应用贝叶斯生物量动态模型评估印度洋黄鳍金枪鱼资源

利用贝叶斯生物量动态模型对印度洋黄鳍金枪鱼(Thunnus albacares)资源进行了评估,并分析了不同标准化单位捕捞努力渔获量(catch per unit effort,CPUE)、内禀增长率(r)先验分布对评估结果的影响。结果表明:(1)模型能较好拟合日本延绳钓渔业的标准化CPUE,但对中国台湾延绳钓渔业的标准化CPUE拟合较差;当模型单独使用日本标准化CPUE时,评估结果显示印度洋黄鳍金枪鱼被过度捕捞;若模型单独使用中国台湾标准化CPUE,则结果相反,显示印度洋黄鳍金枪鱼未被过度捕捞;而当同时使用两个标准化CPUE时,日本标准化CPUE数据获得更大估计权重,因此,评估结果与单独使用日本标准化CPUE的结果类似。(2)当r采用无信息先验时,r估计偏小,而环境容纳量(K)估计则偏大,参数估计不合理;当r采用信息先验时,r与K的后验分布估计相对合理;由于r与K存在显著的负相关关系,生物量动态模型难于同时有效估计这两个参数,特别是在数据质量较差情况下,因而采用信息先验能提高生物量动态模型参数估计的质量。(3)本研究利用偏差信息准则(Deviance Information Criterion,DIC)与均方误差(Mean Square Error,MSE)统计量对模型进行了比较,并选择模型S8用于评价印度洋黄鳍金枪鱼的资源状态。评估结果认为印度洋黄鳍金枪鱼被过度捕捞,既存在捕捞型过度捕捞,也存在资源型过度捕捞,这与资源合成(Stock synthesis version 3,SS3)等模型的评价结果一致。

Assessment of the Indian Ocean yellowfin tuna (Thunnus albacares) using a Bayesian biomass dynamic model

The aim of the present study was to assess the Indian Ocean yellowfin tuna (Thunnus albacares) using a Bayesian biomass dynamic model and to analyze the impacts of two standardized longline CPUE (catch per unit effort) series from Japan and Taiwan and the prior distributions of intrinsic rate of increase (r) on the results of the assessments. (1) The models fit the standardized CPUE from Japan better than that from Taiwan, and the results indicated that the stock was overfished and subject to overfishing when the standardized CPUE from Japan was singly used in the models. The opposite might be achieved using the standardized CPUE from Taiwan. Furthermore, when both standardized CPUEs were used, the weighting of the model-estimated Japan standardized CPUE was greater than that of the Taiwan standardized CPUE, and the results were similar for models where only the Japan standardized CPUE was used. (2) If uninformative prior was assigned to r, the estimate of the parameters seemed unreasonable because the r was likely to be underestimated, and the carrying capacity (K) was overestimated. If informative prior was used for r, the estimates of r and k seemed more reasonable. Because there is often a strong negative correlation between r and K in biomass dynamics models, it is difficult to correctly estimate r and K simultaneously, especially under data-poor situations. However, by using informative priors, estimates of parameters of biomass dynamics models can be improved. (3) Deviance information criterion (DIC) and mean square error (MSE) were used to evaluate model fitness, and model S8 was selected as the best model for assessing stock status. According to model S8, Indian ocean yellowfin tuna are overfished and subject to overfishing, which was identical to the results based on Stock Synthesis.