• 关键词：大气PM₁  重金属  环境磁学  支持向量机

• 基金项目：国家自然科学基金项目（No.42077430， 41771533）

• 肖慧
• 南京师范大学环境学院， 南京 210023

• 许悦
• 江苏环保产业技术研究院股份公司， 南京 210000

• 钱新
• 南京信息工程大学江苏省大气环境与装备技术协同创新中心， 南京210044;南京大学环境学院， 污染控制与资源化研究国家重点实验室， 南京210023

• 邓力刚
• 南京大学环境学院， 污染控制与资源化研究国家重点实验室， 南京210023

• 高翔
• 南京大学环境学院， 污染控制与资源化研究国家重点实验室， 南京210023

• 李慧明
• 南京师范大学环境学院， 南京 210023

• 摘要：在南京仙林地区采集大气PM₁，分析PM₁的磁学参数及其中重金属浓度的季节差异，以PM₁中重金属浓度为输出目标，以气象因子、PM₁浓度和磁学参数作为输入变量，采用支持向量机构建重金属磁学诊断模型.结果显示，PM₁的年均浓度为26.31 μg·m^-3，PM₁和其中绝大部分重金属如：As、Cd、Cu、Pb、Zn等的平均浓度在冬季最高，其次为春季，夏季和秋季较低.PM₁中磁性矿物以亚铁磁性矿物为主，低频磁化率在冬季最高，春季最低，饱和等温剩磁和硬剩磁在夏季最高，秋季最低.主成分分析表明，PM₁中重金属与磁性矿物之间的来源具有一定差异，且同时受到气象因素的影响.当模型输入变量加入了磁学参数时，支持向量机对每个重金属模拟的训练和验证相关系数R值均有所提高，且均>0.8.其中，对As、Pb、Zn的模拟训练和验证R值均>0.85，同时误差也较低.大气颗粒物磁学特征可用于对其中重金属浓度进行模拟，不同重金属与颗粒物磁学参数的内在关联需进一步研究.

• Abstract：Samples of atmospheric PM₁ were collected from Xianlin area of Nanjing. The seasonal variations of magnetic parameters of PM₁ and heavy metal concentrations in PM₁ were analysed. The magnetic diagnosis models of heavy metals were established by support vector machine （SVM） with heavy metal concentration in PM₁ taken as the output target and meteorological factors， PM₁ concentration and magnetic parameters as input variables. The results showed that the average annual concentration of PM₁ was 26.31 μg·m^-3. The average concentrations of PM₁ and most heavy metals such as As， Cd， Cu， Pb and Zn were the highest in winter， followed by spring， and lower in summer and autumn. Magnetic minerals of PM₁ were dominated by ferrimagnetic minerals. The low frequency susceptibility was the highest in winter and the lowest in spring. The saturation isothermal remanence and hard remanence were the highest in summer and lowest in autumn. Principal component analysis showed that sources of heavy metals and magnetic minerals in PM₁ were different， and both affected by meteorological factors. When magnetic parameters were added to the input variables of the model， the simulated correlation coefficients （R） in the training and validation stages for each heavy metal were all enhanced and all were > 0.8. Among them， the R values of As， Pb and Zn in both the training and validation stages were all > 0.85 with relatively lower errors. The magnetic characteristics of atmospheric particulate matter can be used to simulate heavy metal concentrations. The internal relationship between each heavy metal and magnetic parameters of particulate matter needs to be further studied.

• Key words：atmospheric PM₁  heavy metals  environmental magnetism  support vector machine

• 
  摘要点击次数： 147 全文下载次数： 206