THALES SERVICES NUMERIQUES S.A.S (FR)
Context and Motivation:
Volcanic eruptions and landslides represent significant hazards on population safety and health. For instance, exposure to major eruptions is of critical societal and economic concern due to release of large quantities of harmful gases (carbon/sulfur dioxides, hydrogen sulfide, halogens) and ashes. Such emissions impact human health, agriculture, infrastructure, as well as the climate (cooling effects, ozone layer depletion, acid rain). While well-monitored volcanoes benefit from advanced in-situ instruments, many hazardous sites lack sufficient monitoring.
Synthetic Aperture Radar (SAR) imagery is a well-established and effective technology for monitoring of continental surfaces. Its applications span numerous fields, including natural hazard prevention (earthquakes, volcanic eruptions, landslides, floods), agriculture, forestry, geodynamics, and land management. Additionally, Persistent Scatterer Interferometry (PSI) is of great interest since enabling continuous, even near-real-time monitoring of geodynamic processes through fine analysis of surface deformations.
The proposed initiative ultimately aims at developing scalable solutions for the monitoring and early warning of both volcanic unrest (Forecasting Tool Prototype) and landslide risks (Alert System Demonstrator), capable of near-real-time anomaly detection in deformation time series, suitable for generalized application across diverse volcanic and landslide-prone contexts worldwide.
Project Description:
Harnessing the synergy between Earth Observation data, especially PSInSAR products derived from C-Band (Sentinel-1) and L-Band (SAOCOM-1A) SAR, and AI-driven analytics (advanced Machine and Deep Learning (ML/DL)), should offer new opportunities for scalable surface deformation monitoring. Making then this initiative suitable to address critical challenges in timely and spatially extensive geohazard coverage, enhancing risk mitigation strategies and resilience for vulnerable societies.
Schematically, the concept focuses on the implementation of ML/DL algorithms tailored to PSI temporal series analysis applied to several reference sites with well-documented eruptive / landslides histories, to allow for the identification of relevant spatiotemporal features. More precisely, it consists in early-stage detection of anomalies or non-nominal deformation behaviors, such as sudden deformation rate increase or specific movement patterns, improving risk detection sensitivity and strengthening forecasting reliability.
Dealing with the landslide hazards, the project also addresses assimilation of hydro-meteorological data to correlate surface deformation anomalies (e.g., abrupt acceleration patterns) with significant climatic triggers, thus enriching the interpretation of deformation drivers and enhancing forecasting capabilities and risk qualification.
Technological Challenges: