AI for atmospheric composition - how to enhance resolution using satellite data
Webinar with Federico Fierli (EUMETSAT) and Maximilien Houël (MEEO srl)
16 March 2022, 12:00 UTC
This course presents how new Artificial Intelligence methods could be used in Earth Observation and numerical model for monitoring purposes.
The method that will be presented is about super resolution: it consists in the enhancement of the resolution of an image based on the higher resolution of another image. In Earth Observation, a satellite rarely gets the capability to have a same scene at multiple resolutions: in this way the method will use two different data sources, on one side the Copernicus Atmosphere Monitoring Service model for NO2 measurements and on another side the Sentinel-5p NO2 measurements on the tropospheric column.
Both datasets are measuring the same environmental variable, but at different resolution: following the concept of super resolution, we retrieve on one hand our low resolution image and on another hand our high resolution image of a same scene.
Moreover, since Deep Learning is more and more available even without a large amount of knowledge in computer science, the course will also show how to use existing tools as FastAI and Pytorch to use state of the art Deep Learning architectures to build and use an own model.