东南太平洋秘鲁鳀资源量预报模型的构建

Forecasting model to the anchoveta(Engraulis ringens) biomass in the southeastern Pacific Ocean

秘鲁鳀(Engraulis ringens)是栖息于东南太平洋沿岸的一种小型中上层鱼类。有效地对秘鲁鳀资源量进行预报将有助于为我国鱼粉进口企业提供决策支撑。为此,本研究结合秘鲁鳀生物(上一个渔汛季度的资源量、渔获物中的幼鱼比例)和环境(渔场水温和nino1+2区的温度距平)因素及捕捞量为预报因子,利用主成分分析和多元线性回归的方法对17个渔汛季度(2006年至2014年第一渔汛季度)秘鲁鳀的资源量建立预报模型并利用主成分分析的结果对影响秘鲁鳀资源变动的因子进行初步评价。研究表明,随着时间的推移样本量的增加,模型拟合资源量与真实资源量的平均相对误差逐渐下降,拟合资源量序列与真实资源量序列的相关系数逐渐增加。最终模型5(建模数据为2006-2013年第二渔汛季度的数据,验证数据为2014年第一渔汛季度的数据)能够很好地拟合出秘鲁鳀资源量的大小及变动趋势拟合资源量序列与真实资源量序列的相关系数为0.861;拟合资源量序列与真实资源量序列的平均相对误差为12%;预报得到的2014年第一渔汛季度的数据与真实值相比,相对误差为1%。结合主成分分析的结果对影响秘鲁鳀资源变动的因素进行评价结果表明,第一主成分的方差贡献率为46%,其中环境因素占据了最大的载荷;第二主成分的方差贡献率为23%,同样环境因素占据了最大载荷,但是,排名第二和第三的因素分别是上一个渔汛季度的资源量和捕捞量,其载荷相当;第三和第四主成分的方差贡献率分别为9%和7%,其中幼鱼比例占的载荷最大。根据各主成分得分序列与资源量序列的相关系数结果,环境因子对秘鲁鳀资源变动有重要影响。

Anchoveta （Engraulis ringens） is a small pelagic fish living in the coastal area at the Southeast Pacific O- cean. Forecasting anchoveta biomass effectively could provide decision supports for China＇s fishmeal import enter- prises. Therefore, in this study, combining the biological factors （percentage of juvenile and previous biomass）, o- ceanic environmental factors （temperature status in the fishing ground and climatic variations） and catch, we built the anchoveta biomass forecasting model in the seventeen fishing season （from 2006 to the first fishing season of 2014） combining the methods of principal component analysis （PCA） and multiple linear regression. Besides, we evaluated the factors affecting the changes of anchoveta biomass by the result of PCA. With the increase of sample size, the mean of relative error to the predicting sequences and actual sequences became smaller, and the correla- tion coefficient decreased. Finally, model 5 （modeling the data from 2006 to 2013 and validating the data in the first fishing season of 2014） could effectively fit out the changing trends of avchoveta biomass.- the correlation coef- ficient of predicting sequences and actual sequences was 0.86 and the mean of relative error for these two sequences was 12 %. What＇s more, the relative error was only 1% for the predicting value in the first fishing season at 2014. The results of correlation coefficients between the anchoveta biomass sequence and the scores of principal compo- nents sequence indicated that environmental factor could have huge impacts to the anchoveta biomass variations.