Dynamic Source Apportionment Of Ambient Ozone Using Multi-Modeling Approaches Over The Pearl River Delta Region,China

Since the implementation of China’s 13th Five-Year Plan（from 2015 to 2020）,the fine particle matter(PM_2.5)levels have gradually decreased,while ozone（O₃）pollution has emerged as a serious issue that significantly affected the air quality in China,especially in typical city clusters with active photochemical reactions such as the Pearl River Delta（PRD）region.Efficient and accurate quantification of the nonlinear response of O₃to its precursors and the major emission source contributors to O₃is crucial for designing sound O₃control policies.Sensitivity analysis and tracer approaches based on traditional air quality models have been widely used for source apportionment（SA）of O₃;however,due to the high nonlinearity chemistry during the O₃formation and the diversity of emission sources that contribute to O₃,the above SA methods usually suffered from the key bottlenecks of insufficient nonlinear quantification and low efficiency for dynamic multi-source contribution analysis.Consequently,this study established a new SA technology to effectively identify the O₃nonlinearity chemistry and dynamic multiple source contributions,and also developed a flexible tool to comprehensively evaluate the similarities and discrepancies among different O₃SA techniques and the applicability of each method in policy-making in the case study of the PRD.The main results and conclusions are as follows:（1）The novel machine learning-based self-adaptive response surface modeling with the differential method（SARSM-DM）was developed for dynamic source contribution analysis of O₃,and a new DM theory was proposed to allocate the source contributions to O₃directly based on the order ratio of cross-terms in the fitting polynomial function of SARSM.The SARSM-DM system can effectively overcome the bottlenecks of traditional SA methods in nonlinear quantification and dynamic source apportionment.（2）The SARSM-DM was applied for O₃dynamic source contribution analysis in two representative months of PRD,July and October 2015.Model evaluation results suggested that compared with the previous pf-RSM,SARSM can effectively reduce the O₃prediction errors by 73%to 89%under some extreme emission control scenarios（i.e,100%control of NO_xand VOC emissions,respectively）,and the DM can also well address the issues of overestimated or underestimated accumulative source contributions in traditional BFM.SA results demonstrated that NO_xemissions from upwind cities（i.e.,Zhongshan and Jiangmen）and other areas（i.e,Zhaoqing and surrounding areas of the PRD）were the major contributors in July,contributing to 14.21%～30.85%of the O₃formations in different receptor regions of the PRD under the 100%control of all emission sources;while in October,VOC emissions from upwind cities（i.e.,Guangzhou and Dongguan&Shenzhen）dominated the O₃formation in most cities under moderate emission control（e.g.,25%or 50%control of all emission sources）,but when furtherly strengthen the emission reductions,NO_xemissions from upwind and other areas were still the main contributors,contributing to 14.58%～30.24%of the O₃in different receptor regions at the 100%reduction of all emission sources.In addition,on-road mobile,non-road mobile and solvent utilization sources were three major sectoral contributors to O₃in the PRD.（3）A new Flexible Air Quality Scenario Tool-Community Edition（FAST-CE）was developed by integrating the SARSM-DM,Brute-Force Method（BFM）,SARSM-BFM,High-order Decouple Direct Method（HDDM）,and Ozone Source Apportionment Technology（OSAT）.The FAST-CE tool included three major modules of source apportionment input,real-time air quality response and source apportionment analysis to quickly display the analysis results of the selected SA method through various visualization ways.Based on the FAST-CE,the policymakers can efficiently compare the similarities and discrepancies among the results of different SA methods and evaluate the performance and applicability of each method.（4）The FAST-CE was applied for the comprehensive analysis of O₃source apportionment in a typical O₃-polluted month of PRD,September 2019.Evaluations indicated that the sensitivity methods（i.e.BFM,SARSM-BFM and HDDM）were more appropriate for assessing the impact of single-source emissions,while the SARSM-DM and OSAT were more suitable for apportioning the various source contributions to O₃.However,SARSM-DM had an obvious advantage over the OSAT for capturing the nonlinear chemistry during the O₃formation,especially the negative contribution of NO_xemissions to O₃.Combining the results of different SA methods,it was found that under the moderate control scenarios（e.g.,25%or50%control of all emission sources）,VOC emissions from local and upwind cities（i.e.,Guangzhou and Dongguan&Shenzhen）dominated the O₃formations in most cities,but when at strict emission reductions（e.g.,100%reduction of all emission sources）,NO_xemissions from upwind cities were still the major sources with relative high impacts or contributions.Among different emission sectors,on-road mobile source was the major contributor,with an average contribution of 31.5%（estimated by the SARSM-DM）and 29.2%（estimated by the OSAT）to O₃in nine cities of the PRD when at the 100%control of all emission sources.（5）Recommendations of O₃control strategies were provided for policy-makers in the PRD,that is to give priority to the in-depth control of on-road mobile source in center cities（i.e.,Guangzhou and Dongguan&Shenzhen）,and enhance the emission control of industrial source in center cities and non-road mobile source in peripheral cities（e.g.,Zhaoqing）and surrounding areas of the PRD,while also focus on the collaborative control of on-road mobile NO_xand VOC emissions from center cities in heavy O₃-polluted months（e.g.September）.