摘要
PM2.5浓度时空演化分析有助于认知大气污染的发展和现状.由于我国PM2.5浓度监测起步较晚,积累数据短,有必要利用其它已有数据开展PM2.5浓度模型研究.PM2.5浓度变化受到内部因素与外部气象条件的影响.本文以河北省为例,选择PM10为影响PM2.5浓度变化的内部因素、水汽与风速为PM2.5浓度变化的外部气象条件,融合三种要素构建多变量PM2.5浓度模型.开展了PM2.5与PM10、水汽、风速的相关性分析,鉴于水汽值存在季节性差异,利用小波变换对水汽序列分解重构后再开展PM2.5与水汽的相关性分析.采用回归分析方法构建了融合PM10、水汽、风速的PM2.5浓度多变量预测模型,利用PM2.5实测值进行了模型的可靠性检验.研究发现:PM2.5与PM10、小波变换分解重构后的水汽呈正相关,与风速呈负相关;与PM2.5浓度实测值相比,多变量模型PM2.5浓度预测精度优于单变量模型;对于PM2.5浓度分级预测效果统计,在大气空气质量为良、轻度污染、中度污染的情况下,多变量模型PM2.5浓度预测效果较好.基于多变量要素模型反演的PM2.5浓度序列可用于河北省大气污染变化分析.
The analysis about temporal and spatial evolution characteristics of PM2.5 concentration can be used to recognize the development and status of atmospheric pollution.For short accumulated PM2.5 concentration data,it is necessary to use other existing data to establish a new PM2.5 concentration model.PM2.5 concentration change was affected by internal factors and external meteorological conditions.Taking Hebei province as an example,this paper chose PM10 as the internal factor affecting the change of PM2.5 concentration,and water vapor and wind speed as the external meteorological condition for the change of PM2.5 concentration.A multivariate PM2.5 concentration model was constructed by combining three factors.Firstly,the correlation analysis between PM2.5 and PM10,water vapor and wind speed was carried out.In view of the seasonal difference of water vapor value,the correlation analysis of PM2.5 and water vapor was carried out by using wavelet transform to decompose and reconstruct water vapor sequence.A multivariate prediction model of PM2.5 concentration combining PM10,water vapor and wind speed was constructed by regression analysis method.The reliability test was carried out by using the measured values of non-participating modeling and the predicted values of PM2.5 model.The study found that PM2.5 is positively correlated with PM10,water vapor after the wavelet transform decomposition and reconstitution,and negatively correlated with wind speed.Compared with PM2.5 concentration,multivariate model PM2.5 concentration prediction accuracy is better than univariate model.For the PM2.5 concentration grading prediction effect statistics,the multivariate model PM2.5 concentration prediction effect is better when the atmospheric air quality is good,mild pollution,moderate pollution.The PM2.5 concentration sequence based on multivariate factor model inversion can be used to analyze the change of atmospheric pollution in Hebei province.
作者
王勇
任栋
刘严萍
娄泽生
WANG Yong;REN Dong;LIU Yanping;LOU Zesheng(School of Geology and Geomatics,Tianjin Chengjian University,Tianjin 300384,China;School of Economics and Management,Tianjin Chengjian University,Tianjin 300384,China)
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2020年第3期761-770,共10页
Systems Engineering-Theory & Practice
基金
天津市自然科学基金(17JCYBJC21600)
河北省自然科学基金(D2015209024)。
关键词
PM2.5
PM10
水汽
风速
小波变换
回归分析
PM2.5
PM10
precipitable water vapor
wind speed
wavelet transform
regression analysis