摘要
北海市银海区小流域(冯家江流域、三合口江流域和福成河流域)地表水富营养化问题严峻,然而对地表水中铵氮(NH4+-N)浓度的预测研究较少.采用多元线性回归、支持向量机和随机森林3种机器学习模型,利用北海市银海区小流域地表水水质全分析数据,预测了研究区地表水中NH4+-N浓度的空间分布.结果表明:随机森林模型的均方根误差中值最低,拟合效果最佳,预测得到的地表水中NH4+-N浓度空间分布与实际NH4+-N浓度分布高度一致;NH4+-N浓度超过地表水劣V类限值2 mg/L的地表水主要分布在冯家江流域;PO43-、HCO3-和总碱度是研究区地表水中NH4+-N污染最显著的指示因子,这与人类活动密不可分.
Abstract
The issue of eutrophication in surface water of small watersheds(Fengjiajiang River,Sanhekou River,and Fucheng River)in Yinhai,Beihai City,is of grave concern.However,there has been limited re-search on the prediction of surface water ammonium nitrogen(NH4+-N)levels.In this study,three machine learning models,namely multiple linear regression,support vector machine and random forest,were em-ployed to predict the spatial distribution of NH4+-N in small watersheds of Yinhai,Beihai City,using com-prehensive water quality analysis data.The results indicate that in multiple experiments,the random forest model consistently exhibited the lowest median root mean square error and the best fitting performance,showing a high degree of consistency with observed NH4+-N distribution in surface water.Based on the re-sults,areas with NH4+-N concentrations exceeding the Class V water quality standard limit of 2 mg/Lare mainly in Fengjiajiang River.Furthermore,PO43-,HCO3-,and total alkalinity are identified as the most sig-nificant indicator factors contributing to the enrichment of NH4+-N in surface water,highlighting the unde-niable influence of human activities on surface water pollution.
基金项目
中国地质调查局地质调查项目(DD20211385)
中国地质调查局地质调查项目(DD20211139)
中国地质调查局地质调查项目(DD20230104)
维普
万方数据