基于GAM和权重分析的西北太平洋秋刀鱼渔情预报研究

为了提高秋刀鱼(Cololabis saira)渔情预报模型的时空分辨率, 提升生产经济效益, 本研究基于 2013─2016 年 7—11 月中国在西北太平洋公海的秋刀鱼生产数据及海洋环境数据, 利用广义可加模型(generalized additive models, GAM)分别拟合单位捕捞努力量渔获量(catch per unit effort, CPUE)的适宜性指数(suitability index, SI)与各海洋环境变量之间的 SI 模型, 结合提升回归树模型(boosting regression tree, BRT)进行权重分析, 建立以月份为周期的秋刀鱼栖息地适宜性指数(habitat suitability index, HSI)模型。结果表明, (1) GAM 能较好地拟合适宜性指数与环境变量的关系, 获得最优环境变量参数值；(2) 环境变量对 CPUE 影响权重的前 3 位分别为海表温度梯度、海表温度和混合层深度, 其中, 在秋季 9—11 月海表温度梯度的权重值均为最高；(3) HSI 模型的检验和评价总体准确率分别为 82.0%和 73.2%, 秋季可达 87.7%和 77.9%, 在盛渔期 10 月, 预测准确率达 89.4%；(4) HSI 高值区与秋刀鱼实际渔场在空间分布基本一致。研究表明该模型适用于秋刀鱼的渔情预报, 并在每天的速报中具有明显优势。

Forecasting Pacific saury (Cololabis saira) fisheries based on GAM and weighted analysis in the northwest Pacific

To improve the spatial and temporal resolution of fishery forecast models and the resource utilization and economic benefits of Pacific saury (Cololabis saira), a generalized additive model (GAM) was used to fit the suitability index between the catch per unit effort (CPUE) and marine environmental variables, based on Chinese saury fishery and environmental data from the high seas of the northwest Pacific Ocean during July and November from 2013 to 2016. Weighted analysis was also conducted using boosted regression tree models to develop monthly habitat suitability index (HSI) models. The results indicated that the GAM can be reliably used to fit relationships between the suitability index and environmental variables and can obtain optimal environmental variable values. Weighted analysis showed that the three important environmental variables affecting CPUE were sea surface temperature gradient, sea surface temperature, and mixed layer depth. The weight of the sea surface temperature gradient was the highest during September to November (autumn). The overall accuracy of the HSI model test and evaluation stages were 82.0% and 73.2% respectively, reaching 87.7% and 77.9% in autumn, respectively. Furthermore, forecast accuracy was 89.4% in October during the main fishing season. The high-HSI areas were consistent with the fishing grounds of Pacific saury. Thus, the results show that the HSI model is suitable for forecasting the saury fishery and has a significant advantage in daily forecasting.