基于增强回归树的城市PM_(2.5)日均值变化分析:以常州为例  

Variation Analysis of Daily PM_(2.5) Concentrations Based on Boosted Regression Tree: A Case Study in Changzhou

作　　者：葛跃[1] 王明新[1] 孙向武[1] 齐今笛 

机构地区：[1]常州大学环境与安全工程学院,常州213164

出　　处：《环境科学》2017年第2期485-494,共10页Environmental Science

基　　金：国家自然科学基金项目(41641032);江苏省高校"青蓝工程"项目;江苏省研究生培养创新工程项目(KYLX15_1123)

摘　　要：利用2014年12月至2015年11月常州市区6个国控监测站空气污染物浓度逐时数据,分析了PM_(2.5)浓度季节变化特征,采用增强回归树模拟分析了PM10、4种气态污染物和7个气象因子对ρ(PM_(2.5))日变化的贡献.结果表明,常州市区PM_(2.5)污染季节差异明显,冬季污染严重且持续时间长,夏季污染较轻.四季ρ(PM_(2.5))空间分布特征存在一定差异,但各季内不同监测站差异较小.增强回归树对ρ(PM_(2.5))日均值进行模拟和验证得到,训练数据的相关性为0.981,交叉验证的相关性为0.957.此外,模拟值与实测值的标准化平均偏差为1.80%,标准化平均误差为10.41%,可见模型拟合效果较好.PM10、气态污染物、气象因子和区域输送及扩散这4种影响类型对全年ρ(PM_(2.5))日均值差异的贡献率分别为23.4%、28%、36.2%和12.6%,表明在对ρ(PM_(2.5))日均值差异的影响上,气象因子>二次形成>一次源>区域输送及扩散.在对ρ(PM_(2.5))日均值差异贡献率大于5%的因子中,ρ(PM_(2.5))日均值与PM10、相对湿度、CO和O3正相关,与温度、SO2和混合层高度负相关,与大气压和NO2关系较复杂.区域输送及扩散方面,东南风向、偏西风向和偏北风向等上风向周边城市的污染物输送对常州市区PM_(2.5)污染存在较大的负面影响.Based on hourly concentration data from six state-controlled air quality monitoring stations in urban area of Changzhou from December 2014 to November 2015, the seasonal variation of PM2. 5 pollution was analyzed, and the contributions of PM10 , four kinds of gaseous pollutants and seven meteorological factors to daily changes of ρ（PM2. 5 ） were quantified by boosted regression tree （BRT）. The results showed that： the seasonal differences of PM2. 5 pollution were significant, the pollution was serious in winter and the pollution duration was long, while the pollution was light in summer. The spatial distribution of ρ（PM2. 5 ） in four seasons was different, but the six monitoring stations showed similar trends in each season. Daily average ρ（PM2. 5 ） was simulated and verified by BRT. The correlation coefficient of the training data was 0. 981, and the cross-validation correlation coefficient was 0. 957. In addition, the mean deviation between the simulated values and the measured values was 1. 80% , and the standardized mean error was 10. 41% , which showed that the model fitted well. The contribution percentages of four kinds of impact types （PM10 , gaseous pollutants, meteorological factors and regional transport and diffusion） to daily average ρ（PM2. 5 ） changes of four seasons were 23. 4% , 28% , 36. 2% and 12. 6% , respectively. So, the most significant affecting factor was meteorological condition, followed by secondary formation, primary emission, and regional transport and diffusion. In the factors with contribution percentages of more than 5% , the daily averageρ（PM2. 5 ） was positively associated with PM10 , relative humidity, CO and O3 , and was negatively correlated with temperature, SO2 and mixed layer high. In addition, the daily average ρ（PM2. 5 ） had complex relationships with atmospheric pressure and NO2 . For regional transport and diffusion, the polluted air flow from southeast, west and north had a relatively great negative impact on PM2. 5 pollution of

关 键 词：常州市区 PM2.5 季节变化 增强回归树 模拟 验证 贡献率 

分 类 号：X513[环境科学与工程—环境工程]