Introduction
Disaster Pilot Overview
Safe Software demonstrated that enterprise-grade Extract, Transform, Load (ETL) tools can act as a "universal translator" within the DGGS ecosystem, bridging the gap between geospatial and grid-native Artificial Intelligence (AI).
During the last OGC AI-DGGS Disaster pilot, the FME Platform proved that complex grid ingestion does not require specialized spatial code. By showcasing FME Flow’s Data Virtualization, high-volume datasets - including Digital Elevation Models (DEM), NetCDF climate projections, and historical flood models - could be easily integrated and served as OGC-compliant DGGS endpoints.
In case you’re unfamiliar, Data Virtualization turns FME workflows into live, shareable open APIs. This enables on-the-fly transformation and serving of Cloud Optimized GeoTIFFs (COGs) and vector datasets, ensuring multiple sources can be streamed into a DGGS fabric without manual preprocessing.
Within this pilot, we can refer to FME not only as the “Any Data, Any AI” solution, but also as the "Any Data, Any Grid" solution.
Case Study
Safe focused on building foundational DGGS tools and services, while supporting complex Technology Integration Experiments (TIEs). TIEs included powering Hartis’s Flood Impact Index (FII) and enabling TerraFrame’s AI client to query the DGGS for flood level information and then pass that to Spatial Knowledge Graphs in order to identify cascading infrastructure failures throughout affected electrical and transportation networks. The study area was the Red River Basin in Manitoba, Canada.
Terraframe AI client: chat natural language query for flooded zones (FME's DGGS flood levels) followed by a query about which roads are affected (spatial knowledge graph transportation model)
Terraframe AI client: various queries related to power infrastructure and estimated affected populations due to potential outages
Pilot Deliverables
Data Virtualization Approach: Server Side
The architecture consists of an FME Flow Hosted Linux instance. On top, FME Flow’s Data Virtualization environment was used to publish these workflows as RESTful services conforming to OGC API standards. This setup enabled dynamic conversion of traditional geospatial formats into grid-native spatial-temporal representations at multiple resolutions.
The implementation provides a compliant OGC API-DGGS interface supporting grid zone discovery, metadata access, and geometry delivery (GeoJSON and DGGSJSON formats). A key feature is the support for CQL2 (Common Query Language), allowing AI agents and clients to perform attribute, spatial, and temporal range queries.
Data Collection Overview: Collection Metadata
Zone Discovery (ISEA3H, Zoom 12): Zone Query Example
Zone Data Retrieval (ISEA3H, Depth 7): Data Retrieval Example
DGGS collection response in HTML
Custom Transformers: Client Side
Workflow authoring was conducted using FME Form on both MacOS and Windows. To support the mathematical requirements of the grid, Safe Software developed FME Custom Transformers (encoders, decoders, indexers, and queriers) that wrap the open-source DGGAL (Discrete Global Grid Abstraction Library) for use within FME’s PythonCaller.
The following four custom transformers were created as a result:
-
DGGSRelator
- Calculates the relationships between Base and Candidate DGGS grid cell indices. These are often not easily human-discernible from the index values.
-
DGGSIndexer
- Computes and manipulates Discrete Global Grid (DGGS) hierarchical spatial indexes, and enables spatial data to be grouped into grid cells for analysis and visualization.
-
DGGSJSONEncoder
- Encodes point data into the DGGS JSON schema.
-
DGGSJSONDecoder
- Decodes DGGS JSON-encoded data into geometry and attributes
These custom transformers were published on FME Hub to make it easy for all FME users to integrate and explore DGGS.
DGGS zone query response workspace showing use of DGGSIndexer custom transformer
DGGS flood level > 1.5 meters - GeoJSON response in FME Data Inspector
Conclusion
The AI-DGGS Disaster Pilot demonstrated how Discrete Global Grid Systems (DGGS) can serve as a universal geolocation framework to support spatial data integration and AI-driven queries across multiple domains. Using FME's Data Virtualization platform, the pilot successfully created OGC APIs that published no-code FME workflows as open, accessible endpoints.
Orchestrated AI workflows linked multiple AI systems and disaster-related data sources, integrating DGGS with Terraframe's Spatial Knowledge Graph (SKG) to enable seamless data connectivity. Together, these capabilities showcased the potential of DGGS as a foundation for integrated spatial data services in disaster risk assessment and management.