Building an AI-Ready National Dataspace for Exploiting Massive Fiber-Optic DAS Data

Massive Fiber-Optic Distributed Acoustic Data (FO-DAS) streams pose major challenges in archiving, dissemination, and exploitation due to their extreme data rates, long acquisition durations, and high spatial-temporal resolution. Efficient storage is constrained by bandwidth, cost, and metadata standardization, while dissemination is limited by network capacity and interoperability. Scientific exploitation is further hindered by the need for scalable preprocessing, real-time analytics, and robust noise characterization to extract actionable signals from petabyte-scale, heterogeneous datasets.

This contribution targets the presentation of the DATA TERRA (FormaTerre, Odatis, THEIA) approach to describe, store, disseminate and exploit, through the GAIA-Data distributed data and computing infrastructure, massive FO-DAS datasets. Key infrastructure aspects are presented allowing to construct a national and AI-ready FO-DAS dataspace allowing easy and interactive exploitation of massive FO-DAS data for seismological source identification, event characterization and seismic parameter estimation generalizing across volcanoes, glaciers, fault zones, landslides, and urban areas,

FO-DAS bottlenecks are addressed via AI-driven compression (e.g. variational autoencoders), selective archiving, and data augmentation to ensure scalable monitoring. Integration of the dataspace in the DATA TERRA EOSC node will ensure interoperability with other national (NFDI4DEarth) and European research infrastructures (EPOS, EMSO, eLTER).