【摘要】 为了优化稳定性同位素估算消费者营养级的计算过程，降低算法和样品质量对其的影响，应用贝叶斯算法对中国水域已发表文献中的鲢（Hypophthalmichthys molitrix）、鳙（Hypophthalmichthys nobilis）营养级进行了估算和分析。结果显示：贝叶斯算法下鲢、鳙营养级与总磷（TP）、总氮（TN）2种环境因素无显著相关性。部分研究存在采样个数过少、采样频率过低，忽略了基准季节性和种内的变异性等问题，从而无法准确代表目标生物的营养特征，可能导致了营养级估计中出现异常值。贝叶斯算法与传统算法之间无显著性差异，表明其依然无法冲淡采样设计所带来的影响。曲线拟合的结果表明，维持采样数在6个以上有助于提高贝叶斯算法的计算结果的准确性。因此，在试验设计中，建议采样样品数多于6，并考虑季节与空间的差异，避免单次、少量的采集。更多还原

【Abstract】 To improve the calculation process of infer consumer trophic position using stable isotope analysis and to reduce the influence of the current methods and sample quality, this paper applied Bayesian mixing model to estimate the trophic position of silver carp(Hypophthalmichthys molitrix) and bighead carp(Hypophthalmichthys nobilis) in Chinese waters from published literature. The results showed that the trophic position of silver carp and bighead carp had no significant correlation with total phosphorus(TP) and total nitrogen(TN) under Bayesian method. Some studies have the limitation of small sample sizes, low sampling frequency, and ignored the seasonal and spatial variations in baseline, thus failing to accurately represent the trophic signatures of target organisms, which may have led to deviation in trophic position estimates. No significant difference between the Bayesian method and the current method indicates that even more accurate models cannot overcomes the limitations of an irrational sampling design. The results of curve fitting show that the Bayesian approach performs more constrained with the sample size ≥6, and the accuracy is improved. Therefore, in the experimental design, we suggest that the sample size should be ≥6, and that seasonal and spatial variations should be considered and avoid single, small collections.更多还原