Transforming volcanic monitoring: A dataset and benchmark for onboard volcano activity detection

TL;DR

This work tackles the lack of large, annotated datasets for onboard volcanic activity detection by introducing a Sentinel-2 L1C-based dataset with binary anomaly labels, spanning 35 volcanoes across six continents and featuring SWIR-augmented RGB images and 9-channel MSI cubes at multiple GSDs. It benchmarks a range of state-of-the-art and lightweight models, employing transfer learning and score-level knowledge distillation to yield a compact DCNN suitable for onboard deployment, and validates performance on the Intel Movidius Myriad X VPU. Key findings show transformer-based architectures perform strongly, with 10–20 m GSD generally outperforming 75 m, while the distilled DCNN meets stringent onboard size and accuracy requirements (0.8773) for SWIR-RGB inputs. The work demonstrates the feasibility of near-real-time volcanic monitoring directly on satellites, enabling faster alerts and paving the way for constellation-scale onboard disaster management systems, while providing a valuable dataset and baseline benchmarks for future research and deployment.

Abstract

Natural disasters, such as volcanic eruptions, pose significant challenges to daily life and incur considerable global economic losses. The emergence of next-generation small-satellites, capable of constellation-based operations, offers unparalleled opportunities for near-real-time monitoring and onboard processing of such events. However, a major bottleneck remains the lack of extensive annotated datasets capturing volcanic activity, which hinders the development of robust detection systems. This paper introduces a novel dataset explicitly designed for volcanic activity and eruption detection, encompassing diverse volcanoes worldwide. The dataset provides binary annotations to identify volcanic anomalies or non-anomalies, covering phenomena such as temperature anomalies, eruptions, and volcanic ash emissions. These annotations offer a foundational resource for developing and evaluating detection models, addressing a critical gap in volcanic monitoring research. Additionally, we present comprehensive benchmarks using state-of-the-art models to establish baselines for future studies. Furthermore, we explore the potential for deploying these models onboard next-generation satellites. Using the Intel Movidius Myriad X VPU as a testbed, we demonstrate the feasibility of volcanic activity detection directly onboard. This capability significantly reduces latency and enhances response times, paving the way for advanced early warning systems. This paves the way for innovative solutions in volcanic disaster management, encouraging further exploration and refinement of onboard monitoring technologies.

Figures (2)

• Figure 1: Distribution of anomaly and non-anomaly samples across the selected volcanoes in the dataset (2016-2024).
• Figure 2: Sample SWIR-augmented RGB images (high-temperature regions are highlighted in red del2021board) of volcanic activity labelled as anomalies. Images with visible fumes and tephra have also been categorised as anomalies.