FAIR2Adapt | FAIR to Adapt to Climate Change - EOSC HE Technology Group

10 | 12 | 2025 by the FAIR2Adapt Consortium FAIR2Adapt | FAIR to adapt to Climate Change EOSC HE Technology Group Anne Fouilloux, [email protected] •Implement FDOs using RO-Crate •Extend ROHub for CCA communities •Integrate EOSC core components •Integrate new services for the CCA communities (and beyond!) •Provide AI services to support FAIRification & improve user experience FAIR2Adapt, INFRA-EOSC: 36 months | 18 + 1 partners | PSNC EU Node role Enable FAIR data sharing for Climate Change Adaptation through open, distributed architecture integrated with EOSC MISSION TECHNICAL OBJECTIVES Technical Interoperability Challenges in CCA The Problem: Data Silos in Climate Change Adaptation Challenge Specific Problem Cross-domain data needs CCA requires integrating climate, hydrology, ecology, health, agriculture, urban planning, economics — each domain has its own standards and vocabularies Heterogeneous metadata schemas Copernicus uses ISO19115, Climate-ADAPT uses DCAT, weADAPT uses custom schema — different required fields, different vocabularies Fragmented variable descriptions "Sea surface temperature" vs "SST" vs "ocean_temp_surface" — no semantic alignment Multiple data formats NetCDF, Zarr, GeoTIFF, CSV — each requires different tools and libraries Incompatible catalog APIs STAC, CKAN, OGC CSW, GeoNetwork — each needs separate integration Language barriers CCA knowledge in EN, FR, DE, PT... — queries need multilingual support Result: Researchers spend more time finding and reformatting data than analysing it or making decisions. Our approach: “FAIR Digital Objects (FDOs) as interoperability “glue” 3 FAIRification layer Elements: •Persistent Identification •Terminology •Metadata •Provenance •Assessment •Licensing. System Architecture: GFF hourglass model Capacity building (WP5) LAYER 1: Data, tools & services to “FAIRify” (WP6) Copernicus, IS-ENES, Climate-ADAPT, weADAPT, NorESM, RioMar, etc. LAYER 2: FAIRification (WP3) FIPs, PIDs, I-ADOPT, F-UJI Assessment, Metadata CENTRE: FDO (WP2) RO-Crate + DTR + Nanopublications LAYER 3: Abstraction (WP2) ROHub FAIROs, FAIR Connect, STAC, OGC Standards LAYER 4: Services (WP4) Search & discovery, Text & Data Analytics, Knowledge Synthesis LAYER 5: End-User applications (WP4/6) Jupyter, Pangeo, QGIS, JupyterGIS, weADAPT, Connectivity Hub, Dashboards, etc. FDO and supporting technologies at the center of the hourglass Key aspects: •Data and metadata that can be exchanged and rendered as machine-actionable information •Expose and share FAIR data and metadata Core technologies: •RO-Crate as implementation of FDO •ROHub as RO-Crate management platform FDO: The Interoperability Layer RO-Crate + Nanopub Lifecycle FAIR Assessment UI Search & Discovery Python Library USER TOOLS (Consume FDOs via PID - Local/External) Python Pandas, Xarray R Package Statistics QGIS Plugin Geospatial Excel Add-in Spreadsheets Jupyter EGI notebooks, Pangeo@EOSC Dashboards R Shiny, React, etc. API API API API Working Storage EGI Datahub Object Storage Publishing Repositories NIRD Archive CONNECTED SERVICES (External - User Redirected / Aware) Map Visualisation Jackdaw AI LLM from maps Embedded SERVICES (Transparent to user) FIPs I-ADOPT Service Variable Alignment Big Data Pipeline STAC/CKAN Wrappers FAIR Assessment Info Extraction Metadata from publications claims/salient sentence extraction EarthQA NL Search & Discovery Semantic Enrichment Multilingual NLP Generative QA LLM Answering Recommendations FAIRness-aware RO-Crate + DTR + Nanopublication The interoperability Foundation RO-Crate (Research Object Crate) ●Lightweight, JSON-LD based packaging format ●Self-describing: metadata travels with data ●Extensible via profiles (we create CCA-specific profiles) DTR (Data Type Registry) Shared understanding across services ●EOSC DTR alignment (via FAIRCORE4EOSC) Nanopublication: ● metadata and provenance (enrichment, etc.) RO-Crate + DTR + Nanopublication enable: Key output: RO-Crate profiles for CCA Structural Interoperability Through Standards Any Tool → reads RO-Crate → understands via Nanopublication & DTR → processes FDO (Python, R, QGIS, Jupyter...) The Challenge "Same” variable, described 100 different ways: ●"concentration of dissolved oxygen in water" ●"DO_conc_mg_L" ●"Oxygen, dissolved (mg/l)" Machines cannot match these → data remains siloed The Solution: I-ADOPT Framework ●Automatic FAIR variable descriptions from text or human input. ●Automated annotation of description components using controlled vocabularies (NERC, Wikidata, ENVO, etc.) ●Enables semantic faceted search of digital objects in repositories like ROHub Interoperability Impact ●Cross-dataset discovery: "Find all datasets measuring oxygen in water" ●Automatic metadata alignment between Copernicus, PANGAEA, Climate-ADAPT ●Enables semantic faceted search in ROHub Approach & Status: LLM-assisted service in development, pattern library read 6 I-ADOPT Service - Semantic Interoperability 7 EarthQA + Pythia Query Interoperability The Challenge Each knowledge graph has: ●Different ontology ●Different SPARQL endpoint ●Different query patterns Users must learn each system separately The Solution: “Universal” Natural Language Interface Pythia (NKUA) ●Text-to-SPARQL translation ●KG-independent design ●Works on unseen knowledge graphs EarthQA (NKUA) ●Spatiotemporal queries for Earth Observation ●Integrates: YAGO2geo, Sentinel metadata, GADM ●Example: "Find Sentinel-2 images of rivers near ports in France with <20% clouds" Pipeline User Query (any language) ↓ Multilingual NLP (Expert.ai) ↓ Pythia (Text → SPARQL/GeoSPARQL) ↓ EarthQA / GeoQA / TerraQ (execute on YAGO2geo, Sentinel KG, etc.) ↓ Results via ROHub Interoperability Impact ●One query interface → multiple heterogeneous backends ●No need to know underlying schema ●Multilingual access to CCA knowledge 8 Big Data Pipeline Catalog Interoperability The Challenge ●Geospatial data catalogs speak different languages: Standard Used By STAC Copernicus, Destination Earth, etc. CKAN Open data portals, NIRD Archive, Climate-ADAPT GeoNetwork National SDIs, INSPIRE Micka Czech/Slovak infrastructure OGC CSW Legacy catalogs Develop GeoJSON-LD + RO-Crate profiles that: ●Harvest from any catalog type ●Transform to common FDO format ●Preserve provenance and licensing ●Enable unified discovery in ROHub Interoperability Impact ●Researchers discover data across catalogs from single interface ●Case study support: IFREMER (ocean), NERSC (Arctic) ●Aligned with EOSC Interoperability Framework Status: Wrappers in development Each requires separate integration code The Solution: Unified Wrappers