dc.description.abstract | Satellite remote sensing images provide periodical and multispectral information about the atmosphere and ground surface, which empowers various monitoring applications. Among the applications, some of them require high spatial and temporal resolutions, such as disaster management, crop phenology monitoring, and land cover change detection. Recently, with urban development and industrialization, air pollution becomes an important issue to human health. While ground stations can monitor air quality effectively, the spatial resolution of ground stations is usually low to observe local phenomena. Therefore, some studies apply remote sensing techniques on air quality monitoring. However, while air quality changes are local and rapid, the spatial resolution and temporal resolution of a single satellite is not enough to observe such dynamic phenomenon. To address this issue, this research aims at proposing a spatial and temporal image fusion method to synthesize high spatial and temporal resolution images for air quality monitoring. To be specific, this research proposes the TOA-reflectance-based spatial-temporal image fusion method (TOA-STFM), which is a variation of the STARFM method. While the STARFM method focuses on the surface reflectance, to preserve the atmospheric properties in the fused images, the TOA-STFM first uses the band with longer wavelength to estimate the fusion weightings and includes blurring effect adjustment in the image fusion process, and then finally fuses with the green band images to preserve atmospheric effects. In addition, this study aims at examining the feasibility of fusing high spatial resolution and Himawari-8 images. By integrating these two types of satellite images, high spatial resolution and ultra-high temporal resolution images can be generated to observe rapidly-changing phenomena.
To evaluate the proposed solution, we first compared the fused images with actual observation images, and then retrieved aerosol optical depth (AOD) from the fused images and compared with in-situ observations from the AErosol RObotic NETwork (AERONET). The result shows that the fused images from the TOA-STFM can provide more detail information than the STARFM. In addition, the fused images (from both the STARFM and the TOA-STFM) could provide better AOD estimation than the original Himawari-8 images. Furthermore, the purposed TOA-STFM method is more accurate than the original STARFM method on AOD retrieval, where the TOA-STFM can achieve 2% to 14% relative error.
In summary, this research has the following contributions. The TOA-STFM can predict TOA reflectance successfully, and provide ultra-high temporal resolution and high spatial resolution fused images for air quality monitoring. In addition, based on the evaluation result, we can find that the TOA-STFM can provide more stable and accurate result than other methods, which means the TOA-STFM is effective in preserving TOA-reflectance from coarse-resolution images, and is useful for dynamic air quality monitoring.
| en_US |