No abstract

Infrastructures are not inherently durable or fragile, yet all are fragile over the long term. Durability requires care and maintenance of individual components and the links between them. Astronomy is an ideal domain in which to study knowledge infrastructures, due to its long history, transparency, and accumulation of observational data over a period of centuries. Research reported here draws upon a long-term study of scientific data practices to ask questions about the durability and fragility of infrastructures for data in astronomy. Methods include interviews, ethnography, and document analysis. As astronomy has become a digital science, the community has invested in shared instruments, data standards, digital archives, metadata and discovery services, and other relatively durable infrastructure components. Several features of data practices in astronomy contribute to the fragility of that infrastructure. These include different archiving practices between ground-and space-based missions, between sky surveys and investigator-led projects, and between observational and simulated data. Infrastructure components are tightly coupled, based on international agreements. However, the durability of these infrastructures relies on much invisible work -cataloging, metadata, and other labor conducted by information professionals. Continual investments in care and maintenance of the human and technical components of these infrastructures are necessary for sustainability.

The promise of technology-enabled, data-intensive scholarship is predicated upon access to knowledge infrastructures that are not yet in place. Scientific data management requires expertise in the scientific domain and in organizing and retrieving complex research objects. The Knowledge Infrastructures project compares data management activities of four large, distributed, multidisciplinary scientific endeavors as they ramp their activities up or down; two are big science and two are small science. Research questions address digital library solutions, knowledge infrastructure concerns, issues specific to individual domains, and common problems across domains. Findings are based on interviews (n=113 to date), ethnography, and other analyses of these four cases, studied since 2002. Based on initial comparisons, we conclude that the roles of digital libraries in scientific data management often depend upon the scale of data, the scientific goals, and the temporal scale of the research projects being supported. Digital libraries serve immediate data management purposes in some projects and long-term stewardship in others. In small science projects, data management tools are selected, designed, and used by the same individuals. In the multi-decade time scale of some big science research, data management technologies, policies, and practices are designed for anticipated future uses and users. The need for library, archival, and digital library expertise is apparent throughout all four of these cases. Managing research data is a knowledge infrastructure problem beyond the scope of individual researchers or projects. The real challenges lie in designing digital libraries to assist in the capture, management, interpretation, use, reuse, and stewardship of research data.

No abstract

University libraries are partnering with disciplinary data producers to provide long‐term digital curation of research data sets. Managing data set producer expectations and guiding future development of library services requires understanding the decisions libraries make about curatorial activities, why they make these decisions, and the effects on future data reuse. We present a study, comprising interviews (n = 43) and ethnographic observation, of two university libraries who partnered with the Sloan Digital Sky Survey (SDSS) collaboration to curate a significant astronomy data set. The two libraries made different choices of the materials to curate and associated services, which resulted in different reuse possibilities. Each of the libraries offered partial solutions to the SDSS leaders' objectives. The libraries' approaches to curation diverged due to contextual factors, notably the extant infrastructure at their disposal (including technical infrastructure, staff expertise, values and internal culture, and organizational structure). The Data Transfer Process case offers lessons in understanding how libraries choose curation paths and how these choices influence possibilities for data reuse. Outcomes may not match data producers' initial expectations but may create opportunities for reusing data in unexpected and beneficial ways.