基于随机森林的南京市PM_(2.5)和O_(3)对减排的响应  

Response of PM_(2.5) and O_(3) to Emission Reductions in Nanjing Based on Random Forest Algorithm

作　　者：尚永杰 茅宇豪 廖宏[1,2] 胡建林 邹泽庸 SHANG Yong-jie;MAO Yu-hao;LIAO Hong;HU Jian-lin;ZOU Ze-yong(Jiangsu Key Laboratory of Atmospheric Environment Monitoring and Pollution Control,Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology,School of Environmental Science and Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China;Key Laboratory of Meteorological Disaster,Ministry of Education(KLME),Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters(CIC-FEMD),International Joint Research Laboratory on Climate and Environment Change(ILCEC),Nanjing University of Information Science and Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学环境科学与工程学院,江苏省大气环境监测与污染控制重点实验室,江苏省大气环境与装备技术协同创新中心,南京210044 [2]南京信息工程大学气象灾害教育部重点实验室,气象灾害预报与评估协同创新中心,气候与环境变化国际联合研究实验室,南京210044

出　　处：《环境科学》2023年第8期4250-4261,共12页Environmental Science

基　　金：江苏省自然科学基金项目(BK20220031)。

摘　　要：自2013年《大气污染防治行动计划》实施后,南京市大气污染有所改善,但仍面临着细颗粒物(PM_(2.5))和臭氧(O_(3))污染问题.为探究污染物浓度对其前体物减排的响应,获得有效的减排策略,常利用大气化学模式进行多组基于排放扰动的敏感性试验,而这需要消耗大量计算时间和计算资源.应用随机森林算法对2015年大气化学传输模式(GEOS-Chem)模拟结果进行机器学习,高效地预测了南京2019年PM_(2.5)浓度日均值和日最大8 h臭氧(MDA8 O_(3))浓度对不同人为源排放控制情景的响应.随机森林结果表明2019年中国人为排放每减少10%,南京ρ(PM_(2.5))季节平均值下降2~4μg·m^(-3).当2019年中国人为源减排比例高于20%时,南京ρ(PM_(2.5))年均值将低于国家二级限值(35μg·m^(-3)).若仅对中国地区O_(3)前体物氮氧化物(NO_(x))和挥发性有机污染物(VOCs)同比例减排,反而可能导致南京MDA8 O_(3)浓度季节平均值上升.2019年中国地区人为排放同等比例减少10%~50%,南京ρ(MDA8 O_(3))季节平均值在春、秋和冬季分别比基准试验增高约1~3、1~4和3~11μg·m^(-3).而当中国地区NO_(x)减排10%且VOCs减排20%时,南京各季节的ρ(MDA8 O_(3))平均值均有所下降(3~6μg·m^(-3));在此基础上,进一步加大VOCs减排比例(30%),南京ρ(MDA8 O_(3))年均值将减少7μg·m^(-3).若是仅进行南京本地人为源减排,南京O_(3)浓度年均值将出现增加.因此,为有效缓解南京O_(3)污染,中国地区NO_(x)和VOCs减排比需小于1∶2.结合随机森林和GEOS-Chem模式可高效地得到污染物对前体物减排的响应,为大气污染防治策略的制定提供有效的科学支撑.High levels of fine particulate matter(PM_(2.5)) and ozone(O_(3)) in ambient air affect climate change and also endanger human health and ecosystems.Air pollution in Nanjing has been improving since the implementation of the“Air Pollution Prevention and Control Action Plan”in 2013.However,Nanjing still faces PM_(2.5) and O_(3) pollution.Evaluating the response of pollutant concentrations to the reductions in precursor emissions is helpful to obtain effective strategies of emission reduction to improve pollution levels.The sensitive simulations of emission perturbation in atmospheric chemistry models directly demonstrate the response of pollution to the reductions in emissions.Nevertheless,these sensitive simulations are limited in computing time and resources.The random forest algorithm was trained by using the simulation results of the atmospheric chemical transport model(GEOS-Chem)in 2015.The changes in daily PM_(2.5) and daily maximum eight-hour O_(3)(MDA8 O_(3))concentrations in Nanjing in 2019 were efficiently predicted under different reduction scenarios of anthropogenic emissions.The simulations showed that the seasonal average ofρ(PM_(2.5)) in Nanjing would decrease by 2-4μg·m^(-3)with the reduction in anthropogenic emissions of 10%in 2019 in China.In the case of controlling only local emissions in Nanjing,the concentrations of PM_(2.5) in Nanjing decreased significantly without local anthropogenic emissions.Additionally,the simulations showed that the annual average ofρ(PM_(2.5)) in Nanjing could be lower than the national secondary limit(35μg·m^(-3))when the anthropogenic emission reduction in China was higher than 20%in 2019.For ozone,the equal proportional emission reductions in nitrogen oxides(NO_(x)) and volatile organic pollutants(VOCs)of O_(3) precursors in China likely led to the increase in seasonal average concentrations of O_(3) in Nanjing.For the proportional reduction of anthropogenic emissions by 10%-50%in China,the seasonal average ofρ(MDA8 O_(3)) in Nanjing in 2019 would increa

关 键 词：细颗粒物(PM_(2.5)) 臭氧(O_(3)) 随机森林 减排情景分析 GEOS-Chem模式 

分 类 号：X51[环境科学与工程—环境工程]