Characterization of Volcanic Cloud Components Using Machine Learning Techniques and SEVIRI Infrared Images

Area of Science:

• Earth and Space Sciences
• Remote Sensing
• Volcanology

Background:

• Volcanic explosive eruptions generate atmospheric clouds posing risks to human health and air travel.
• Satellite infrared (IR) sensors are crucial for monitoring volcanic clouds.
• Existing methods require significant effort for volcanic cloud characterization.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) approach for detecting and classifying volcanic clouds using satellite infrared imagery.
• To automate the monitoring and mapping of volcanic ash clouds in near real-time.

Main Methods:

• Implemented a supervised support vector machine (SVM) algorithm within Google Earth Engine (GEE).
• Utilized a combination of thermal infrared (TIR) bands from the geostationary MSG-SEVIRI sensor.
• Applied the ML algorithm to data from Mt. Etna's explosive events (2020-2022).

Main Results:

• The ML approach achieved an accuracy of 0.86 in detecting and classifying volcanic clouds.
• Demonstrated the efficiency of using combined TIR bands from geostationary satellites.
• Successfully automated the detection, tracking, and mapping of volcanic ash clouds.

Conclusions:

• Machine learning, particularly SVM with TIR bands, offers an efficient solution for volcanic cloud monitoring.
• The developed method enables near real-time detection and mapping of volcanic ash clouds.
• This approach enhances safety for populations near volcanoes and for air traffic.