1 Introduction

The Pacific saury Cololabis saira is widely distributed in the North Pacific Ocean (Hubbs and Wisner, 1980) and caught commercially by Japan, Russia, Korea, China, and Vanuatu mainly in the western area of 165°E (Tseng et al., 2013). They migrate northward from the Oyashio (subarctic cold) waters to the Kuroshio (subtropical warm) waters in May and July, moving through the Kuroshio-Oyashio transitional waters, and then start southward migration between July and August (Kurita et al., 2004). Pacific saury is characterized by short life span (2 years), high growth rate, and early maturity (Suyama et al., 2010; Suyama et al., 2019), exhibiting marked population fluctuations (Tian et al., 2003; Watanabe et al., 2003). The total landings fluctuated over the last 20 years from 187898 tons in 1999 to 260178 tons in 2017 (The Food and Agriculture Organization: http://www.fao.org/fishery/statistics/global-capture-production/query/en, accessed 15 October, 2019). Despite the increase in total fisheries catches, the standing stock of Pacific saury has declined recently (Tohoku National Fisheries Research Institute, Fisheries Research Agency TNFRI, 2017). Thus stock assessment and management aiming to achieve the sustainable exploitation of Pacific saury resources are urgently needed. Therefore, a regional fisheries management organization, North Pacific Fisheries Commission (www.npfc.int), was established in 2015 to take timely actions.

One essential part of fish stock assessment and fisheries management is to identify distinct populations that have unique demographic properties and life history patterns as well as divergent responses to fishery exploitation (Begg et al., 1999; Stephenson, 1999), and to manage them separately (Hutchinson, 2008). For Pacific saury, Suyama et al. (2012a found two spatially separated groups of one year old individuals through comparing the radius of otolith annual rings (ROA) of samples collected during prefishing season (June and July): one distributed in the west of 160°E with large ROAs (western group) and the other in the east of 170°E with small ROAs (eastern group). However, the two groups are mixed in the high-seas fishing ground during the fishing season, which makes them difficult to be distinguished by ROA.

The different growth and migration patterns of distinct groups can result in different dynamics. The western group, a main stock captured by Japanese fleets, has declined recently (Miyamoto et al., 2019). However, this group is also captured along with the eastern group in the high-seas fishing ground, which makes management and allocation of fishing quota difficult because of lacking effective tools to separate these two groups.

Due to the effects of both genetics and environment, fish with different life histories often vary in otolith morphology (Vignon, 2012). This variation has led to the development and wide application of otolith shape analysis as an ideal tool to identify groups of fish that may have been spatially or temporally discrete at some stages in their life history (Stransky et al., 2008; Agüera and Brophy, 2011; Bacha et al., 2014; Keating et al., 2014; Chi et al., 2016). If the separated habitats for the two groups of the Pacific saury during the first year result in differences in ROAs, it could also cause differences in otolith shape. Such differences could last during spawning migration because of the different environmental conditions they experienced, which thus allow the quantification of group composition in the fishing ground. In this study, we tested the potential of using otolith shape analysis combined with ROAs to identify the two groups of one year old Pacific saury. The aims of this study were 1) to separate the two groups and quantify the spatial and temporal dynamics of group composition of Pacific saury in the fishing ground; and 2) to provide information on the timing of migration to facilitate sustainable fisheries management.

2 Materials and Methods 2.1 Sampling

Pacific saury were randomly sampled from commercial catch (using stick-held net) in the fishing ground of highseas (147.7°-163.1°E, 37.8°- 47.6°N) during the fishing season (July-November) from 2016 to 2018 with the vessels of Qingdao Zhongtai Oceanic Fishery Co., Ltd., China. Fish samples were frozen on board. After being transported to the laboratory, all fish samples were measured by knob length and weighted. Otoliths were extracted, washed with ultrapure water, air-dried, and then stored in plastic tubes. The left undamaged otolith of each individual was placed on glass slides and photographed using a micrographic system (Olympus BX53 and DP74) at a magnification of 40. For Fourier analysis, the otoliths were placed along the sulcus side, facing down and rostrums resting horizontally to the left. Polarized light was used to produce high-contrast images and to enhance the clarity of translucent zones, where otoliths were bright objects on a black background.

All animal experiments were conducted in accordance with the guidelines and approval of the respective Animal Research and Ethics Committees of Ocean University of China (Permit Number: 20141201. http://www.gov.cn/gongbao/content/2011/content_1860757.htm). The studies did not involve endangered or protected species.

2.2 Radius of Annual Ring Measurement and Otolith Shape Analysis

Only one year old individuals were selected for subsequent analysis based on the otolith classification method recommended by Suyama et al. (2009 to exclude age-related variability in otolith shape. Only the otoliths enclosed by a broad translucent area (type Ⅱ) or containing a complete hyaline zone within the otolith (type Ⅲ) were selected.

The ROA of each selected otolith, which is defined as the distance from the otolith core to the inner edge of the first annual ring (Suyama et al., 2012b), was measured with ImageJ and recorded. Then, the images were analyzed using the R package 'shapeR' specifically designed for otolith shape analysis (Libungan and Pálsson, 2014). This package 'shapeR' has built-in functions that allow users to extract the otolith outlines from images, visualize the mean shape, smooth the outline by eliminating pixel noise, and transform the outlines into independent coefficients using either normalized elliptic Fourier or discrete wavelet. Elliptic Fourier analysis was chosen and 48 normalized elliptic Fourier coefficients were produced for each otolith. The first three coefficients were degenerated and constant for all outlines and thus were omitted, leaving 45 (48-3 = 45) normalized elliptic Fourier coefficients left.

2.3 Statistical Analysis

Samples were identified according to sampling time and location to demonstrate the spatial and temporal changes of group compositions (Table 1). K-means cluster analysis was carried out on the Fourier coefficients and ROAs. This method allows a priori assumptions on the number of clusters to compute, and two morphotypes were specified based on a previous study (Suyama et al., 2012b). Then, the algorithm estimates iteratively the cluster means and assigns each case to its respective cluster until the sum of squares of the assigned cluster centers is minimized. Cluster was performed for each year respectively to exclude potential effects of year-class shape differences. ROA frequency distribution histograms and Gauss fitting curves of the clustered groups in three years were plotted using OriginPro 9.1. Random forest was also used to compare the otolith shapes of the clustered groups. Out-of-bag error was estimated internally in the model runs, which is conceptually similar to cross-validation, and used to estimate classification success (Breiman, 2001).

3 Results

K-means cluster dendrograms identified two otolith morphotypes (Fig. 1). The ROA cluster centers of Morphotype Ⅰ were 0.5173778, 0.52516270, and 0.5283049 in the three years, respectively, whereas those of Morphotype Ⅱ were 0.6084395, 0.6296290, and 0.6033788. The ROA frequency distribution histograms and Gauss fitting curves of the two clusters across the three years showed consistent patterns with a smaller ROA of Morphotype Ⅰ than Morphotype Ⅱ (Fig. 2). The average contours of the two morphotypes also showed high consistency across the three years (Fig. 3). The major differences lie on the dorsal and ventral margins, namely, the otolith shape of the large ROA group (Morphotype Ⅱ) is smaller and slender than that of the small ROA group.

The spatial and temporal distribution patterns of Pacific saury showed that the two groups were well mixed from July to November in the three years. In the similar longitude (150°E in 2016, Fig. 4), the proportion of individuals from Morphotype Ⅰ gradually increased with prolonged sampling time. Meanwhile, individuals from Morphotype Ⅰ accounted for a large proportion in the west sampling locations in September 2017 (Fig. 4). With the passing of sampling time and the westward movement of the fleets, the proportion of Morphotype Ⅰ generally increased, except in August 2016 (Fig. 4).

Random forest showed high classification success (ranging from 86% to 98%) in assigning individuals to their clustered group (Table 2).

4 Discussion

In this study, otolith shape analysis identified two groups of one year old Pacific saury that were captured from the high-seas fishing ground. The high classification success of random forest further supports the existence of the two groups. Previous studies (Suyama et al., 2012b; Miyamoto et al., 2019) reported that fishes with small ROAs (< 0.54 mm) and large otolith (Morphotype Ⅰ) were from east of 170°E (eastern group), whereas individuals with large ROAs (> 0.6 mm) and small otolith (Morphotype Ⅱ) originated from west of 160°E (western group). The fish samples were collected with a small space coverage from an assemblage of fish during their migration; therefore, the result may merely provide a 'snapshot' of the temporal and spatial distribution of the two groups. However, the revealed distribution trend is consistent with the previous observation (Miyamoto et al., 2019). The western group accounted for the majority in July and August at about 160°E. With the passage of sampling times and westward movement of the fishing locations, the proportion of eastern group increased gradually except in August 2016 and finally reached more than 60% in November at about 150°E in these three years.

Otolith shape is a species-specific mark that is determined by the genetics of the stock; however, it can also change under the influences of ontogeny and environment (Vignon, 2012). Environmental factors such as temperature (Chi et al., 2014; Keating et al., 2014), water depth (Lombarte and Castellón, 1991), salinity (Capoccioni et al., 2011), substrate type (Mérigot et al., 2007), diet (Gagliano and Mccormick, 2004), and hydrogeomorphic factors (Ding et al., 2019) affect otolith shape. The response of otolith shape to temperature and food availability is mediated by the effects of these variables on growth rate (Campana and Casselman, 1993; Hüssy, 2008). Pacific saury migrates between the Oyashio and Kuroshio waters and passes through the Kuroshio-Oyashio transition zone of complex oceanic structures (Watanabe et al., 1997; Liu et al., 2019). Moreover, migration is considered to be driven by water temperature and food resources (Huang et al., 2007; Tseng et al., 2014). Therefore, the variation in otolith shape might be attributed to the complex and changeable environmental conditions during this long-distance migration. The observed otolith shape differences between the two groups might be ascribed to temperature, considering that high temperature contributes to large sizes (Lombarte and Lleonart, 1993). Two dominant Pacific saury fishing grounds were recorded in the fishing season, and the temperature of eastern ground is higher than that of the western one (Tseng et al., 2011). The otolith shape differences might also ascribe to different water velocities, as Ding et al. (2019 suggested that fast water flow coincides with long and slender otolith. Coincidently, westward velocities are highly correlated with the distribution and migration of Pacific saury (Miyamoto et al., 2019). Further understanding the origins of otolith shape variance and response to longitudinal environmental gradients will be the focus of future work.

Using otolith shape analysis and ROA measurement, two groups of Pacific saury were identified in the high-seas fishing ground in this research. The one year old individuals, which are winter cohort in the Kuroshio region, play a crucial role in the recruitment of this stock because of the stable high growth rates (Watanabe et al., 2003). Tian et al. (2004 indicated that the decadal-scale variation in Pacific saury abundance is strongly affected by oceanic region shifts in the Kuroshio region. However, the shortperiod fluctuation of saury abundance remains to be studied, especially when inter-regional management is involved. The result can provide information for stock assessment and for understanding the migration route of Pacific saury. As shown in this study, ROAs contribute to the most significant differences. Future studies will focus on early life information, such as retrospective analysis using otolith microstructure and microchemistry, which can build a baseline for separating the two groups and provide useful information on group origin and migration route.

Acknowledgements

We acknowledge the crew of Zhongtai Oceanic Fishery Co. for their great help in sample collection. We thank Dr. Caihong Fu from Pacific Biological Station, Fisheries and Oceans Canada for proof reading the article, and appreciate Prof. Yoshiro Watanabe from University of Tokyo for his valuable discussions and suggestions. This work was supported by the National Natural Science Foundation of China (No. 41930534) and the Fundamental Research Funds for the Central Universities to Ocean University of China (Nos. 201762015 and 201822027).