基于贝叶斯二项式模型分析棕头鸥种群多度与环境因子间的关系

传统的种群多度调查方法由于默认观察率 p 等于 1,因此极有可能低估种群大小,进而误判种群多度与环境因子间的关系。为了了解禽流感爆发后青海湖棕头鸥种群与环境因子间的关系,为棕头鸥管理提供有效依据,于 2010 年和 2011 年的 4 ～6 月调查了青海湖保护区 23 个观测点的棕头鸥种群数量及环境因子。通过包含观察率的贝叶斯二项式混合模型分析棕头鸥种群多度与环境因子间的关系,采用 DIC 准则进行因子筛选。结果表明: 种群数量亚模型包含取样面积、放牧强度、距公路距离和植被盖度 4 个参数,种群数量随取样面积、距公路距离和植被盖度的增加而增加,随放牧强度的增加而减少; 观察率亚模型包含观察经验和棕头鸥的行为月节律 2 个参数,观察率随观察月的递增而降低,随观察经验的增加而升高,高经验观察者平均每千米观察到 18. 1 只棕头鸥,低经验观察者可以平均观察到 13. 7 只。天气状况不影响观察率,这可能与棕头鸥的觅食栖息地距岸边较近,不影响观察者的观察有关。

Bayesian Binomial Mixture Model to Study of Larus brunnicephalus Population Sizes in Relation to Local Environmental Characteristics

Count-based indices are widely used to study the relationship of bird population and environmental characteristic.But indices are often confounded by variation in detection probability.To characterize environmental conditions that affect breeding distributions,we analyzed count data on brown-headed gulls(Larus brunnicephalus) that were collected around Qinghai Lake in 2010 and 2011.We modeled count data for brown-headed gulls using Bayesian hierarchical model including five local-scale habitat covariates(area of lake,distance to the nearest road,distance to Qinghai Lake,level of grazing,vegetation cover) and four variables for detection probability(wind,rain,month,experience).We used DIC to choose the best model.Our state model for abundance contained four independent log-linear Poisson regressions on area of lake,distance to the nearest road,level of grazing and vegetation cover.The observation model for detection of an individual brown-headed gull contained factors of month and observer experience.Result showed that brown-headed gulls populations increased with area of lake,distance to the nearest road,vegetation cover and decreased with level of grazing.Detectability increased with observer experience and decreased with month.The weather conditions didn’t significantly effect on the detection probability for brown-headed gulls,suggesting that the habitat condition has affect on detectability.