主办单位:中国气象局沈阳大气环境研究所
国际刊号:ISSN 1673-503X
国内刊号:CN 21-1531/P

气象与环境学报 ›› 2020, Vol. 36 ›› Issue (4): 59-66.doi: 10.3969/j.issn.1673-503X.2020.04.008

• 论文 • 上一篇    下一篇

重庆市臭氧污染及其气象因子预报方法对比研究

韩余(),周国兵,陈道劲,杨春,闵凡花   

  1. 重庆市气象台, 重庆 401147
  • 收稿日期:2019-05-23 出版日期:2020-08-30 发布日期:2020-06-16
  • 作者简介:韩余,女,1978年生,副研级高级工程师,主要从事环境气象研究, E-mail:hanyubox@126.com
  • 基金资助:
    重庆市气象局2019年业务技术攻关团队项目(ZHCXTD-201905)

Characteristics of ozone pollution and forecasting technique based on meteorological factors in Chongqing

Yu HAN(),Guo-bing ZHOU,Dao-jin CHEN,Chun YANG,Fan-hua MIN   

  1. Chongqing Meteorological Observatory, Chongqing 401147, China
  • Received:2019-05-23 Online:2020-08-30 Published:2020-06-16

摘要:

利用2014年1月1日至2018年12月31日的重庆市空气质量日均值资料,分析了重庆近5 a臭氧污染的特征。发现重庆市臭氧是除PM2.5以外的第二大大气污染物,具有较强的季节变化特征,主要污染时段位于夏半年,在7—8月臭氧污染程度明显超过了PM2.5。臭氧年平均浓度呈现逐年增加的趋势,首要污染物为臭氧的日数在2018年首次超过PM2.5,臭氧成为2018年重庆市的第一大污染物,表明重庆正在由一个以颗粒物污染为主的城市转变为臭氧污染为主的城市。通过对同期逐日气象资料与臭氧8 h滑动平均日最大值相关性分析发现,大气温度、湿度及气压均为影响臭氧污染的重要气象因子。利用气象影响因子,采用逐步回归、支持向量机、神经网络方法对臭氧8 h滑动平均日最大值进行预报实验表明,三种预报模型均具有较强的预报能力,但总体来看预报均比实况略偏小。支持向量机方法的预报效果要稍好于逐步回归和神经网络方法,可为重庆市臭氧浓度预报提供参考。

关键词: 臭氧污染, 气象因子, 逐步回归, 支持向量机, 神经网络

Abstract:

Characteristics of ozone pollution in Chongqing were analyzed based on daily air quality data from 2014 to 2018.The results indicated that O3 is the second primary air pollutants following PM2.5, and has a distinct seasonal variation.O3 pollution mostly occurs in summer and is more severe than PM2.5 pollution in July and August.The annual mean O3 concentration tends to increase year by year, and O3 has become the first primary pollutant in Chongqing since 2018, with the day number of O3 as the daily primary pollutant exceeding that of PM2.5 for the first time.This suggests Chongqing is turning from a city mainly affected by particulate matter pollution to the one dominantly affected by ozone pollution.Air temperature, humidity, and air pressure are important meteorological factors influenc the O3 level in Chongqing based on correlation analysis between daily meteorological factors and daily maximum O3 8-h moving average (O3-8H) concentration.The daily maximum O3-8H concentration is predicted using methods of stepwise regression, support vector machine, and neural network based on meteorological factors.All the three methods perform well and have small underestimations on average.The support vector machine method performs better in prediction of the daily maximum O3-8H concentration than the other two methods and has a good implication for O3 concentration prediction in Chongqing.

Key words: Ozone pollution, Meteorological factor, Stepwise regression, Support vector machine, Neural network

中图分类号: