Leveraging Machine Learning techniques and SEVIRI data to detect volcanic clouds composed of ash, ice, and SO2 during the 2020-2022 Etna eruption activity

Volcanic eruptions inject large amounts of particles and gases into the atmosphere. The detection of volcanic constituents is crucial to support aviation safety and to quantify their impact on human health, environment and climate. Detection of volcanic clouds represents a key input for retrieval algorithms such as VPR (Volcanic Plume Retrieval) and LUT (Look-Up Tables), which are applied to get information on particles and SO2 total mass. 

The detection of volcanic clouds using satellite data is challenging, particularly in the presence of high quantities of water vapor. This latter, in combination with ash particles, can turn into water droplets and ice. This physical phenomenon supposes a limitation for the detection of volcanic clouds.

Mount Etna (Italy), between 2020 and 2022, has produced 66 lava fountain events. These events have generated volcanic clouds mixed with ash, ice and SO2, with a top height ranging between 4 and 13 km. In this work a Machine Learning-Based approach to detect the volcanic clouds generated during these Etna’s lava fountain events is carried out. The models have been trained and validated by exploiting a dataset that covers the 66 lava fountains observed by the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) instrument, on board of Meteosat Second Generation (MSG) geostationary satellite, aiming to get insights for the discrimination of ash, ice, and SO2 in the volcanic clouds. The results are promising for the automatic detection of volcanic clouds in near-real time.