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

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 increase by 1-3μg·m^(-3)in spring,1-4μg·m^(-3)in autumn,and 3-11μg·m^(-3) in winter,respectively,compared with that in the base simulation.With the reduction in anthropogenic NO_(x) emission by 10%and VOCs by 20%,the seasonal average ofρ(MDA8 O_(3)) in Nanjing would decrease by 3-6μg·m^(-3).On this basis,further increasing the proportion(30%)of VOCs emission reduction could reduce the annual average ofρ(MDA8 O_(3)) in Nanjing by 7μg·m^(-3).However,the annual average ofρ(MDA8 O_(3)) of Nanjing in 2019 increased by 1μg·m^(-3),with the local emission reduction of NO_(x) by 10%and VOCs by 30%.Therefore,this showed that the key to alleviate ozone pollution in Nanjing is a reasonable control ratio of ozone precursor emissions and the implementation of regional joint prevention and control.In order to effectively reduce the O_(3) pollution in Nanjing,the emission reduction ratio of NO_(x) and VOCs in China should be less than 1∶2.The response of pollutant concentrations to reductions in precursor emissions were efficiently obtained by the random forest algorithm and GEOS-Chem model.The simulations would provide the scientific basis for the emission control strategy to alleviate air pollution.