| 英文摘要 |
In this study, we applied generalized additive model (GAM) to analyse the annual recruitment variability of Ommastrephes bartramii in the North Pacific. The abundance of squid was presented by the catch per unit effort (CPUE) from Taiwanese squid fishery data, and the predictor variables were latitude, longitude, month and seas surface temperature (SST). The model was performed firstly for low-CPUE (1997 and 2001) and high-CPUE (1995 and 2000) years. The optimal model was constructed by cubic spline smoother with Gaussian error distribution and identity link. Then, this optimal model was applied to describe annual variability for 18 years (1986 - 2004). The results showed that the effects of predictor variables on the squid abundance differed for low- and high-CPUE years. There was no consistence pattern for spatial variables, while the abundance peak occurred in September and October. The SST variable showed significant effect for low-CPUE years. This may suggest that O. bartramii would concentrate at favourable-SST regions during low abundance period. The models with SST could describe the variability of squid abundance better than models without SST did, though the explained variance was limited. The GAM illustrated the variability of squid abundance better for high abundance periods than for low abundance periods. |