首页|基于机器学习算法的高海拔地区臭氧影响因素重要性分析

基于机器学习算法的高海拔地区臭氧影响因素重要性分析

扫码查看
臭氧(O3)是表征大气氧化性和衡量光化学污染程度的重要指标,长期暴露在高浓度O3环境中,生物体会受到严重危害。银川市位于西北高海拔地区,夏季持续高温且紫外线辐射强烈,在光化学反应下极易生成O3,导致O3污染频发。因此,亟待开展O3污染的研究,探明影响O3浓度变化的关键因子。本研究依托宁夏银川城市生态系统国家定位站,以银川市凤凰公园为研究对象,开展野外同步定位观测试验,获取O3浓度、气象因子和大气污染物等数据,应用机器学习算法中的随机森林模型,研究影响O3浓度变化的关键气象因子和大气污染物。结果表明:(1)建立的随机森林模型,其方差解释率在 88%以上,且观测值与拟合值间的决定系数R2 为 0。974,均方根误差为85。8,拟合效果良好;(2)通过对模型筛选出的影响O3浓度的关键因子进行重要性排序,对模型贡献较大的4个变量分别为相对湿度(27。8)、NO(20。1)、NO2(16。1)和PM2。5(12。7);(3)各变量与O3浓度之间存在显著的非线性关系,其中氮氧化物(NO、NO2)对O3浓度影响的阈值效应最大,其次为相对湿度和温度。因此,应用随机森林模型可从非线性角度阐明O3浓度与气象因子和大气污染物的关系,明确影响O3浓度的关键因子及其阈值效应,从而为高海拔地区O3污染防治提供科学依据和技术支撑。
Importance Analysis of Ozone Influencing Factors in High-altitude Regions Based on Machine Learning Algorithms
Ozone(O3)is a crucial indicator of atmospheric oxidizing capability and photochemical pollution,pos-ing severe risks to organisms due to prolonged exposure to elevated O3 concentrations.Yinchuan City,located in the high-altitude region of Northwest China,experiences persistent high temperatures and intense ultraviolet radiation in summer,facilitating photochemical reactions that lead to frequent O3 production.Therefore,it is imperative to study O3 pollution and identify the key factors influencing O3 concentration changes.This study relies on data from the National Positioning Station of Yinchuan Urban Ecosystem in Ningxia,and focuses on Yinchuan Phoenix Park for field synchronous positioning observation experiments.Data on O3 concentration,meteorological factors and air pollutants were collected and analyzed using the random forest model,a machine learning algorithm,to identify the key meteorological factors and air pollutants affecting O3 concentration changes.The results indicate that:(1)The variance interpretation rate of the random forest model exceeds 88%,with a determination coefficient(R2)of 0.974 between observed and fitted values,and a root mean square error(RMSE)of 85.8,indicating a good fit.(2)The importance ranking of key factors influencing O3 concentration,as identified by the model,shows that the four variables with significant contributions are relative humidity(27.8),NO(20.1),NO2(16.1),and PM2.5(12.7).(3)There is a significant nonlinear relation-ship between each variable and O3 concentration,with nitrogen oxides(NO,NO2)having the largest threshold effect on O3 concentration,followed by relative humidity and temperature.Thus,the application of the random forest model provides a nuanced understanding of the nonlinear relationships between O3 concentration and its influencing factors,clarifying the critical factors and their threshold effects.These findings offer scientific basis and technical support for the prevention and control of O3 pollution in high-altitude regions.

machine learning algorithmsozonerandom forest modelmeteorological factorsatmospheric pollutants

施光耀、杨思琪、张劲松、杜慧慧、庞丹波

展开 >

宁夏大学 生态环境学院/西北土地退化与生态恢复省部共建国家重点实验室培育基地,宁夏 银川 750021

宁夏银川城市生态系统国家定位站,宁夏 银川 750021

中国林业科学研究院林业研究所,北京 100091

银川市生态环境监测站,宁夏 银川 750001

展开 >

机器学习算法 臭氧 随机森林模型 气象因子 大气污染物

国家自然科学基金资助项目宁夏自然科学基金资助项目宁夏高层次人才科研启动项目银川市科技支撑项目宁夏高等学校自然科学优秀青年项目

322016312023AAC031422023BSB030732023SFZD04NY2022006

2024

宁夏大学学报(自然科学版)
宁夏大学

宁夏大学学报(自然科学版)

CSTPCD
影响因子:0.377
ISSN:0253-2328
年,卷(期):2024.45(2)