Satellites have brought new opportunities for monitoring atmospheric NO₂, although the results are limited by clouds and other factors, resulting in missing data. This work proposes a new process to obtain reliable data products with high coverage by reconstructing the raw data from multiple satellites. The results are validated in terms of traditional methods as well as variance maximization and demonstrate a good ability to reproduce known polluted and clean areas around the world.