基于最大熵和栖息地指数模型预测东、黄海日本鲭渔场分布

最大熵模型（Maximum Entropy Model，Maxent）和栖息地指数（Habitat Suitability Index，HSI）模型均广泛应用于渔情预报研究中。为比较两模型渔情预报效果以提升日本鲭（Scomber japonicus）资源的科学管理水平，本研究利用2003−2012年东、黄海日本鲭的渔业数据以及海表温度、海面高度、海表盐度、海表温度梯度等海洋环境数据，构建最大熵模型和HSI模型，以分析、比较两模型对东、黄海日本鲭栖息地的预测效果，并利用受试者工作特征（Receiver Operating Characteristic，ROC）曲线下面积（Area Under Curve，AUC）、模型预测的渔场概率与实际渔获量百分比之间的对应关系对两模型渔情预报效果进行了定量评价。结果表明：（1）最大熵模型预测的渔场发生高概率位置与捕捞位置基本重合，在无历史捕捞数据海域预测渔场发生的概率较低；HSI模型预测的高栖息地指数位置与捕捞位置部分重合，在无历史捕捞数据海域也可获得较高的栖息地指数，将非渔场预测为渔场的概率较高；（2）最大熵模型和HSI模型的月平均AUC值分别为0.95和0.66，故最大熵模型的预测结果相对较好；（3）使用HSI模型时，应在模型中加入非渔场数据，并加强对此类数据的收集，否则该类模型预报渔场时有扩大化的可能；使用最大熵模型时，必须提高渔业数据的空间覆盖率，否则无法全面反映渔场时空分布动态。本文研究结果可为提升东、黄海日本鲭渔情预报精度提供参考。

Prediction of chub mackerel fishing ground distribution in the East China Sea and Yellow Sea based on maximum entropy model and habitat suitability index model

The maximum entropy model (Maxent) and habitat suitability index (HSI) model are widely used in fishery forecasting studies. To compare the forecasting performance of these two models on fishing grounds and improve the scientific management of chub mackerel (Scomber japonicus) resources, this study used the fishery data of chub mackerel in the East China Sea and Yellow Sea from 2003 to 2012, and marine environmental data, including sea surface temperature, sea surface height, sea surface salinity and sea surface temperature gradient, to construct the Maxent model and HSI model. The aim was to analyze and compare the effectiveness of these two models in predicting the habitat of chub mackerel in the East China Sea and Yellow Sea. The quantitative evaluation of the prediction performance of the two models was conducted using the area under curve (AUC) of the receiver operating characteristic (ROC), and the correspondence between the probability of fishing grounds predicted by the models and the percentage of the actual catches. The results showed that: (1) locations predicted by the maximum entropy model to have a high probability of fishing occurrence coincided with actual fishing locations. The probability of predicting fishery occurrence in the sea area without historical fishing data was lower. Locations predicted to have a high habitat index by the HSI model partially overlapped with actual fishing locations. A high habitat index was obtained in the sea area without historical fishing data. The probability of the HSI model predicting non-fishing grounds as fishing grounds was higher than that of the Maxent model; (2) the monthly average AUC values of the Maxent and HSI model were 0.95 and 0.66, respectively, indicating that the Maxent had relatively better predictive results; (3) when using the HSI model, non-fishing grounds data should be added to the model, and the collection of such data should be strengthened otherwise, there is a possibility of overestimation when such models forecast fishing grounds. When using the Maxent, the spatial coverage of fishery data must be improved otherwise, it cannot fully reflect the spatial and temporal distribution dynamics of the fishery. The results of this study provide a reference for improving the accuracy of forecasting for the chub mackerel fishery in the East China Sea and Yellow Sea.