A Data‑Centric Approach to Data Provenance in Seismic Imaging Data

Seismic imaging data is noisy and visually complex and thus its interpretation requires human intuition. However, we do not often have a single feature but multiple views of a feature. Managing multi-user and multi-version interpretations, combined with version tracking, with the existing application-centric software is challenging since feature objects are stored separately from their seismic surveys and possibly in different local machines. In this work, we restructure the seismic volume data and the interpreted features into a combined central storage, implemented in a relational database. Features, by multiple user sessions, are presented as a metadata. A parallel data fetching mechanism is linked to this structure to produce an intermediate object that can be graphically rendered on the user screen. We present case studies that illustrate our system's ability to reproduce users' modifications to the interpretations of others and the ability to retrace the history of modifications to a visual feature. We believe that the implications are significant in that a data provenance architecture can be imposed on multiple remote, visually-rich, user interactions with data and derived data products. This can enable workflow logic to be transferred from the application domain to the data management platform.