决策树模型在水环境监测网络中选取代表性样点的应用

A Study of Representative Sampling Point Selection in Water Monitoring Network via a Decision Tree Model

地表水体中的硝酸盐污染已经成为全球关注的热点环境问题之一。现今,国内外均建立了相关的监测网络对地表水体的水质实施长期监测,但是却导致大量的监测数据累积,给后续的科学研究工作带来了不便,尤其是在庞大的监测网络中如何选取有代表性样点的研究点则成为急需解决的问题之一。以比利时弗拉芒地区地表水的长期监测物理化学指标为例,利用决策树模型评估地表水样点的硝酸盐污染来源专家分类的有效性,为点位优化提供理论依据。原有监测点位的污染源专家分类和模型输出的可匹配率为80%,优化后监测点位从原有47个点降低到30个点,提高了监测工作效率。

Nitrate pollution in surface water has been becoming an environmental issue worldwide. The establishment and development of water environment monitoring programs results in a large amount of data and relative scientific work, especially representative sampling point selection. This paper set up a decision tree model based on physic-chemical parameters to evaluate expert classification. The results demonstrated that the correct classification percentage is 80% compared to expert knowledge. The number of sampling points was reduced from 47 to 30 and promotes work efficiency.