基于LUR和空间杜宾模型耦合的PM2.5浓度时空精细模拟研究

Urban air pollution is increasingly becoming a major environmental concern in cities around the world. Because of the frequent occurrence of severe haze pollution in China, it is necessary to analyze the spatio-temporal pattern of PM2.5 concentrations with high precision. Land Use Regression modeling (LUR) is a classical model for urban-scale atmospheric pollution concentration analysis and simulation. It is a statistical approach that uses spatial explanatory parameters to estimate pollutant concentrations at specific locations, and has been widely used in air quality studies. Its advantages are limited needs for explanatory parameters, a short computing time, and ease of development, making his empirical method capable to explain the spatial distributions of urban PM2.5 concentrations for entire cities. However, the existing LUR research has shortcomings. First, it rarely considers the spatial spillovers and autocorrelation of PM2.5 concentrations due to wind transmission, and possibly other influencing factors. Second, the LUR model variables are mostly two-dimensional and static, with little consideration of three-dimensional urban structure and dynamic factors. Finally, the time scale of existing research is mostly focused on the year or month, thus ignoring higher-frequency time disaggregation (day or hour). These shortcomings lead to biased estimations and inaccurate simulation results. The present study addresses all these issues. .Combining the LUR with the Spatial Durbin Model (LUR-SDM) to account for spatial diffusion spillovers and autocorrelation, this new coupling model will enhance the timeliness and accuracy of PM2.5 concentrations simulation. First, a prototype of the LUR-SDM coupling model will be explored, and the spatial weight matrix of 'wind direction-wind speed-distance' will be developed. Second, the multi-dimensional dynamic-static factor system and indicators dataset will be built. Finally, the proposed coupling model will be applied to Zibo city, and the precise PM2.5 concentration analysis and modeling will be carried out, based on a high-frequency time scale and multi-dimensional dynamic factors. This study has both theoretical and practical value for optimizing PM2.5 concentration modeling, revealing pollution effect mechanisms, and simulating and predicting future pollution for different scenarios of changes in land-use, transportation, and other factors.

在我国霾污染天气频发背景下，开展高精度PM2.5模拟及影响机制研究迫在眉睫。LUR模型是城市尺度大气污染模拟的经典模型，但已有研究中模型较少考虑PM2.5污染空间依赖性(尤其随风传输)，缺乏污染物本身及影响要素的空间溢出研究；其次模型指标多聚焦二维静态因子，三维结构、动态因子重视不够；此外模型时间尺度多基于年或月，忽略高频时间差异性；以上问题导致模拟结果精准性、时效性及真实性较差。而具有时间动态空间杜宾模型SDM的引入可弥补上述不足。本课题开展基于LUR-SDM耦合的PM2.5精细模拟研究，首先探索并搭建LUR-SDM耦合模型原型，研究‘风向-风速-距离’空间权重矩阵；其次构建模型多维动静态因子体系；最后应用至淄博市，开展时尺度面向多维动态时空因子的PM2.5精细模拟及污染机制分析。本研究对优化PM2.5模拟、揭示影响机制、预测不同情景PM2.5浓度及针对性污染治理具有重要理论和实用价值。

城市PM2.5浓度空间分布及影响机制的研究对于演技区域大气污染防治具有重大意义。本项目通过建立BRT模型，得到不同季节各土地利用因子对PM2.5的作用特征及贡献大小，并深入分析影响PM2.5变化的影响因素，结果表明，土地利用类型在不同季节中对PM2.5浓度影响存在显著差异，其中主导因素分别为支路、次干路、植被覆盖面、建设用地。本课题注重空间权重矩阵的构建，空间权重矩阵是一种有效表达空间关系的方式，本研究在假设全部空间中风速是均匀的、稳定的前提下提出了顾及“风速—风向—距离”约束的空间权重矩阵。通过耦合LUR与空间杜宾模型SDM，结合气象站数据以及多维城市因子来监测空气质量监测站点之间的PM2.5污染的传输及溢出效应，最终完成实验区域PM2.5污染浓度的实时定量评估与精准模拟。.本课题执行期间共发表学术论文6篇（SCI/SSCI检索5篇），培养硕士研究生6人，其中4人均已毕业。本项目的研究成果对城市PM2.5污染浓度的研究提供参考，为环保部门针对性治理大气污染提供决策支持。