摘要:构建精度高，适用性强的水生生物生境适宜性评价模型，为河流生境质量评价和生境修复提供理论依据。以长江一级支流桥边河为例，采用典范对应分析和独立性分析，筛选出影响桥边河大型底栖无脊椎动物优势种河蚬生境质量的关键化学因子和物理因子，利用广义加性模型与多项式拟合模型构建生境因子与生物选择的关系。结果表明：（1）影响桥边河河蚬生境质量的主要因子为COD_Mn、TN、DO、Tur（浊度）和Dep（水深）；其中COD_Mn、TN、Tur、Dep与大型底栖无脊椎动物分布成负相关，DO与大型底栖无脊椎动物分布成正相关；（2）桥边河流域河蚬的最适宜COD_Mn含量为1.228 mg/L，最适宜TN含量为0.269 mg/L，最适宜DO含量为11.170 mg/L，最适宜Dep为0.3 m，最适宜Tur为1.130 NTU。（3）线性拟合情况下两种方法均适用；非线性拟合情况下广义加性模型较优于传统多项式拟合模型，且GAM模型在处理离散程度大的数据集时，可以很好的规避多项式拟合过程中出现的过拟合现象。研究显示，运用GAM模型可以更加精确、合理的模拟生境因子与生物选择之间的关系。



Abstract:The health of the Yangtze River Basin is threatened by increasing hydropower development and human activities. How to properly diagnose river has become a global concern. Constructing high accuracy and strong applicability habitat suitability assessment model to evaluate river habitat health statue can provide theoretical basis for habitat quality assessment and habitat restoration in rivers. In this paper, the habitat suitability of macroinvertebrates in Qiaobian River, a tributary of the Yangtze, China was evaluated by investigating the distribution of macroinvertebrates, chemical and physical parameters during January and April, 2019. Base on macroinvertebrate habitat suitability, the key habitat factors were screened by Canonical correlation analysis (CCA) and Independent analysis (Pearson correlation analysis and Spearman correlation analysis). Suitability curves were determined by Generalized Additive Model (GAM) and Polynomial Fitting Model (PFM). The results indicated that: (1) five factors, including COD_Mn, total nitrogen (TN), dissolved oxygen (DO), turbidity and water depth, had significant effects on Corbicula fluminea, the dominant species of Qiaobian River. The Corbicula fluminea was negatively correlated with COD_Mn, TN, turbidity and water depth and positively correlated with DO. (2) Most suitable COD_Mn of 1.228 mg/L, TN of 0.269 mg/L, DO of 11.170 mg/L, water depth of 0.300m and turbidity of 1.130 NTU. (3) While both methods were applicable in the case of linear fitting, Generalized Additive Model is better than Polynomial Fitting Model in the case of nonlinear fitting, and the GAM can avoid the over-fitting phenomenon in the process of polynomial fitting when dealing with data sets with large discretization degree. The research shows that the GAM can more accurately and reasonably simulate the relationship between habitat factors and biological selection, which has important practical significance for river habitat quality evaluation and habitat restoration.





PDF全文下载地址:

https://www.ecologica.cn/stxb/article/pdf/stxb201909201973