Extreme atmospheric events detection using IASI


Definition of an extreme atmospheric event

Many types of extreme atmospheric events can be identified:

  • high pollution episode generated by large fires
  • high pollution peaks on large urban areas
  • volcanic plumes
  • agricultural pollution
  • major industrial accident


Objective and approach

This study aims to define a method for the detection and the characterisation of exceptional atmospheric events, like those listed above, and to develop a software able to process IASI data in near real-time.
Two methods based on Principal Component Analysis (PCA) are implemented and assessed: i) standard PCA of spectra based upon reference database optimized for detection of extreme events; ii) innovative approach combining PCA and technique of principal angles, exploiting jointly simulated and measured reference databases.
The prototype is evaluated against two IASI datasets: i) few cases of atmospheric events for validating the method; ii) one year of global IASI data, including clearly identified episodes of reference, for assessing the software performance.



Monthly average cosine below the surface (characterize the angular structure of the light field) for May 2011 from MERIS.

Funder: CNES
Duration: 2017-2018
Contact at Hygeos: Dominique Jolivet