引用本文:白燕,杨建斌,孙俊奎,李建云.基于GAM模型的昆明地面臭氧与气象要素拟合分析[J].环境监控与预警,2023,15(4):38-44
BAI Yan,YANG Jianbin,SUN Junkui,LI Jianyun.Fitting Analysis of Surface Ozone and Meteorological Elements in Kunming Based on GAM Model[J].Environmental Monitoring and Forewarning,2023,15(4):38-44
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 2122次   下载 633 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于GAM模型的昆明地面臭氧与气象要素拟合分析
白燕1,杨建斌2*,孙俊奎3,李建云2
1.中国海洋大学,海洋与大气学院气象学系,山东 青岛 266100;2.云南省疾病预防控制中心,云南 昆明 650022; 3.昆明市气象局,云南 昆明 650034
摘要:
基于昆明2018—2021年O3日最大8h滑动平均值[ρ(O3-8 h)]和气象要素数据,采用广义相加模型(GAM)中的平滑样条函数拟合单要素、交互项的平滑回归函数拟合多要素与ρ(O3-8 h)的影响关系。引入相对危险度概念,用分布滞后非线性模型(DLNM)分析气象要素和ρ(O3-8 h)的滞后效应。构造滞后项和交互项的GAM模型,进行ρ(O3-8 h)拟合预测。结果表明:当地面气压>818 hPa或平均风速<2.0 m/s时,ρ(O3-8 h)出现1~3 d的滞后效应;GAM模型的交互项平滑回归函数优于单要素平滑样条函数的效果;干冷、湿热、低压大风、高压小风天气以及适当的气温和适中的水汽压有利于ρ(O3-8 h)的增加;纳入交互项(包含滞后项)的GAM模型的拟合效果好于其他模型。该模型的判定系数达到0.672,广义交叉验证得分为352,拟合误差为137μg/m3 ,准确率达71.1%,特别在拟合因变量峰值和谷值时优势明显。
关键词:  气象要素  臭氧浓度  广义相加模型  交互作用  滞后效应
DOI:10.3969/j.issn.1674-6732.2023.04.006
分类号:X513
基金项目:云南省应用基础研究计划项目(青年项目)(2017FD004)
Fitting Analysis of Surface Ozone and Meteorological Elements in Kunming Based on GAM Model
BAI Yan1 ,YANG Jianbin 2*,SUN Junkui 3,LI Jianyun2
1. Department of Meteorology, Ocean University of China,Qingdao,Shandong 266100,China; 2.Yunnan Center for Disease Control and Prevention, Kunming,Yunnan 650022,China; 3.The Meteorological Bureau of Kunming,Kunming,Yunnan 650034, China
Abstract:
Based on the O3 concentration (O3-8 h) and meteorological elements data of the same period of each day in Kunming from 2018 to 2021, the smooth spline function in the GAM model was used to fit the single factor and the smooth regression function of the interaction term was used to fit the influence relationship between multiple factors and O3 concentration. The concept of relative risk was introduced,and the lag effect of meteorological elements and O3 concentration was analyzed by DLNM model. The GAM model of lag term and interaction term was constructed for O3 concentration fitting prediction. The results showed that: When the surface pressure was greater than 818 hPa or the average wind speed was less than 2.0 m/s, a lag effect of 1~3 d appears. The interaction term smoothing regression function of GAM model was better than the single factor smoothing spline function. Dry and cold weather, hot and humid weather, low pressure and strong wind, high pressure and weak wind weather, appropriate air temperature and moderate water pressure were conducive to the increase of O3 concentration. The GAM model with lag term and interaction term had better fitting effect than other models. The determination coefficient of the lag term and interaction term GAM model was 0.672, the generalized cross validation score was 352, the fitting error was 13.7 μg/m3 , and the accuracy was 71.1 %, especially in the fitting of the peak and valley values of the dependent variable.
Key words:  Meteorological factor  Ozone concentration  Generalized additive model  Interaction  Lag effect