NOvel cOmputational methoDs for reLiablE SAteLlite-based Air quality Data (NOODLESALAD)

NOODLESALAD aims to develop computational methods for improving the satellite-based air quality estimates. More specific, it will concentrate on improving the air quality key indicator PM2.5, which is the dry mass concentration of fine particulate matter with an aerodynamic diameter of less than 2.5 micrometers (micrograms per cubic meter of air).

This activity will be developing a novel artificial intelligence approach for retrieving PM2.5 from earth observation data. The innovative strategy will be based on machine learning post-process correction that we recently developed. The novel approach will utilize an innovative fusion of Sentinel-3 satellite data, simulation model information, ground-based observations, traditional satellite retrieval techniques, and machine learning to produce satellite-based PM2.5.

In this development work, data from the year 2019 will be used and select Central Europe as region of interest. The project will produce and validate PM2.5 estimates with a high spatial resolution of 300 meters for the Sentinel-3 satellites overpasses. In addition, this prototype approach will be used to create high temporal resolution air quality datasets for 5-10 European cities. Finally, the PM2.5 datasets produced will be publicly shared together with an open-source code package for Sentinel-3 PM2.5 retrieval.