| 摘要: |
| 西南大西洋阿根廷滑柔鱼(Illex angentinus)是世界上重要的经济柔鱼类，也是我国远洋鱿钓的主要捕捞对象之一。单位努力量渔获量(CPUE)是渔业中广泛使用的表达种群丰度的指标，但CPUE易受到其他因素的影响，需对其进行标准化。本研究利用2012—2017年1—4月中国大陆西南大西洋阿根廷滑柔鱼鱿钓生产统计数据以及对应区域的环境数据，构建了20种误差反向传播人工神经网络(error backpropagation network, EBP)模型以标准化CPUE。模型以月份(month)、经度(Lon)、纬度(Lat)、海表面温度(SST)、95 m深层水温(PT95)、叶绿素a浓度(Chl-a)、海表面盐度(SSS)为输入因子，隐含层结点数从1~20个逐步增加，输出层为CPUE，以决定系数(R2)、最小均方误差(MSE)和平均相对方差(ARV)作为模型评价标准。结果显示，7-18-1结构模型为最优模型，输入层因子权重从大到小依次为SST、SSS、Month、PT95、Lon、Lat和Chl-a。研究表明，最优BP神经网络模型能较好地预测CPUE时空变化趋势，可以尝试用来作为阿根廷滑柔鱼CPUE标准化的新方法。 |
| 关键词: 阿根廷滑柔鱼 BP神经网络 CPUE标准化 环境因子 |
| DOI：10.19663/j.issn2095-9869.20210421002 |
| 分类号: |
| 基金项目: |
|
| CPUE Standardization of Illex angentinus Based on BP Neural Network |
|
ZHANG Xiancheng1, WANG Jintao1,2,3,4,5, CHEN Xinjun1,2,3,4,5
|
|
1.College of Marine Sciences, Shanghai Ocean University Shanghai 201306, China;2.Key Laboratory of Oceanic Fisheries Exploration, Ministry of Agriculture and Rural Affairs;3.National Engineering Research Center for Oceanic Fisheries;4.Key Laboratory of Sustainable Exploitation of Oceanic Fisheries Resources, Ministry of Education;5.Scientific Observing and Experimental Station of Oceanic Fishery Resources, Ministry of Agriculture and Rural Affairs, Shanghai 201306, China
|
| Abstract: |
| Illex angentinus is an important economic cephalopod worldwide, as well as an important fishing target for China’s mainland and Taiwan Province. Catch per unit effort (CPUE) is a widely used index to express stock abundance in fisheries. However, CPUE is susceptible to many factors; therefore, it must be standardized. In this research, the statistical data of squid fishing production and the corresponding environmental data from January to April, 2012 to 2017 in mainland China were selected, and the BP neural network method was adopted to establish a model to standardize CPUE. The model uses month, longitude (Lon), latitude (Lat), sea surface temperature (SST), potential temperature of the -95 m layer (PT95), chlorophyll-a (Chl-a), and sea surface salinity (SSS) as input factors. There were 12 hidden layers, from 4 to 15, and the output layer was CPUE. R2, mean squared error (MSE), and average relative variance (ARV) were used as the evaluation criteria of the model. The results showed that a 7-18-1 structure was the optimal model, and the input layer factors in order from high to low weights were SST, SSS, month, PT95, Lon, Lat, and Chl-a. The temporal and spatial distribution predictions for the same sea area indicated that although the BP neural network model could not accurately predict the specific values of CPUE, it could predict the temporal and spatial variations of CPUE, which could be used for the CPUE standardization of I. angentinus. |
| Key words: Illex angentinus BP neural network CPUE standardization Environmental factors |
