From Sea to Servers: Temporalities of Data Management and the Limits of Availability in Oceanography

Abstract

A persistent challenge across scientific fields is determining what research data to keep, why, how, and for how long. This dissertation examines how research data is managed and shared in oceanography, focusing on the impact of data policies introduced by the United States (U.S.) National Science Foundation (NSF). Since 2011, the NSF has required data management plans (DMPs) as part of grant proposals with the aim of making research data available over time. While DMPs may be a relatively recent requirement, NSF data sharing policies for oceanographers can be traced back to the World Ocean Circulation Experiment (WOCE), 1990 - 2002. Oceanographers participating in WOCE were required to make their data available two years after collection. A requirement that at the time, was noted as a departure from traditional research practices. Nevertheless, this time norm for data sharing has endured, as it can be found in early NSF Division of Ocean Sciences (OCE) data policies from the 1990s and remains in OCE data policies today. And yet, oceanographers still report difficulties managing and sharing research data. This study explores what happens to data, researchers, and infrastructures under the requirements mandated in federal data policies. Through interviews and document analysis, this dissertation foregrounds the temporal dimensions of managing and making data available and makes the following three contributions. First, this study shows that data do not all follow 4 the same lifecycle as understood in prescriptive research data lifecycle models. Moreover, while data policy imagines universal availability, in practice, data persists unevenly over time. From DMPs, I identify three forms of planned data afterlives: secluded, splintered, and speculative, to describe how material conditions, disciplinary norms, and institutional arrangements shape what data endure, and how. Second, I explore the ways that researchers struggle to meet policy expectations for data availability. Researchers describe data management as tedious, timeconsuming, and hard to prioritize, even for those who understand its importance. Following Elizabeth Shove's argument that practices have distinct temporal characteristics, I argue that data management and sharing lack an established "practice-time profile". This absence leads to endloaded, last-minute efforts during the project's sunset phases, which can create minor to substantial delays in data sharing. At the same time, some researchers are developing practicetime profiles to better manage their data in preparation for sharing and preservation. Third, I introduce the concept of temporal paradox to describe how data infrastructures built for longterm access are often marked by short-term fragility. Building on Marisa Cohn's "convivial decay," I describe how researchers and data managers preemptively anticipate infrastructural demise, not only in aging systems, but also in relatively new ones. I articulate how this practice of planning for the end, paradoxically, supports the long-term persistence of data. Together, these findings contribute to information science, STS, and infrastructure studies through an empirical account of the temporal dynamics between data practices, infrastructures, and policy in the shortterm availability and long-term stewardship of scientific data.