Fusion of Sentinel-2 and Sentinel-3 data for high-resolution daily aerosol advection mapping in urban areas

We present a real-time unsupervised procedure for generating maps of advected aerosols, mainly dust and smoke from biomass burning, that combines the 10m scale, suitable for urban environments, allowed by Sentinel-2 imagery, and the almost daily revisit capabilities of Sentinel-3. The map is generated at 10m scale from the bands 2, 3, 4, and 5 of Sentinel-2, L1C, and L2A formats, every five days, in the absence of cloud cover. Then it is extended daily via spatial modulation by an analogous map, calculated from the visible and near-infrared (VNIR) observations of the OLCI instrument onboard Sentinel-3, available at 300 scale. The bands of the two satellites are separately processed and fusion occurs at feature level, by combining the maps of Sentinel-3 with the maps of Sentinel-2. So, the average aerosol content in a 300m×300m cell is updated almost daily; the spatial distribution of the aerosols (spatial modulation) within the 300m×300m cell is inferred for the missing dates of the Sentinel-2 map. The spatial index is an upgrade of a preliminary study aimed at discriminating dust and smoke based on their scattering properties. The overall chain is validated through correlations with the point measurements of fine- and coarse-mode AOD performed by an AERONET station available at the test site.