地球信息科学学报 ›› 2022, Vol. 24 ›› Issue (8): 1459-1474.doi: 10.12082/dqxxkx.2022.210824

• 地球信息科学理论与方法 • 上一篇    下一篇

基于移动监测数据的不同城市场景下PM2.5浓度精细模拟与时空特征解析

谢晓苇1,2,3(), 李代超1,2,3,*(), 卢嘉奇1,2,3, 吴升1,2,3, 许芳年1,2,3   

  1. 1.福州大学 空间数据挖掘与信息共享教育部重点实验室,福州 350002
    2.福州大学 数字中国研究院(福建),福州 350002
    3.福州大学 地理空间信息技术国家地方联合工程技术研究中心,福州 350002
  • 收稿日期:2021-12-23 修回日期:2022-02-08 出版日期:2022-08-25 发布日期:2022-10-25
  • 通讯作者: *李代超(1989—),女,河南郑州人,博士,助理研究员,主要从事地理信息资源环境分析、时空数据分析与可视化 研究。E-mail: lidc@fzu.edu.cn
  • 作者简介:谢晓苇(1997— ),男,福建永泰人,硕士生,主要从事地理信息资源环境分析研究。E-mail: n195520021@fzu.edu.cn
  • 基金资助:
    中国科学院战略性先导科技专项(A类)(XDA23100502)

Fine Simulation and Analysis of Temporal and Spatial Characteristics of PM2.5 Concentration Distribution in Different Urban Scenarios based on Mobile Monitoring Data

XIE Xiaowei1,2,3(), LI Daichao1,2,3,*(), LU Jiaqi1,2,3, WU Sheng1,2,3, XU Fangnian1,2,3   

  1. 1. Key Laboratory of Spatial Data Mining & Information Sharing of Ministry of Education, Fuzhou University, Fuzhou 350002, China
    2. The Academy of Digital China, Fuzhou University, Fuzhou 350002, China
    3. National Centre for Local Joint Engineering Research on Geospatial Information Technology, Fuzhou University, Fuzhou 350002, China
  • Received:2021-12-23 Revised:2022-02-08 Online:2022-08-25 Published:2022-10-25
  • Contact: LI Daichao
  • Supported by:
    Strategic Priority Research Program of the Chinese Acadamic of Science(XDA23100502)

摘要:

城市内部PM2.5浓度分布具有明显的空间异质性,而传统方法基于遥感数据或监测站点数据进行分析,难以揭示高时空分辨率下城市内部的PM2.5浓度分布特征,缺少不同时刻城市场景(如:道路、工业区、住宅区等)对PM2.5浓度复杂非线性影响的解析。本研究将移动监测传感器安装于快递车上,采集福州市主城区南部不同类型场景的PM2.5浓度,然后融合地理加权回归(Geographical Weighted Regression, GWR)和梯度提升决策树(Gradient Boosting Decision Tree, GBDT)方法,提出一种基于GWR-GBDT的PM2.5模拟与场景解析模型,能够较好地拟合气象、场景因素与PM2.5浓度的非线性关系,提升了城市PM2.5污染精细监测能力;并结合部分依赖图解析不同时段不同场景因素对PM2.5浓度的非线性作用影响。结果表明:① 基于移动PM2.5浓度监测数据,利用GWR-GBDT模型能够较好地模拟城市场景、气象和PM2.5浓度之间的非线性关系,能够有效精细模拟PM2.5浓度的空间分布,十折验证R2结果为0.52~0.94;② 通过部分依赖图分析同一场景在不同时段对PM2.5浓度响应的异质性,发现各类场景对PM2.5浓度提升或抑制作用并不稳定;③ 解析不同时段人类活动与城市场景对PM2.5浓度的交互作用发现,教育医疗单位和住宅区两类场景对PM2.5浓度的提升作用都与人类通勤有密切关系,高污染场景中的建筑工地在采取的洒水降尘措施后能在数小时内有效缓解PM2.5污染,公园文体服务区在多数时段对PM2.5浓度具有抑制作用,工业区和道路多数时段会致使对PM2.5浓度提升;④ 从PM2.5浓度的空间分布来看,福州市主城区南部PM2.5浓度总体呈现东南高-西北低的分布趋势,建筑工地、道路和工业区场景轻度以上污染面积占比明显高于其他场景,公园场景总体PM2.5浓度较低,山体公园傍晚会受到周边工业区的影响而导致PM2.5浓度升高,而城市陆地外围水域对沿岸PM2.5浓度具有抑制作用;⑤ 研究结果可为不同场景下PM2.5污染精细化治理、城市规划以及老人、儿童等高危人群的PM2.5污染暴露风险防范提供支持。

关键词: PM2.5模拟, 移动监测, 城市场景, GWR, GBDT, 部分依赖图, 时空分析, 福州市主城区

Abstract:

The distribution of PM2.5 concentration has obvious spatial heterogeneity in the inner city. However, traditional analysis methods based on remote sensing data or monitoring station data are difficult to reveal the distribution characteristics of PM2.5 concentration in the inner city at high spatial-temporal resolution, and there is also a lack of analysis on the complex nonlinear effect of urban scenes (e.g., roads, industrial areas, residential areas, etc.) on PM2.5 concentration. In this study, we installed the mobile monitoring sensor on the express van to collect PM2.5 concentration in different urban scenes in the south of Fuzhou main urban area. The PM2.5 simulation and scene analysis model based GWR-GBDT method was proposed by fusing Geographical Weighted Regression (GWR) and Gradient Boosting Decision Tree (GBDT). The model can fit the nonlinear relationship between meteorological factors, scene factors, and PM2.5 concentration, and enhance the fine-scale monitoring ability of PM2.5 pollution in city. Combined with partial dependency plot, the nonlinear effect of different urban scenes on PM2.5 concentration in different periods was analyzed to provide support for urban PM2.5 pollution control. The results show that: (1) Based on the mobile PM2.5 concentration monitoring data, the GWR-GBDT model can well simulate the nonlinear relationship between urban scene factors, meteorological factors, and PM2.5 concentration, and simulate the fine spatial distribution of PM2.5 concentration. The results of cross-validation R2 was between 0.52 and 0.94; (2) The heterogeneity of the response of the same scene to PM2.5 concentration in different time periods was analyzed by the partial dependence plots, and we found that the effect of various scenes on PM2.5 concentration was different; (3) By analyzing the interaction of human activities and urban scenes on PM2.5 concentration in different periods, we found that the effect of urban scenes on PM2.5 concentration was related to human commuting between schools, hospital, and residential areas. As the high pollution scene, construction site can effectively reduce PM2.5 pollution in several hours after taking watering measures. In the park and sports service area, PM2.5 concentration was low in most periods. For industrial area and roads, PM2.5 concentration was high in most periods; (4) For the spatial distribution of PM2.5 concentration, PM2.5 concentration in the south of Fuzhou main urban area presented a general trend of high pollution in the southeast and low pollution in the northwest. The proportion of slightly polluted areas in construction sites, roads, and industrial areas was significantly higher than that in other scenes. The overall PM2.5 concentration in the park scene was low, however, the park in mountains was affected by the surrounding industrial areas at nightfall, resulting in increased PM2.5 concentration. The urban outer waters have mitigation effect on PM2.5 concentration around them. This study can provide support for fine-scale PM2.5 pollution treatment, urban planning, and PM2.5 pollution exposure risk prevention of high-risk groups such as the elderly and children in different scenarios.

Key words: PM2.5 simulation, mobile monitoring, urban scene, GWR, GBDT, partial dependency plot, spatiotemporal analysis, main urban area of Fuzhou