POPCORN is a novel data processing approach for satellite data post-process correction. POPCORN combines the best aspects of the conventional physics-based retrieval algorithms and state-of-the-art machine learning techniques.
The aim of the POPCORN project is to develop novel machine-learning-based post-process correction methodology that will improve the Earth observation satellite data accuracy. The methodology will be general and applicable to most satellite data. In POPCORN, we will use Sentinel-3 atmospheric aerosol data as an example data to develop and test the new methodology.
The main users of the POPCORN methods and data are the atmospheric scientists and satellite algorithm developers. Satellites are the only way to obtain near real-time global daily information of Earth’s atmosphere and there is a widespread need for accurate satellite-based information about the atmosphere. With POPCORN methods it may be possible to significantly improve the accuracy of existing satellite data with reasonable computational costs.
The main product of POPCORN is the novel methodology that combines the best aspects of conventional physics-based satellite retrieval algorithms and state-of-the-art machine learning algorithms. In addition, we have applied the POPCORN methodology to Sentinel-3 high-resolution atmospheric aerosol data and this data is available as open data for year 2019 and five regions of interest (Central Europe, East USA, West USA, Southern Africa, India).
POPCORN shows the true potential of satellite data in atmospheric remote sensing and improves the current operational atmospheric aerosol satellite data product accuracy