Ship space to database: emerging infrastructures for studies of the deep subseafloor biosphere

Knowledge infrastructures

The term infrastructure is often used in reference to large-scale systems with multiple social and technical components that provide services, resources, and facilities (Edwards et al., 2009). Infrastructures, in the senses used herein, are “best understood as ecologies or complex adaptive systems” (Borgman, 2015, p. 33). Infrastructures are complex in the sense that they comprise multiple systems that may have been devised, built, and configured by different actors with varying objectives, but which function together. They are continuously adaptive in the sense that components change or are introduced (Edwards et al., 2013). A particular type of infrastructure is “knowledge infrastructures,” which Edwards (2010, p. 17) defined as “robust networks of people, artifacts, and institutions that generate, share, and maintain specific knowledge about the human and natural worlds.”

Infrastructure is a “fundamentally relational concept” in the sense that a sociotechnical configuration “becomes infrastructure in relation to organized practices” (Star & Ruhleder, 1996, p. 113). From the point of view of a particular domain, what is considered to be a knowledge infrastructure is precisely those configurations of social and technical components that provide resources to support a community’s shared objectives and practices. Indeed, as a domain’s interests and practices change, the knowledge infrastructure must adapt if the domain is to remain relevant (Ribes & Polk, 2015).

Single vs. multi-domain infrastructures

In some instances, a component of a knowledge infrastructure may serve one scientific domain alone. In other cases, a component serving one scientific domain may also serve as part of a knowledge infrastructure for others. Over time, these relationships tend to shift, requiring both infrastructures and domains to adapt, sometimes willingly, sometimes less so.

The Large Synoptic Survey Telescope (LSST), a telescope under construction with a projected budget of $1.1 billion, is an important example of a multi-domain infrastructure. LSST promises to provide data unprecedented in scale and scope for multiple domains within astronomy (LSST Science Collaboration et al., 2009). The process of building LSST requires negotiation between domains on decisions such as how to allocate scarce observing time during the planned ten years of data collection. This multiple-domain infrastructure, in turn, interacts with infrastructures serving single domains within astronomy. These include the Dark Energy Survey, and the Gaia Telescope, which focuses on the Milky Way (Dark Energy Survey, 2014; European Space Agency et al., 2016). These single-domain infrastructures were planned and began operations after significant funds were invested in LSST. The sociotechnical configurations of each project are shaped by LSST and vice versa (European Space Agency et al., 2016; LSST Science Collaboration et al., 2009). The authors of this paper also are involved in a case study of LSST (Borgman et al., 2015).

Examples abound of interactions between single- and multiple-domain knowledge infrastructures. One example is between infrastructure that serves researchers studying earthquakes in Southern California only (Southern California Earthquake Center , 2016) and infrastructure that serves multiple domains related to the formation of volcanoes in the continent of North America (Earthscope, 2016). A second example is between infrastructure intended to support a single domain studying coastal dynamics and multi-domain infrastructure that collects and distributes data about oceans, coastal regions, and the US Great Lakes (Center for Coastal Margin Observation & Prediction, 2015; NOAA, 2016).

Domains sharing infrastructure

While infrastructures that serve individual domains have received the most study, other important infrastructures span multiple, and often competing domains. Domain-spanning infrastructures often represent significant investment of public research funding and are critical sources of data for the research communities they serve. Among the few shared infrastructures receiving scholarly attention is Argos, whose study by Benson (2012) revealed how marine biologists were able to negotiate a share of its infrastructural resources. Argos is a satellite-based environmental surveillance system that provides data for oceanography, meteorology, and marine biology (Ortega, 2003). Marine biologists persuaded Argos leaders to collect data elements important to them by appealing to the interests of their commercial partners. In return, the biologists compromised other elements of their data collection methods to satisfy Argos partners in oceanography and meteorology.

Another perspective on negotiations in shared infrastructure is offered by Ribes & Finholt (2008), in their study of building infrastructure for studying the water environment. The forerunner to this infrastructure served a single community of researchers in environmental engineering, but the new infrastructure was intended to serve hydrological scientists as well. Ribes and Finholt show that the spokespeople of these domains negotiate infrastructure building more effectively when they represent a strong, clearly defined, and cohesive community. Various mechanisms for defining, canvassing the opinions of, and representing a community—such as organizing forums and conducting surveys—play an important role in facilitating these spokespeople’s effectiveness. Unfortunately, the effort to build this infrastructure fell apart before it began scientific operations (Jackson & Buyuktur, 2014).

These studies suggest the importance of studying how domains negotiate processes of building and accessing infrastructural resources shared with other domains, and how the configurations of these infrastructures shape the work and organization of individual communities.

Data scarcity and abundance

As science became more professionalized in the early twentieth century, more emphasis was placed on mathematical methods. These methods, in turn, require data generation and statistical confirmation (Bowler & Morus, 2010). That shift in focus endowed science with greater cultural authority as the primary knowledge-generating institution within society (Porter, 1996). The tight coupling of mathematization, data, and cultural authority helps to explain why domains that experience data scarcity are often so concerned with increasing their volume of data (Sawyer, 2008). For example, in molecular biology, increasing quantification and statistical inference were driven by “an ever-present methodological anxiety manifested in the constant search for an increased objectivity –or in its converse: the avoidance of subjectivity” (italics in original). These methodological changes both require and accommodate increasing quantities of data (Suárez-Díaz, & Anaya-Munoz, 2008, p. 452). By increasing quantification and data-intensive practices, communities can increase their scientific credibility (Hagen, 2003; Kay, 2000). Indeed, this increasing quantification frequently shapes, and is shaped by, hypothesis testing with laboratory-generated data (Lenoir, 1999; Paul, 2009).

Least studied are situations where “no data” exist, whether because no data were collected about a particular phenomenon, data may have existed but were not curated for longer term use, or data still exist but cannot be discovered or retrieved for a variety of reasons (Borgman, 2015).

Increasing data production

Social mechanisms can encourage greater production of data and use of statistical methods. As domains develop norms that promote data-intensive research, those who eschew such approaches may be marginalized (Keller, 1984). A domain’s quest for enhanced status may drive changes at the institutional level, even leading to the development of new domains. Natural historians in the early 20th century, in the face of profound challenges to their domain’s status, made alliances with the more data-intensive and mathematically based discipline of population genetics, forming the new discipline of evolutionary biology (Ceccarelli, 2001).

Domains sometimes address scarcity by producing large volumes of data. A notable example is cosmology, traditionally regarded as the poor relation to other domains of study in astronomy (such as lunar and solar astronomy) because they lacked sufficient data to support fundamental conjectures (Kragh, 1996). Cosmologists’ concern with this lowly status motivated the emergence of large telescope projects, known as sky surveys, that collect vast quantities of data about astronomical phenomena (Sloan Digital Sky Survey, 2016). Cosmology is now characterized by the use of data-intensive statistical methods (Strauss, 2014). Similarly, molecular biology addressed data scarcity by developing the Human Genome Project (Lenoir & Hays, 2000).

Gaining access to extant data

One way for a domain to address its data scarcity is to negotiate access to extant infrastructures, which often occurs in parallel with adopting computational- and data-intensive epistemologies (Chow-White & García-Sancho, 2012). Scientific databases are a means to organize and classify information, providing “a contemporary key to both state and scientific power” (Bowker, 2005, p. 108).

Infrastructure projects, in turn, may increase data production and circulation by shaping the behavior of scientists. Interoperability strategies include imposing or embedding common standards, such as what counts as reliable evidence, acceptable research methods, and data management practices (Bowker, 2005; Leonelli & Ankeny, 2012). Infrastructure can also foster norms of behavior that encourage greater openness within a scientific community, including openness around data (Leonelli, 2010). Contributions to a database can, in turn, encourage greater willingness to contribute to a database, resulting in a self-reinforcing cycle of increased data availability and normative shifts towards greater data openness. Databases can help encourage the circulation of types of data that the database does not support by creating expectations that scientists will share data when asked (Leonelli & Ankeny, 2012). Kelty (2012) discusses how scientific newsletters in biology not only constituted an infrastructure to build communities around particular organisms, but also promoted sharing of research objects by requiring openness of researchers as a condition of receiving the newsletter (and thus continued membership of the community).

Lastly, and perhaps most importantly for knowledge infrastructures, is to address scarcity by building infrastructure that aggregates extant data or databases (Meyer, 2009), or that integrates existing sites of data production into a single infrastructure (Aronova, Baker & Oreskes, 2010). These databases play critical roles in supporting research and in fostering data-intensive methods (Bowker, 2000; Leonelli, 2012). One example is the Long Term Ecological Research Network, whose infrastructure attempts to improve the management and accessibility of data produced by distributed ecological research stations. This network originated in the efforts of biologists to leverage “Big Science” funding opportunities by collecting and organizing large-scale datasets (Aronova, Baker & Oreskes, 2010).

However, a major barrier to data circulation and integration is the use of disparate standards by individual scientists, a problem particularly acute when scientists come from different disciplines or communities of practice (Baker, Jackson & Wanetick, 2005; Bietz & Lee, 2009; Borgman et al., 2015; Borgman et al., 2007). Scientists’ concerns about control of data, authorship rights, and incentives to undertake the work necessary to make their data shareable also limit the adoption of these infrastructures (Borgman, 2015; Borgman, Wallis & Enyedy 2007).

Research questions

Data scarcity is a critical problem for many scientific communities. Much richer accounts are needed of how domains employ infrastructural approaches to data scarcity. One approach is to emphasize scarce resources when negotiating access to infrastructure. As discussed above, competition arises when an infrastructure attempts to accommodate all types and forms of data present in the participating communities, particularly in the face of conflicting data standards and formats. A second approach is to consider relationships between an infrastructure and others with which it overlaps. A single domain may rely upon multiple infrastructures simultaneously. A domain that builds an infrastructure to increase its flow of data may be expressing dissatisfaction with the existing infrastructure. New infrastructure is likely to address the shortcomings and exploit affordances of existing infrastructures. These considerations motivate our three research questions:

1. How do scientific domains experience data scarcity?

2. How does a scientific domain address data scarcity through developing knowledge infrastructures?

3. What interactions occur between infrastructure specific to a domain and infrastructure shared with other domains?

Data collection

A key feature of this case study is long-term ethnographic observation of C-DEBI (Hammersley & Atkinson, 2007). We were embedded for eight months in a laboratory headed by a leading figure in C-DEBI at a large US research university, which involved one of the authors visiting the laboratory for two or three days per week to observe bench work and laboratory meetings. The first author also conducted weeklong observational work in two other participating laboratories in the US and joined researchers on a three-day field research expedition. We made extensive notes about what we observed, including the physical layout of the laboratories and laboratory benches, tools and methods used, and patterns of collaboration. Our informants told us about their backgrounds, aspirations, and experiences in the laboratories and on scientific cruises.

These organizations are distributed across multiple institutions and countries, which posed issues of scalability for the researcher (Star, 1999). The work of C-DEBI and IODP/IODP2 spans more sites than a small team of researchers can visit, much less meet face-to-face with all personnel. We focused on the techniques and technologies—the “scalar devices”—employed by our research subjects to understand C-DEBI, IODP/IODP2, and their domain (Ribes, 2014, p. 158).

One scalar device that we observed was gatherings such as the C-DEBI All-Hands’ Meeting and research workshops. A larger gathering was the American Geophysical Union Fall Meeting 2013 in San Francisco, a major conference for the deep subseafloor biosphere community and IODP/IODP2; we also presented our early findings at this conference. These events enabled our research subjects to take stock of the scale of the communities and infrastructures in which they are embedded, in terms of the people involved, organizational hierarchies, and the range of scientific work conducted. Another form of scalar devices is reports and workplans, which we studied as part of document analysis (see below).

The distributed nature of C-DEBI and IODP/IODP2 also means that work in these organizations often takes place through communications media. By using multiple forms of media, we could establish “co-presence” when “co-location” was not possible (Beaulieu, 2010). Co-presence involves the researcher witnessing how the work of scientific collaborations is conducted even when they are not physically (or necessarily temporally) collocated with the subjects of research. Lacking the opportunity to observe practices on board an IODP cruise, given the expense and limited places available, we studied IODP work conducted elsewhere. We attended online meetings and seminars where participation and data collection were planned, and watched a feature-length documentary that used footage from a deep subseafloor biosphere-focused cruise. Other online observations included workshops, meetings where key C-DEBI personnel planned infrastructure to coordinate data management across the project, and websites of organizations and people.

We assembled a corpus of documents for analysis. Documents such as instruction manuals for laboratory equipment help to explain the work conducted by C-DEBI-affiliated scientists in their laboratories. Other documents help us to interpret contexts in which C-DEBI scientists operate. These include official C-DEBI documents such as the funding proposal, Annual Reports to the NSF, and operating documents for C-DEBI and IODP/IODP2. These documents function as scalar devices by giving details and metrics about activities, plans, and available infrastructural resources.

Our research also draws heavily on semi-structured interviews; the sample for this article consists of 49 people, which includes C-DEBI-affiliated scientists, curators and managerial staff involved in IODP/IODP2, as detailed in Table 1. The column Involved in IODP indicates which C-DEBI interviewees are involved in decision- or policy-making in IODP/IODP2. These interviewees are further split into two groups: those in cruise operations, and those with the Consortium for Ocean Leadership, responsible for administering US involvement in IODP (but not IODP2). Interviews ranged in length from 35 min to two and one-half hours, with the majority being between one and two hours long. With the consent of the interviewees, interviews were recorded and sent to an external company for transcription.

C-DEBI interviewees were initially recruited from those scientists being observed in the laboratory, and were typically interviewed after an extended period of observation. Other C-DEBI interviewees were recruited from those who had been awarded C-DEBI-funded grants, with these interviews typically taking place over Skype. We have interviewed undergraduate and graduate students, postdoctoral researchers, faculty members, and non-scientists with senior level roles in administering and operating C-DEBI. IODP interviewees were identified and approached through a range of methods, including personal introductions from C-DEBI-affiliated scientists and other IODP personnel, and from public websites.

Our interviews covered a range of topics, including interviewees’ backgrounds and career trajectories. We asked scientists and technical staff detailed questions about the scientific work they are undertaking and the importance and role of data in their work. We also asked IODP/IODP2 technical staff and C-DEBI scientists who have participated in IODP cruises about their work on board expeditions, how they negotiate for access to cruise resources, and how they transfer data, methods, techniques, and collaborative networks between cruises and their onshore laboratories. Non-scientists were asked about their roles in building and administering C-DEBI and IODP/IODP2 infrastructure.

Where interview quotations are used in this paper, we add a code in parentheses after each quote indicating whether they are IODP or C-DEBI, their career stage, and a number unique to each interviewee: (IODP curator, #1) or (C-DEBI faculty, #3) etc.

Data analysis

Our initial data analysis involved close reading of our ethnographic notes, interview transcripts, and documents. Based on these readings, we identified emerging themes about the relational, complex, and dynamic nature of knowledge infrastructures, and coded our data accordingly. In particular, we focused on themes relating to how those we interviewed described their own work (scientific, organizational, building infrastructure); how they identified and defined communities of which they considered themselves members; what resources, both current and anticipated, they identified as necessary to their own work and to deep subseafloor biosphere research as a whole; what they considered to be infrastructure; and how they and their community engaged with other scientific communities to negotiate, access, and build infrastructure. We refined our coding scheme iteratively, going back and forth between our scheme and the data. Using a range of sources enables us to triangulate, cross-checking our data to validate our findings (O’Donoghue & Punch, 2004).

We began our data analysis after completing approximately six months of laboratory observation and fifteen interviews. We have thus been able to strike a balance between ensuring that our observations have not been biased by preconceived ideas and being able to assess our emerging findings and tentative hypotheses against further observations. We presented our emerging findings to the deep subseafloor biosphere research community at major scientific meetings (see above) for feedback and clarification.

Data scarcity

Complaints by deep subseafloor biosphere scientists about the “dearth of data” for their core research questions led to the founding of C-DEBI (Edwards, 2009: 5). Relevant data still exist only for a few sites in the ocean and, compared to studies of microbial ecology in other environments, is about relatively basic phenomena.

Two reasons emerged for the community’s continuing concern with data scarcity. The first is constraints on scientists’ ability to pursue their field’s immediate research objectives, which are to characterize deep subseafloor microbial communities in terms of the quantity and types of microbes that exist, how these microbes interact with the physical environment they inhabit, and how microbial communities vary between geographic sites on the seafloor (Orcutt et al., 2013). As a C-DEBI report stated, “Evidence for microbial alteration [of the physical environment] exists, yet scientists lack robust molecular, biochemical, or physiological data so needed” (Center for Dark Energy Biosphere Investigations, 2011, p. 11).

The second concern about data scarcity relates to the status of deep subseafloor biosphere research relative to studies of microbial ecology in other environments, and thus for their ability to attract future resources and funding. In the words of many scientists that we studied, research on the deep subseafloor biosphere is largely exploratory or discovery-driven, while research on microbial ecology in other environments is largely hypothesis-driven:

“Our work in the lab in general tends to be classified as rather exploratory as opposed to hypothesis-driven. This is something…that I met researchers who take issue with because they insist that to be a true science, proper science, you need to have a question and then you need to have a methodology that will either answer ‘yes’ or ‘no’, or some number. Whereas, our kind of science is oftentimes, it’s more like, ‘I wonder if...’ And then you try something and the results are occasionally interesting, and then you go, ‘Look what I found.’ You didn’t know what you were looking for, you just cast a big net out.” (C-DEBI graduate student, #1)

As some C-DEBI scientists stated the problem, if the approach of deep subseafloor biosphere researchers has a lower scientific status than studies of microbial ecology in other environments, they will receive fewer or smaller grants because “funding agencies will rarely fund basic discovery science” (Teske, Biddle & Schrenk, 2011, p. 9).

As the deep subseafloor biosphere emerged as a domain of study, researchers adopted a strategy of building and leveraging infrastructure to acquire more data. One of their strategies was to build infrastructure specifically for deep subseafloor biosphere researchers, first by establishing C-DEBI. The second strategy was to use C-DEBI as a means to gain greater access to, and to reconfigure, IODP2 for their advantage. Notably for our research questions, IODP/IODP2 is an infrastructure that deep subseafloor biosphere research shares with other scientific domains.

IODP/IODP2 provides the requisite infrastructure for the C-DEBI community to access the geographic sites and to acquire the physical samples necessary to produce data about the deep subseafloor biosphere. C-DEBI was initially established as a way to consolidate the position of the deep subseafloor biosphere within IODP and to recruit new researchers into the domain. Over time, C-DEBI also began to build its own infrastructure to respond to the limitations of IODP in providing data, and to reconfigure the IODP2 infrastructure so that a greater share of IODP2 resources will be allocated to deep subseafloor biosphere researchers, thus increasing their supply of data.

Ocean drilling meets deep subseafloor biosphere research

Interest in deep subseafloor microbial life that emerged in the late 1990s coincided with institutional challenges facing scientific ocean drilling programs. The predecessors to IODP that ran from 1968 to 2003 focused almost exclusively on physical science research. These programs facilitated major scientific successes. Best known is the evidence for the theory of plate tectonics and continental drift ( Committee on the Review of the Scientific Accomplishments and Assessment of the Potential for Future Transformative Discoveries with US-Supported Scientific Ocean Drilling, 2012). Many scientists were concerned that funding for ocean drilling would cease with the anticipated end of the Ocean Drilling Program, IODP’s immediate predecessor, in 2003. Expanding the drilling mission to include the deep subseafloor biosphere provided the momentum necessary to secure funding for IODP to launch in 2003. One of our interviewees, a senior administrator within IODP, quoted Admiral Watkins, who then headed the Joint Oceanographic Institutions, “I can remember …him saying, ‘Give me bugs and I can give you a new program” (IODP policy, #1).

C-DEBI as a single-domain infrastructure

To reconstruct the origins of C-DEBI as an infrastructure for the emergent community of deep subseafloor biosphere researchers, we drew heavily upon documentary sources to complement our interviews and ethnography. The round of proposals for the 2003 launch of IODP was a critical inflection point. Deep subseafloor biosphere research was one of three major scientific foci for IODP ( Integrated Ocean Drilling Program (IODP) Planning Sub Committee (IPSC) Scientific Planning Working Group, 2001), which planned four to five expeditions per year. Proposals were required to state the scientific objectives of the cruise and to identify the sites a cruise would visit. Three separate teams of scientists independently, and successfully, submitted proposals (in 2003, 2005, and 2007 respectively) for expeditions focused primarily on the deep subseafloor biosphere (although an IODP/IODP2 cruise will typically have a focus on one particular scientific domain, it will nevertheless involve scientists from all domains of study represented within IODP/IODP2).

Rather than work independently, the successful teams joined forces to coordinate the three biosphere-focused IODP cruises, which were planned for 2010 and 2011. To consolidate the position of deep subseafloor biosphere research within IODP, in 2008 they established the Dark Energy Biosphere Investigations Research Coordination Network. This network had four specific goals (Edwards & Amend, 2008, p. 3):

• “Develop an interactive community of deep-biosphere researchers

• Facilitate coordination of science between deep-biosphere drilling projects

• Stimulate interaction and education among disparate disciplines

• Enable synthesis and integration of data and technology advances”

This network brought together scientists in regular face-to-face meetings, with the proposal to the NSF for C-DEBI in 2009 arising from one such meeting. C-DEBI, as introduced above, is a Science and Technology Center (STC) funded by the US National Science Foundation (NSF), initially for five years from 2010–2015, and subsequently renewed for another five years from 2015–2020. The proposal to establish C-DEBI set out two major goals. One is “to coordinate, integrate, support, and extend the science” of deep subseafloor biosphere research, and the second to help “foster and educate an interdisciplinary community of deep subseafloor biosphere researchers, with a focus on students and junior researchers” (Edwards, 2009: 1). The focus on students and junior researchers highlights the aspiration to establish an enduring community. The significance of this long-term view was highlighted, sadly, by the deaths of two of the five founding co-investigators during the first five years of C-DEBI. It is a tribute to their vision that the collaboration was sufficiently robust to reorganize with new investigators and to win the second five-year award.

C-DEBI has developed social, technical, and policy means to pursue its aims, including funding support for scientists, a website and meetings to circulate knowledge and to bring together globally distributed researchers, and an infrastructure for data management that continues to evolve. In its early years, C-DEBI focused more on recruiting scientists to study the deep subseafloor biosphere than on building technical or policy infrastructure. Recruiting occurs by distributing grants to support small-team (usually two or three persons) laboratory-based research projects (typically one to three years in length), or graduate and postdoctoral fellowships. These grants typically support projects in which scientists use cores collected from scientific ocean drilling cruises to produce new data in their onshore laboratories, and to support projects that develop new techniques to study the deep subseafloor biosphere. To date, nearly 90 such grants have been distributed across more than 40 institutions in the USA (Center for Dark Energy Biosphere Investigations, 2016). However, in the early years of C-DEBI, these grants were not accompanied by a strategy for managing the data produced by the funded projects.

C-DEBI data portal

Although the C-DEBI proposal aspired to build data management infrastructure, stating that “C-DEBI will develop and maintain a website for public access and data sharing among the C-DEBI research community” (Edwards, 2009: 24), little was accomplished toward this goal in the first two years of operation. The first five-year Strategic Implementation Plan 2010–2015 claimed “technical difficulty” as the barrier to establishing the data management infrastructure (Center for Dark Energy Biosphere Investigations, 2010; 11).

However, by 2012, data management infrastructure became part of the work of C-DEBI. This plan stresses that the “C-DEBI STC is committed to open access for all information and data gathered during scientific research that is conducted as part of C-DEBI” (Center for Dark Energy Biosphere Investigations, 2012a: 1). Starting in 2012, C-DEBI responded to new NSF requirements by mandating that participating scientists must make data publicly available after a moratorium, typically two years. C-DEBI developed their data management strategy as part of their application for renewal of NSF funding for the second five-year period, 2015–2020. The most critical strategic challenge for C-DEBI during its first years of operation was to navigate this NSF renewal process successfully.

The C-DEBI Data Portal, comprising a registry and repository, is a central element of C-DEBI’s strategic goals for their second five-year project (Center for Dark Energy Biosphere Investigations, 2014a). C-DEBI requires participating teams to register all associated datasets that support published results on the portal, and to deposit them in a relevant, online, publicly accessible database. The portal contains metadata about each dataset, including details about the provenance of the cores used (including the name of the research site, the cruise number, and the specific drill hole(s) where the samples originated), and a link to the dataset if it is hosted externally.

In mid-2013, C-DEBI assembled a team to build the data portal, which included one co-PI, a scientific database expert from another C-DEBI institution, and the C-DEBI Administrative Assistant. Since the Center’s inception, the Administrative Assistant had responsibility for building and maintaining the C-DEBI project website. His job was expanded to include leading the development of the data portal. C-DEBI allocated substantial funding to portal development: $95,000 in 2013 and an additional $287,000 for 2014 (Center for Dark Energy Biosphere Investigations, 2014b). The first phase, which included prototyping and site architecture, was completed for the NSF visit to C-DEBI in January 2014. Subsequently, the job title of the Administrative Assistant was changed to Data Manager, reflecting the shift in the importance to C-DEBI of data management infrastructure from a marginalized aspiration to a central goal.

Alternative approaches to single-domain infrastructure

The existence of the data portal, and the form it takes, is only partially determined by NSF requirements. To determine the degree to which the form of C-DEBI data management infrastructure is driven by NSF data management requirements, we examined how other NSF Science and Technology Centers have addressed these requirements. C-DEBI is one of 11 current STCs, all of which are subject to NSF data release requirements. However, the STCs interpret these requirements in a variety of ways. Hence, only three of the other ten STCs operate an online, publicly-accessible registry or repository to make these data accessible, whether by downloading datasets, by providing links to other sources, or by providing contact information for people associated with datasets (Center for Coastal Margin Observation & Prediction, 2015; Center for Microbial Oceanography Research and Education, 2015a; Center for Remote Sensing of Ice Sheets, 2015).

The data registries or repositories of these STCs vary in the types of datasets they contain and in the scope of metadata they capture about each dataset. The Center for Microbial Oceanography Research and Education, or C-MORE, is the STC most scientifically comparable to C-DEBI, in that they study microbial ecology in the ocean and scientists use samples collected from ocean research cruises operated directly by C-MORE.

C-DEBI is distinguished from other STCs with online data infrastructure by requiring the most comprehensive set of metadata for datasets in its registry and repository. For example, while both the C-DEBI and C-MORE registries have a category to name the research cruise from which each dataset was derived (Center for Microbial Oceanography Research and Education, 2015b), C-DEBI also has a category for the precise geographic location from where the sample was drawn. Another example is that the C-DEBI registry has metadata categories detailing the publication(s) in which a particular dataset has been used; by contrast, the C-MORE registry does not.

The current STCs have implemented the NSF requirements in different ways, despite their scientific and organizational similarities. In other words, the existence of the NSF requirements do not completely account for what data management policies or infrastructure are implemented by C-DEBI, nor do they account for the specific form and function of these policies and infrastructure.

Shared infrastructure and community building

IODP/IODP2 expeditions are the primary source of physical samples and data used in the onshore laboratories of the deep subseafloor biosphere researchers we studied. Some scientists also participate in cruises operated by other organizations (University-National Oceanographic Laboratory Services, 2015). However, these non-IODP/IODP2 cruises usually revisit sites drilled during previous IODP/IODP2 expeditions.

International ocean drilling programs

IODP/IODP2 is infrastructure, consisting of ships, scientific equipment and personnel, and an organizational structure shared between the deep subseafloor biosphere community and several other domains. Consequently, these domains compete for scarce resources. Their representatives are involved in decision-making processes within IODP/IODP2 about the scope of their programmatic research and the organization of specific cruises. The expansion of IODP scientific activities to include deep subseafloor biosphere research required other domains to concede some of their share of IODP resources. While our C-DEBI participants generally reported a collegial atmosphere among scientists on IODP/IODP2 cruises, many deep subseafloor biosphere researchers nevertheless encounter resistance, such as objections to the number of cores allocated to deep subseafloor biosphere researchers:

“We encounter resistance [from other domains] when we apply to sail. We encounter it when we apply for sample requests. We encounter it when we set up for post-cruise fundings, and also for regular grant writing.” (C-DEBI faculty, #1)

Increased competition for places on cruises is but one of the ways in which other domains concede resources to accommodate research about the deep subseafloor biosphere. A wider variety of domains also compete for allocation of cores, the precious sources of physical and microbiological data from cruises. Initial decisions made on board the ships determine the core’s value to these competing communities. The most critical initial decision for the microbiology community is the temperature at which cores are stored. While samples for physical analysis are typically stored at −4 °C, samples for microbiological analysis are typically stored at −80 °C to avoid contamination and to stabilize biological material. Other handling decisions include the ways in which the cores are cut and distributed to participating scientific teams. The number of cores per cruise is finite, therefore producing more cores suitable for microbiological analysis results in fewer cores suitable for physical science analysis:

“What you do with this core you just split these one meter sections lengthwise, open it up so you have two halves of it…. For microbiology and geochemistry you do it somewhat differently. You take the core, you’re not cutting it up lengthwise but you cut out short sections...So you lose the entire stratigraphic information from that core.” (C-DEBI faculty, #2)

The allocation of cores to research teams requires intensive negotiation. Pre-cruise meetings to negotiate allocations result in a Sampling Plan for the voyage. During a cruise, Sampling Plans frequently are adapted to changing conditions and to the relative success of core sampling, leading to further negotiation of resources:

“Sometimes …we don’t receive that much [biological] core material. Sometimes you may have a small piece and 15 people want something from the small piece so then that has to go through iterations of compromise.” (IODP curator, #1)

As a new area of science, a particular challenge for deep subseafloor biosphere researchers is their own methodological diversity. The lack of agreement within the deep subseafloor biosphere research community on standard practices for data handling works against them in negotiating for more cores. As discussed in more depth elsewhere (Darch & Sands, 2015), workflows to characterize microbiological communities in cores vary significantly, even between scientists working on adjacent benches in the same laboratory. Scientists from other domains sometimes use this variation to argue against allocation of cores to deep subseafloor biosphere researchers, as one of our interviewees encountered:

“Those that are competing with us for sediment material, the hard rocks guys, the sedimentologists, those guys that are then lobbying for the same sediment samples that we’re going after, they can turn to us and say, ‘Well, you know what, if we handed you half of this and you half of this, you guys would come back with two different datasets, so what’s the point of handing it to any of you because you guys can’t describe it in the same way anyway?’ ” (C-DEBI faculty, #1)

As IODP approached the end of its funding period in 2013, concerns arose among stakeholders about continued government funding for scientific ocean drilling cruises, and indeed, funding was reduced for IODP’s successor, IODP2 (Committee on Guidance for NSF on National Ocean Science Research Priorities: Decadal Survey for Ocean Sciences, 2015). However, the physical infrastructure (cruise ships, core repositories, data management systems) of IODP2 is largely that of IODP.

In addition to maintaining their position within IODP2, deep subseafloor biosphere researchers also aspire to secure more resources for deep subseafloor biosphere research in the future. One aspiration is to produce more standardized data on board the ocean drilling cruises, akin to the standardized set of analyses of physical properties that are routinely conducted and made publicly available through an IODP2 database. As no comparable set of standards exists for the microbiological properties, cruises neither conduct nor report basic microbiological descriptions of cores. Many of the C-DEBI-affiliated scientists interviewed mentioned this lack of agreement around standardized methods as a serious constraint on their scientific progress (Orcutt et al., 2013). Instead, individual scientists devote much effort to basic microbiological analyses in their home laboratories. The time and expense of basic description limits the resources available to conduct more advanced analyses, as explained by one of our interviewees:

“Post-expeditions Awards provide $15,000 worth of support for up to two years, for you to do the research that you proposed to do while you were at sea …The difference between what we can use that money for, and say what a sedimentologist can use that money for, is grossly different. Because a sedimentologist, the geochemist, the petrologist, the paleo-mag guys, all of them pretty much have all the data. And so, they’re looking at the $15,000 as seed money to maybe do some analysis that they maybe pay for a grad student, maybe pay for a technician, maybe pay for somebody’s time, to analyze it, to maybe take it another direction…For the biologist, we have $15,000 to now process all of our samples, do all the sequence analysis, do the bulk labor of all of our work on the equipment that we already have to have in our lab versus what everybody else is using on the ship.” (C-DEBI faculty, #1)

However, to include microbiological description of cores on board all cruises would require major reconfiguration of existing cruise practices. These practices were established over many years before microbiology’s inclusion in cruises:

“It took them decades to come up with the system... standard protocols, standard procedures, standard storage. It makes it a little bit rigid like I said, when you do something new and novel, like the living things don’t really have a place yet.” (C-DEBI faculty, #4)

Consequently, attempts to change IODP2 practices in support of deep subseafloor biosphere research would require a significant amount of effort and would diminish the resources available for more physical science-oriented domains in IODP2. To date, these attempts have faced considerable resistance:

“Geochemical and geophysical research objectives represented on IODP expeditions are routinely provided by dedicated shipboard scientists and technicians assigned to completing standard procedures on all core material. The call for an additional biological workload on these individuals is typically met with an argument claiming a lack of time and resources on board.” (Orcutt et al., 2013, p. 8)

One of our interviewees told us that much of the resistance from these physical scientists is that, “as a community [of deep subseafloor researchers], we can’t agree on anything” regarding methodology. While this resistance on the part of physical science disciplines appears to be motivated by a fear of losing IODP resources, their arguments are that the lack of standardization of microbiological workflows means that microbiological analyses cannot be included as part of the workflow on IODP cruises.

Consequently, many C-DEBI scientists recognize that the deep subseafloor biosphere community must undertake the work necessary to standardize microbiological workflows.

Infrastructure convergence

The development of the C-DEBI data management infrastructure can be understood in the light of the aspirations of deep subseafloor biosphere researchers to improve exploitation of currently-available resources from IODP/IODP2 cruises, and to reconfigure IODP2 infrastructure to secure a greater share of drilling cruise resources. Those developing C-DEBI infrastructure are working towards three goals: better curation and circulation of extant data; building a community with norms of open data; and explicating the roles of IODP/IODP2 data in C-DEBI research. Here we examine how C-DEBI is addressing these three goals, and how the goals both shape, and are shaped by, the relationships between C-DEBI and IODP/IODP2 infrastructure.

Data scarcity

Terms like “big data” and “little data” are commonly employed to denote the scale of data to which scientific domains have access (Borgman, 2015). “Data scarcity” is a more poignant term as it suggests a state that is unsatisfactory for a domain’s practitioners (Sawyer, 2008). The degree of data scarcity can be understood only in relation to that domain’s scientific objectives.

As emergent scientific domains such as the deep subseafloor biosphere struggle to survive and thrive, they must attract resources to support researchers and infrastructure. To justify support for public funding in highly competitive environments, scientific domains must demonstrate their utility by contributing to one or both of the following:

• 1.

  Issues of major political and social concern (Mukerji, 2014), such as those relating to the environment, agriculture, and national defense;

• 2.

  Existential questions about humanity, such as the origins and evolution of life, and the origins and extent of the universe (Bowler & Morus, 2010).

C-DEBI makes both claims: study of the deep subseafloor biosphere will contribute significantly to understanding the effects of global environmental change, and to the origins and evolution of life. The deep subseafloor biosphere domain faces data scarcity because it aspires to pursue its scientific objectives in a more statistically intensive manner than is afforded by the data to which it currently has access. For instance, domain scientists would like to answer questions about the global distribution of microbes, which is essential to understanding the role of the deep subseafloor biosphere in important environmental processes. To answer these questions, scientists must be able to perform meta-analyses that involve aggregating datasets about the size and composition of microbial communities in many different sites of the ocean. At present, insufficient data currently exists to make accurate estimates about the global distribution of microbes.

Leaders of the deep subseafloor biosphere community wish to shift the research emphasis from discovery to hypothesis-driven science, bringing their domain into line with other domains of microbiology. Hypothesis-driven science, involving statistical methods to test hypotheses, is generally regarded as more scientifically credible than discovery-driven science (Lenoir, 1999; Paul, 2009). These scientists wish to test the effects of environmental changes on the ability of different types of microbes to survive and thrive. Others wish to study microbes’ abilities to survive and adapt to extreme conditions, such as those that may have been present when life on earth began.

Knowledge infrastructures

Our account of C-DEBI infrastructure is enriched by understanding relationships between this single-domain infrastructure and the complexities of IODP/IODP2, and by examining how infrastructure negotiations influence access to resources. We observe a mutual shaping of the single-domain and shared infrastructure, driven by the deep subseafloor biosphere researchers’ desire to address their data scarcity.

The case presented in this paper contributes to studies of knowledge infrastructures (Edwards, 2010; Edwards et al., 2013) in two respects. First, although infrastructure components often are shared and negotiated between multiple domains, surprisingly few studies have paid close attention to the difficulties of sharing infrastructural resources (Benson, 2012; Ribes & Bowker, 2008; Ribes & Finholt, 2008). Our study pays close attention to how scientists from the deep subseafloor biosphere have negotiated a share of scarce resources of IODP/IODP2, thus extending research on negotiating shared infrastructure.

Second, very few studies offer accounts of how an infrastructure emerges in relation to other infrastructures upon which a domain may depend. The case presented in this article is an exemplar of configurations of knowledge infrastructure common to many domains of science. Here, the knowledge infrastructure of the deep subseafloor biosphere includes a major component shared with other domains, and another major component that it wholly controls. Further, this case demonstrates how building a single-domain infrastructure unfolds both in response to, and as a significant intervention in, the configuration of shared infrastructure. Although building the single-domain C-DEBI infrastructure may appear intended to provide immediate resources to scientists, it is also a means to access a greater portion of the shared infrastructure of IODP/IODP2.

How domains share or build their own infrastructures

Infrastructural approaches are commonly used to gain access to data resources, and to facilitate shifts toward more data-intensive epistemologies (Bowker, 2000; Bowker, 2005; Chow-White & García-Sancho, 2012). Scientific domains can engage in infrastructure-building activities to increase opportunities for producing, analyzing, aggregating, accessing, circulating, and long-term curating of data. Existing studies suggest these activities follow one of two strategies. The strategy most widely studied involves a domain that builds infrastructure intended exclusively for itself. The second strategy, documented in a few studies, involves a domain that constructs infrastructure to share with other domains, either by building new multi-domain infrastructure or by negotiating access to extant infrastructure that already serves other domains (Benson, 2012; Ribes & Bowker, 2008; Ribes & Finholt, 2008).

One contribution of our study is to demonstrate that a single scientific domain may pursue both strategies simultaneously. The deep subseafloor biosphere scientists constructed infrastructure specific to the domain (C-DEBI and its Data Portal) and negotiated greater access to infrastructure shared with other domains (ocean drilling programs such as IODP/IODP2).

The single-domain C-DEBI infrastructure was built in response to the constraints and opportunities of the multi-domain IODP/IODP2 infrastructure. When multiple domains share infrastructure, they compete for elements of design and operation that serve their needs, such as choices of what data are to be collected and what standards are to be applied. Infrastructure is built on an installed base, so that adaptations are both afforded and constrained by the configurations of extant infrastructure (Star & Ruhleder, 1996). Hence, when a scientific domain first seeks access to an infrastructure controlled by other domains, they may need to adapt their scientific practices accordingly (Benson, 2012). This was the situation faced by deep subseafloor biosphere scientists in gaining access to the IODP/IODP2 infrastructure, which had established procedures for collecting, curating, and accessing samples and data; for ship-based facilities; and had favored geographic locations for scientific ocean-drilling cruises. These pre-existing practices helped to shape and constrain the scientific research opportunities of deep subseafloor biosphere scientists.

One way to acquire more data is to distribute funding accordingly. Other studies have observed cases where infrastructure is designed to increase data production (Lenoir & Hays, 2000; Strauss, 2014) or to coordinate distributed sites of data collection (Aronova, Baker & Oreskes, 2010). The motivations of C-DEBI are similar, with top priority assigned to exploiting cores whose allocation between domains is hotly contested.

A second way to maximize scientific work is to aggregate and integrate existing data more effectively (Leonelli & Ankeny, 2012; Meyer, 2009), for instance by embedding common standards within or across domains (Bowker, 2005; Leonelli & Ankeny, 2012) or by building links between members of the domain and fostering norms of data sharing and openness across this domain (Kelty, 2012; Leonelli, 2010). The C-DEBI data infrastructure exists to foster collaboration among other deep subseafloor biosphere researchers, and to assemble and circulate data produced by distributed domain scientists.

Building and converging infrastructures

The single-domain C-DEBI infrastructure was designed in response to scarce IODP/IODP2 resources for deep subseafloor biosphere science. However, C-DEBI also can be understood as a means to negotiate a greater share of the IODP/IODP2 infrastructure.

Negotiations between domain representatives are a typical feature of projects to build shared infrastructure (Ribes & Bowker, 2008; Ribes & Finholt, 2008). Ribes & Finholt (2008) focus on the importance of defining and representing the interests of the communities involved. C-DEBI, and its associated data infrastructure, exemplifies such a strategy. It was formed to build a strong, enduring community that speaks with one voice, creating a larger presence of deep subseafloor biosphere scientists in IODP2 negotiations.

One way that C-DEBI increases the domain’s status is by making explicit the contribution of IODP/IODP2 resources to their research. The C-DEBI Data Portal, and the choices of metadata categories, provide evidence that deep subseafloor biosphere researchers can use to negotiate further involvement in IODP2. A second way is to enable meta-analyses and cross-comparisons of methods that promote methodological standardization across the domain, with the goal of making microbiological workflows standard practice on future cruises. This standardization, in turn, is deemed necessary by many C-DEBI scientists to reconfigure how IODP2 cruises operate in the future, and thus to secure more data for their community.

Our case also demonstrates how relationships between single-domain and multi-domain infrastructure change over time. The infrastructure studied by (Ribes & Finholt, 2008), namely WATERS, fell apart even before it began its scientific operations. Our study, on the other hand, exposes how the configuration of C-DEBI, and the priorities of those building C-DEBI infrastructure, has shifted. In the early years of C-DEBI, distributing grants to enable data production and to recruit more scientists into the community was very much the priority. Over time, C-DEBI’s priorities changed, following from experiences in negotiating with other domains for resources during the three biosphere-focused IODP expeditions in 2010 and 2011. Biosphere scientists realized both that standardizing microbiological workflows and making explicit how they are using cores to produce data are critical challenges for acquiring more data from future drilling cruises. In turn, this awareness promoted a greater focus of C-DEBI on building and configuring online data management infrastructure. Single-domain infrastructure is subject to change and reconfiguration as domain scientists gain more experience of, and become increasingly sophisticated operators in, using shared infrastructure. In turn, these shifts will change the configurations of resources available to scientists as they seek to go about their day-to-day work.