集成多源遥感数据的PM2.5浓度空间分布制图

针对传统地面稀疏站点监测PM_2.5浓度以点带面的缺陷,本研究拟借助多源遥感数据开展了地面大气细颗粒物PM_2.5浓度空间分布模拟研究。以京津冀地区2013年的年均、季均PM_2.5浓度模拟图为例,用简化的气溶胶反演算法（SARA）反演了 1 km高分辨率AOD,并结合高分辨率遥感提取污染相关地理要素,对研究区PM_2.5浓度空间分布进行地理统计模拟及优选。结果表明：① SARA算法反演的AOD与地基AERONET相关性达0.99,能准确地反映研究区AOD的时空分布特征;② 集成多源遥感数据的地理加权回归模型拟合度高（平均R²-0.66）,其空间模拟显示研究区平均PM_2.5污染南部城镇最重,中东部城区次之,西北山区较轻;③ 研究区PM_2.5污染程度高,全年平均模拟浓度高达75 μg/m³,在气候环境及主要污染源季节性差异驱动下,浓度分布季节性特征显著,冬季污染最严重,而夏、秋季相对较轻。该成果对于精细把握PM_2.5污染特征,指导污染防控具有重要意义。

Mapping of PM2.5 Fine Particulates Using Multi-source Remote Sensing Data

Chronic exposure to fine particulate matter with the aerodynamic diameter less than 2.5 μm (PM_2.5) is a major risk factor for cardiovascular and respiratory morbidity. Environmental problems of PM_2.5 in China have aroused great concerns within both the government and the public. However, studies of this subject have been limited due to the spare distribution of national PM_2.5 monitoring network. Satellite-derived aerosol optical depth (AOD) measurements can provide the spatiotemporally resolved monitoring and have been widely applied to estimate the ground level PM_2.5 concentration. So far, the AOD measurements only have moderate spatial resolutions (3 to 17.6 km) which are insufficient for resolving the city level aerosol features. In order to address this issue, a 1 km resolution AOD retrieval algorithm-SARA (for MODIS instruments) was applied. Additionally, based on the high-spatial resolution fusion images, the PM_2.5 related factors (such as the land use data, dusty area and major roads information) had been extracted. This study then proposed a novel geostatistical modeling method that integrates the fine resolution retrieved AOD and the high resolution area-source emission data to investigate the spatial variability of PM_2.5 levels. Results show that the SARA retrieved AOD can obtain an average correlation coefficient R of 0.99 and an RMSE of 0.08, when comparing with AERONET measurements, which indicates that SARA has good accuracy and reliability in representing the aerosol conditions over the study area. Results show that, the annual average value of PM_2.5 even exceeded 75μg/m³ and a discrepant spatial heterogeneity occurred between seasons. The worst conditions occurred in the winter, and then in the autumn. The GWR-predicted PM_2.5 concentrations (with an average R² of 0.66) are strongly correlated with the in situ observations. The spatial patterns revealed in the mapping results show that the high-level PM_2.5 pollution clustered in southern parts of the study domain, while the low-level mainly appears in the rural or mountain land within the northern areas. Furthermore, these findings recommend the combination superiority of GWR model and multi-source remote sensing data, as well as introduced the great potential of fine-scale PM_2.5 simulation in distinguishing the spatial patterns of PM_2.5 and guiding the air pollution control strategies.