Satellite remote sensing has been reported to be a promising approach for the monitoring of atmospheric PM<sub>2.5</sub>. However, the satellite-based monitoring of ground-level PM<sub>2.5</sub> is still challenging. First, the previously used polar-orbiting satellite observations, which can be usually acquired only once per day, are hard to monitor PM<sub>2.5</sub> in real time. Second, many data gaps exist in satellitederived PM<sub>2.5</sub> due to the cloud contamination. In this paper, the hourly geostationary satellite (i.e., Harawari-8) observations were adopted for the real-time monitoring of PM<sub>2.5</sub> in a deep learning architecture. On this basis, the satellite-derived PM<sub>2.5</sub> in conjunction with ground PM<sub>2.5</sub> measurements are incorporated into a spatio-temporal fusion model to fill the data gaps. Using Wuhan Urban Agglomeration as an example, we have successfully derived the real-time and seamless PM<sub>2.5</sub> distributions. The results demonstrate that Harawari-8 satellite-based deep learning model achieves a satisfactory performance (out-of-sample cross-validation R<sup>2</sup>&thinsp;=&thinsp;0.80, RMSE&thinsp;=&thinsp;17.49&thinsp;&mu;g/m<sup>3</sup>) for the estimation of PM<sub>2.5</sub>. The missing data in satellite-derive PM<sub>2.5</sub> are accurately recovered, with R<sup>2</sup> between recoveries and ground measurements of 0.75. Overall, this study has inherently provided an effective strategy for the realtime and seamless monitoring of ground-level PM<sub>2.5</sub>.
Supplementary material to "Using Deep Learning to Fill Spatio-Temporal Data Gaps in
Hydrological Monitoring Networks"