EARTH-AID delivers GEODE, a trustworthy agentic AI that turns plain-language questions from experts and non-experts into local, uncertainty-aware impact assessments on top of DestinE-related Digital Twins (DTs). GEODE couples an generative AI interface (grounded in curated data lakes) with reasoning-and-planning and a probabilistic belief state, then actuates DT workflows and fast surrogate models to produce explainable maps, metrics, and recommendations in minutes. The project integrates assets from DT-GEO, GeoTwins, EPOS, and EuroHPC: harmonised workflow orchestration, interoperable metadata/APIs, and scalable HPC/Cloud deployment. A high-resolution exposure model (building-level, dynamic population) and citizen-science streams (e.g. felt reports) enable robust consequence analysis. Learning Local Emulators accelerate complex simulations while preserving quantified uncertainty and provenance.
EARTH-AID is user-centric: domain WPs (earthquakes, volcanoes, tsunamis, geoenergy/resources) co-design actionable user stories with operators, civil protection, and industry; cross-domain pilots address compound hazards (e.g. earthquake→tsunami). A transversal Responsible AI package implements ethics-by-design (EU AI Act), transparency, audit trails, bias testing, and multilingual, accessible interfaces. Outcomes meet the call’s Expected Outcomes: (i) innovative AI aligned with GenAI4EU/Apply AI delivering powerful local assessments fused with real-time data; (ii) novel interfaces translating non-technical needs into tailored scenarios; and (iii) best-practice guidelines for DT-AI impact assessment and risk communication. Target TRL ≥7 demonstrators will be validated with end users at representative European sites. EARTH-AID’s open standards, training, and uptake plan (DestinE/EPOS integration; standardisation with ISO/OGC/CEN) provide a clear path to operational adoption, helping save lives, reduce losses, and strengthen public trust in Europe’s geohazard services.
Project Leader at SED
Stefan Wiemer
SED Project Members
Maren Böse, Carlo Cauzzi, John Clinton, Florian Haslinger, Federica Lanza, Laura Laurenti, Michèle Marti, Romano Meier, Leila Mizrahi, Vanille Ritz, Danjiel Schorlemmer, Nadja Valenzuela, Cédric Zubler
HORIZON EUROPE 1.3 - Research infrastructures (HORIZON-INFRA-2025-01 – TECH-03)
Duration
September 2026–September 2029
Keywords
Agentic AI, Digital Twins, Impact simulations, geohazards, georesources
Research Field
Earthquake processes (Statistical seismology, induced seismicity, Geomechanical modelling); Monitoring, alerting and analysis (Data processing, products and alerting, real-time seismology), Earthquake Hazard and risk (hazard and risk modelling, geo risks), Support (Communication, IT and technology)