Commercial logbook data for Chinese squid-jigging fisheries of I. argentinus grouped by 0.5° × 0.5° grid cell and by month were obtained from the National Data Center for Distant-water Fisheries of China, Shanghai Ocean University. The peak fishing season of I. argentinus was from January to April in each year (Chen and Chen, 2016). Thus, data from January to April during 2013–2017 were used in the analysis. The data information contained fishing effort (days fished), the location of fishing ground (latitude and longitude in degrees), and catch (unit: tons). The catch per unit effort (CPUE) [tons (t)/days (d)] in a certain fishing unit for I. argentinus were determined by total catch divided by the total fishing effort (Cao et al., 2009). Fishing locations for the I. argentinus fisheries were distributed in the high seas outside the exclusive economic zone of Argentina and the Malvinas Islands within 41°–49°S and 55°–61°W.

Monthly SIE indices were achieved from National Snow and Ice Data Center (https://nsidc.org/data/seaice_index/). Monthly sea water temperature at different depths data were obtained from the Asia-Pacific Data-Research Center, University of Hawaii (http://apdrc.soest.hawaii.edu/las_ofes/v6/dataset?catitem=71). The spatial resolution was 0.1° ×0.1° for sea water temperature data at different depths. Before the environmental data were compiled for analysis, they were resampled on a 0.5° ×0.5° latitude/longitude grid to match the fisheries data.

The first step in construction of the HSI model was to determine the suitability index (SI) based on each environmental factor, which quantified the probability of species availability. Fishing effort was used to develop SI model (Tian et al., 2009). All the SI models were established using the method from Brown et al. (2000) and Li et al. (2014). The calculation formula of SI value was as follows:

where, Effort was the fishing effort in different sea water temperature and Max(Effort) was the maximum fishing effort in different sea water temperature. The estimated SI values and the sea water temperature in different depths were regarded as the input factors to fit the SI curve based on the least square method; the fitting formula was shown as follows:

where a and b were the model parameters estimated by the least square method by reducing residuals between observed and predicted SI values to a minimum; SI_T was the SI values of sea water temperature in different depths; SI values ranged from 0 to 1; and T was the sea water temperature in different depths.

The following step was to construct the integrated HSI model. On the basis of the above SI model, 10 different weight scenarios are assigned to sea water temperature at different depths ( Table 1 ) (Yu et al., 2018). The calculation formula of HSI value was as follows:

where SI_50m , SI_100m , and SI_200m were the SI values of sea water temperature at 50, 100, and 200 m, respectively; k_50m , k_100m , and k_200m were the weight coefficients of SI_50m , SI_100m , and SI_200m , respectively. HSI ≤ 0.2, 0.2 < HSI < 0.6, and HSI ≥ 0.6 were defined as unsuitable habitat, common habitat, and suitable habitat, respectively (Yu et al., 2016; Yu et al., 2019). The proportion of catch, fishing effort, and CPUE within the abovementioned three-type habitat were calculated under different weighting scenarios and used to select the optimal HSI model by comparing the model performance. In general, the proportion of catch, fishing effort, and CPUE showed an increasing trend with the increasing HSI class interval. Finally, the data in 2017 were used to validate the optimal HSI model by determining the proportion of catch, fishing effort, and CPUE within the three-type habitats of I. argentinus. The fishing efforts were overlapped with the predicted HSI values to examine the reliability of the HSI models.

To examine the relationship between vertical water temperature and SIE, spatial correlation between sea water temperature at different depths and SIE from January to April in 1979–2017 was initially determined and drawn (Yu et al., 2017).

In addition, the data of sea water temperature at 50, 100, and 200 m from January to April in 1979–2017 were included into the optimal HSI model, which only need to input sea water temperature in different depths to quantify the proportion of suitable habitat (areas with HSI ≥ 0.6) accounting for the fishing ground in the high seas. The relationship between the proportion of suitable habitat and the SIE was evaluated by correlation analysis.

Furthermore, the years with high and low SIE were selected and divided into two groups from 1979 to 2017 (including 1982, 1983, 1987, 2003, 2008, 2013, 2014, and 2015 with the high SIE; 1980, 1984, 1992, 2005, 2006, 2007, 2010, and 2016 with the low SIE). The suitable habitat and the latitudinal gravity centers (LATG) of HSI were calculated and compared between the two year groups (Yu et al., 2016). The boxplot analysis was used to compare the difference between the two year groups. The LATG was calculated using the following equation:

where LAT_is was the latitude within the fishing unit i in month s and HSI_is was the total HSI within the fishing unit i in month s.

Two representative years with the highest and the lowest SIE (2014 and 2017) were specifically selected from each year group. Between the two years, the sea water temperature, SI, HSI, and the optimal temperature isotherm at different water depths for I. argentinus were compared and analyzed. Spatial distribution maps were created to show the habitat characteristics and specific differences between the two years.

The SI curves of I. argentinus were all significant (P < 0.05) for each environmental factor with high correlation coefficients from January to April during 2013–2016. The specific fitted curve, corresponding equation, and the statistical assessments of the SI models were referred to Figure 1 and Table 2 .

With regard to the selection of the optimal HSI model, 10 different weighting-based scenarios were compared one by one, the proportion of fishing effort, catch, and average CPUE were determined. According to the results, the optimal HSI model with the best prediction performance from January to April was case 4 ( Tables 3 , 4 ). The weights of water temperature at depths of 50, 100, and 200 m were 0.1, 0.1, and 0.8 ( Table 1 ), respectively, implying that sea water temperature at 200 m contributed most to the habitat variation of I. argentinus. Fisheries data in 2017 were used to validate the optimal HSI model. It showed that high catch, fishing effort, and CPUE were consistent with the high HSI class interval in each month, indicating that this model could effectively predict the potential habitat for I. argentinus ( Table 5 ). Moreover, the spatial distribution maps of HSI and effort in 2013–2016 and 2017 also showed the same results ( Figures 2 , 3 ).

According to the spatial correlation analysis between SIE and sea water temperature at 50, 100, and 200 m on the fishing ground in the high seas of the Southwest Atlantic Ocean from January to April in 1979–2017, SIE was significantly positively correlated with sea water temperature at different depths ( Figure 4 ). Regarding the relationship between the proportions of suitable habitats of I. argentinus and SIE from January to April in 1979–2017, it showed that the proportions of suitable habitats was significantly positively correlated with SIE (r = 0.17, P < 0.05), implying that the larger the SIE, the higher the proportions of suitable habitats ( Figure 5 ).

According to the annual average SIE, 1979–2017, 1982, 1983, 1987, 2003, 2008, 2013, 2014, an 2015 were the years with the high SIE; 1980, 1984, 1992, 2005, 2006, 2007, 2010, and 2016 were the years with the low SIE. Those years were divided into two year groups with the high and low SIE and compared the difference in habitat pattern including the proportion and latitudinal location of suitable SI_50m (SI_50m ≥0.6), suitable SI_100m (SI_100m ≥ 0.6), suitable SI_200m (SI_200m ≥ 0.6), and suitable habitats of I. argentinus (HSI ≥ 0.6) ( Figures 6 , 7 ). It showed that the proportions of abovementioned parameters were higher during the years with high SIE, whereas they were lower in the years with low SIE ( Figure 6 ). Furthermore, latitudinal gravity centers of suitable SI_50m (LATG_SI_50m), suitable SI_100m (LATG_SI_100m), and suitable SI_200m (LATG_SI_200m), and suitable habitats of I. argentinus (LATG_HSI) were distributed in the northern waters on the fishing ground in the years with high SIE, whereas they were located in the southern waters in the years with low SIE ( Figure 7 ).

The year 2014 with the highest SIE and year 2017 with the lowest SIE were selected as two representative years corresponding to the year group with high and low SIE, respectively. The spatial distribution for the difference in sea water temperature at various depths illustrated that the water temperature was high on the high-sea fishing ground at different depths. The average water temperature in each month from January to April was higher in 2014 than that in 2017 ( Figure 8 ). Similarly, it was clearly showed that SI values at different water layers were much higher in 2014 than those in 2017 ( Figure 9 ). In addition, the spatial distribution for difference in HSI was consistent with the results of SI comparison; meanwhile, LATG_HSI was located in the northern waters in 2014 comparing to 2017 ( Figure 10 ). On the basis of the fitted formula of the SI model for each environmental factor, the optimum temperature at each water depth from January to April was determined [the b value in formula (2), which was calculated in Table 2 ]. The isotherms of the most suitable temperature at depths of 50, 100, and 200 m in different months of 2014 and 2017 were drawn in the Southwest Atlantic Ocean ( Figure 11 ). According to the spatial distribution of the isotherms, the most suitable temperature isotherms in 2014 with high SIE moved northward comparing to the year of 2017 with lower SIE. The results indicated that the water temperature at different depths was warmer in 2014 combined with higher SI and HSI, and the isotherm of LATG_HSI and optimum water temperature in each water depth shifted northward.

The integrated HSI model was established on the basis of the fisheries data combined with sea water temperature at three crucial water layers from 2013 to 2016, and the accuracy and feasibility of the model were validated using the data in 2017. Moreover, according to the consistency in spatial distribution of HSI and fishing effort, the integrated HSI model was realized with satisfactory prediction performance. HSI model was generally applied to describe the relationship between species and their habitats (Hyun et al., 2020). At present, HSI model has been widely used in the detection and prediction of fishing ground based on different environmental factors (Yu et al., 2020). Previous studies suggested that fishing effort was the optimal indicator of species occurrence among fishing effort, catch, and CPUE for squids (Abrahams and Healey, 1993; Rijnsdorp et al., 2000; Swain and Wade, 2003). Therefore, the fishing effort was used to calculate SI value in this study. Compared the HSI model established by two species occurrence indicators, the importance of fishing vessel operation position was highlighted in this study (Chen et al., 2012; Yu et al., 2020). In addition, the prediction performance in January 2017 was relatively poor compared with other months, the reasons might be as follows: (i) the total fishing effort in January was less; (ii) the fishing effort was scattered at the beginning of fishing season; and (iii) January was the transition period of the habitat distribution for I. argentinus from shallow to deep water (Bazzino et al., 2005).

SST, Chla, and SSH were often used as environmental factors to build HSI model. However, most marine organisms including I. argentinus migrated horizontally and vertically day and night during their whole life cycle and inhabited at different water layers (Koronkiewicz, 1986). Therefore, the HSI model based on surface environmental factors was difficult to accurately describe the relationship between the habitat and organisms. Very few studies investigated the correlation between environmental factors at deep water layers and I. argentinus except surface environmental factors. For other squid species such as Ommastrephes bartrammi in the Northwest Pacific Ocean, model prediction performance showed large difference based on surface environmental conditions and vertical water temperature. For example, four surface environmental factors (SST, sea surface salinity, SSH anomaly, and Chla) from August to October in 1999 to 2004 were used to establish HSI models for O. bartramii, and the prediction success rate was 65.17% (Chen et al., 2010). Whereas the prediction accuracy was 82.5% for the HSI model developed by fishing effort as well as SST, vertical structure of water temperature in 0–50 m, and sea water temperature at 200 m and 300 m (Chen et al., 2012). These findings implied that the HSI model should consider sea water temperature at deep-sea layers to obtain a more accurate HSI model.

According to the optimal HSI model selecting from the 10 weighting scenarios, it illustrated that sea water temperature at 200 m was the most important factor in regulating habitat distribution of I. argentinus. Variability in areas of suitable HSI and its latitudinal location was consistent with suitable SI_200m and its distribution ( Figures 6 , 7 ). The difference of proportion of suitable habitats and LATG_HSI for the two year groups with high and low SIE were most obvious at depth of 200 m.

Illex argentinus is a pelagic economic cephalopod with vertical migration behavior (Bazzino et al., 2005). The vertical water temperature will affect the abundance and distribution of fishing ground of I. argentinus (Chen et al., 2012). There were studies suggested that the ratio of temperature difference at different water depth (sea water temperature gradient △t/△d) also played an important role in fish habitat distribution (Song and Zhou, 2010; Chen et al., 2012; Yan et al., 2016). In this study, an integrated HSI model was established on the basis of sea water temperature at different depths. However, inclusive of water temperature gradient in the establishment of HSI model might be more accurate. The results showed that sea water temperature at 200 m contributed the most important influences to I. argentinus habitat. The important relationship between the distribution of I. argentinus and water temperature at 200 m water layers may be related to thermocline (Yatsu and Watanabe, 1996). The formation, rupture, and spatial heterogeneity of the thermocline in different seasons had a significant impact on marine life in Patagonia Gulf in the northern Argentina of the Southwest Atlantic Ocean (Luzenti et al., 2021). Therefore, for the marine species living above the thermocline, it prevented them from entering the deeper water, which was favorable for the formation of a productive fishing ground above the thermocline. On the contrary, for the marine species inhabiting below the thermocline, it was difficult to attract the fish to rise through the thermocline by means of lighting and other trapping methods, which might lead to lower catch (Wang et al., 2019). Thus, the water layer of the upper and lower boundary of thermocline was important for the formation of fishing grounds (Luzenti et al., 2021). At present, there were few studies on the seasonal and interannual changes on the thermocline position on the fishing ground of I. argentinus in the Southwest Atlantic Ocean. The relationship between the thermocline position and the water depth of 200 m should be further explored.

Antarctic sea ice change impacts global biology and environment. Antarctic Circumpolar Wave (ACW) (White and Peterson, 1996) and Antarctic dipole (ADP) (Yuan and Martinson, 2000) induced by Antarctic sea ice will lead to global large-scale climate variability. With the Antarctic sea ice change-driven global climate change as well as the changing ocean salinity structure, the thermohaline in the oceans will significantly change and yield a huge impact on the distribution and migration of marine species (Shin et al., 2003). In this study, SIE was used as an environmental parameter to indicate the change of Antarctic sea ice; sometimes, sea ice area and SIC have been used as indicators to describe the change condition of Antarctic sea ice (Ma et al., 2004; Dai et al., 2012). To explore the impacts of Antarctic sea ice on the organisms outside the sea ice extend area, we can use the SIE or some other indicators as a bridge to connect the Antarctic Ocean with the marine species in the adjacent waters. For instance, the northern boundary of Antarctic sea ice (Ma et al., 2004) and the Antarctic Sea Ice Oscillation Index (Cheng et al., 2002) were used as parameters to indicate the change of Antarctic sea ice to explore the impact on the oceanic environment outside the Antarctic continent. Sea ice distribution of the Indian Ocean, Atlantic Ocean, and Pacific Ocean was estimated on the basis of the quantitative study of diatoms and radiolarians (Gersonde et al., 2005).

In this study, the spatial correlation coefficients between SIE and sea water temperature at different depths were determined to explore the relationship between them in the high-sea fishing ground ( Figure 4 ). The results showed that SIE was positively correlated with sea water temperature at different depths in the main fishing ground of I. argentinus in the Southwest Atlantic Ocean. Moreover, overviewing the whole Southwest Atlantic Ocean at 100- and 200-m water layers, the influence mechanism of SIE on BC and MC was different, and the SIE was negatively correlated with BC in the northern waters and positively correlated with MC in the southern waters near the entrance of La Plata-Parana River. It was speculated that global warming would lead to the melting of Antarctic sea ice and consequently the decreasing of SIE. The sea waters of MC in the south were affected by the deep cold waters because of the melting sea ice, whereas the northern BC was less affected by the sea ice melting. Therefore, the correlation between the sea water temperature and SIE was different in the northern and southern regions. Further analysis should be performed to explore the specific reasons why the cold and warm currents in different directions were affected by Antarctic sea ice in different ways. Antarctic sea ice was characterized by large interannual variability and an increase in areal coverage since 1979 (Eisenman et al., 2014; Comiso et al., 2017) despite an abrupt reduction in SIE since 2016. It may be the turning point of maintaining low SIE (Ludescher et al., 2018), moreover, the influences of the sustaining low SIE to I. argentinus required continuous observation and study.

Except Antarctic sea ice, because of the increase of carbon dioxide emissions and human-induced global warming, the global climate was undergoing dramatic changes, which affected short-lived cephalopods. For example, the North Atlantic Oscillation dramatically affected Illex illebrosus inhabiting in the North Atlantic Ocean (Dawe et al., 2000), whereas the Antarctic Oscillation (AAO) (Chang et al., 2015) and the Southern Oscillation (Chen et al., 2007) led to the lag of resource change for I. argentinus. In addition, SST anomalies associated with the El Niño events in the Pacific Ocean propagated around the globe via ACW; thus, predicting anomalous oceanic events via teleconnections between SST anomalies in the Pacific and Atlantic Ocean would appear to have the potential to predict the recruitment strength of I. argentinus in the Southwest Atlantic Ocean (Waluda et al., 1999). The relationship between the environment of the Southern Ocean and cephalopods living near the Antarctic continent has also been explored; ocean acidification may have a greater impact on cephalopods (Rodhouse, 2013; Constable et al., 2014; Xavier et al., 2016), especially affecting the hatching rate of cephalopods in summer (Rosa et al., 2014). The expansion of the cold waters in the Southern Ocean will increase the carbon absorption capacity (Bakker et al., 1997) with the declined exchange of carbon dioxide between the atmosphere and ocean (Stephens and Keeling, 2000), resulting in the enhancement of ocean acidity.