Item

Abstract

Explosive volcanic eruptions can disperse significant quantities of ash over large areas with consequences for human health, air traffic security, and critical infrastructure. Consequently, the detection and monitoring of volcanic ash clouds is of crucial importance to enhance the safety of human settlements and air traffic. The latest generation of high resolution satellite sensors (e.g. EUMETSAT MSG Spinning Enhanced Visible and InfraRed Imager) provide radiometric estimates to observe and to monitor volcanic clouds at global scale in an efficient and timely manner. However, the satellite data volume is too large to be manually processed and analyzed on a daily basis and global scale. Deep learning (DL), a fastest-growing technique of artificial intelligence in remote sensing data analysis applications, has an excellent ability to learn massive, high-dimensional image features and have been widely studied and applied in classification, recognition, and detection tasks involving satellite images. Here, we evaluate the ability of deep Convolutional Neural Networks (CNNs), a DL class, to detect and track the dispersion of volcanic ash clouds into the atmosphere, exploiting a variety of spatiotemporal information mainly coming from satellite sensors. We train some popular deep CNN models through transfer learning. We show that even when using transfer learning the size of the dataset plays a key role in the performance of the models, with those trained on the smaller datasets showing a strong tendency to overfit. These models were applied to the paroxysmal explosive events that occurred at Mt. Etna between 2020 and 2022.