| Abstract: | We developed an approach that integrates generalized additive model(GAM) and neural network model(NNM)for projecting the distribution of Argentine shortfin squid(Illex argentinus). The data for this paper was based on commercial fishery data and relevant remote sensing environmental data including sea surface temperature(SST), sea surface height(SSH) and chlorophyll a(Chl a) from January to June during 2003 to 2011. The GAM was used to identify the significant oceanographic variables and establish their relationships with the fishery catch per unit effort(CPUE). The NNM with the GAM identified significant variables as input vectors was used for predicting spatial distribution of CPUE. The GAM was found to explain 53.8% variances for CPUE. The spatial variables(longitude and latitude) and environmental variables(SST, SSH and Chl a) were significant. The CPUE had nonlinear relationship with SST and SSH but a linear relationship with Chl a. The NNM was found to be effective and robust in the projection with low mean square errors(MSE) and average relative variances(ARV).The integrated approach can predict the spatial distribution and explain the migration pattern of Illex argentinus in the Southwest Atlantic Ocean. |
