基于紫外荧光水质传感器和机器学习的总氮和总有机碳估算  

Estimation of total nitrogen and total organic carbon based on UV fluorescence water quality sensor and machine learning

作　　者：闫丹丹 黄玉 王冬梅 陈求稳 王智源 刘东升 祝秋恒 隗岚琳 洪艺铭 YAN Dandan;HUANG Yu;WANG Dongmei;CHEN Qiuwen;WANG Zhiyuan;LIU Dongsheng;ZHU Qiuheng;WEI Lanlin;HONG Yiming(Center for Eco-Environmental Research,Nanjing Hydraulic Research Institute,Nanjing 210029;College of Water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210098;Jiangsu Hydraulic Research Institution,Nanjing 210017;Yangtze Institute for Conservation and Development,Nanjing 210098)

机构地区：[1]水利部交通运输部国家能源局南京水利科学研究院生态环境研究所,南京210029 [2]河海大学水利水电学院,南京210098 [3]江苏省水利科学研究院,南京210017 [4]长江保护与绿色发展研究院,南京210098

出　　处：《环境科学学报》2023年第10期155-165,共11页Acta Scientiae Circumstantiae

基　　金：国家自然科学基金(No.52121006,52070132,U2040209);江苏省优秀青年基金(No.BK20200053);江苏省水利科技项目(No.2019002);技术服务类项目(No.Hj221086-4)。

摘　　要：溶解性有机质(DOM)是表征水生态环境演变和退化的重要参考物质,其光谱参数可为水体污染提供稳定的信号,最新研发的以深紫外LED为光源的紫外荧光水质传感器,可实现对DOM主要组分的便携检测.为了更加便捷、及时判断水体污染程度,基于DOM检测数据与机器学习结合的研究方法,探究了长江下游不同水体水质及DOM紫外吸光度、蛋白类和腐殖质类荧光的时空分布特征,并构建了可被DOM荧光特征反映的总氮(TN)和总有机碳(TOC)模拟模型.研究结果表明,受降雨及温度等因素影响,丰水期水质优于枯水期和平水期;DOM浓度受人类活动及沿程富集作用影响,入海口处腐殖化程度高于其他地区;构建的TN、TOC模拟模型,验证期R2分别为0.78和0.83,展现出对两种指标良好的模拟及估算能力.紫外荧光水质传感器检测的DOM荧光特征与TN、TOC具有显著相关性,可作为水质监测的指示性指标,未来以期搭载于智能化监测设备上,实现大江大河DOM实时监测和水质预测预警.Dissolved organic matter(DOM)is an important reference substance to characterize the evolution and degradation of water ecosystems,making its spectral parameters an essential tool for monitoring water pollution.The newly developed UV fluorescence sensor,utilizing deep UV LED as light source,now can realize portable detection of the main components of DOM.In order to enhance the convenience and timeliness of assessing water pollution,the combination of DOM detection and machine learning were applied to investigate the spatial and temporal distribution characteristics of water quality in the lower reaches Yangtze River.This involved the analyzing ultraviolet absorbance of DOM,the fluorescence of protein-like and humic-like in different types of water.Additionally,and the simulation models were constructed to represent total nitrogen(TN)and total organic carbon(TOC)based on DOM fluorescence characteristics.The results showed that the water environment in the basin was generally favorable,with better water quality during the wet season due to the factors like rainfall and temperature,etc.However,The DOM concentration was influenced by human activities,leading to enrichment along the river,and the degree of humization was higher in coastal area than in other areas.The constructed simulation models demonstrated promising performance,with an R2 of 0.78 and 0.83 for TN and TOC,during the model validation period,indicating strongsimulation and estimation ability for these two indicators.The fluorescence characteristics of DOM,detected by the UV fluorescence sensor,were significantly correlated with TN and TOC,making them valuable indicative indexes for water quality monitoring.In future,integrating this technology into intelligent monitoring equipment holds promise for realizing real-time DOM monitoring,water quality prediction and early warning of major rivers.

关 键 词：长江下游 溶解性有机质 紫外荧光 支持向量回归 响应关系 

分 类 号：X522[环境科学与工程—环境工程]