Response of Sthenoteuthis oualaniensis to marine environmental changes in the north-central South China Sea based on satellite and in situ observations

In the South China Sea (SCS), Sthenoteuthis oualaniensis (S. oualaniensis) generally has the highest stock density in spring and occupies an important position in fisheries. The responses of S. oualaniensis to marine environments in the north-central SCS in spring (March to May) from 2006 to 2010 were analyzed using satellite and in situ observations, with generalized additive models (GAMs). A high proportion variation in catch per unit effort (CPUE) was explained by environmental variables, including sea surface temperature (SST; explaining 13.8%) and the interaction between SST and chlorophyll a (Chl-a) concentration (explaining 16.9%). SSTs within the range of 24–28°C and Chl-a concentrations within 0.10–0.35 mg/m3 had positive effects on S. oualaniensis CPUE, and SST within 28–29.5°Cand Chl-a concentrations within 0.05–0.20 mg/m3 had negative effects. In addition, the response time of the maximum standardized catch per unit effort (SCPUE) in May to the maximum Chl-a in March was approximately six ten-day time step. The higher Chl-a and smaller stock size of S. oualaniensis in early March 2008 were partly associated with climatic anomalies caused by La Niña in spring and the limitation of S. oualaniensisby low temperature in 2008. The findings in this study can help better protect and manage S. oualaniensis resources in the SCS.


Introduction
Sthenoteuthis oualaniensis (S. oualaniensis) is in the family Ommastrephidae [1], is widely distributed in the tropical and subtropical areas of the Pacific Ocean and Indian Ocean, exhibits strong phototropism and is one of the major target species of large-scale light falling-net fishing in the central South China Sea (SCS) [2]. Research  in the SCS is concentrated in the north-central area and contains the maximum stock density in spring [3,4]. This species migrates from offshore to coastal waters in spring to breed [5,6], primarily consumes fishes, cephalopods and crustaceans in the high trophic levels of 3 and 4, and breeds from March to May [7,8]. S. oualaniensis has a short lifecycle, a rapid growth rate, and high fecundity, thus occupying an important position in the marine ecosystem of the SCS [8].
S. oualaniensis is one of the major species in the SCS, with the number caught by the automatic squid jigging gear on the west coasts of Philippines of at a water depth of 50-100 m ranging from 0. 25-9.11 squids/line hour [9][10][11]. In Vietnamese waters, S. oualaniensis is found at night in 18-30˚C water at a depth from 125 m to the surface; the central habitat of this species is located at 14˚N, 112˚E (9.11 squids/line hour), where offshore upwelling exists [10,11]. In the north-central SCS, studies on S. oualaniensis have focused on its biological characteristics [12,13], yield, exploitation status [8,14,15], and environmental characteristics [3,5]. The suitable sea surface temperature (SST) of the S. oualaniensis habitat was found to be 25.6-29.6˚C and the optimum SST was 28.5-29.5˚C in spring [5,6,8]. These studies focused on the impacts of a single environmental factor on the distribution of S. oualaniensis, but multifactorial interactions and the weights of each factor remain unclear. In addition, the extent to which marine environments influence the distribution of S. oualaniensis habitat in the north-central SCS has rarely been reported.
As squid are short-lived ecological opportunists, the distribution and abundance of their stocks are extremely sensitive to changes in environmental conditions [13]. The distribution of habitat and abundance of S. oualaniensis are closely related to SST, Chl-a, sea surface height, the occurrence of El Niño and La Niña events [16]. Large-scale environmental variability might lead to significant fluctuations in the spatiotemporal distribution of habitat and squid stock level, especially when anomalous environmental conditions occurred. For example, Yu et al. suggested that the La Niña event resulted in more favorable habitats for Ommastrephes bartramii in the Northwest Pacific [17]. Nigmatullin et al. found that the El Niño phenomenon resulted in a reduced population size of Dosidicus gigas and subsequently a large decline in fishery yields [18]. Hu et al. concluded that the changes fishing grounds of tuna are strongly correlated with the occurrence of La Niña and El Niño events in the Western and Central Pacific, and the index of La Niña and El Niño can be used to predict fishing grounds at the year and month scales [19]. To understand the impacts of the abundance of S. oualaniensis on ocean ecosystems, especially in the north-central SCS, it is important to analyze the spatiotemporal distribution of SST, Chl-a and catch per unit effort (CPUE) in the area influenced by La Niña, as well as the mechanism of these changes.
The north-central SCS, with a water depth of more than 1000 m, is far from land; therefore, survey data are difficult to obtain. Satellite remote sensing can provide overall and spatiotemporal information about sea surface wind (SSW), SST, and Chl-a concentration, which the limited number of ship stations and surveys cannot provide [20][21][22]. This study analyzed the spatiotemporal variability and its possible mechanism in S. oualaniensis habitat in the northcentral SCS in spring with satellite remote sensing and in situ investigations. The results of this study will help better protect S. oualaniensis resources and provide an ecosystem-based approach to S. oualaniensis management in the SCS.

Research area
The north-central SCS, located at 13-19˚N, 110-116˚E (black box in Fig 1), including the Xisha-Zhongsha deep waters, is influenced by the South Asia monsoon [23]. The area has favorable climatic conditions with abundant marine organisms near the islands and is one of the major tropical fishing grounds [24]. S. oualaniensis data were derived from monitoring records of the large-scale light falling-net fishing ship "Qiongwenchang 33180". Specifications of the ship are as follows: the material is wood, full length is 32 m, molded breadth is 5.8 m and molded depth is 2.9 m; there is one main engine with 220.5 kW power and one auxiliary engine with 183.8 Kw power. The fishing ship is equipped with 238 metal halide fishing lamps (×1 kW), the outboard effective length of the jackstay is 26.8 m, the ground rope length of fishing net is 208 m, the straightened net height is 66 m and the shallowest operating water depth is 40 m. The Ministry of Agriculture and Rural Affairs, Chinese Government, granted a research permit for the Xisha-Zhongsha Expedition to work within the whole sea area. The research area is not privately owned or protected, and S. oualaniensis is not protected species. No permission was required to collect the fishery data in the Xisha-zhongsha waters.

Catch and effort data
Daily catch and effort data of S. oualaniensis were obtained from the net-cap fishing ship ( Table 1). All the data were grouped by 0.5˚× 0.5˚grid cells (Fig 1B). The survey stations and time are shown in Table 1, including the operating date, voyage number, longitude, latitude, and yield. The CPUEof a fishing grid of S. oualaniensis was calculated as follows: where, the summed catches and fishing days were obtained for all the fishing vessels within a fishing grid. Ten days was chosen as the time for grouping CPUE values within each grid. Response of Sthenoteuthis oualaniensis to marine environmental changes in the north-central South China Sea

Environmental data
Satellite remote sensing SST and Chl-a data were derived from the MODIS Aqua products of NASA (http://oceancolor.gsfc.nasa.gov), with a temporal resolution of one day and a spatial resolution of 4 km. SSW data from 2006-2009 were derived from the QuikSCAT products. As The QuikSCAT sensor had not been in service since 2009, the SSW data in 2010 were derived from WindSat level-3 products (http://www.remss.com). All SSW data were ascending orbit data with a temporal resolution being day and spatial resolution of 0.25˚× 0.25˚.

Generalized additive model fitting procedures
General additive models (GAMs) [25] were used to investigate the influence of environmental variables on the abundance and distribution of fishery resources. This flexible class of mathematical models allows the incorporation of smoothing functions to model the nonlinear effect of continuous explanatory variables [26]. GAMs were constructed in R (Version 3.3.0) (R development Core Team, 2016), using the GAM function of the mgcv package [25], with CPUE as the response variable and time (year and month), location (latitude and longitude) and environmental (SST and Chl-a) variables as the explanatory variables. The formulation of this model was as follows: where, s(x) denotes a spline smoothing function of the covariate x or the interaction between two covariates. SST and Chl-a were treated as interaction terms to account for interactive effects between these drivers of variation in the species, probably driven by marine environmental variables. Logarithmic transformation of the CPUE was used to normalize its asymmetrical frequency distribution, and a value of one was added to all CPUE values to account for zero-value CPUE data. A Gaussian model with an identity link function was the most appropriate and reliable fit for the transformed CPUE data compared with models with other possible GAM error distributions and link functions [27,28]. The modeling approach, based on information theory [29], was chosen to build sets of candidate models of increasing complexity and to select the best model based on minimizing an information criterion. In all cases, decreasing the generalized cross validation (GCV) score coincided with decreasing values of Akaike's information criterion (AIC) and increasing percentages of explanatory deviance [29]. The best GAM was obtained with a backward stepwise procedure by selecting significant P values for each variable. The significance of the factor and the nonlinear contribution of the factor to the nonparametric effect were evaluated by an F test and a chi-square test, respectively [27][28][29].

Data processing
MATLAB 2015b software was used to read the satellite remote sensing SST, Chl-a and SSW data in the research area, and invalid values were eliminated. To reduce the impact of the lack of fishery data in this study, all environmental variables had a basic time unit of 10 days. The mean value of factors was also computed over a time unit of 10 days being the unit, consistent with the CPUE time scale. GrADS software was used to draw spatiotemporal distribution diagrams of SSW, SST, Chl-a and standardized CPUE (SCPUE).

CPUE standardization
Before GAMs were used to standardize the CPUE of S. oualaniensis, it was necessary to determine the statistical distribution of this variable. The histogram of CPUE presented a partially normal distribution (Fig 2A). After logarithmic transformation of CPUE, its partially normal distribution was improved and generally was consistent with a normal distribution ( Fig 2B). Therefore, logarithmic transformation of CPUE was conducted before standardization.

GAM analysis
The results of GAM fitting indicated that the best model for S. oualaniensis included six explanatory variables ( Table 2, Fig 3). The deviance explained by this model was 59.9%, with an R 2 of 0.48 ( Table 2). The relationship between CPUE and each variable in the GAM allowed us to examine the contribution of each variable separately (Table 3). For S. oualaniensis, variables accounting for the greatest deviance in the univariate GAM analysis were the interaction between SST and Chl-a (explaining16.9%), SST (explaining13.8%) and longitude (explaining 8.7%). As indicated by the ANOVA F-ratio test, all the factors in the model were significant (P < 0.05), except for year. When year was added to the model, the AIC and GCV values of the model continued to decrease, and the cumulative deviation interpretation of the model increased, indicating an increase in the fit and generalization of the model. Therefore, year was kept in the model. The chi-squared test indicated nonparametric smoothing effects of the predictive variables. The most significant effect in the nonparametric smoothing was that of the interaction between SST and Chl-a, as indicated by the chi-square test values of each predictive variable shown in Table 3.
Relationships between CPUE and the explanatory variables from GAMs showed that the highest S. oualaniensis CPUE was in May (Fig 3). A significant increase was observed in the period from March to May, and the maximum was reached in May (Fig 3B). Interannual patterns showed a linear decrease in abundance from 2006 to 2010 (Fig 3A). In terms of spatial factors, the CPUE slowly decreased from 14-18˚N (Fig 3C). Longitude and S. oualaniensis CPUE showed a negative linear relationship at 110-112˚E, and a positive linear relationship at 112-116˚E (Fig 3D). SST and S. oualaniensis CPUE increased from 24˚C to 29˚C, declined sharply at 27˚C, and peaked at 26˚C (Fig 3E). The effect of Chl-a on S. oualaniensis CPUE increasedfrom 0.05-0.20 mg/m 3 and decreased from 0.20-0.40 mg/m 3 (Fig 3F).
Interaction effect of SST, Chl-a and SCPUE. The interaction effect of SST and Chl-a on the SCPUE was analyzed by GAMs, which revealed a positive effect when SST was24-27˚C and Chl-a was0.10-0.35mg/m 3 , and a negative effect when SST was28-29.5˚C and Chl-a was 0.05-0.20 mg/m 3 (Fig 4). The interaction effect on SCPUE decreased gradually with increasing SST and decreasing Chl-a (Fig 4). Analysis of spatial trend surface interpolation showed effects of Response of Sthenoteuthis oualaniensis to marine environmental changes in the north-central South China Sea SST and Chl-a on S. oualaniensis SCPUE (Fig 5). In the ranges of 25-28.5˚C and 0.10-0.16 mg/m 3 Chl-a, the SCPUE increased slowly with an increase in SST and a decrease in Chl-a, reaching its maximum in May. In addition, the maximum SCPUE and Chl-a appeared in May and March, respectively, suggesting that the response of S. oualaniensis SCPUE to Chl-a may have lagged by approximately six ten-day time steps (Fig 5).
The climatological averages of SSW, SST, Chl-a and SCPUE in normal years (2006-2007 and 2009-2010) and anomalous year (2008) are shown in Fig 7. In the anomalous year (2008), changes in the marine environment (SSW, SST, Chl-a) were greater than those in normal years (2006-2007 and 2009-2010), especially in early March (Fig 7). Specifically, SSW and Chl-a were significantly higher (Fig 7A and 7C) and SST was lower than those in normal years from March to May (Fig 7B). Furthermore, in the anomalous year (2008), the SCPUE was slightly lower in March and increased significantly in late April (Fig 7D).

Discussion
The landing statistics of S. oualaniensis was used as a proxy for their abundance in the northcentral SCS. The index of CPUE is generally assumed to be proportional to stock size [17]. However, CPUE data can be variable or compromised by environmental changes or management through time. In addition, changes in the abundance of pelagic species from commercial data were often difficult because of the high variability of these resources and their extremely aggregated geographical distribution [16]. However, as S. oualaniensis is the major target species of the north-central SCS and catches are not restricted by management strategy, the CPUE was considered as a proxy of their abundance in the resrarch area. By matching catch statistics with environmental factors, this study provided new insight into the response of S. oualaniensis to marine environmental changes in the north-central SCS.

SST effects on S. oualaniensis
Temperature is one of the major environmental factors affecting squid activities, including aggregation, breeding, and emigration [19,30]. S. oualaniensis is a warm-water oceanic squid species with a habitat closely associated with SST, and the ability to adapt to SST [12,16]. This research indicated that the SST of S. oualaniensis habitat was 25-28.5˚C in the north-central  (Figs 2E, 5 and 6). In the south-central SCS, the SST of S. oualaniensis habitat was 25.6-29.6˚C in spring, and the most appropriate SST was 28.5-29.5˚C [6]. Exploratory fishing production showed that the appropriate SST of the S. oualaniensis central habitat was within 27-29˚C in the northern Arabian Sea [31] and was within 25-29˚C in the northwest Indian Ocean [32,33]. The appropriate SST of S. oualaniensis habitat varied among watersbecause the latitude of cental S. oualaniensis habitat was different and the appropriate SST of S. oualaniensis was higher compared to those of other developed cephalopods [6]. In this study, the research area was at a latitude similar to that of the investigated area in the northern Arabian Sea; therefore, the appropriate SST for S. oualaniensis was consistent with that in the Arabian Sea. Response of Sthenoteuthis oualaniensis to marine environmental changes in the north-central South China Sea CPUE standardization based on GAMs reflect variations in S. oualaniensis resources more objectively [34] and also allows the influences of different spatiotemporal and environmental factors on CPUE to be measured [35]. In this research, SST explained more deviance in the CPUE, reaching 13.80% of deviance explained (Table 3). Thus, S. oualaniensis was greatly influenced by water temperature [6,8,30]. As low SST in March was inappropriate for the growth of S. oualaniensis, the species entered the peak breeding season in April [3], when water temperature rose gradually, and higher net primary production (NPP) occurred during March-April. Larvae and juvenile fish of S. oualaniensis experienced a suitable growth environment, and the number of individuals in April and May gradually increased due to theirshort life cycle, fast growth and breeding migration, peaking in May (Figs 3,6, and 7). The total accumulative deviation explained by theGAM was 59.90% (Table 2), which was similar to previous results [36][37][38]. The GAM had limited ability to explain deviance for S. oualaniensis CPUE, mainly because the growth of S. oualaniensis is influenced by multiple factors, including water temperature, currents, nutrients, and prey biomass [3,31,39]. Environmental factors obtained by satellite remote sensing have limited ability to explain the abundance of S. oualaniensis (Tables 2 and 3). Moreover, the number of survey data points affects model calculation and statistical analysis ( Table 2, Fig 2). In follow-up studies, factors such as ship type, salinity, currents, water depth, etc. will be considered to improve model accuracy.

Chl-a effects on S. oualaniensis
Chl-a can reflect the standing crop of phytoplankton, which is the most important constituent of NPP [1,40]. This measure also directly reflects the biomass and distribution of zooplankton, which are closely related to the distribution of the central habitat of S. oualaniensis [39,40]. GAM analysis showed that Chl-a had secondary significance to S. oualaniensis CPUE (Table 3). This result occurred because S. oualaniensis primarily consumes fishes, cephalopods and crustaceans, with high trophic levels of 3 and 4 [3,7,8], rather than directly feeding on phytoplankton. The effect of Chl-a on the spatiotemporal distribution of S. oualaniensis was indirect and delayed [8,41]. In the north-central SCS, the phenomena of "high Chl-a and small stock size" and "low Chl-a and large stock size" in spring were observed (Fig 5), partly related to the delay in stock growth caused by the biological habit of S. oualaniensis [12]. This species has a short lifecycle, with the maximum age of larva and juvenile squid being 100-110 days [41], and enters the peak breeding season for the growth of larva and juvenile squid in April [7]. Chl-a was at a high level in March, providing prey allowing an increase in S. oualaniensis resources from April to May (Figs 6 and 7). However, S. oualaniensis migrated from the offshore area to the northern shallow waters of the SCS to breed in spring [6,42]. Water temperature gradually increased, reaching the optimum habitat temperature for S. oualaniensis and the multi-island topography (Xisha-Zhongsha Islands) provided favorable conditions for the breeding migration and larvae aggregation of S. oualaniensis in the north-central SCS [43,44].
Larval and juvenile S. oualaniensis grew rapidly with high primary production and appropriate water temperature, resulting inthe appearance of maximum stock size in May (Figs 5, 6, and 7). The delay between the peak value of S. oualaniensis and that of Chl-a was approximately six tenday time steps (Fig 5). In addition, the SCPUE of S. oualaniensis exhibited small fluctuations during 2009-2010, which might have been related to the small number of survey data pointsin this period (Fig 2C). Further investigation will be conducted to improve the size of the survey dataset.

Low S. oualaniensis density induced by climatic anomalies in 2008
Climatic anomalies, such as El Niño and La Niña, influence the distribution and stock density of fishery resources [19]. As S. oualaniensis is a squid with a short life cycle, its abundance was extremely vulnerable to anomalous environmental conditions. SSW was at its maximum in early March 2008, when SST was the lowest (expressed by dotted boxes in Fig 6), which might have been related to the La Niña event in early 2008 and local circulation anomalies in the north-central SCS [45,46]. The La Niña event formed in August 2007 and ended in April 2008, during which time sea temperature in the middle-eastern equatorial Pacific Ocean and the SCS was partially low [45]. In addition, the tropical convective activity in the Western Pacific warm pool in spring of 2008 was strong, and the north-central SCS waters are located at the edge of the Western Pacific warm pool [42,47]. Severe convective weather resulted in increased wind speed, decreased water temperature and large amounts of rainfall in the observation area [47,48]. La Niña increased upwelling and SSW in the Western Pacific warm pool (Fig 6), resulting in adifference in the heat distribution at the sea surface and an increase in vertical seawater exchange in this area [48,49]. As a result, nutrients from bottom water were brought to the surface. In tropical area, this bottom water with low temperature and large amounts of nutrients will increase surface Chl-a and local alga blooms (Fig 6) [47][48][49]. Therefore, the breeding, growth and migration of S. oualaniensis might have been limited by the lower SST in early March 2008 [50], resulting in the appearance of the lowest SCPUE observed in the five years (March-May during 2006-2010; Fig 2C). The high Chl-a in this period provided high primary production for the growth of juvenile S. oualaniensis (Fig 6), promoting the maximum abundance of S. oualaniensis in May, 2008 ( Fig 2C).

Conclusions
In this research, satellite remote sensing data of SSW, SST, Chl-a and fishery resource production were used to analyze the relationship between S. oualaniensis and marine environments in the north-central SCS in spring.
1. Positive effects on S. oualaniensis CPUE were observed for SSTs of 24-28˚Cand Chl-a concentrations of 0.10-0.35 mg/m 3 , and negative effects were observed for SSTs of 28-29.5˚C and Chl-a concentrations of 0.05-0.20 mg/m 3 .
2. The SCPUE of S. oualaniensis gradually increased from March to May, reaching its maximum in May. The response time of maximum SCPUE to maximum Chl-a was approximately six ten-day time steps in the north-central SCS in spring.
3. The higher Chl-a and smaller stock size of S. oualaniensis in early March in 2008 were caused by higher SSW and lower SST, which were partly associated with climatic anomalies induced by La Niña in spring 2008.