作为化学品生产和消费大国，我国现存环境风险企业数量众多，总体呈现区域型、结构型和布局型环境风险特征。近年来，突发大气污染事故事件频频发生、后果触目惊心，严重威胁到人民健康和生态环境安全。城市作为人口高度密集、风险众多、工业区与人群聚集区混杂的典型区域，布局型突发大气污染事故风险突出。因此，进行城市突发大气污染事故风险评价与针对性防控工作刻不容缓。本研究着眼于城市突发大气污染事故风险特征与形成机理，综合考虑城市尤其是城区范围内风险源与风险受体之间的风险叠加和传递效应，构建基于风险链的城市固定源突发大气污染事故贝叶斯网络风险评价方法，进一步选择大气环境风险企业分布密集的长春市中心城区开展案例研究，研究内容与主要结论如下：

（1）明确城市固定源突发大气污染事故风险要素。本研究通过风险链深入分析风险形成的机理与风险要素之间的关联性，构建包括风险源暴露范围、源-受体急性暴露-效应关系、风险受体脆弱性、企业风险防控能力四个层面的城市突发大气污染事故风险分析要素集合。

（2）开展城市固定源突发大气污染事故扩散模拟与暴露评估研究。通过前期调查勘探，识别、筛选出研究区范围内的主要涉气风险企业23家与对应的典型风险物质，在此基础上确定最大可信事故类型，并根据风险物质的具体泄漏情形选择扩散模式（SLAB/AFTOX）与急性暴露浓度标准（PACs），开展长春市中心城区突发大气污染事故扩散模拟与暴露评估。

（3）构建基于贝叶斯网络的城市固定源突发大气污染事故风险评价方法。首先，划定城市风险评价单元，并依据突发大气污染事故扩散模拟结果，确定各风险源的暴露范围、评估风险评价单元的暴露情况；其次，将城市突发大气污染事故风险要素构成的风险链转化为贝叶斯网络模型，将事故扩散模拟的结果作为贝叶斯网络的重要节点输入；最后，采用贝叶斯网络Netica软件，通过计算各个风险节点先验概率和构建条件概率表，完成风险传递推理，汇总各条风险链的评价结果，并通过叠加分析得到各风险评价单元和风险源的风险水平。

（4）案例研究表明：长春市中心城区固定源突发大气污染事故风险空间分布异质性较大，高风险区呈现集群分布的特点。在23家企业风险源中，4家为高风险水平，5家为中风险水平，其余14家为低风险水平；在84个风险评价单元中，13个为高风险水平，19个为中风险水平，其余52个为低风险水平。在风险源视角下，硫酸、液氨危险性较高；在风险受体视角下，硫酸、液氨、甲基叔丁基醚对受体威胁最大，其中对评价单元风险水平贡献最大的风险源为S2、S4和S15。敏感性分析的结果表明：在所有风险水平情景下，风险源急性暴露水平、暴露范围和污染扩散范围都是敏感性最高的风险节点。由此分析，长春市中心城区应按照优先序，从消减风险源危险性、阻断风险暴露和保护风险敏感受体三个方面着手，采取差异化风险防控措施，提升城市大气环境风险防控水平。

As a large country in the production and consumption of chemicals, China has a large number of existing environmental risk enterprises, which are generally characterized by regional, structural and layout-type environmental risks. In recent years, sudden air pollution incidents have occurred frequently with alarming consequences, seriously threatening people's health and ecological safety. Cities, as highly populated areas, have a prominent risk of layout-type sudden air pollution. Therefore, it is urgent to carry out the risk evaluation and targeted prevention and control of urban sudden air pollution. This study focuses on the characteristics and formation mechanism of urban sudden atmospheric environmental risks, and comprehensively considers the risk superposition and transmission effects between risk sources and risk receptors within the urban area. On this basis, this study constructs a Bayesian network evaluation method based on the risk chain of urban sudden air pollution, and further selects the central city of Changchun, which is densely distributed by atmospheric environmental risk enterprises, to conduct a case study. The content and main conclusions of this study are as follows:

(1) This study clarifies the risk elements of urban air pollution emergencies. Based on the risk chain of “risk source-transmission pathway-risk receptor”, this study analyzes in depth the mechanism of the formation of urban atmospheric environmental risks and the correlation between risk elements, and on this basis constructs a collection of risk analysis elements, including four dimensions: risk source exposure range, source-receptor acute exposure-effect relationship, risk receptor vulnerability, and enterprise risk prevention and control capability.

(2) This study has carried out dispersion modeling and exposure assessment of urban air pollution emergencies. Through preliminary investigation and exploration, the study identified and screened out 23 major gas-related risky enterprises and corresponding typical risky substances within the study area. Accordingly, the maximum credible accident types were determined, and the dispersion model (SLAB/AFTOX) and acute exposure concentration criteria (PACs) were selected according to the specific leakage of the risk substances to conduct the dispersion simulation and exposure assessment of air pollution accidents in the central city of Changchun.

 (3) This study constructed a Bayesian network-based risk evaluation method for urban air pollution emergencies. Firstly, the urban risk evaluation unit is delineated, and the exposure range of each risk source is determined and the exposure of the risk evaluation unit is assessed based on the results of the diffusion simulation of the sudden air pollution accident. Secondly, the risk chain composed of the risk elements of the sudden air pollution accident in the city is transformed into a Bayesian network model, and the results of the accidental diffusion simulation are taken as the inputs to important nodes of the Bayesian network. Finally, this study adopts the Bayesian network Netica software to complete the risk transfer inference by calculating the a priori probability of each risk node and constructing the conditional probability table, summarizing the evaluation results of each risk chain, and obtaining the risk level of each risk evaluation unit and risk source through superposition analysis.

(4) The case study shows that the spatial distribution of the risk of sudden air pollution in the central city of Changchun is heterogeneous, and the high-risk areas are characterized by cluster distribution. Among the 23 enterprise risk sources, 4 are at high risk level, 5 are at medium risk level, and the remaining 14 are at low risk level; among the 84 risk evaluation units, 13 are at high risk level, 19 are at medium risk level, and the remaining 52 are at low risk level. Under the risk source perspective, sulfuric acid and liquid ammonia are more dangerous; under the risk receptor perspective, sulfuric acid, liquid ammonia, and methyl tertiary butyl ether are the most threatening to the receptor, and among them, the risk sources that contribute the most to the risk level of the risk evaluation units are S2, S4, and S15. The results of the sensitivity analysis show that the acute exposure level of the risk source, the range of exposure, and the range of contamination spreading are the risk nodes with the highest sensitivities in all scenarios. From this analysis, the relevant departments in Changchun City should adopt differentiated risk prevention and control measures in accordance with the order of priority, starting from the three aspects of abating the danger of risk sources, blocking the risk exposure and protecting the risk-sensitive receptors, so as to improve the level of prevention and control of urban atmospheric environmental risks.