Detection of extreme events with IASI observations

The 3 IASI instruments on-board the Metop satellites have been sounding the atmospheric composition since 2006. Up to ~30 atmospheric gases can be measured from IASI spectra, allowing monitoring of weather, atmospheric chemistry, and climate.

Extreme events such as fires, high pollution episodes, volcanic eruptions, industrial accidents, etc., that impact on the population and the environment have become a major political issue. With IASI providing global observations twice a day in near real time, a new way for the systematic and continuous detection of exceptional atmospheric events to support operational decisions is possible.

In this work, we explore and improve an automatic system for the detection and characterization of extreme events, which relies on the principal component analysis (PCA) method. We assess this PCA-based system by analysing IASI raw and compressed spectra along with their differences (residuals) for various past and documented extreme events. The benefits and limitations of this method will be discussed. A new method based on the refined analysis of residuals for the whole year 2019 is proposed, that could be used as an automatic detection method for unexpected events. Finally, we investigate the potential of deep learning methods as a way to compare residuals with a database of extreme event in order to better characterize detected events.