基于机器学习的深圳大气颗粒物消光来源识别

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (1396 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要利用浊度计于2021年11月~2021年12月在深圳市开展颗粒物光散射在线观测,并同步利用相关在线仪器开展PM_2.5及其化学组分观测,结合受体模型和可解释性机器学习模型探究深圳市秋冬季颗粒物散射的主要来源.结果显示,深圳市颗粒物525nm散射系数在观测期间平均值为119.6 1/Mm.相关性分析表明,颗粒物散射系数与相对湿度相关性低,与PM_2.5相关性高(R²>0.8),其中,与PM_2.5中NH₄⁺、NO₃^-、m/z 44、OM、SO₄^2-、黑碳的相关性在0.55~0.83,表明PM_2.5中二次组分及黑碳是影响深圳市颗粒物散射的重要因素.基于相对湿度和受体模型识别的污染源,利用XGBoost模型建立了颗粒物散射系数预测模型并利用SHAP模型识别其主要来源,研究表明二次硝酸盐、二次硫酸盐、机动车排放和二次有机气溶胶是影响深圳市颗粒物散射系数的主要源类,燃煤和相对湿度也有一定贡献,生物质燃烧、船舶排放、工业排放、扬尘和建筑尘的影响小.

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	李仕平
	彭杏
	姚沛廷
	林晓玉
	云龙
	黄晓锋
	何凌燕

关键词 ： PM_2.5, 散射系数, 来源解析, PMF模型, 机器学习

Abstract：High time resolution observations of scattering coefficient of particulate matter (PM) were conducted using turbidimeter, while PM_2.5 and its chemical compositions were analysed using others corresponding online instruments from November 2021 to December 2021 in Shenzhen, China. Receptor model and interpretable machine learning models were employed to identify the major sources of PM scattering coefficient. The average PM scattering coefficient at 525nm was 119.6 1/Mm during the observation period in Shenzhen. Correlation analysis revealed that PM scattering coefficient had low correlations with relative humidity (RH) and high correlations with PM_2.5 and its components, such as NH₄⁺, NO₃^-, m/z 44, OM, SO₄^2-, and black carbon, indicating the important influence of secondary components and black carbon in PM_2.5 on PM scattering coefficient in Shenzhen. The XGBoost model was employed to predict the PM scattering coefficient based on RH and PM_2.5 sources identified by the receptor model, and then the SHAP model was used to identify the major factors of PM scattering coefficient. The results indicated that secondary nitrate, secondary sulfate, vehicle emissions, and secondary organic aerosols were the major sources of PM scattering coefficient in Shenzhen during fall and winter, followed by coal combustion and RH. Biomass burning, ship emissions, industrial emissions, fugitive dust and construction dust had small contributions.

Key words： PM_2.5 scattering coefficient source apportionment PMF model machine learning

收稿日期: 2023-10-16

PACS:

X513

基金资助:国家自然科学基金资助项目(42375101);深圳市科技计划项目(GXWD20201231165807007-20200808165742001)

通讯作者: 彭杏,副研究员,pengxing@pku.edu.cn E-mail: pengxing@pku.edu.cn

作者简介: 李仕平(1965-),男,高级工程师,本科,主要研究方向为大气环境监测.发表论文10篇.953685146@qq.com.

引用本文:

李仕平, 彭杏, 姚沛廷, 林晓玉, 云龙, 黄晓锋, 何凌燕. 基于机器学习的深圳大气颗粒物消光来源识别[J]. 中国环境科学, 2024, 44(6): 2977-2984. LI Shi-ping, PENG Xing, YAO Pei-ting, LIN Xiao-yu, YUN Long, HUANG Xiao-feng, HE Ling-yan. Source apportionment of extinction of ambient particulate matter in Shenzhen using machine learning. CHINA ENVIRONMENTAL SCIENCECE, 2024, 44(6): 2977-2984.

链接本文:

http://www.zghjkx.com.cn/CN/ 或 http://www.zghjkx.com.cn/CN/Y2024/V44/I6/2977

[1] Wang X Y, Zhang R H, Yu W. The effects of PM_2.5 concentrations and relative humidity on atmospheric visibility in Beijing[J]. Journal of Geophysical Research-Atmospheres, 2019,124(4):2235-2259.
[2] 龚天宜,孙兆彬,张小玲,等.人群的自然和社会属性分类下PM_2.5对心血管疾病的影响[J].中国环境科学, 2019,39(3):1289-1298. Gong T Y, Sun Z B, Zhang X L, et al. Natural and social factor as modifiers of the effects of PM_2.5 on daily cardiovascular mortality in Beijing, China[J]. China Environmental Science, 2019,39(3):1289-1298.
[3] 徐北瑶,王体健,李树,等."双碳"目标对我国未来空气污染和气候变化的影响评估[J].科学通报, 2022,67(8):784-794. Xu B Y, Wang T J, Li S, et al. Assessment of the impact of "dual-carbon" goal on future changes in air pollution and climate in China[J]. Chinese Science Bulletin, 2022,67(8):784-794.
[4] Pitchford M, Malm W, Schichtel B, et al. Revised algorithm for estimating light extinction from IMPROVE particle speciation data[J]. Journal of the Air&Waste Management Association, 2007,57(11):1326-1336.
[5] 曹阳,王陈婧,景宽,等.北京城区不同组分PM_2.5散射特性及来源分析[J].环境科学, 2023,44(2):658-669. Cao Y, Wang C J, Jing K, et al. Characteristics and sources of PM_2.5 and its light-scattering properties with different chemical compositions in urban area of Beijing[J]. Environmental Science, 2023,44(2):658-669.
[6] Chow J C, Watson J G, Kuhns H, et al. Source profiles for industrial, mobile, and area sources in the Big Bend Regional Aerosol Visibility and Observational study[J]. Chemosphere, 2004,54(2):185-208.
[7] 韦莲芳,杨复沫,谭吉华,等.大气气溶胶消光性质的研究进展[J].环境化学, 2014,33(5):705-715. Wei L F, Yang F M, Tan J H, et al. Research progress on aerosol extinction properties[J]. Environmental Chemistry, 2014,33(5):705-715.
[8] Tang M, Chan C K, Li Y J, et al. A review of experimental techniques for aerosol hygroscopicity studies[J]. Atmospheric Chemistry and Physics, 2019,19:12631-12686.
[9] Kanakidou M, Seinfeld J H, Pandis S N, et al. Organic aerosol and global climate modelling:A review[J]. Atmospheric Chemistry and Physics, 2005,5:1053-1123.
[10] Zou J N, Liu Z R, Hu B, et al. Aerosol chemical compositions in the North China Plain and the impact on the visibility in Beijing and Tianjin[J]. Atmospheric Research, 2018,201:235-246.
[11] 李倩,吴琳,张进生,等.廊坊市夏季大气气溶胶消光特性及其来源[J].中国环境科学, 2019,39(6):2249-2257. Li Q, Wu L, Zhang J S, et al. Extinction characteristics of aerosol and the contribution of pollution source to light extinction in Langfang[J]. China Environmental Science, 2019,39(6):2249-2257.
[12] 夏瑞,谭健,汪琼琼等.基于PMF的武汉市大气细颗粒物消光来源解析[J].中国环境科学, 2023,43(1):7-19. Xia R, Tan J, Wang Q Q, et al. PMF model based atmospheric aerosol light extinction budget source apportionment study in Wuhan[J]. China Environmental Science, 2023,43(1):7-19.
[13] Peng X, Xie T T, Tang M X, et al. Critical Role of Secondary organic aerosol in urban atmospheric visibility improvement identified by machine learning[J]. Environmental Science&Technology Letters, 2023,10(11):976-982.
[14] Gui K, Che H Z, Zeng Z L, et al. Construction of a virtual PM_2.5 observation network in China based on high-density surface meteorological observations using the Extreme Gradient Boosting model[J]. Environment International, 2020,141:105801.
[15] Ivatt P D, Evans M J. Improving the prediction of an atmospheric chemistry transport model using gradient-boosted regression trees[J]. Atmospheric Chemistry and Physics, 2020,20(13):8063-8082.
[16] Ma N, Aviv D, Guo H S, et al. Measuring the right factors:A review of variables and models for thermal comfort and indoor air quality[J]. Renewable&Sustainable Energy Reviews, 2020,135:110436.
[17] Si M, Du K. Development of a predictive emissions model using a gradient boosting machine learning method[J]. Environmental Technology&Innovation, 2020,20:101028.
[18] Lundberg S, Lee S. A unified approach to interpreting model predictions[C]//In Proceedings of the 31st International Conference on Neural Information Processing Systems, 2017,30:4768-4777.
[19] Su C P, Peng X, Huang X F, et al. Development and application of a mass closure PM_2.5 composition online monitoring system[J]. Atmospheric Measurement Techniques, 2020,13(10):5407-5422.
[20] Yang K, Chen D H, Ding X, et al. Different roles of primary and secondary sources in reducing PM_2.5:Insights from molecular markers in Pearl River Delta, South China[J]. Atmospheric Environment, 2022,294:119487.
[21] Peng X, Huang X F, Wei F H, et al. Identifying the key drivers in retrieving blue sky during rapid urbanization in Shenzhen, China[J]. Journal of Cleaner Production, 2022,356:131829.
[22] Crenn V, Sciare J, Croteau P L, et al. ACTRIS ACSM intercomparison-Part 1:Reproducibility of concentration and fragment results from 13 individual Quadrupole Aerosol Chemical Speciation Monitors (Q-ACSM) and consistency with co-located instruments[J]. Atmospheric Measurement Techniques, 2015,8(12):5063-5087.
[23] 林楚雄,姚沛廷,彭杏,等.深圳市城区冬季大气PM_2.5在线来源解析[J].中国环境科学, 2023,43(2):506-513. Lin C X, Yao P T, Peng X, et al. Online source apportionment of PM_2.5 during winter at urban site in Shenzhen[J]. China Environmental Science, 2023,43(2):506-513.
[24] Chen T, Guestrin C. Xgboost:a scalable tree boosting system[C]. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016:785-794.
[25] Wong P Y, Lee H Y, Chen Y C, et al. Using a land use regression model with machine learning to estimate ground level PM_2.5[J]. Environmental Pollution, 2021,277:116846.
[26] Tao J, Zhang L M, Cao J J, et al. A review of current knowledge concerning PM_2.5 chemical composition, aerosol optical properties and their relationships across China[J]. Atmospheric Chemistry and Physics, 2017,17(15):9485-9518.
[27] Li Z, Sun Y, Wang Q, et al. Nitrate and secondary organic aerosol dominated particle light extinction in Beijing due to clean air action[J]. Atmospheric Environment, 2022,269:118833.
[28] Cheng Y F, Wiedensohler A, Eichler H, et al. Relative humidity dependence of aerosol optical properties and direct radiative forcing in the surface boundary layer at Xinken in Pearl River Delta of China:An observation based numerical study[J]. Atmospheric Environment, 2008,42(25):6373-6397.
[29] Izhar S, Gupta T, Qadri A M, et al. Wintertime chemical characteristics of aerosol and their role in light extinction during clear and polluted days in rural Indo Gangetic plain[J]. Environmental Pollution, 2021, 282:117034.