Neon flying squid, Ommastrephes bartramii, is widely distributed around the world’s oceans extending from the subtropics to temperate waters in the northern and southern hemispheres except in equatorial waters (Rodhouse 2005, Lefkaditou et al. 2011). Despite its huge potential abundance, it only supports a commercial fishery in the northwest Pacific Ocean. The exploitation of this epipelagic species was started in 1974 (Chen et al. 2009), with roughly 200000 to 300000 t landed during the 1980s by Japan, Korea and Taiwan (Burke et al. 1993). Mainland China started fishing this squid in 1993, and mainly fished in the western waters 160° east of the North Pacific. Recently, commercial squid jigging vessels of mainland China have mainly fished in the areas 150°E to 170°W and 40° to 46°N from May to October with annual catches of between 50000 and 100000 t.

The population structure of O. bartramii has been studied by many researchers for putative seasonal cohorts (Murata 1990, Murata and Hayase 1993), rates of infection by helminth parasites (Bower and Margolis 1991, Nagasawa et al. 1998) and mantle length distribution (Murata 1990, Yatsu et al. 1998a). Murakami et al. (1981) identified the four stocks based on the squid body sizes extra-large (LL), large (L), small (S) and extra-small (SS). Chen et al. (2002) found that two populations existed in the waters of 165°E westward through nine body variables based on the Grey System Theory. A geographic pattern of population genetic variability was observed in O. bartramii, with major genetic differentiation attributable to inconsistency in allele frequency distribution and in levels of genetic variation between the squid from the western and eastern parts of the species, which covers a wide area in the North Pacific Ocean (Katugin 2002). Generally, in the North Pacific, the population of O. bartramii mainly comprises two cohorts: a) the autumn cohort hatching from September to February and b) the winter-spring cohort hatching from January to May (Katugin 2002, Chen and Chiu 2003, Ichii et al. 2004, Bower and Ichii 2005, Chen et al. 2011). The autumn cohort consists of the central stock and east stock, which separate near 160°W. The winter-spring cohort also comprises the west stock and central-east stock, which separate near 170°E (Ichii et al. 2004, Bower and Ichii 2005). Although these two cohorts overlap geographically, they have been caught in different areas at separate times (Bower and Ichii 2005). The autumn cohort (NE stock) distributed in the waters east of 170°E from May to June and the winter-spring cohort (NW stock) distributed in the waters west of 165°E from July to November have been the main fishing target for Chinese squid jigging fleets since 1998 (Chen and Chiu 2003, Wang and Chen 2005).

Stock identification is critical for an effective fishery management to avoid overfishing and promote the sustainable development of fisheries (Cadrin and Silva 2005). Stock structures are often identified and verified on the basis of their different life history strategies and genetic structures. However, morphometric traits are still often used in this field (Sajina et al. 2011). Traditional measurements are based on the conventional orthogonal method, which uses length and width to describe the variables for species with rigid body forms. Unlike fish and many crustaceans, squid have a flexible soft body without a hard surface structure. The cylindrical mantle cavity varies during locomotion and respiration, and the stretched arms and tentacles are also broken easily, often by hooks during the jigging capture process (Cabanellas-Reboredo et al. 2011, Kurosaka et al. 2012). Thus, measurements based on the soft parts of squid are challenging and frequently contain errors, and an alternative structure should be used to separate stocks on the basis of body morphology. However, these measurements are fairly reliable if the approach is correct, as in the studies of Loliginidae by Pierce et al. (1994) and of Ommastrephidae by Martínez et al. (2002).

Hard structures, including the statolith, beak (mandibular) and gladius (pen), which contain a series of ecological information during its mysterious life history, have gradually been used for their stable and constant configuration (Bizikov and Arkhipkin 1997, Piatkowski et al. 2001, Ikeda et al. 2003, Jackson and Domeier 2003, Guerra et al. 2010, Ruiz-Cooley et al. 2013). As calcified structures embedded in cartilage, a pair of statoliths is an indispensable part of the acceleration receptor system that controls the movement and direction of the cephalopod (Hanlon and Messenger 1996, Arkhipkin and Bizikov 2000). They are also used in studies of species identification (Clarke 1978, Dommergues et al. 2000, Lombarte et al. 2006), age estimation (Villanueva 1992, Sánchez 1995, Arkhipkin and Shcherbich 2012), growth pattern (Yatsu 2000, Yatsu et al. 1997) and trace elements (such as strontium) (Durholtz et al. 1997, Ikeda et al. 1996, 1997, Yatsu et al. 1998b) to determine life history and relevant environmental conditions. As the main feeding organ, the beak can be easily preserved, and is non-corroding, so it has been used in the studies of chemical structure (Miserez et al. 2010), aging and growth (Yatsu et al. 1997, Raya and Hernández-González 1998, Yatsu and Mori 2000, Raya et al. 2010, Castanhari and Tomás 2012, Perales-Raya et al. 2014), species identification (Smale et al. 1993), biomass estimation (Lu and Ickeringill 2002), trophic dynamics (Cherel and Hobson 2005, Cherel et al. 2009) and paralarval ontogeny (Uchikawa et al. 2009).

The measurement of hard structures has been the basis of these studies. Radial measurement is a simple way to analyse the relationship between statolith shape and growth pattern (Arkhipkin 2003, Ma et al. 2009, Chen et al. 2010). The beak also has specific characteristics and has been used in species identification (Clarke 1986, Ogden et al. 1998, Lu and Ickeringill 2002, Xavier and Cherel 2009). It has been proved that hard structures with a stable form can perform better than soft tissues (mantle, arm, tentacle, etc.) in squid population division (Martínez et al. 2002). Some new methods (landmarks and outline study) have also revealed that hard structures give good results in identity analysis (Dommergues 2000, Neige and Dommergues 2002, Lombarte et al. 2006, Neige 2006, Crespi-Abril et al. 2010). Some researchers may be cautious about using the beak as an identification material (Xavier et al. 2007), but it is still important for stock identification (Xavier et al. 2011). Therefore, the phenotypic characteristics, such as statoliths and beaks, can play a supportive role for stock identification and investigation.

A morphological variation of cephalopods has occurred among the species due to different genetic structures and populations, induced mainly by oceanographic environmental factors (Adkison 1995). Multivariate analyses of morphometric data have been used in some cephalopods at intraspecies (Martínez et al. 2002) and interspecies (Wolff 1984) levels for the evaluation of taxonomic and geographic variations. Stocks inhabiting different environments can impact their growth patterns, including the shapes of statolith and beak. It is common that the separation of stocks is only based on one material and then compared with the results obtained from different materials (Martínez et al. 2002), and the approach of integrating two materials in the stock separation is rarely used. In this study we integrated the morphological variables of statolith and beaks to compare their differences between stocks and between the sexes and to establish a discriminant function for two seasonal cohorts in the North Pacific Ocean. Our aims were to identify different cohorts of O. bartramii by hard structures (statolith and beaks). This integrated approach can also be used for other species when different body structures are available in the stock identifications.

The survey was carried out in the waters of 150°E-177°W and 38°-44°N from May to October in 2010 to 2012 by the Chinese squid jigging vessel Jinhai 827. The samples were collected randomly from the daily catches and frozen on board immediately for future analysis. The sampling station is shown in Figure 1.

A total of 570 individuals were collected. The mantle length (ML) was measured to the nearest 0.1 mm after thaw in the laboratory, and the sex was identified by their entirely different gonad structure. The sexual maturity stages were determined according to Lipinski and Underhill (1995). Statoliths and beaks were dissected according to Raya et al. (2010) and Chen et al. (2013). As a result, a subsample of 406 pairs of statoliths and beaks were prepared for the analysis. A pair of beaks were precisely paired with a pair of statoliths which had been picked from the same squid sample (Table 1).

Photos were taken using a charge-coupled device (CCD, connecting device) for right statolith under a 50× optical microscope (Olympus) and nine parameters of statolith morphology, i.e. total statolith length (TSL), maximum width (MW), dorsal dome length (DDL), lateral dome length (LDL), dorsal lateral length (DLL), rostrum lateral length (RLL), rostrum length (RL), rostrum width (RW) and wing length (WL), were measured by using the image analysis software WT-Tiger3000 (Fig. 2A). The parameters of each statolith were measured to the nearest 0.01µm. Beaks were also measured by vernier caliper, which included upper hood length (UHL), upper crest length (UCL), upper rostrum length (URL), upper rostrum width (URW), upper lateral wall length (ULWL), upper wing length (UWL), lower hood length (LHL), lower crest length (LCL), lower rostrum length (LRL), lower rostrum width (LRW), lower lateral wall length (LLWL), and lower wing length (LWL) (Fig. 2B). All the data for beaks were measured to the nearest 0.01 mm. Measures of these variables were obtained independently by the two readers. The average was used if the range of the counts for the same sample was within 5% of the error; otherwise it was measured again and the average of the variables of each reader was used for the same sample (Francis and Mattlin 1986, Chen et al. 2013).

A Student t-test was conducted to compare differences between the geographic stock and sexes. All of the variables were subjected to normal distribution (Kolmogorov-Smirnov test, P>0.05). Considering the impact of allometric growth (Moltschaniwskyj 1995, Lombarte et al. 1997, O’Dor and Hoar 2000), raw data standardization should be done before the analysis. The normalization method, as introduced by Lleonart (2000), was used to standardize morphological variables of statoliths and beaks. The accuracy of this method has been demonstrated in related investigations (Pineda et al. 2002, Vega et al. 2002, Lefkaditou and Bekas 2004, Chen et al. 2012). The TSL in statolith and UHL in beak were chosen as the independent variables to standardize the other variables of beaks (Chen et al. 2012). The standardized morphometric variables were represented by adding a lower case letter “s” after each variable, i.e. MWs, DDLs, LDLs, DLLs, RLLs, RLs, RWs, WLs; or UCLs, URLs, ULWLs, UWLs, LHLs, LCLs, LRLs, LLWLs and LWLs.

A stepwise discriminant analysis (SDA) was performed to select the significant standardized morphological variables based on the statolith, beak and the combination of the two structures (P<0.05; Rencher 2002), and the classification functions were developed for the three different materials (i.e., statolith, beak, and their combinations). Finally, a leave-one-out cross-validation (the Jackknife method) was used to determine rates of successful classification of squid from the two stocks for different uses of the hard structures.

Since the 1960s, the structures and morphological characteristics of the statolith and beak of cephalopods have attracted much interest (Clarke 1962, 1978, 2003, Villanueva 1992, Arkhipkin 2003, Chen et al. 2012). For this reason we chose the statolith and beak as materials in this study for their rigid characteristics and wide usage. Meanwhile, data standardization can effectively remove the influence of allometric growth. There are also other sorts of data standardization, such as log-transformation. Some studies have already proved that the method of data standardization used in this study is more effective than the original data, and there are other methods for identification with a low rate of misclassification (Chen et al. 2012, Fang et al. 2012)

Sexual dimorphism usually occurs in cephalopods (Mercer et al. 1980, Pineda et al. 2002, Vega et al. 2002, Chen et al. 2012). Our study also found this difference in the NE stock in both the statolith and beak (P<0.05), and this was mainly affected by the sex-segregated migration whereby males are separated from northward migrating females, which stay at the spawning/nursery ground to avoid cannibalism (Yatsu et al. 1997, O’Dor and Dawe 1998, Ichii et al. 2009). This meant that we could not find more male individuals in our study areas. With the disparity of latitude distribution, environmental conditions can eventually affect the morphology of hard structures. This was also reflected in the results from the discriminant analysis (Tables 2 and 3).

The NW stock revealed a different situation in morphometric characteristics. Most beak characteristics varied according to sex (P<0.05); however, none of the statolith variables showed differences between the sexes (P>0.05). Both sexes of the NW stock had lived in the same area, and shared a similar oceanographic environment with a single migration pattern. Therefore, the difference in the beaks might have been caused by asynchronous maturation, which is a type of reproductive strategy commonly found in other cephalopods (Rocha et al. 2001). However, this sexual dimorphism was small, as is shown by the overlapped dots in the discriminant analysis (Figs 3 and 4).

The population structure of O. bartramii in the Pacific Ocean has been discussed in previous studies (Murakami et al. 1981, Murata 1990, Bower and Margolis 1991), and the two main seasonal cohorts (the NE stock with large size located east of 170°E and the NW stock with small size located west of 170°E, Chen and Chiu 2003) lived in different habitats. One of the reasons for the difference in statolith (both sexes) and beak (female) morphology is a separated migration trajectory. The NW stock occurs in the subtropical frontal zone (STFZ) with rich productivity, whereas NE stock occurs in the subtropical domain which is less productive because it is far from the transition zone chlorophyll front (TZCF). When the TZCF shifts northward in spring, the migration pattern of the NW stock moves with the change of the TZCF, but the NE stock remains southward to the north of the TZCF until summer or autumn (Ichii et al. 2009). Therefore, the NW stock is of large size in the enhanced productivity water, feeding on myctophid (Symbolophorus californiensis, Ceratoscopelus warmingi) and squid (Onychoteuthis borealijaponica, Gonatus berryi). The small-sized NE stock feed on euphausiids, amphipods and fish (Maurolicus imperatorius) (Ichii et al. 2004). This disparity of feeding may induce variation in the beak morphology (Uchikawa et al. 2009). The high rate of classification also demonstrated the difference between the two stocks (Fig. 5).

Many studies have focused on interspecies identification, with relatively high classification rates using beaks (Martínez et al. 2002, Pineda et al. 2002, Chen et al. 2012). The total correct classification rate was nearly 50% for both statoliths and beaks, but increased to about 70% for the combination of both hard parts. Thus, adding more relevant variables can increase the success of classification, using SDA to improve the accuracy of classification by retaining some highly correlated variables although the results of correct classification are still low compared with other studies of cephalopods (Martínez et al. 2002, Chen et al. 2012). The result was credible as the first three variables entered the combination group, which all entered the beak-based and statolith-based SDA (Table 8).

There are still two sub-stocks in the northwest Pacific (central stock and east stock) and northeast Pacific (west stock and central-east stock) (Bower and Ichii 2005). The relatively high successful classification rates demonstrate that morphology of hard parts in sub-stocks showed less variation and it is hard to separate them using only morphology of hard structures.

In conclusion, hard structures, such as statolith and beak in cephalopod, are credible materials to identify at an intraspecies level. The NE stock of O. bartramii showed sexual dimorphism in hard parts, but the NW stock showed no significant difference by sex. The among-stock differences were greater in the female group than in the male group. The classification approach could separate the two stocks, especially the NE stock, with more than 80% accuracy. Meanwhile, data standardization and variable combination could help improve the accuracy of the classification. As the overlapped dots showed in the results, we must be cautious of over-interpretation of phenotype-based analysis, which is more reliable when combined with genetic methods.