环境科学  2023, Vol. 44 Issue (2): 670-679   PDF    
排放和气象对疫情前后武汉不同类型点位大气污染物的影响
熊江荷1, 孔少飞1,2, 郑煌1, 肖婉1, 刘傲1, 朱明铭1     
1. 中国地质大学(武汉)环境学院, 武汉 430078;
2. 湖北省大气复合污染研究中心, 武汉 430078
摘要: 采用随机森林算法剥离了排放和气象对6种大气污染物(SO2、NO2、CO、PM10、PM2.5和O3)浓度的贡献, 识别了疫情前后武汉市中心城区、郊区、工业区、三环线交通点和城市背景点这5种类型点位的大气污染物浓度变化.结果表明, 与管控前相比, 管控期间PM2.5/CO、PM10/CO和NO2/CO分别减小了10.8~21.7、9.34~24.7和14.4~22.1倍, 表明排放对PM2.5、PM10和NO2贡献减小; O3/CO增加了50.1~61.5倍, 表明二次生成增加明显.解除管控后排放对各类污染物的贡献均增加.管控期间, 受一些不可间断工序的运行影响, 工业区PM2.5降幅最小(20.5%).与管控期间相比, 解除管控后居民生活、交通出行和工业生产等基本恢复, 使5种类型站点PM2.5排放贡献量的降低值减小.管控期间O3排放贡献量的升高与NO和颗粒物浓度降低有关, O3升高部分抵消了NO2和PM2.5浓度降低带来的空气质量改善效果.解除管控后郊区和城市背景点气象贡献的ρ(O3)分别升高了16.2 μg·m-3和16.1 μg·m-3, 与气温升高和相对湿度降低有关.三环线交通点和中心城区交通、工业减排导致的PM2.5浓度降低及O3浓度增加, 可为当前开展分区域PM2.5和O3协同精准管控提供参考.
关键词: 新冠疫情      随机森林(RF)      大气污染物      时空变化      功能区     
Impacts of Emission and Meteorological Conditions on Air Pollutants at Various Sites Around the COVID-19 Lockdown in Wuhan
XIONG Jiang-he1 , KONG Shao-fei1,2 , ZHENG Huang1 , XIAO Wan1 , LIU Ao1 , ZHU Ming-ming1     
1. School of Environmental Studies, China University of Geosciences, Wuhan 430078, China;
2. Research Centre for Complex Air Pollution of Hubei Province, Wuhan 430078, China
Abstract: The random forest algorithm was used to separate the mass concentrations of six air pollutants (SO2, NO2, CO, PM10, PM2.5, and O3) contributed by emissions and meteorological conditions. Their variations for five types of sites including Wuhan's central urban, suburb, industrial, the third ring road traffic, and urban background sites were investigated. The results showed that the values of PM2.5/CO, PM10/CO, and NO2/CO during the lockdown period decreased by 10.8-21.7, 9.34-24.7, and 14.4-22.1 times compared with the period before the lockdown, indicating that the contributions of emissions to PM2.5, PM10, and NO2 were reduced. O3/CO increased by 50.1-61.5 times, implying that the secondary formation increased obviously. The contributions of emissions to various types of pollutants all increased after the lockdown. During the lockdown period, affected by the operation of some uninterrupted industrial processes, PM2.5 concentrations in industrial areas dropped the least (20.5%). Compared with the lockdown period, residential activities, transportation, and industrial production were basically restored after the lockdown, resulting in the alleviation of the reduction in PM2.5 emission-related concentrations. The increase in emission-related O3 concentrations could be associated with the decreased NO and PM2.5 concentrations during the lockdown period. The elevated O3 partially offset the improved air quality brought by the reduced NO2and PM2.5 concentrations. After the lockdown, ρ(O3) related with meteorology at the suburban and urban background sites increased by 16.2 μg·m-3 and 16.1 μg·m-3, respectively, which could be attributed to the increased ambient temperature and decreased relative humidity. The decrease in PM2.5 and increase in O3 concentrations caused by reduced traffic and industrial emissions at the third ring road traffic and central urban regions can provide reference for the current coordinated and precise control of PM2.5 and O3 in subregions.
Key words: COVID-19      random forest(RF)      air pollutant      spatial-temporal variation      functional region     

为控制新冠肺炎疫情(COVID-19)的传播, 2020年1月23日, 武汉市启动了公共卫生事件一级应急响应, 采取了管控措施.管控期间, 公共交通运输停止, 公共场所关闭, 居民居家隔离[1].以上管控措施使移动源、工业源、扬尘源和生活源等[1]排放的大气污染物减少, 环境空气质量显著改善[2].管控措施可看作一次对武汉市一次污染源减排的极限测试[1, 2].定量评估疫情期间(管控期间和解除管控后)污染源减排对武汉市大气污染物浓度变化的影响, 可为后续空气质量改善和科学精准管控提供支撑.

已有学者研究了管控期间大气污染物浓度的变化.周亚端等[3]研究发现, 疫情期间武汉市电力、水泥和陶瓷行业对PM、SO2和NOx减排总量的贡献超过50.0%.减排使PM2.5浓度降低25.4%, O3浓度升高52.5%.陈楠等[4]研究发现由于管控期间禁止机动车通行, 武汉市NO2和PM2.5浓度与2019年同期相比分别下降了53.2%和35.6%.Zheng等[2]研究发现, 管控期间武汉市ρ(PM2.5)和2019年同期(72.9 μg·m-3)相比下降了27.0 μg·m-3, 其中减排贡献了92.0%.Yao等[5]探究了管控期间武汉市PM2.5浓度降低和能见度提高的协同效益, 发现管控期间ρ(PM2.5)较管控前1个月下降了39.0 μg·m-3.Sulaymon等[6]研究发现, 管控期间武汉市NO2、PM2.5、PM10和CO浓度和管控前相比分别降低了50.6%、41.2%、33.1%和16.6%, 而O3浓度增加了149%.前人研究普遍关注武汉市整体的污染物浓度变化, 并未对不同类型站点和不同区域污染物浓度的变化趋势进行讨论, 不利于后续分区域精准管控措施的制定.由于不同类型污染源空间分布的差异, 以及疫情管控措施对不同排放源的影响程度不同[2], 不同类型站点的污染物浓度变化在疫情前后的差异值得研究.

排放和气象是影响污染物浓度变化的关键因素, 但气象变化可能掩盖排放对污染物浓度变化的影响[7~10].Wang等[7]研究发现, 疫情期间交通运输和工业排放等人为排放减少促进了PM2.5浓度的降低, 但高相对湿度不利于PM2.5浓度改善.Zhao等[8]研究发现, 减排导致管控后大多数城市污染物浓度降低, 但气象条件对空气质量的改善也有明显贡献, 为20.0%~60.0%.Fu等[9]研究发现, 疫情期间气象因素(如低温、较高相对湿度、较高风速、能见度)和人为排放减少分别驱动广西ρ(O3)下降了7.84 μg·m-3和4.11 μg·m-3.Matthias等[10]研究了欧洲中部地区COVID-19管控期间减排和气象条件对改善空气质量的作用, 发现气象条件对O3和PM2.5浓度的影响大于减排.因而在评估排放对空气质量的影响时, 需要剥离气象条件的影响(即气象归一化), 以准确评估减排对大气污染物浓度变化的影响.

剥离排放和气象对空气污染物影响的方法主要有数据挖掘、数值模拟和数学统计分析这3种.数据挖掘可解释大气污染物排放和气象的影响[11, 12], 但其对极端事件预测能力较低[11, 13], 受到历史数据库信息的准确性和完善性制约[11], 不适用于疫情期间一次污染源极限减排情境下人为排放和气象影响的评估.数值模拟依赖输入排放清单的准确性以及物理化学机制的完善[14, 15], 且计算时间较长, 耗费资源较多, 更受到排放清单时效性、清单的时间和空间分辨率等多种因素制约[11, 14], 不适用于精细尺度大气污染物浓度实时变化的模拟和评估.数学统计分析是一种剥离排放和气象影响的常用方法[16], 包括KZ(Kolmogorov-Zurbenko)滤波[17]和深度神经网络[18, 19]等.但这些统计方法不能分析输入变量的影响[20], 不能用于评估管控期间人为排放和气象条件对污染物浓度变化的贡献.机器学习提供了一种可靠的替代方法, 可用于量化人为排放和气象条件引起的空气质量变化[21~24].随机森林(random forest, RF)算法作为机器学习方法的一种, 能较好解决变量间的多重共线性及非线性问题, 不受预测因子和污染物之间复杂关系的限制[25, 26], 具有更好的预测性能[21~23].在空气质量相关研究中, RF可用于构建回归模型以剥离气象参数, 排除其对污染物的影响, 进而分析排放对大气污染物浓度变化趋势的影响[21~23], 适用于多站点污染物浓度变化的快速评估.

本研究以2019年12月23日至2020年4月30日武汉市10个国控空气质量监测点(国控点)和11个市控空气质量监测点(市控点)的6种大气污染物(SO2、NO2、CO、PM10、PM2.5和O3)的小时浓度数据和气象数据为基础, 采用随机森林方法剥离人为排放和气象条件对污染物浓度变化的影响, 分析疫情管控措施对武汉市5种不同类型点位(中心城区、郊区、工业区、三环线交通点和城市背景点)大气污染物浓度的影响, 以期为分区域分点位的大气污染物精准管控提供科学依据.

1 材料与方法 1.1 研究区域和时间段

武汉市5种不同类型点位分散在不同辖区, 点位分布信息如图 1所示.参考点位所处区域, 将汉阳月湖、汉口江滩、东湖梨园、汉口花桥、武昌紫阳、蔡甸区汉阳大街站、洪山区鲁磨路站、江汉区大兴路站和江汉区后襄河南路站划分为中心城区; 沉湖七壕和汉南区纱帽正街站划分为郊区; 沌口新区、青山钢花和化工区化工大道站划分为工业区; 东湖高新、吴家山、东西湖区环湖路和硚口区解放大道站划分为三环线交通点; 黄陂区黄陂大道站、江夏区纸坊大街站和新洲区红旗街站划分为城市背景点.

图 1 武汉市环境空气质量监测点分布示意 Fig. 1 Distribution of ambient air quality monitoring sites in Wuhan

武汉市于2020年1月23日采取管控措施, 并于2020年4月8日解除管控.本研究将2019年12月23日至2020年1月22日定义为“管控前”, 2020年1月23日至4月7日定义为“管控期间”, 2020年4月8~30日定义为“解除管控后”.

1.2 数据来源

逐小时的6种大气污染物(SO2、NO2、CO、PM10、PM2.5和O3)浓度监测数据下载自中国环境监测总站全国城市空气质量实时发布平台(http://www.cnemc.cn/).逐小时气象要素数据(风速、相对湿度、气压和气温)下载自国家气候数据中心(NCDC)ISD(integrated surface dataset)数据集.大气污染物浓度监测数据和气象数据的缺失值用上一小时污染物浓度和气象要素数值填充.填充完毕后, 大气污染物浓度监测数据有效率为99.1%, 气象数据有效率为93.8%.

1.3 研究方法

为分析排放和气象对疫情期间大气污染物浓度变化的影响, 本研究使用R语言rmweather包, 分别为武汉市每个监测站点的每种大气污染物建立了RF模型.RF模型的输入变量包括时间变量[日期、Unix时间、儒略日(在儒略周期内以连续日数计算时间的计时法)、一周中天数和小时数]和气象数据(风速、相对湿度、气压和气温).除一周中天数为分类变量外, 其余输入变量均为数字变量.RF模型参数如下:森林中树的数目(n_tree)为300, 单个模型预测次数(n_sample)为300, 每个节点可能分裂的变量数(mtry)为3, 终端节点的最小大小(min_node_size)为5.

随机选取70%的输入数据集作为训练数据集, 用于构建RF模型.而测试数据集(输入数据集的剩余30%), 用于测试模型的性能.RF模型建立后, 使用气象归一化方法, 随机选取100次气象参数, 预测某一实测时间点(例如:2019年12月23日08:00)的大气污染物浓度[2, 22].在每个预测中, 对输入变量(不包括时间变量)进行重新采样以代替解释变量, 并随机分配给响应变量预测.然后计算100个预测值的算术平均值, 得到污染物气象归一化浓度(即排放贡献量)[2, 27].计算污染物原始浓度与气象归一化浓度的差值, 得到气象贡献量[1, 2, 28].

对测试数据集进行可视化分析, 发现RF模型在21个监测站点的拟合精度较高, 相关系数(R2)为0.68~0.76.利用模型遍历并预测所有站点的每种污染物浓度后, 将数据处理为日时间尺度, 进行分析和作图.

2 结果与讨论 2.1 不同类型点位疫情前后污染物浓度变化

图 2所示, 管控期间大部分站点除O3外其余5种污染物浓度出现了不同程度的降低(0.21 μg·m-3~0.23 mg·m-3).解除管控后, 由于居民生活、工业生产和交通逐渐恢复正常, 除PM2.5外其余5种污染物浓度出现了不同程度的升高(2.09 μg·m-3~1.78 mg·m-3).图 2(a)所示, 管控期间5种类型站点ρ(SO2)变化值为-0.21~0.57 μg·m-3, 这与Shen等[29]和Wang等[30]的研究结果一致.在5种类型站点中, 郊区ρ(SO2)降低值最大, 为0.21 μg·m-3, 工业区ρ(SO2)升高值最大, 为0.57 μg·m-3.解除管控后, 5种类型站点ρ(SO2)升高了0.16~2.09 μg·m-3, 这与Sulaymon等[6]和Hu等[31]的研究结果一致.其中, 城市背景点ρ(SO2)升高值最小, 为0.16 μg·m-3, 三环线交通点ρ(SO2)升高值最大, 为2.09 μg·m-3.污染物浓度与CO浓度的比值可反映人为排放的影响[32, 33].管控期间5种类型站点SO2/CO增加了1.20~3.05, 解除管控后SO2/CO增加了2.70~5.56, 表明疫情期间人为排放对SO2贡献增加.

图 2 6种大气污染物管控前、管控期间和解除管控后浓度平均值及变化 Fig. 2 Average mass concentrations of six air pollutants and their variations before, during, and after the lockdown

图 2(b)所示, 管控期间5种类型站点ρ(CO)降低了0.14~0.22 mg·m-3, 这与Barua等[34]和Zhou等[35]的研究结果一致.在5种类型站点中, 郊区ρ(CO)降低值最小, 为0.14 mg·m-3, 中心城区ρ(CO)降低值最大, 为0.22 mg·m-3.解除管控后, 5种类型站点ρ(CO)变化值为-0.23~1.78 mg·m-3, 大部分站点ρ(CO)升高, 这与Hu等[31]和Liu等[36]的研究结果一致.其中, 三环线交通点ρ(CO)降低值最大, 为0.23 mg·m-3, 中心城区ρ(CO)升高值最大, 为1.78 mg·m-3.管控期间CO浓度降低, 人为排放贡献减少.而解除管控后中心城区CO浓度增加, 人为排放贡献增加.郊区、工业区、三环线交通点和城市背景点CO浓度降低, 人为排放贡献减少.

图 2(c)所示, 管控期间5种类型站点ρ(PM10)降低了22.8~30.1 μg·m-3, 这与Sulaymon等[6]和Ju等[37]的研究结果一致.在5种类型站点中, 城市背景点ρ(PM10)降低值最小, 为22.8 μg·m-3, 中心城区ρ(PM10)降低值最大, 为30.1 μg·m-3.解除管控后, 5种类型站点ρ(PM10)升高了3.80~13.2 μg·m-3, 这与Sulaymon等[6]的研究结果一致.其中, 城市背景点ρ(PM10)升高值最小, 为3.80 μg·m-3, 工业区ρ(PM10)升高值最大, 为13.2 μg·m-3.管控期间5种类型站点PM10/CO减小了9.34~24.7, 表明人为排放对PM10贡献降低.而解除管控后PM10/CO增加了16.2~37.0, 表明工业生产、居民生活和交通等人为活动逐渐恢复正常, 人为排放对PM10贡献增加.

图 2(d)所示, 管控期间5种类型站点ρ(PM2.5)降低了15.1~28.5 μg·m-3, 该结果与Sulaymon等[6]、Liu等[28]和Ju等[37]的研究结果一致.在5种类型站点中, 郊区ρ(PM2.5)降低值最小, 为15.1 μg·m-3, 三环线交通点ρ(PM2.5)降低值最大, 为28.5 μg·m-3.解除管控后, 5种类型站点ρ(PM2.5)降低了3.80~7.30 μg·m-3, 这与Sulaymon等[6]和Kotsiou等[38]的研究结果一致.其中, 工业区ρ(PM2.5)降低值最小, 为3.80 μg·m-3, 三环线交通点ρ(PM2.5)降低值最大, 为7.30 μg·m-3.管控期间5种类型站点PM2.5/CO减小了10.8~21.7, 表明人为排放对PM2.5贡献降低.而解除管控后PM2.5/CO变化范围为-0.49~5.10, 体现了解除管控后除中心城区外其余4种类型点位人为排放对PM2.5贡献增加.

图 2(e)所示, 管控期间5种类型站点ρ(NO2)降低了16.5~26.7 μg·m-3, 该结果与Collivignarelli等[39]和Chu等[40]的研究结果一致.在5种类型站点中, 郊区ρ(NO2)降低值最小, 为16.5 μg·m-3, 三环线交通点ρ(NO2)降低值最大, 为26.7 μg·m-3, 且三环线交通点ρ(NO2)的降低值远大于城市背景点(17.1 μg·m-3), 表明管控期间NO2浓度的降低主要由疫情交通管制所驱动[21].管控期间5种类型站点ρ(NO2)降低值(16.5~26.7 μg·m-3)远大于ρ(SO2)的变化值(-0.21~0.57 μg·m-3).该结果表明, 管控措施对车辆排放的影响大于对发电厂和重工业(例如钢铁和石化工业)排放的影响[28].解除管控后, 5种类型站点ρ(NO2)升高了8.60~22.6 μg·m-3, 这与Keller等[41]和Wang等[42]的研究结果一致.其中城市背景点ρ(NO2)升高值最小, 为8.60 μg·m-3, 工业区ρ(NO2)升高值最大, 为22.6 μg·m-3.管控期间5种类型站点NO2/CO减小了14.4~22.1, 表明人为排放对NO2贡献降低.解除管控后NO2/CO升高了16.4~40.0, 体现了解除管控后车辆出行恢复正常, 移动源排放的NO2增加明显.

图 2(f)所示, 管控期间5种类型站点ρ(O3)升高了35.2~42.3 μg·m-3, 该结果与Ding等[43]和Zhu等[44]的研究结果一致.在5种类型站点中, 郊区ρ(O3)升高值最小, 为35.2 μg·m-3, 三环线交通点ρ(O3)升高值最大, 为42.3 μg·m-3.解除管控后, 5种类型站点ρ(O3)升高了13.6~21.5 μg·m-3, 这与Sulaymon等[6]的研究结果一致(解除管控后武汉市O3浓度增加了25.3%).其中, 三环线交通点ρ(O3)升高值最小, 为13.6 μg·m-3, 城市背景点ρ(O3)升高值最大, 为21.5 μg·m-3.管控期间5种类型站点O3/CO增加了50.1~61.5, 解除管控后O3/CO增加了32.1~53.4, 表明疫情期间人为排放对O3的贡献显著增加, 控制前体物排放是控制O3污染的关键.

2.2 排放和气象对疫情前后PM2.5浓度变化的影响

管控期间三环线交通点和中心城区PM2.5排放贡献量分别降低21.3 μg·m-3和18.4 μg·m-3, 降幅分别为37.1%和34.2%, 与其余类型站点相比下降显著[图 3(a)].管控期间工业区PM2.5排放贡献量降幅最小(20.5%), 与一些不可间断工序(如大型重工业、燃煤电厂和家庭烹饪等[1, 28])在管控期间运行有关.

图 3 5种类型站点管控前、管控期间和解除管控后排放和气象贡献的ρ(PM2.5)及变幅 Fig. 3 Variations in PM2.5 concentrations related to emission and meteorology before, during, and after the lockdown at the five types of sites

PM2.5/CO可用于剥离气象条件的影响, 反映排放和化学转化的贡献[21, 45].管控期间中心城区、郊区、三环线交通点和城市背景点PM2.5/CO分别减小了9.90、8.10、14.2和6.40, 表明其排放或二次颗粒物生成减少.而工业区PM2.5/CO增加了0.91, 表明其排放增加.解除管控后, 中心城区和工业区PM2.5/CO分别减小了15.0和2.51, 郊区、三环线交通点和城市背景点PM2.5/CO分别增加了0.90、4.90和2.50, 表明中心城区和工业区PM2.5排放减少, 郊区、三环线交通点和城市背景点排放和二次反应贡献增加.此外, 疫情期间工业区PM2.5/CO平均值为3.93, 远小于其余站点, 这与工业区一次污染物排放高的事实相符[46, 47].

图 4所示, 管控期间5种类型站点粗颗粒物排放贡献的ρ(PM10-PM2.5)[48, 49]降低了3.31~11.2 μg·m-3.与管控期间禁止车辆出行和建筑施工[50, 51], 使烟尘、粉尘和扬尘等[52, 53]一次粗颗粒物的排放减少有关. 同时这些源的变化, 也促进了PM2.5的源贡献降低[54], 导致PM2.5浓度显著降低.

图 4 5种类型站点管控前、管控期间和解除管控后粗颗粒物排放贡献的ρ(PM10-PM2.5) Fig. 4 Coarse particulate matters (PM10-PM2.5) concentrations related to emissions before, during, and after the lockdown at the five types of sites

SO2和NO2是PM2.5主要的气态前体物[21, 55, 56].管控期间车辆出行大幅减少和建筑工地停工, 使SO2和NO2排放贡献量分别降低了8.80%~22.0%和29.2%~56.6%[图 5(a)5(b)], 导致其对PM2.5的贡献降低[21, 57].管控期间一次粗颗粒物源贡献的降低和气态前体物排放减少导致了PM2.5浓度降低, 并有效改善了武汉市的空气质量.疫情期间武汉市5种类型站点PM2.5排放贡献量降低, 平均降幅为21.8%, 但郊区、三环线交通点和城市背景点气象归一化后的PM2.5/CO增加, 表明二次污染的增长抵消了部分一次污染物减排带来的浓度降低[21].

向下箭头表示减小, 向上箭头表示增加, 下同 图 5 5种类型站点管控前、管控期间和解除管控后排放贡献的ρ(SO2)和ρ(NO2)及变幅 Fig. 5 Variations in SO2 and NO2 concentrations related to emissions before, during, and after the lockdown at the five types of sites

图 3(a)所示, 解除管控后5种类型站点PM2.5排放贡献量的降低值(3.80~7.30 μg·m-3)低于管控期间(15.1~28.5 μg·m-3).Zheng[58]提出工业用电量由于与生产活动直接相关, 适合作为衡量解除管控后复工的指标, 其研究表明, 截止2020年4月21日, 解除管控后中国复工率为73.1%. Liu等[59]和Wang等[42]指出, 解除管控后企业生产逐渐恢复, 人员和货物大规模流动, 能耗上升, 交通出行逐渐恢复.上述研究证实解除管控后PM2.5降低值的减小与居民生活和交通逐渐恢复正常、工业生产基本恢复秩序有关[6, 31, 40].

总体而言, 管控期间人为活动减少使一次污染物排放量减少, 但部分站点二次污染加剧.解除管控后, 人为活动逐渐增加, PM2.5排放贡献量逐渐恢复.分析管控期间和解除管控后气象参数的变化, 发现海平面气压和风速的变化值小, 故本研究只考虑相对湿度和气温对污染物气象贡献量的影响.

管控期间5种类型站点PM2.5气象贡献量分别下降了84.0%、69.7%、90.1%、94.6%和84.5%[图 3(b)].管控期间相对湿度较高(平均值为68.7%), PM2.5沉降到地面的概率增大[60].与管控前相比, 气温升高(平均升高6.46℃), 有利于对流扩散稀释PM2.5[61, 62].高相对湿度和气温升高可以解释管控期间PM2.5气象贡献量的降低.解除管控后中心城区、郊区和工业区PM2.5气象贡献量升高了36.4%、7.63%和37.7%, 与气温升高(平均升高7.27℃)和相对湿度降低(平均降低9.26%)有关.解除管控后三环线交通点和城市背景点PM2.5气象贡献量分别下降了65.6%和80.3%, 与气温升高(平均升高7.30℃)和相对湿度较高(平均值为62.7%)有关.

2.3 排放和气象对疫情前后O3浓度变化的影响

管控期间5种类型站点O3排放贡献量分别升高了25.6、21.0、18.4、29.0和23.1 μg·m-3, 升幅为76.1%、49.7%、38.2%、76.8%和53.1%, 其中中心城区和三环线交通点O3排放贡献量升幅显著高于其余站点[图 6(a)].

图 6 5种类型站点管控前、管控期间和解除管控后排放和气象贡献的ρ(O3)及变幅 Fig. 6 Variation in O3 concentrations related to emissions and meteorology before, during, and after the lockdown at the five types of sites

为分析O3浓度大幅升高的原因, 计算发现5种类型站点O3排放贡献量与PM2.5排放贡献量(R2:0.62~0.75, P < 0.001)及NO排放贡献量存在显著负相关关系(R2:0.76~0.88, P < 0.001), 与Zhang等[63]和Li等[64]的研究结果一致.

因此, 管控期间5种类型站点O3排放贡献量的升高一方面与NO浓度降低、O3滴定效应减弱有关[6, 21, 24], 另一方面与颗粒物浓度降低有关.管控期间颗粒物浓度降低, 会导致地表接收更多辐射, 促进NO2光化学反应生成O3[65].此外, PM2.5排放贡献量降低, 使空气中可被其清除的过氧化氢自由基(HO2·)和氮氧化物自由基(NOx·☰NO+NO2)增加, 导致氢氧自由基(HOx· ☰·OH+HO2· +RO2·)和氮氧化物自由基浓度升高, 促进NOx·和HOx·催化挥发性有机化合物(VOC)光化学氧化产生O3, 导致O3积累[66].管控期间中心城区和三环线交通点PM2.5(降幅为34.2%和37.1%)和NO2排放贡献量(降幅为50.6%和56.6%)在5种类型站点中降低显著, 故其O3排放贡献量升幅(76.1%和76.8%)明显高于其余站点.

有研究表明[24, 28, 45, 67], NOx·排放减少会导致O3和其它大气氧化剂(HO2·、NO3·和·OH)的形成增强.在此情况下, 大气氧化能力显著增加, 二次气溶胶生成增强[24, 45], 并强化大气氧化产物(H2SO4、HNO3、N2O5和含氧有机化合物)的非线性响应, 增强二次颗粒物组分的生成[45]. NOx·排放量的大幅降低会增加O3污染[68], 而O3浓度的升高可能部分抵消NO2和PM2.5浓度降低带来的空气质量改善效果[24].未来需要采取更全面的污染防控措施, 进一步控制一次污染物和PM2.5、O3等气态前体物的排放, 尤其是在中心城区和三环线交通点位.

图 6(b)所示, 与管控期间相比, 解除管控后郊区和城市背景点O3气象贡献量升高了16.2 μg·m-3和16.1 μg·m-3, 升幅为1 421%和1 451%, 与其余类型站点相比升高明显.有研究表明, O3的生成条件为晴天少云、紫外辐射较强、气温较高、相对湿度较低和风速较小的天气[69~72].在本研究中, 由于气压和风速变化较小, 只考虑相对湿度和气温的影响.分析可知, O3气象贡献量与相对湿度呈负相关关系(R2:0.76~0.85, P < 0.001), 与气温呈正相关关系(R2:0.61~0.67, P < 0.001).故解除管控后郊区和城市背景点气温升高(升幅为57.1%和63.7%)和相对湿度降低(降幅为12.2%和13.8%), 导致了O3气象贡献量的升高.

2.4 不确定性讨论

随机森林方法与数据挖掘、数值模拟和数学统计分析等方法相比, 具有不受预测因子和污染物之间复杂关系限制[25, 26]、计算速度更快和预测性能更好[21~23]等优良特性, 但其仍具有不确定性与局限性.Sun等[73]和Grimm等[74]的研究表明, 在RF模型中, 当森林中树的数目(n_tree)为1 000时, 模型模拟结果将趋于稳健.在本研究中, 受制于计算机性能和计算时间, n_tree取值为300, 不足以使模型产生稳健的模拟结果.故本研究中应用RF模型进行模拟后, 相关系数(R2)为0.68~0.76, 表明模拟结果存在不确定性.此外, 在RF模型模拟过程中, 未考虑大气中具体存在的物理和化学过程, 且无法对森林中的每棵树检查预测因子和响应变量之间的关系[74], 导致RF模型可解释性有限.在未来评估中, 应结合研究数据规模、计算机性能和计算时间, 综合选择n_tree的取值.同时RF模型中节点可能分裂变量数(mtry)取值越高, 模型中每棵树之间的强度越强, 模型性能越好[75].随着mtry取值的增加, 森林中每棵树之间的相关性增强, 反而削弱模型性能[75].在未来评估中, 还应综合考虑RF模型中mtry的取值, 使模型具有最优性能.

3 结论

(1) 管控期间武汉市郊区ρ(SO2)、中心城区ρ(CO)、ρ(PM10)、三环线交通点ρ(PM2.5)和ρ(NO2)降低值最大, 分别为0.21 μg·m-3、0.22 mg·m-3、30.1 μg·m-3、28.5 μg·m-3和26.7 μg·m-3.三环线交通点ρ(O3)升高值最大(42.3 μg·m-3).管控期间人为排放对SO2和O3贡献增加, 对PM2.5、PM10和NO2贡献减小.

(2) 解除管控后武汉市三环线交通点ρ(SO2)、中心城区ρ(CO)、工业区ρ(PM10)、ρ(NO2)以及城市背景点ρ(O3)升高值最大, 分别为2.09 μg·m-3、1.78 mg·m-3、13.2 μg·m-3、22.6 μg·m-3和21.5 μg·m-3.三环线交通点ρ(PM2.5)降低值最大(7.30 μg·m-3).解除管控后, 人为排放对各类污染物贡献增大.

(3) 管控期间不可间断工序的运行, 使工业区PM2.5排放贡献量降幅最小(20.5%).管控期间PM2.5浓度降低, 但部分站点二次污染加剧.解除管控后PM2.5降低值减小与居民生活、交通及工业生产逐渐恢复正常有关.在高相对湿度和气温升高条件下, 管控期间PM2.5气象贡献量降低.

(4) 管控期间O3排放贡献量升高, 与NO和颗粒物浓度降低有关.解除管控后郊区和城市背景点气温升高和相对湿度降低导致O3气象贡献量升高.O3升高部分抵消了NO2和PM2.5浓度降低带来的空气质量改善效果.

参考文献
[1] 代兴良, 宋国君, 姜晓群, 等. 新冠肺炎疫情对咸阳市空气质量的影响[J]. 中国环境科学, 2021, 41(7): 3106-3114.
Dai X L, Song G J, Jiang X Q, et al. Impacts of the COVID-19 pandemic on air quality in Xianyang[J]. China Environmental Science, 2021, 41(7): 3106-3114. DOI:10.3969/j.issn.1000-6923.2021.07.012
[2] Zheng H, Kong S F, Chen N, et al. Significant changes in the chemical compositions and sources of PM2.5 in Wuhan since the city lockdown as COVID-19[J]. Science of the Total Environment, 2020, 739. DOI:10.1016/j.scitotenv.2020.140000
[3] 周亚端, 朱宽广, 黄凡, 等. 新冠肺炎疫情期间湖北省大气污染物减排效果评估[J]. 环境科学与技术, 2020, 43(3): 228-236.
Zhou Y D, Zhu K G, Huang F, et al. Emission reductions and air quality improvements during the COVID-19 pandemic in Hubei Province[J]. Environmental Science & Technology, 2020, 43(3): 228-236. DOI:10.19672/j.cnki.1003-6504.2020.03.032
[4] 陈楠, 张周祥, 李涛, 等. 武汉地区疫情管控期间空气质量变化及改善措施研究[J]. 气候与环境研究, 2021, 26(2): 217-226.
Chen N, Zhang Z X, Li T, et al. Air quality change and improvement measures during the COVID-19 epidemic in Wuhan[J]. Climatic and Environmental Research, 2021, 26(2): 217-226.
[5] Yao L Q, Kong S F, Zheng H, et al. Co-benefits of reducing PM2.5 and improving visibility by COVID-19 lockdown in Wuhan[J]. npj Climate and Atmospheric Science, 2021, 4(1). DOI:10.1038/s41612-021-00195-6
[6] Sulaymon I D, Zhang Y X, Hopke P K, et al. COVID-19 pandemic in Wuhan: ambient air quality and the relationships between criteria air pollutants and meteorological variables before, during and after lockdown[J]. Atmospheric Research, 2021, 250. DOI:10.1016/j.atmosres.2020.105362
[7] Wang P F, Chen K Y, Zhu S Q, et al. Severe air pollution events not avoided by reduced anthropogenic activities during COVID-19 outbreak[J]. Resources, Conservation and Recycling, 2020, 158. DOI:10.1016/j.resconrec.2020.104814
[8] Zhao Y B, Zhang K, Xu X T, et al. Substantial changes in nitrogen dioxide and ozone after excluding meteorological impacts during the COVID-19 outbreak in mainland China[J]. Environmental Science & Technology Letters, 2020, 7(6): 402-408.
[9] Fu S, Guo M X, Fan L P, et al. Ozone pollution mitigation in Guangxi (south China) driven by meteorology and anthropogenic emissions during the COVID-19 lockdown[J]. Environmental Pollution, 2021, 272. DOI:10.1016/j.envpol.2020.115927
[10] Matthias V, Quante M, Arndt J A, et al. The role of emission reductions and the meteorological situation for air quality improvements during the COVID-19 lockdown period in central Europe[J]. Atmospheric Chemistry and Physics, 2021, 21(18): 13931-13971. DOI:10.5194/acp-21-13931-2021
[11] 熊亚军, 徐敬, 孙兆彬, 等. 基于数据挖掘算法和数值模拟技术的大气污染减排效果评估[J]. 环境科学学报, 2019, 39(1): 116-125.
Xiong Y J, Xu J, Sun Z B, et al. Air pollution reduction effect evaluation based on data mining algorithm and numerical simulation technology[J]. Journal of Environmental Science, 2019, 39(1): 116-125. DOI:10.13671/j.hjkxxb.2018.0271
[12] González-Pardo J, Ceballos-Santos S, Manzanas R, et al. Estimating changes in air pollutant levels due to COVID-19 lockdown measures based on a business-as-usual prediction scenario using data mining models: a case-study for urban traffic sites in Spain[J]. Science of the Total Environment, 2022, 823. DOI:10.1016/j.scitotenv.2022.153786
[13] 周凌峰, 孟耀斌, 逯超, 等. 天气发生器MulGETS和k-NN对区域历史气象场特征重现能力的比较[J]. 中国农业气象, 2019, 40(6): 341-349.
Zhou L F, Meng Y B, Lu C, et al. Reproducibility evaluation of multi-site stochastic weather generators: a comparison between a typical parametric model and a non-parametric model[J]. Chinese Journal of Agrometeorology, 2019, 40(6): 341-349. DOI:10.3969/j.issn.1000-6362.2019.06.001
[14] Li M, Liu H, Geng G N, et al. Anthropogenic emission inventories in China: a review[J]. National Science Review, 2017, 4(6): 834-866. DOI:10.1093/nsr/nwx150
[15] Hallquist M, Munthe J, Hu M, et al. Photochemical smog in China: scientific challenges and implications for air-quality policies[J]. National Science Review, 2016, 3(4): 401-403. DOI:10.1093/nsr/nww080
[16] Liang X, Zou T, Guo B, et al. Assessing Beijing's PM2.5 pollution: severity, weather impact, APEC and winter heating[J]. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 2015, 471(2182). DOI:10.1098/rspa.2015.0257
[17] Rao S T, Zurbenko I G. Detecting and tracking changes in ozone air quality[J]. Air & Waste, 1994, 44(9): 1089-1092.
[18] Wise E K, Comrie A C. Meteorologically adjusted urban air quality trends in the Southwestern United States[J]. Atmospheric Environment, 2005, 39(16): 2969-2980. DOI:10.1016/j.atmosenv.2005.01.024
[19] Hogrefe C, Vempaty S, Rao S T, et al. A comparison of four techniques for separating different time scales in atmospheric variables[J]. Atmospheric Environment, 2003, 37(3): 313-325. DOI:10.1016/S1352-2310(02)00897-X
[20] Henneman L R F, Holmes H A, Mulholland J A, et al. Meteorological detrending of primary and secondary pollutant concentrations: method application and evaluation using long-term (2000-2012) data in Atlanta[J]. Atmospheric Environment, 2015, 119: 201-210. DOI:10.1016/j.atmosenv.2015.08.007
[21] Shi Z B, Song C B, Liu B W, et al. Abrupt but smaller than expected changes in surface air quality attributable to COVID-19 lockdowns[J]. Science Advances, 2021, 7(3). DOI:10.1126/sciadv.abd6696
[22] Grange S K, Carslaw D C, Lewis A C, et al. Random forest meteorological normalisation models for Swiss PM10 trend analysis[J]. Atmospheric Chemistry and Physics, 2018, 18(9): 6223-6239. DOI:10.5194/acp-18-6223-2018
[23] Vu T V, Shi Z B, Cheng J, et al. Assessing the impact of clean air action on air quality trends in Beijing using a machine learning technique[J]. Atmospheric Chemistry and Physics, 2019, 19(17): 11303-11314. DOI:10.5194/acp-19-11303-2019
[24] Dai Q L, Hou L L, Liu B W, et al. Spring festival and COVID-19 lockdown: disentangling PM sources in major Chinese cities[J]. Geophysical Research Letters, 2021, 48(11). DOI:10.1029/2021gl093403
[25] Hu X F, Belle J H, Meng X, et al. Estimating PM2.5 concentrations in the conterminous United States using the random forest approach[J]. Environmental Science & Technology, 2017, 51(12): 6936-6944.
[26] Tella A, Balogun A L, Adebisi N, et al. Spatial assessment of PM10 hotspots using random forest, K-nearest neighbour and Naïve Bayes[J]. Atmospheric Pollution Research, 2021, 12(10). DOI:10.1016/j.apr.2021.101202
[27] Li K, Jacob D J, Liao H, et al. Anthropogenic drivers of 2013-2017 trends in summer surface ozone in China[J]. Proceedings of the National Academy of Sciences of the United States of America, 2019, 116(2): 422-427. DOI:10.1073/pnas.1812168116
[28] Liu L, Zhang J, Du R G, et al. Chemistry of atmospheric fine particles during the COVID-19 pandemic in a megacity of Eastern China[J]. Geophysical Research Letters, 2021, 48(2). DOI:10.1029/2020gl091611
[29] Shen N C, Zhao X, Li L J, et al. Spatial and temporal variation characteristics of atmospheric NO2 and SO2 in the Beijing-Tianjin-Hebei region before and after the COVID-19 outbreak[J]. Air Quality, Atmosphere & Health, 2021, 14(8): 1175-1188.
[30] Wang Q, Li S Y. Nonlinear impact of COVID-19 on pollutions-Evidence from Wuhan, New York, Milan, Madrid, Bandra, London, Tokyo and Mexico City[J]. Sustainable Cities and Society, 2021, 65. DOI:10.1016/j.scs.2020.102629
[31] Hu M, Chen Z B, Cui H Y, et al. Air pollution and critical air pollutant assessment during and after COVID-19 lockdowns: evidence from pandemic hotspots in China, the Republic of Korea, Japan, and India[J]. Atmospheric Pollution Research, 2021, 12(2): 316-329. DOI:10.1016/j.apr.2020.11.013
[32] Zhang L, Wang S X, Wang L, et al. Atmospheric mercury concentration and chemical speciation at a rural site in Beijing, China: implications of mercury emission sources[J]. Atmospheric Chemistry and Physics, 2013, 13(20): 10505-10516. DOI:10.5194/acp-13-10505-2013
[33] Hudman R C, Murray L T, Jacob D J, et al. Biogenic versus anthropogenic sources of CO in the United States[J]. Geophysical Research Letters, 2008, 35(4). DOI:10.1029/2007gl032393
[34] Barua S, Nath S D. The impact of COVID-19 on air pollution: evidence from global data[J]. Journal of Cleaner Production, 2021, 298. DOI:10.1016/j.jclepro.2021.126755
[35] Zhou M Q, Jiang J Y, Langerock B, et al. Change of CO concentration due to the COVID-19 lockdown in China observed by surface and satellite observations[J]. Remote Sensing, 2021, 13(6). DOI:10.3390/rs13061129
[36] Liu Q, Harris J T, Chiu L S, et al. Spatiotemporal impacts of COVID-19 on air pollution in California, USA[J]. Science of the Total Environment, 2021, 750. DOI:10.1016/j.scitotenv.2020.141592
[37] Ju M J, Oh J, Choi Y H. Changes in air pollution levels after COVID-19 outbreak in Korea[J]. Science of the Total Environment, 2021, 750. DOI:10.1016/j.scitotenv.2020.141521
[38] Kotsiou O S, Saharidis G K D, Kalantzis G, et al. The impact of the lockdown caused by the COVID-19 pandemic on the fine particulate matter (PM2.5) air pollution: the Greek paradigm[J]. International Journal of Environmental Research and Public Health, 2021, 18(13). DOI:10.3390/ijerph18136748
[39] Collivignarelli M C, De Rose C, Abbà A, et al. Analysis of lockdown for CoViD-19 impact on NO2 in London, Milan and Paris: what lesson can be learnt?[J]. Process Safety and Environmental Protection, 2021, 146: 952-960. DOI:10.1016/j.psep.2020.12.029
[40] Chu B W, Zhang S P, Liu J, et al. Significant concurrent decrease in PM2.5 and NO2 concentrations in China during COVID-19 epidemic[J]. Journal of Environmental Sciences, 2021, 99: 346-353. DOI:10.1016/j.jes.2020.06.031
[41] Keller C A, Evans M J, Knowland K E, et al. Global impact of COVID-19 restrictions on the surface concentrations of nitrogen dioxide and ozone[J]. Atmospheric Chemistry and Physics, 2021, 21(5): 3555-3592. DOI:10.5194/acp-21-3555-2021
[42] Wang Q, Su M. A preliminary assessment of the impact of COVID-19 on environment-A case study of China[J]. Science of the Total Environment, 2020, 728. DOI:10.1016/j.scitotenv.2020.138915
[43] Ding J, Dai Q L, Li Y F, et al. Impact of meteorological condition changes on air quality and particulate chemical composition during the COVID-19 lockdown[J]. Journal of Environmental Sciences, 2021, 109: 45-56. DOI:10.1016/j.jes.2021.02.022
[44] Zhu S Q, Poetzscher J, Shen J Y, et al. Comprehensive insights into O3 changes during the COVID-19 from O3 formation regime and atmospheric oxidation capacity[J]. Geophysical Research Letters, 2021, 48(10). DOI:10.1029/2021GL093668
[45] Huang X, Ding A J, Gao J, et al. Enhanced secondary pollution offset reduction of primary emissions during COVID-19 lockdown in China[J]. National Science Review, 2020, 8(2). DOI:10.1093/nsr/nwaa137
[46] Alyuz U, Alp K. Emission inventory of primary air pollutants in 2010 from industrial processes in Turkey[J]. Science of the Total Environment, 2014, 488-489: 369-381. DOI:10.1016/j.scitotenv.2014.01.123
[47] Lei Y, Zhang Q, Nielsen C, et al. An inventory of primary air pollutants and CO2 emissions from cement production in China, 1990-2020[J]. Atmospheric Environment, 2011, 45(1): 147-154. DOI:10.1016/j.atmosenv.2010.09.034
[48] Janssen N A H, Fischer P, Marra M, et al. Short-term effects of PM2.5, PM10 and PM2.5-10 on daily mortality in the Netherlands[J]. Science of the Total Environment, 2013, 463-464: 20-26. DOI:10.1016/j.scitotenv.2013.05.062
[49] Pateraki S, Asimakopoulos D N, Flocas H A, et al. The role of meteorology on different sized aerosol fractions (PM10, PM2.5, PM2.5-10)[J]. Science of the Total Environment, 2012, 419: 124-135. DOI:10.1016/j.scitotenv.2011.12.064
[50] Cui Y, Ji D S, Maenhaut W, et al. Levels and sources of hourly PM2.5-related elements during the control period of the COVID-19 pandemic at a rural site between Beijing and Tianjin[J]. Science of the Total Environment, 2020, 744. DOI:10.1016/j.scitotenv.2020.140840
[51] Pathak A K, Sharma M, Nagar P K. A framework for PM2.5 constituents-based (including PAHs) emission inventory and source toxicity for priority controls: a case study of Delhi, India[J]. Chemosphere, 2020, 255. DOI:10.1016/j.chemosphere.2020.126971
[52] Wheeler A J, Dobbin N A, Lyrette N, et al. Residential indoor and outdoor coarse particles and associated endotoxin exposures[J]. Atmospheric Environment, 2011, 45(39): 7064-7071. DOI:10.1016/j.atmosenv.2011.09.048
[53] Jain S, Sharma S K, Vijayan N, et al. Seasonal characteristics of aerosols (PM2.5 and PM10) and their source apportionment using PMF: a four year study over Delhi, India[J]. Environmental Pollution, 2020, 262. DOI:10.1016/j.envpol.2020.114337
[54] Huang R J, Zhang Y L, Bozzetti C, et al. High secondary aerosol contribution to particulate pollution during haze events in China[J]. Nature, 2014, 514(7521): 218-222. DOI:10.1038/nature13774
[55] Wang Y H, Gao W K, Wang S, et al. Contrasting trends of PM2.5 and surface-ozone concentrations in China from 2013 to 2017[J]. National Science Review, 2020, 7(8): 1331-1339. DOI:10.1093/nsr/nwaa032
[56] Qin M M, Hu A Q, Mao J J, et al. PM2.5 and O3 relationships affected by the atmospheric oxidizing capacity in the Yangtze River Delta, China[J]. Science of the Total Environment, 2022, 810. DOI:10.1016/j.scitotenv.2021.152268
[57] Huang K, Zhuang G, Lin Y, et al. Typical types and formation mechanisms of haze in an Eastern Asia megacity, Shanghai[J]. Atmospheric Chemistry and Physics, 2012, 12(1): 105-124. DOI:10.5194/acp-12-105-2012
[58] Zheng Y. Air pollution and post-COVID-19 work resumption: evidence from China[J]. Environmental Science and Pollution Research, 2022, 29(12): 17103-17116. DOI:10.1007/s11356-021-16813-y
[59] Liu Y X, Pei T, Song C, et al. How did human dwelling and working intensity change over different stages of COVID-19 in Beijing?[J]. Sustainable Cities and Society, 2021, 74. DOI:10.1016/j.scs.2021.103206
[60] Katata G, Kajino M, Matsuda K, et al. A numerical study of the effects of aerosol hygroscopic properties to dry deposition on a broad-leaved forest[J]. Atmospheric Environment, 2014, 97. DOI:10.1016/j.atmosenv.2013.11.028
[61] Han Y, Qi M, Chen Y L, et al. Influences of ambient air PM2.5 concentration and meteorological condition on the indoor PM2.5 concentrations in a residential apartment in Beijing using a new approach[J]. Environmental Pollution, 2015, 205: 307-314. DOI:10.1016/j.envpol.2015.04.026
[62] Chan A T. Indoor-outdoor relationships of particulate matter and nitrogen oxides under different outdoor meteorological conditions[J]. Atmospheric Environment, 2002, 36(9): 1543-1551. DOI:10.1016/S1352-2310(01)00471-X
[63] Zhang Z Y, Zhang X L, Gong D Y, et al. Evolution of surface O3 and PM2.5 concentrations and their relationships with meteorological conditions over the last decade in Beijing[J]. Atmospheric Environment, 2015, 108: 67-75. DOI:10.1016/j.atmosenv.2015.02.071
[64] Li L, Lu C, Chan P W, et al. Tower observed vertical distribution of PM2.5, O3 and NOx in the Pearl River Delta[J]. Atmospheric Environment, 2020, 220. DOI:10.1016/j.atmosenv.2019.117083
[65] Zhang L, Wang T, Zhang Q, et al. Potential sources of nitrous acid (HONO) and their impacts on ozone: a WRF-Chem study in a polluted subtropical region[J]. Journal of Geophysical Research: Atmospheres, 2016, 121(7): 3645-3662. DOI:10.1002/2015JD024468
[66] Li K, Jacob D J, Liao H, et al. A two-pollutant strategy for improving ozone and particulate air quality in China[J]. Nature Geoscience, 2019, 12(11): 906-910. DOI:10.1038/s41561-019-0464-x
[67] Lv Z F, Wang X T, Deng F Y, et al. Source-receptor relationship revealed by the halted traffic and aggravated haze in Beijing during the COVID-19 lockdown[J]. Environmental Science & Technology, 2020, 54(24): 15660-15670.
[68] Grange S K, Lee J D, Drysdale W S, et al. COVID-19 lockdowns highlight a risk of increasing ozone pollution in European urban areas[J]. Atmospheric Chemistry and Physics, 2021, 21(5): 4169-4185. DOI:10.5194/acp-21-4169-2021
[69] 陈菁, 彭金龙, 徐彦森. 北京市2014~2020年PM2.5和O3时空分布与健康效应评估[J]. 环境科学, 2021, 42(9): 4071-4082.
Chen J, Peng J L, Xu Y S. Spatiotemporal distribution and health impacts of PM2.5 and O3 in Beijing, from 2014 to 2020[J]. Environmental Science, 2021, 42(9): 4071-4082.
[70] Deng T, Wang T J, Wang S Q, et al. Impact of typhoon periphery on high ozone and high aerosol pollution in the Pearl River Delta region[J]. Science of the Total Environment, 2019, 668: 617-630. DOI:10.1016/j.scitotenv.2019.02.450
[71] 张莹, 倪长健, 冯鑫媛, 等. 基于GAMs模型分析成都市气象因子交互作用对O3浓度变化的影响[J]. 环境科学, 2021, 42(11): 5228-5238.
Zhang Y, Ni C J, Feng X Y, et al. Interactive effects of the influencing factors on the changes of O3 concentrations based on GAMs model in Chengdu[J]. Environmental Science, 2021, 42(11): 5228-5238.
[72] 王雨燕, 杨文, 王秀艳, 等. 淄博市城郊臭氧污染特征及影响因素分析[J]. 环境科学, 2022, 43(1): 170-179.
Wang Y Y, Yang W, Wang X Y, et al. Characteristics of ozone pollution and influencing factors in urban and suburban areas in Zibo[J]. Environmental Science, 2022, 43(1): 170-179.
[73] Sun H W, Gui D W, Yan B W, et al. Assessing the potential of random forest method for estimating solar radiation using air pollution index[J]. Energy Conversion and Management, 2016, 119: 121-129. DOI:10.1016/j.enconman.2016.04.051
[74] Grimm R, Behrens T, Märker M, et al. Soil organic carbon concentrations and stocks on Barro Colorado Island-Digital soil mapping using Random Forests analysis[J]. Geoderma, 2008, 146(1-2): 102-113. DOI:10.1016/j.geoderma.2008.05.008
[75] Ließ M, Glaser B, Huwe B. Uncertainty in the spatial prediction of soil texture: comparison of regression tree and Random Forest models[J]. Geoderma, 2012, 170: 70-79. DOI:10.1016/j.geoderma.2011.10.010