Long-term reconstruction of daily global OMI NO2 product between 2005–2023 with spatiotemporally constrained compressive sensing

A novel compressive sensing combined with the spatiotemporal discrete cosine transform and a selective total variation denoising (STCS) framework to recover the missing pixels in the daily global ozone monitoring instrument (OMI) NO2 total vertical column density (TotVCD) data from 2005–2023, which can exploit the spatiotemporal autocorrelation properties of data. Validation results reveal a satisfactory recovery of the OMI NO2 TotVCD missing data through the proposed method. The OMI recovered dataset achieves the root-mean-square error (RMSE) = 0.25 DU, and mean bias (MB) = -0.085 DU against Pandora measurements, compared with RMSE of 0.242 DU and MB of -0.096 DU for the original OMI dataset. These results indicate that our STCS model achieves meaningful reconstruction that remained broadly consistent with original OMI observations. The recovered results also show a high consistency with the NO2 TotVCD data from TROPOspheric Monitoring Instrument (TROPOMI), with the correlation coefficient (R) = 0.752.Furthermore, the spatial distributions of the recovered NO2 TotVCD on multi-temporal scales provide detailed information, clearly revealing regional and global spatial patterns and variations. This dataset offers long-term daily seamless NO2 TotVCD information across the globe, which can facilitate the detection of short-term NO2 changes to identify transitory emission sources and support the analysis of long-term NO2 concentration patterns to help continuously improve air quality and other relevant research.