyond 60 m, the probability of occurrence was negatively correlated with water depth, i.e., the probability of occurrence was higher in shallow waters within the depth of 60 m. Although the probability of occurrence was greatest in shallow waters at 40 m, there was a high uncertainty due to the limited number of shallow water survey stations with depths smaller than 40 m. The inconsistency between the optimum occurrence areas and the peak abundance areas may be attributed to individuals’ choice of different habitats for their development, i.e., the ecological niche differentiation of different sized individuals drives the adaptation of occurrence and abundance to local areas. The differences in occurrence and abundance were adjusted through the competition for resources, which, in turn, drives the migrations between different areas (Bijleveld et al., 2018). In macroecology, different mechanisms have been proposed to explain the interspecific abundance-occurrence relationships. The ecologically based explanations focus on the availability and distribution of habitat and food resources, individual size and the corresponding migration ability, species specificity, and adaptation to the habitat (Borregaard and Rahbek, 2010; Miranda and Killgore, 2019). As in the description of macrobenthos in the Wadden Sea made by Bijleveld et al. (2018), there is no general relationship between the biomass of a species and its occupancy. Friedland et al. (2021) studied the interannual variation in the occurrence and biomass gravity center of fish and macroinvertebrates in the northeastern U.S. shelf ecosystem and found that habitats differed in depth and that habitats with the center of gravity were in shallower waters and the biomass gravity center was shifting to deeper waters. Since the occurrence and abundance characteristics of species distribution offered valuable information for the management process but showed inconsistent patterns, occurrence and abundance distributions need to be balanced in the spatial planning of resources.

Thermal conditions constrain species’s distributions through physiological processes (Stuart-Smith et al., 2017). All fish species are characterized by their temperature tolerance and the preferred temperature in which they demonstrate optimal growth (Kır et al., 2017). Juveniles are expected to be adapted to a relatively narrow range of temperatures due to relatively weak thermal tolerance (Turko et al., 2020). Both the occurrence and abundance of juvenile largehead hairtail showed optimal temperature ranges, juveniles showed a preference for warmer waters to colder waters, confirming previous studies (Mi, 1997; Wang et al., 2011) that suggested that warmer waters promoted largehead hairtail growth. The results in the present study indicated a narrower temperature range of the bottom water temperature than that was reported by You and Xu (1984), which was 16–22°C in summer in the central fishing ground. Our results suggested that adult and juvenile largehead hairtail inhabited a similar thermal niche, but juveniles that occupied nursery grounds were restricted to these habitats because of narrower physiological thermal tolerance compared to tolerances of adults. Sun et al. (2020b) studied daily growth of young-of-the-year largehead hairtail by otolith microstructure analysis, suggested that daily growth of juvenile largehead hairtail was influenced by the interactive effects of water temperature and nutrient supplies. For the October-spawned and November-spawned cohort, otolith daily increment showed maximum growth at temperatures 18 and 21°C, respectively, which were approximate optimal temperature in the nursery areas of our present study. Temperature-dependent physiological mechanism may contribute to persistent existence of largehead hairtail nursery grounds. However, the mechanism of the effects of increased thermal variability on largehead hairtail physiology and stress are not well understood, many experimentally manipulated thermal regimes and found that thermal variance affected the biochemical reaction rates and metabolic rates of fish species (Brown et al., 2004; Steel et al., 2012). Moreover, the formation of largehead hairtail nursery grounds may largely depend on those that maximize a particular physiological performance, but will also depend on food and habitat availability and various other ecological factors (Payne et al., 2015).

Largehead hairtail spawn almost year round with peak spawning periods spanning from April to October in the East China Sea. In late spring, the spawning stock migrates from the southern Zhejiang offshore to the northern Zhejiang offshore drove by the Taiwan warm current (TWC). In autumn, the spawning stock spawns offshore at the mouth of the Yangtze River and gradually migrates southward (Li, 1982). The waters around the nursery grounds and spawning grounds are dominated by two current systems with different features, i.e., the low-temperature water along the coast of Jiangsu and Zhejiang in nearshore areas, and the high-temperature water of the TWC in offshore areas. Thus the primary productivity is high in the frontal zone (Jiao et al., 1998). The location of nursery grounds changes with the advance and retreat of the Taiwan warm current. As the temperature decreases, the aggregation area of largehead hairtail moves south; as the temperature increases, the aggregation area moves north. The catch of largehead hairtail is also closely related to the coastal current and the TWC. When the coastal current is stronger, the catch in the ECS is higher, and vice versa. The catch in the ECS is inversely proportional to the strength of the TWC (Chen et al., 2004). The effects of coastal current and the TWC on catch may be related to the transport and survival rate of juvenile largehead hairtail. TWC originating from the northward Kuroshio flows along the Sea of ECS coast, has poor nutrients and low primary production (Yatsu et al., 2013; Karu et al., 2020). In contrast, the coastal current has rich nutrients and high primary productivity (Ichikawa and Beardsley, 2002). In some years, strong TWC and weak coastal current can change the nursery grounds location of largehead hairtail, and the quality of largehead hairtail nursery grounds declines, thus reducing the early survival rate. The effects of NPPV are consistent with those of temperature on the abundance. Increases in temperature within a certain range promote increases in primary productivity. Guan et al. (2014) describes the relationship between the maximum photosynthetic rate and temperature exhibits a peak near 20°C, followed by a decline with further increases in temperature through the VGPM (Vertically Generalized Production Model) algorithm, indicating that although higher temperature usually increased photosynthesis, it was often associated with the nutrient limitation that inhibited phytoplankton growth. In addition, increasing temperature can increase the respiration of phytoplankton and limit the accumulation of photosynthetic products. Although the relationship between temperature and NPPV was not analyzed in this paper, we found the abundance of juvenile fish began to decrease at temperatures above 20°C and NPPV above 120 mg/m³, which shown a similar response trend. Du et al. (2020) concluded that low salinity and high temperature were conducive to the growth of phytoplankton, which promote the migration, feeding, aggregation of largehead hairtail, and the formation of fishing grounds in the ECS.

The non-linear relationship between NPPV and the abundance of juvenile largehead hairtail showed that the effect of NPPV on the abundance was positive over a wide range but shifted to negative beyond a threshold. Guan et al. (2014) found the same pattern in the relationship between Scomber japonicus and NPPV in the ECS and concluded that interspecific competition between Scomber japonicus and largehead hairtail limited the biomass increase of Scomber japonicus. In addition, a negative feedback mechanism exists between zooplankton and phytoplankton (Goldyn and Kowalczewska-Madura, 2008; Daewel et al., 2014; Vallina et al., 2014). As a result, the abundance of zooplankton decreases with the increases in NPPV. Thus the abundance of zooplankton-feeding fish is regulated through the bottom-up mechanism. Juvenile largehead hairtail mainly feed on plankton and small pelagic fish (Zhang, 2004), thus having a closer trophic relationship to the NPPV. Chen et al. (2004) analyzed the relationship between the feedstuff and fishing grounds location of largehead hairtail in the ECS using the distribution data of zooplankton from 1960 to 1981, and found that the distribution of zooplankton biomass, particularly krill was closely related to the changes of largehead hairtail fishing ground location and the migration of fish schools. However, largehead hairtail is highly cannibalistic during its juvenile and adult stages, and the weight of juvenile largehead hairtail could reach 25.2% of their stomach contents (Liu et al., 2009b). Based on the available results, the non-linear response of juvenile largehead hairtail to NPPV may be attributed to the complex ecosystems, and it is difficult to determine which mechanism controls the dynamics of juvenile largehead hairtail abundance.

Marine spatial planning is an important tool for marine fisheries management and an important means to govern marine space and promote sustainable use of marine fisheries. Currently, the understanding of the spatial ecology of marine species is limited, which is necessary for the conservation of marine species and the development of marine spatial planning approaches. As tools to support management and conservation, habitat suitability models could help identify priority conservation and management areas. This study extends the knowledge of the relationship between juvenile and the environment in the highly exploited species of largehead hairtail. Considering the commercial and economic importance of the species, the results of our present study provided priority conservation areas for permanent nursery grounds in early life history stages and that can be used to achieve sustainable and adaptive fisheries management and spatial management of human activities.

The small size and proximity of the identified permanent nursery grounds and the fact that they did not cover multiple administrative units brought the feasibility of implementing effective marine spatial planning and connected conservation area networks. Marine conservation areas are set up to manage human activities rather than fish or biodiversity. Since fishing is the most important human activity, and the distribution of trawl fishing grounds is highly overlapping with nursery grounds (Zhang et al., 2016). Thus, resistance to spatial planning stems mainly from fisheries actions, especially trawling. The smaller area of permanent nursery grounds may be a convenient feature, which could help the negotiation with fishery administrations and reducing the impacts on fisheries and fishermen (Paradinas et al., 2015). An effective assessment of permanent nursery grounds should include the spatial and temporal dynamics of the species throughout the year (Crowdera and Norse, 2008; Paradinas et al., 2015), even involving the knowledge on environmental regimes. Therefore, future studies may focus on the analysis of the spatial distribution of different months and supplement other information, such as the fishery-dependent catch and fishing effort. If a persistent pattern exists throughout the year, a permanent fishery restriction area may be recommended. However, fish reproduction is seasonal and cyclical, and concentrating on protecting the space of peak reproductive and nursery periods could prove a more flexible management approach. In addition to strict protection of existing fish distribution hotspots, new suitable habitats could be added to expand the range of hotspots. For example, analysis in this study showed potentially suitable areas were outside the permanent nursery habitats. Although most of them were at the periphery of the hotspots, those areas were even the core of distribution in specific years. At present, those potentially suitable areas are disturbed by human activities and lack of adequate protection. Due to the complex system behavior of marine populations and ecosystems, there is no guarantee that removing the stresses will lead to population recovery at times of populations decline. Therefore, precautionary action is a more robust management strategy than restoration measures under population collapse.

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

Ethical review and approval was not required for the animal study because all the fish samples were collected by scientific investigation and no specific licenses for the studied populations were needed.

ZL, YJ, and LLY conceived and designed the study. ZL, YZ, HZ, and CS collected the data. ZL, YJ, and LLY analyzed the data. ZL, YJ, LPY, LLY, and JC developed the initial version of the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.