Interoperable Infrastructure in the AI Era

Interoperability in Climate and Atmospheric sciences
FAIR NetCDF Data

Problem

How can Ash train an extreme-rainfall prediction model using years of radar and atmospheric data without correcting each dataset manually?

Previously, Ash compared two IDRA radar datasets. She now wants to build an extreme-rainfall prediction model using several years of radar observations, atmospheric observations, and reanalysis data. A human can sometimes rename variables, inspect a README, or download individual files, but a large automated pipeline cannot reliably depend on those interventions.

AI is not a fourth interoperability layer. It makes the existing structural, semantic, and technical requirements more visible as the workflow becomes more automated and data-intensive.

What does “AI-ready” mean?

AI-ready infrastructure enables a defined machine-learning workflow to discover, access, interpret, retrieve, transform, and trace the data it needs with minimal dataset-specific manual intervention.

• AI readiness is task-dependent.

• No single data format or technology automatically makes a dataset AI-ready.

• AI readiness therefore builds on interoperability, but extends beyond it.

AI changes the scale of the interoperability problem

Interoperability does not guarantee trustworthy AI

• Interoperability is an important foundation for automated and reproducible AI workflows, but it should not be confused with scientific validity or model quality.

• Datasets can be structurally, semantically, and technically interoperable while still contain measurement errors.

• Interoperability does not address questions of data quality, representativeness, modelling methodology, validation, uncertainty, and scientific interpretation.

Example : AI for Earth Observation

FAIR-EO: AI for Earth Observation

https://eodata.bvlabs.ai/ai4eo/#/

Anemoi European framework for AI

Anemoi: European framework for AI

https://www.ecmwf.int/en/about/media-centre/aifs-blog/2026/anemoi-european-framework-ai

Key takeaways

• AI readiness is task-dependent. No single data format or technology automatically makes a dataset AI-ready.

• AI does not introduce a new interoperability layer. Instead, automated AI workflows increase the importance of structural, semantic, and technical interoperability.

• Interoperable source data provide a foundation from which task-specific AI training datasets can be constructed reproducibly.

• Cloud-native layouts such as Zarr can improve selective and parallel access for suitable large-scale workflows, but they do not automatically improve semantic interoperability or data quality.

• Persistent identification, provenance, and versioning complement interoperability by connecting models and derived training data to the exact source data and transformations from which they were produced.

• Interoperability supports reproducible and inspectable AI workflows, but it does not guarantee that training data are representative, scientifically appropriate, unbiased, or sufficient for a particular model.