Climate research produces petabytes of heterogeneous data that must remain accessible and citable for decades despite frequent migrations across HPC and federated infrastructures. This contribution presents how FAIR Digital Objects (FDOs) can separate persistent identity and machine-actionable semantics from physical storage, ensuring long-term referability and interpretability. Combined with IPFS as a decentralized, content-addressed replication layer, the approach strengthens technical resilience and reliable data access. A prototype implementation in the ORCESTRA campaign demonstrates the benefits for distributed climate data workflows and highlights prospects for resilient, interoperable research infrastructures.