| Due to the complex changes and influential factors of the atmospheric components in the atmospheric environment,how to effectively analyze the changes of these substances,and then study the changes in air quality are of great significance.In this context,the long-term or short-term variations of atmospheric substances of different phase in the atmosphere were studied based on the machine learning algorithm,and fully considering the relevant factors such as meteorological conditions and source emissions,to achieve accurate analysis of air quality evolution trends.To this end,we have chosen three scenarios to conduct our study:(1)Studying about the change rule of atmospheric substances in short term under the scenario of pollutants burst reduction.This study was based on the observation data of 35 air quality monitoring sites in Beijing,with PM2.5,PM10,SO2,NO2,O3 and CO as research objects.During the lockdown,the study found that most of air pollutants in Beijing had declined significantly due to the remarkable reduction in anthropogenic emissions,of which PM2.5,PM10,SO2,NO2 and CO concentrations decreased by 6.5%?42.3%,48.4%?52.5%,48.6%?54.4%,37.1%?51.0%,and 33.5%?37.8%,respectively.There had the largest decrease for PM2.5 and SO2 at the transport site,for CO and NO2 at background sites,and for PM10 at the urban site.During the epidemic,most of the pollution levels of most air pollutants reached the lowest level in 2020,but there were still many serious haze events during the Spring Festival,primarily due to unfavorable meteorological conditions resulting in increasing in PM2.5 by 92%?126.4%.(2)Studying about the change rule of atmospheric substances in short term under the scenario of pollutants burst emission.This study was based on the observation data of the Xianghe site,with NH3 as the research object.During the study,the study found that most of the NH3concentrations were mostly below 65 ppb.However,unexpected burn events result in significant increase in NH3 concentrations from November 18 to 21,2017,with a peak exceeding 600 ppb.Using a machine-learning technique,we quantified that this burn event caused a significant increase in NH3 concentrations by 411%compared with the scenario without the burn event.Notably,?NH3/?CO ratio is 0.016 during the burning period,indicating that biomass burning may be the dominant source.In addition,the NH3 emissions from the burn event could be transported to downwind regions far away(?300 km),such as Tianjin,Tangshan,and the Bohai Bay.(3)Studying about the change rule of atmospheric substances in long term in the atmospheric environment.This study was based on the observation data of Nanyue site,with p H and electric conductivity in the precipitation as the research object.During the 2007–2020,the study found that the average annual rainfall showed a"M"trend and ranged from 1508.8?2364.9 mm,with an average rainfall of 1876.5 mm.The average annual p H presented a change trend of first decreasing,then slowly increasing,and quickly increasing,with the change ranges from 4.4 to 5.4 and the average is 4.9.The average annual electric conductivity showed a decline trend in the overall,with the change ranges from 26.9 to 74.8?s cm-1 and the average was 46.7?s cm-1.The influence of meteorological conditions on the p H did not significant,and the relative changes within±5%.While for the electrical conductivity,meteorological conditions caused a decrease of 6.6%to 24.0%during the 2013–2020.After the decoupling meteorological effect,the p H is still below 5.6.In summary,the machine learning algorithm can analyze the long-term or short-term variations of the atmospheric components of different phase in the atmosphere,and effectively supplemented the method of atmospheric chemistry research,with broad application prospects. |
PDF Full Text Request