Support Vector Machine for volcano hazard monitoring from space at Mount Etna

Abstract

Lava fountains produced from explosive eruptions at Etna volcano are characterized by the development of several km high ash plumes and lava overflows from the crater rim. The formation of lava flows and tephra deposits and the dispersion of volcanic clouds have the potential to impact on population, aviation and infrastructure. For this reason, mapping and quantifying the areal extension of the products erupted is of great importance for forecasting volcanic hazards. The huge amount of multispectral satellite data provides new perspectives for the near real-time monitoring of volcanic hazards, which needs new machine learning (ML) techniques for the automatic process of data. Support Vector Machine (SVM) and satellite data have been shown can be used to track lava flows and volcanic clouds. Here, we show the potentiality of using SVM and satellite data with different spatial and temporal resolutions to fully characterize the volcanic products erupted during the lava fountain event occurred at Etna volcano on 10 February 2022. Two SVM models have been adopted using Google Earth Engine platform to map and quantify (a) the extension of the lava flows and tephra deposits, and (b) the volcanic cloud dispersed into the atmosphere. Exploiting the potentiality of machine learning and cloud computing to process satellite remote sensing data, we obtained a global quantitative analysis of products erupted during lava fountaining with a high accuracy level.