Managing Research Data in Big Science

The project which led to this report was funded by JISC in 2010–2011 as part of its ‘Managing Research Data’ programme, to examine the way in which Big Science data is managed, and produce any recommendations which may be appropriate. Big science data is different: it comes in large volumes, and it is shared and exploited in ways which may differ from other disciplines. This project has explored these differences using as a case-study Gravitational Wave data generated by the LSC, and has produced recommendations intended to be useful variously to JISC, the funding council (STFC) and the LSC community. In Sect. 1 we define what we mean by ‘big science’, describe the overall data culture there, laying stress on how it necessarily or contingently differs from other disciplines. In Sect. 2 we discuss the benefits of a formal data-preservation strategy, and the cases for open data and for well-preserved data that follow from that. This leads to our recommendations that, in essence, funders should adopt rather light-touch prescriptions regarding data preservation planning: normal data management practice, in the areas under study, corresponds to notably good practice in most other areas, so that the only change we suggest is to make this planning more formal, which makes it more easily auditable, and more amenable to constructive criticism. In Sect. 3 we briefly discuss the LIGO data management plan, and pull together whatever information is available on the estimation of digital preservation costs. The report is informed, throughout, by the OAIS reference model for an open archive.

💡 Research Summary

The report “Managing Research Data in Big Science” presents the findings of a JISC‑funded project (2010‑2011) that examined how large‑scale scientific enterprises handle their data and formulated recommendations for funders, the Science and Technology Facilities Council (STFC), and the LIGO Scientific Collaboration (LSC). The authors begin by defining “big science” as research that generates data at petabyte‑scale, involves extensive international collaborations, and requires rapid, often real‑time, data sharing. These characteristics set big‑science data apart from more traditional disciplines, creating unique challenges in storage, transfer, metadata management, and access control.

The core of the analysis is anchored in the OAIS (Open Archival Information System) reference model, which structures the lifecycle of digital information into three packages: the Submission Information Package (SIP), the Archival Information Package (AIP), and the Dissemination Information Package (DIP). By mapping LIGO’s workflow onto this model, the authors illustrate how raw strain‑time series, calibrated data, and derived parameter estimates are each treated as distinct AIPs, allowing the collaboration to satisfy both long‑term preservation requirements and the immediate needs of analysis teams.

A major contribution of the paper is its discussion of the benefits of a formal data‑preservation strategy. The authors argue that well‑preserved data underpin scientific reproducibility, enable future discoveries, and support open‑data mandates. They note that, despite the massive scale, the LSC already practices a level of data stewardship that aligns closely with international best practices (e.g., ISO 16363 certification criteria). Consequently, the report recommends a “light‑touch” policy approach: funders should require that projects produce a documented Data Management Plan (DMP) that formalizes existing good practices, makes them auditable, and invites constructive criticism, rather than imposing heavy new procedural burdens.

The third section provides a concise overview of LIGO’s DMP. It details automated metadata capture at acquisition, integrity‑checking pipelines, multi‑site replication, and scheduled format migration. The authors also present a cost‑estimation framework that breaks preservation expenses into hardware (storage media, servers), software (archival management systems), and personnel (data curators, system administrators). By leveraging existing high‑performance computing infrastructure, LIGO minimizes the need for dedicated preservation hardware, achieving economies of scale that significantly reduce the overall budget.

Policy recommendations are distilled into five actionable points: (1) formalize DMPs at project inception and update them at key milestones; (2) adopt and enforce community‑wide metadata standards through automated pipelines; (3) design replication and migration strategies in advance to mitigate technology obsolescence; (4) quantify preservation costs and negotiate long‑term funding with sponsors; and (5) publish transparent access policies, applying controlled‑access mechanisms only where necessary.

In conclusion, the authors assert that big‑science data management, while demanding due to volume and collaboration complexity, can be effectively addressed by applying the OAIS framework and by codifying the already‑robust practices of projects like LIGO. A modest regulatory overlay—focused on documentation, auditability, and cost transparency—will enhance data longevity and reuse without stifling scientific productivity. This balanced approach promises to safeguard the immense scientific value embedded in big‑science datasets for future generations.