| 非参数模型在河湖富营养化研究领域应用进展 |
| |
| 引用本文： | 豆荆辉,夏瑞,张凯,邹磊,宋进喜,塔拉.非参数模型在河湖富营养化研究领域应用进展[J].环境科学研究,2021,34(8):1928-1940. |
| |
| 作者姓名： | 豆荆辉 夏瑞 张凯 邹磊 宋进喜 塔拉 |
| |
| 作者单位： | 1.西北大学城市与环境学院, 陕西省地表系统与环境承载力重点实验室, 陕西 西安 710127 |
| |
| 基金项目： | 中国科学院“美丽中国”生态文明建设科技工程专项XDA230405国家自然科学基金面上项目51879252国家重点研究计划项目2019YFC0408902 |
| |
| 摘 要： | 河湖富营养化过程受流域水污染、生境破坏和闸坝控制等多因素非线性叠加影响，在一定程度上限制了常规水生态机理模型的模拟精度.非参数模型以其强大的数据分析能力在河湖水生态问题诊断和预测方面得到了广泛应用，该文系统梳理了国内外近20年来河湖富营养化非参数模型的相关研究成果，通过Citespace开展基于WoS与CNKI数据库的相关文献大数据可视化分析，全面阐明了结构方程模型（SEM）、贝叶斯网络（BN）、支持向量机（SVM）、人工神经网络（ANN）、随机森林（RF）、梯度推进机（GBM）、广义相加模型（GAM）等主流非参数模型在河湖富养化营研究中的适用性与局限性，对具有相似特征的模型进行对比分析并提出展望，以期为水生态模拟相关研究提供科学有效的方法支撑.结果表明:非参数模型在河湖富营养化研究领域中的应用呈指数增长趋势，其中SEM、BN、RF、GBM和GAM模型适用于河湖富营养化问题的诊断和驱动要素识别，BN、ANN、SVM、RF、GBM和GAM具有良好的非线性拟合预测能力.非参数模型将是今后一段时期内开展水生态大数据分析诊断和预测管控的关键技术手段.综合考虑区域异质性与多重环境因子在不同时空尺度上响应关系及强人类活动干扰下的河流水生态退化风险，利用生态机理模型与非参数模型耦合求解与优化算法引入，精准识别水生态健康退化的环境压力阈值，开展变化环境下的水生态退化风险预测预警，将是未来非参数模型在河湖富营养化应用研究的重要方向.
|
| 关 键 词： | 富营养化 非参数模型 问题诊断 预测预警 对比分析 |
| 收稿时间： | 2020-11-12 |
Application Progress of Non-Parametric Models in the Field of River and Lake Eutrophication Research |
| |
| Abstract: | The eutrophication process of rivers and lakes is affected by the non-linear superposition of many factors such as water pollution, habitat destruction, and dam control, which limits the simulation accuracy of conventional aquatic ecological mechanism models to a certain extent. Non-parametric models have been widely used in the diagnosis and prediction of river and lake water ecological problems with their powerful data analysis capabilities. This paper systematically summarizes the relevant research with the visual analysis of big data in related literature based on WoS and CNKI databases, and comprehensively clarified the applicability and limitations of mainstream non-parametric models, such as, structural equation model (SEM), Bayesian network (BN), support vector machine (SVM), artificial neural network (ANN), random forest (RF), gradient propulsion machine (GBM), and generalized additive model (GAM) in the study of eutrophication of rivers and lakes. Comparative analysis of models with similar characteristics and prospects is put forward. It is expected to provide scientific and effective method support and application progress summary for research related to water ecological simulation. The results show that the application of non-parametric models in the field of river and lake eutrophication research is increasing exponentially. Among them, SEM, BN, RF, GBM and GAM models are suitable for the diagnosis of river and lake eutrophication problems and the identification of driving factors. BN, ANN, SVM, RF, GBM and GAM have good nonlinear fitting and prediction capabilities. Non-parametric models will become the key technical means for the development of aquatic ecological big data analysis and diagnosis and prediction control in the future. Comprehensive consideration of the regional heterogeneity and the response of multiple environmental factors on different time and space scales and the risk of river water ecological degradation under the interference of strong human activities, the use of ecological mechanism model and non-parametric model coupling solution and optimization algorithm introduction, accurately identify the environmental pressure threshold of water ecological health degradation, and carry out the prediction and early warning of water ecological degradation risk under changing environment, which will be the general direction of the future application of non-parametric models in the eutrophication of rivers and lakes. |
| |
| Keywords: | |
|
| 点击此处可从《环境科学研究》浏览原始摘要信息 |
|
点击此处可从《环境科学研究》下载全文 |
|
