Summary
This technical report presents MDIO v1, an open-source format designed for the storage and processing of seismic data, particularly in the context of artificial intelligence (AI) training and data visualization. The report outlines the architecture of MDIO, which is built on established open-source projects such as Zarr, Blosc, Xarray, and TensorStore, facilitating efficient data access and management in both local and cloud environments. It describes the structured dataset model introduced in MDIO v1, which includes components like Metadata, Variables, Dimensions, and Coordinates, enhancing the organization and usability of seismic data. The interoperability with Python’s Xarray and support for Dask allows for parallel processing, making MDIO suitable for large-scale geophysical analysis. The report includes case studies demonstrating MDIO's application in seismic workflows, showcasing its advantages in data ingestion, storage efficiency, and accessibility. Overall, MDIO v1 is positioned as a significant advancement in seismic data management, offering cost-effective solutions for the industry.
