两种GAM模型对海州湾短吻红舌鳎（Cynoglossus joyneri）资源分布预测效果的比较研究

根据2011年及2013-2018年春、秋两季在海州湾及其邻近海域进行的底拖网调查数据,研究该海域短吻红舌鳎(Cynoglossus joyneri)的资源分布特征及其受环境因子和饵料生物的影响,并比较了两种模型(普通GAM模型和PCA-GAM模型)对其资源分布的预测效果,采用交叉验证的方法对模型的预测能力及拟合效果进行评价。结果显示:PCA-GAM模型的拟合度及预测效果均优于普通GAM模型。春、秋两季海州湾短吻红舌鳎资源丰度均呈现南高北低、近岸浅水区大于深水区的分布特征,因为海州湾南部近岸海域较高的水温利于春、秋季短吻红舌鳎产卵群体性腺发育,较低的盐度利于其鱼卵及仔鱼的生长发育,同时,近岸海域丰富的饵料资源为产卵后的亲体提供大量食物供给。分别应用两种模型预测了2018年春季和秋季短吻红舌鳎在海州湾的资源分布,结果显示,PCAGAM模型的预测值与实际调查的结果更为吻合,预测效果要优于普通GAM模型。本研究为今后开展渔业生物空间分布的研究提供了一种新的方法。

Evaluation of the prediction effect of two GAMs on the distribution of Cynoglossus joyneri in the Haizhou Bay

Based on the bottom trawl surveys in the Haizhou Bay and adjacent waters during spring and autumn of 2011 and 2013-2018, the performance of regular GAM and PCA-GAM was compared, and the distribution of Cynoglossus joyneri in this area was predicted. The predictive ability and fitting effect of the two GAMs were evaluated by cross-validation. The results showed that the goodness of fit and prediction effects of PCA-GAM were better than those of regular GAM. In spring and autumn, the abundance of C. joyneri in the southern waters was higher than that in the northern waters, and the abundance in the near-shore shallow waters was larger than that in the deep waters. The higher water temperature in the coastal waters of the southern Haizhou Bay was conducive to the development of gonads for the spawning groups during spring and autumn. The lower salinity was conducive to the growth and development of fish eggs and larvae. At the same time, the abundant prey resources in the coastal waters provides a large amount of food for it after spawning period. In this study, two GAMs were used to predict the resource distribution of C. joyneri in the Haizhou Bay in spring and autumn of 2018. The results showed that the predicted abundance by PCA-GAM were more consistent with the actual catches, and the performance of PCA-GAM was better than the regular GAM. This study provides a new method for studying the spatial distribution of marine organisms in the future.