AI赋能的鱿鱼渔情预报:进展、挑战与展望OA

Cephalopods are major targets of global ocean fishing.To understand the application of artificial intelligence(AI)in squid fishery forecasting,this study addresses on three core scientific questions in distant-water fisheries:where the fishing grounds are(spatial distribution),what the resource abundance is(quantity),and when the peak season occurs(temporal dynamics).The paper systematically reviews the key technological advances of AI in squid fishery forecasting:1)In fishing ground prediction,multi-environmental factor and asymmetric spatiotemporal scaling techniques have significantly improved the accuracy of identifying central fishing grounds and enabled spatiotemporally coupled short-and medium-term forecasting;2)In resource abundance assessment,the integration of deep learning with interpretability analysis has effectively addressed challenges such as"pseudo-zero"bias and spatiotemporal heterogeneity in spatial CPUE prediction;3)In fishing season forecasting,deep learning models based on hybrid attention mechanisms have successfully captured key environmental features to achieve precise predictions of peak fishing periods.Furthermore,this study presents the translational outcomes from algorithms to practical applications,including the global oceanic squid fishery forecasting system"Canglu"and the domestic edge-computing fishery forecasting chip"Aoxin".Finally,it analyzes the current challenges including multi-source heterogeneous data fusion and model uncertainty quantification,and discusses future directions such as vertical-domain large models,intelligent agent interaction,full-link edge computing,and embodied AI.