Understanding Earth observation data without being a geospatial data specialist.

Challenge: Valuable Data, High Barriers to Access

Earth observation data is highly relevant for many organizations. Climate adaptation, land sealing, urban planning, infrastructure development—all of these could be planned significantly better with spatially resolved satellite data. In practice, however, its use often fails due to technical complexity. Data selection, preprocessing, geospatial analysis, and interpretation of results require specialized expertise that simply isn’t available in municipalities, NGOs, or smaller planning firms.
The result is a paradoxical imbalance: The data exists, is available, and would improve important decisions. But only a few can use it.

Solution: Natural Language as a Gateway to Geodata Analysis

As part of the ESA InCubed project actinia-copilot URBAN, or ac-URBAN for short, a fully open-source-based AI framework is being developed that combines large language models with geospatial data processing. VAGO Solutions is developing the AI components for retrieval-augmented generation and agent systems, which act as an intelligent interface between user queries, knowledge sources, and the geospatial analysis environment.
Users submit their queries in natural language. The agent system translates these into transparent, reproducible geospatial workflows, selects appropriate data sources, and passes the processing steps to the actinia geoprocessing engine, which is based on the GRASS GIS ecosystem. The results are presented in the form of interactive maps, analyses, and reports.

Technical Highlights: Transparent, Verifiable, Reproducible

A key feature of ac-URBAN is the traceability of the generated analyses. Instead of “black-box” outputs, the system provides explicit process chains that can be validated and reproduced by users and institutions. The RAG component ensures that queries are precisely understood based on documentation, data catalogs, and project-specific knowledge. The entire architecture is open, extensible, and designed so that public contracting authorities and regulated application contexts can trust the solution in the long term.

• Explicit, reproducible geospatial process chains
• RAG component with data catalog and document integration
• Fully open-source-based system architecture

Result: Complex analyses. Easy access. Reliable results.

With ac-URBAN, municipalities, engineering firms, environmental monitoring organizations, and NGOs gain low-threshold access to high-quality Earth observation analyses. The system handles the technical complexity of geospatial processing. What remains are reliable, locally applicable insights for urban climate adaptation and planning that can be interpreted and reused even without in-depth expertise in geodata.