文章摘要
应用STAR模型研究海州湾小黄鱼春季资源量的时空分布
Modelling spatio-temporal distribution of Larimichthys polyactis in Haizhou Bay based on STAR model
投稿时间:2020-09-19  修订日期:2021-01-07
DOI:
中文关键词: 小黄鱼  海州湾  结构化加性回归  空间效应  交叉验证  模型比较
英文关键词: Larimichthys polyactis  Haizhou Bay  Structured additive regression  spatial effect  Cross-validation  model comparison
基金项目:山东省支持青岛海洋科学与技术试点国家实验室重大科技专项(2018SDKJ0501-2);国家重点研发计划(2018YFD0900904);国家自然科学基金(31772852)
作者单位邮编
赵伟 中国海洋大学水产学院

青岛海洋科学与技术试点国家实验室

海州湾渔业生态系统教育部野外科学观测研究站
海州湾渔业生态系统教育部野外科学观测研究站 
266003
任一平 中国海洋大学水产学院 
徐宾铎 中国海洋大学水产学院 
薛莹 中国海洋大学水产学院 
张崇良 中国海洋大学水产学院
海州湾渔业生态系统教育部野外科学观测研究站
海州湾渔业生态系统教育部野外科学观测研究站 
266003
摘要点击次数: 125
全文下载次数: 0
中文摘要:
      根据2011年、2013—2016年春季在海州湾进行的渔业资源调查数据,应用结构化加性回归(Structured Additive Regression , STAR)框架,结合delta方法,根据对空间数据的不同处理方式构建了5种物种分布模型,并通过AIC(akaike information criterion)和交叉验证比较了模型应对两种数据类型的拟合效果和预测性能。结果表明,加入空间因子后的模型残差空间自相关性显著降低,且positive模型和delta模型在加入空间因子后模型的拟合效果的提升较binomial模型更明显。空间加性模型(Geoadditive Models)的AIC值最低;变系数模型(Varying Coef?cient Models)的决定系数和AUC最高,模型拟合效果最佳。预测性能上,空间加性模型准确度最高。在最优模型的基础上,本研究根据FVCOM模拟环境数据,利用delta空间加性模型预测了海州湾小黄鱼春季资源的空间分布,结果表明小黄鱼资源分布主要集中于南部和西部近岸地区,随着水深的增加而逐渐减少,且年间变动明显。本研究旨在为海州湾小黄鱼渔业资源的开发和保护提供科学依据。
英文摘要:
      A major challenge for species distribution modeling is to account for the high complexity inherent in the survey data in terms of spatial and temporal variation and eliminate modeling error due to the presence of excessive zeros. In this context, this study was conducted to evaluate performance of five models for estimating abundance and occurrence of small yellow croaker (Larimichthys polyactis) from fishery independent bottom trawl surveys data in Haizhou Bay and adjacent waters during the spring of 2011 and 2013-2016. Five models were formulated that differed in how spatial covariates were represented under structured additive regression framework. Two commonly used response variables were analyzed: occurrence and abundance. Performance metrics includes goodness-of-fit (AUC and AIC for binomial model, R2 and AIC for positive model, R2 for delta model), predictive ability based on cross validation (AUC for binomial model, RMSE for delta model) and independence of residuals (gobal Moran’s index). Result shows, models with spatial covariates have slightly higher fiting AUC and significantly higher R2 in binomial model and delta model respectively. Models with spatial covariates have lower residual spatial correlation, positve and delta models show better improvement than binomial model. Geoadditive model outperformed other mdoels in the predictive performance and gain a better balance between goodness-of-fit and predictive ability. Therefore, we predicted the spatial distribution of small yellow croaker across years using delta-geoadditive model based on FVCOM simulation data. Prediction result reflects the distribution and variations of small yellow croaker. Small yellow croaker mainly distributes in the southern and western coastal areas, in accordance with the water depth in Haizhou Bay. There is an obviously interannual variation in the distribution of small yellow croaker. The scope of distribution also shrinked gradually across years. Our study provided scientific basis for the application of ef?cient management strategies and ensurance of the sustainability of small yellow croaker.
HTML     下载PDF阅读器
关闭

手机扫一扫看