南太平洋长鳍金枪鱼渔场重心预测研究

Study on center of gravity prediction of albacore tuna(Thunnus alalunga)fishing grounds in the south Pacific

长鳍金枪鱼(Thunnus alalunga)是大洋洄游性鱼类,其渔场分布与多种环境因子存在一定联系。根据2016−2021年的南太平洋长鳍金枪鱼延绳钓渔捞日志,利用南方涛动指数(Southern Oscilllation Index,SOI)、海表温度(Sea Surface Temperature,SST)、叶绿素(chlorophyll-a,CHLA)浓度、溶解氧(Dissolved Oxygen,DO)浓度等海洋遥感资料,把时空和环境因子作为输入层,单位捕捞努力量渔获量(Catch per unit effort,CPUE)作为输出层,采用遗传算法(genetic algorithm,GA)优化结构为9-13-1的BP神经网络。GA-BP神经网络预测的均方误差(MSE)和R2分别为0.0074和0.397,均优于BP神经网络。预测分析显示,SST和DO浓度是长鳍金枪鱼渔场重心变动的主要环境因子,CPUE较高的渔场SST范围为18~20℃,DO浓度范围为210 mmol/m^(3)以上。预测与实际渔场重心基本一致,且标准化的CPUE与名义CPUE的时间变化趋势总体一致。研究表明,GA-BP神经网络能够较好地预测长鳍金枪鱼的渔场,为金枪鱼渔业生产与管理提供参考依据。

The albacore tuna(Thunnus alalunga)is an oceanic migratory fish species,whose fishing ground distribution is linked to various environmental factors.Based on the longline fishing logs of albacore tuna in the south Pacific from 2016 to 2021,oceanic remote sensing data including the Southern Oscillation Index(SOI),Sea Surface Temperature(SST),Chlorophyll-a(CHLA)concentration and Dissolved Oxygen(DO)concentration were utilized.With spatiotemporal and environmental factors as input layer and catch per unit effort(CPUE)as output layer,a genetic algorithm(GA)was utilized to optimize the structure of a 9-13-1 BP neural network.CPUE standardization was carried out for albacore based on GA-BP neural network to investigate the spatiotemporal variation and influencing factors on the abundance of albacore tuna resources.The model evaluation results indicated that the GABP neural network had better effect than the BP neural network.Predictive analysis showed that SST and DO concentration were the primary environmental factors influencing the shifts in the core area of albacore tuna fishing grounds,with the most productive fishing areas having SSTs between 18℃and 20℃,and DO concentration levels above 210 mmol/m³.The predicted and actual center of gravity for fishing grounds largely coincided,and the standardized CPUE trends over time were generally consistent with nominal CPUE.The research has demonstrated that the utilization of GA-BP neural networks can effectively predict the distribution of fishing grounds for albacore tuna,providing a reference basis for tuna fishery production and management.