###
气象:2022,48(8):1032-1042
本文二维码信息
码上扫一扫!
基于Bayes判别分析的大气污染潜势定量预报
周须文,高旭旭,于长文,韩世茹,许启慧
(河北省气候中心,河北省气象与生态环境重点实验室,石家庄 050021)
Prediction Method of Air Pollution Potential Based on Bayes Discriminant Analysis
ZHOU Xuwen,GAO Xuxu,YU Changwen,HAN Shiru,XU Qihui
(Key Laboratory of Meteorology and Ecological Environment of Hebei Province, Hebei Climate Center, Shijiazhuang 050021)
摘要
图/表
参考文献
相似文献
本文已被:浏览 255次   下载 835
投稿时间:2021-09-28    修订日期:2022-05-24
中文摘要: 依据大气污染物质量守恒方程,构建大气污染气象因子,并以空气质量指数日增量为对象,对大气污染潜势进行了定量化分级划分。采用Q型聚类分析方法,把秋、冬季大气环流背景分为冷空气型、混合型、暖空气型三种环流类型,并研究了区分三种大气环流类型的指标因子及其阈值。采用Bayes判别分析方法,分别建立了冷空气型、混合型和暖空气型大气环流背景的污染潜势五分级预测模型。对邢台2017—2019年秋、冬季资料建立的预测模型,各年判别正确率分别为80.0%、71.0%、74.7%,综合正确率为75.2%。采用2015—2017年秋、冬季资料对该模型进行检验,综合正确率为63.6%。对2019—2021年秋、冬季大气自净能力指数和污染潜势五分级预测结果与空气质量指数(AQI)日增量实况进行对比分析,污染潜势五分级预测结果与AQI日增量的变化趋势有较高的一致性,相关系数在0.67以上,明显好于大气自净能力指数计算结果;污染潜势五分级预测对极利于扩散等级和极不利于扩散等级判别正确次数明显高于大气自净能力指数。
Abstract:Based on the equation of atmospheric pollutant mass conservation, the meteorological factors which affect air pollution are constructed and daily increment of air quality index (AQI) is taken as the object to quantitatively classify the air pollution potential. The atmospheric circulation background is divided into cold, mixed and warm-air circulations in autumn and winter by the Q-type cluster analysis method. Then the meteorological factor and its thresholds for distinguishing three types of atmospheric circulation are studied. Based on autumn-winter atmospheric data during 2017-2019, the five-grade prediction models of pollution potential are established, and the discriminant accuracies are 80.0%, 71.0% and 74.7% for cold, mixed and warm-air circulations respectively by the use of the Bayes discriminant analysis method. The mean accuracy of three air types reaches 75.2%. When the five-grade prediction models of pollution potential are tested with autumn-winter atmospheric data during 2015-2017, the mean accuracy of the three air types can reach 63.6%. Through comparison of atmospheric self-purification index (ASI) and the results from the five-grade prediction models of pollution potential with daily increment of AQI in autumn and winter from 2019 to 2021, the five-grade prediction models of pollution potential discriminant results are more consistent with the daily increment trend of AQI than ASI, and the correlation coefficient exceeds 0.67. The correct times of the five-grade prediction models of pollution potential for extremely favorable and extremely unfavorable pollutant diffusions are significantly more than by AQI.
文章编号:     中图分类号:    文献标志码:
基金项目:河北省科技计划项目(17275303D)和河北省气象局科研开发项目(19kyd01)共同资助
引用文本:
周须文,高旭旭,于长文,韩世茹,许启慧,2022.基于Bayes判别分析的大气污染潜势定量预报[J].气象,48(8):1032-1042.
ZHOU Xuwen,GAO Xuxu,YU Changwen,HAN Shiru,XU Qihui,2022.Prediction Method of Air Pollution Potential Based on Bayes Discriminant Analysis[J].Meteor Mon,48(8):1032-1042.