【摘要】 截止到2020年，全球78亿人中仍有20亿人无法获得或只能获得有限的安全饮用水。地质成因劣质地下水(GCG)的广泛存在是造成这种严酷现实的重要原因之一，因此识别GCG已成为全球关注的热点。近年来，基于树的机器学习方法不仅成为揭示GCG空间分布和防范公共健康风险的有力工具，而且能帮助我们更好地理解地下水中劣质组分的水文生物地球化学行为。为促进基于树的机器学习方法在水文地质尤其是地下水水质与健康领域更为广泛的运用，综述了近20年来分类和回归树、随机森林和增强回归树等基于树的机器学习方法在GCG研究中的应用，讨论了如何应对正确优化模型超参数、细心选择强有力的预测变量和合理评估模型性能等诸多挑战。更多还原

【Abstract】 By 2020,two billion people out of a global population of 7.8 billion still have had no or only limited access to safe drinking water.The widespread distribution of geogenic contaminated groundwater(GCG) is one of the major causes of this serious situation, and identifying GCG has thus become a global concern.In recent years, tree-based machine learning methods have not only become powerful tools for revealing GCG and preventing public health risks, but helped us better understand the hydrobiogeochemical behavior of geogenic contaminants.To promote the wider application of tree-based machine learning in hydrogeological, especially in groundwater quality and health studies, this paper reviews the application of tree-based machine learning methods such as Classification and Regression Trees, Random Forest, and Boosted Regression Trees in GCG researches over the past 20 years, and discusses how to address the challenges of optimizing model hyperparameters, selecting powerful predictors, and evaluating model performance.更多还原