3.2 Overwintering habitat suitability of L. crocea in the ECS
We used both fitting-based and regression-based methods to construct SI models of each environmental variable and employed both the arithmetic mean model and the geometric mean model under different environmental variable combinations to calculate HSI values. Fitting-based arithmetic mean model with two variables (e.g. depth and SST) yielded the maximum\(R^{2}\) and the minimum AICc value (Supporting information), thus was selected as the final HSI model. The statistical analysis of fitting-based SI models (Supporting information) shows they were all significant (P < 0.05). As shown by the SI curves (Supporting information, Fig. S1), the optimal range for depth, SST and SSS during winter in our study area was 36 – 72 m, 18.2 – 20.5℃ and 33.89 – 34.27, respectively.
The recent five decades’ cooling trend in winter is remarkable, the reduced average SST (–0.028℃/year, R2=0.31, P<0.05) between 1982 and 2019 in the mid-southern ECS in winter (Fig. 4). The cooling trend in our study area may influenced by the Kurushio extension, specifically, in the latest IPCC report (2019), the Kurushio extension exhibit long-term cooling, which is consisted with our result. Also, another study also revealed the cooling trend along China and Japan coast (−0.69± 0.44 °C/decade), opposing to the overarching global warming trend, especially in winter season due to the extreme cold events (Bindoff et al., 2019; Liao et al., 2015). Consisted with the cooling trend of SST during overwintering phase of L crocea , the results of HSI models show the mean habitat overwintering suitability of the 1970s (1971 – 1980), 1980s (1981 – 1990), 1990s (1991 – 2000), 2000s (2001 – 2010) and 2010s (2011 – 2019) shifting in our study area. Fig. 5A shows that there was no significant change (P > 0.05) in the average and optimal habitat area from the 1970s to the 1990s. However, the percentage of optimal habitat decreased significantly (P< 0.05) from 13%, 12% and 13% in 1970s, 1980s and 1990s to 4% and 5% in 2000s and 2010s. Fig. 5B shows that the spatial distribution of habitat suitability also changed: optimal area has moved toward a southeast direction, with suitable habitats became offshore-oriented. Unfortunately, regarding data availability, the HSI models conducted in our study may biased because we used catch data during 1971-1982 as a measure of abundance (e.g. highly dependent on the effort). Hence, abundance data obtained from scientific cruise is more convincing than using catch data as abundance data and should be encouraged in future study.