Comparison between different modeling techniques for assessing the role of environmental variables in predicting the catches of major pelagic fishes off India's north-west coast

The contribution of four variables, namely Chlorophyll-a (Chl-a), Sea Surface Temperature (SST), diffuse attenuation coefficient (Kd_490 or Kd), and Photosynthetically Active Radiation (PAR), in predicting the catches of major pelagic fish species (Indian mackerel, horse mackerel, Bombay duck, oil sardine, and other sardines) was evaluated using Canonical Correlation Analysis (CCA). The outcome of the analysis was compared with those obtained by using the following models and methods: the Generalized Linear Model (GLM), the Generalized Additive Model (GAM), connection weight methods, and the explanatory methods of Artificial Neural Networks (ANNs). Both the sets of results were in agreement. Neither the GAM nor the ANNs method showed any clear advantage over each other, although the GAM performed better than the GLM.