首页|High-resolution PM2.5 retrieval using Gaofen-1 WFV camera data
High-resolution PM2.5 retrieval using Gaofen-1 WFV camera data
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Elsevier
To characterize the distribution characteristics of PM2.5 pollution within a small scale within the city, PM2.5 concentration estimation data with high spatial resolution is required. However, the existing PM2.5 data mostly have a resolution of more than 1 km, and the spatial scale is relatively coarse, which makes it difficult to meet the actual needs of fine pollution monitoring and governance in cities. To address this limitation, this study based on the deep neural network (DNN) model, used multi-source auxiliary data such as the top atmospheric reflectance (TOAR) data and meteorological data of the Gaofen-1 (GF-1) satellite to achieve near-ground PM2.5 concentration retrieval with a spatial resolution of 30 m in the Beijing-Tianjin-Hebei (BTH) region. The model's performance was evaluated by a cross-validation (CV) method based on samples and sites. The results showed that the sample-based R2 was 0.912 with an RMSE of 10.326 mu g/m3, while the site-based R2 was 0.780 with an RMSE of 16.375 mu g/m3. The high-resolution PM2.5 data retrieved in this study can reveal the spatial differences in pollution at the urban block and even building scale. Combined with the analysis of a typical pollution event, the practical application potential of this high-resolution data in identifying local pollution sources and analyzing pollution changes within the city was verified. This study provides a new technical path for high-resolution PM2.5 retrieval and provides important data support for refined urban air quality monitoring and scientific governance.
NETL
NSTL
Elsevier
