基于频度统计和神经网络的北太平洋柔鱼渔场预报模型比较

Comparisons Between two Forecasting Models of Fishing Ground Based on Frequency Statistics and Neural Network for Ommastrephes bartramii in the North Pacific Ocean

利用2004～2010年北太平洋鱿钓船队生产数据和海洋环境数据，以海表温度（SST）1℃、海面高度（SSH）为1cm、叶绿素a浓度（CHL-a）为0．1mg／m^3的间距，分析作业产量、CPUE与SST、SSH、CHL-a的关系，得到柔鱼渔场适宜环境因子范围，并将生产数据和环境数据匹配组成样本集，建立北太平洋柔鱼空间分布BP神经网络模型；利用2011年环境数据预报柔鱼渔场，并与2011年实际生产数据进行对比。结果表明，6～10月各月实际作业位置落入基于频度统计方法预报渔场的概率达90％以上；而BP模型预报的平均精度为79．2％，最低精度为52．5％。基于多环境因子的频度统计柔鱼渔场预报模型优于神经网络模型。

Ommastrephes bartramii is an important target species for Chinese squid jigging fleets in the North Pacific, and the accurate forecasting of fishing ground can provide better scientific guidance for fishing activities. Based on the Chinese squid fishing production data and remote sensing environmental data, the article analyzed the relations of the productions, daily catch （CPUE） and the sea surface temperature （SST）, sea surface height （SSH）, chlorophyll-a （CHL-a）, to get the suitable range of environment factors of suitable fishing ground according to the 1° interval of SST, lcm interval of SSH, 0.1 mg/m^3 interval of CHL-a. Also the spatial distribution model of BP neural network was established.The catch data in 2011is used to verify two models of forecasting fishing ground. It is found that the probability of fishing sites in the potential fishing ground during June to October reached more than 90%. However, the average accuracy rate of BP model reached 79.24% and the minimum accuracy rate was only 52.46%. It is concluded that the forecasting model of fishing ground based on frequency statistics is better than that based on neural network for O. bartramii in the North Pacific Ocean.