Changes in Ecology: Expanding the Scope of Ecological Research

The history of ecology is marked by an expansion in the scope of ecological research. Bocking (2010) presents the history of the concept of ecosystem science as a new approach in 1935 to the study of ecology, considering the flow of energy and nutrients in natural systems as complementary to the earlier studies of population and communities. It was followed by the development of an ethics-oriented notion of land stewardship. Further expansion of ecological insight occurred with the recognition of “coupled systems,” particularly human and natural systems (e.g., Liu et al., 2007). New funding opportunities were developed to support collaborative research including multi-institutional research, centers of excellence, and research coordination networks. Big ecology became evident with global-scale programs such as the International Biological Program (IBP), a program running from 1964–1974 that coordinated internationally (Aronova, Baker, & Oreskes, 2010; Kwa, 1987) as well as through development of ecological modeling (Jopp, Reuter, & Breckling, 2011; Shugart, 1976) and satellite remote sensing of the environment (Kwa, 2005; Kwa & Rector, 2010). The IBP was followed by various programs with new approaches that emphasized interdisciplinary, large-scale, and long-term perspectives in addition to collaboration and technology. In the United States these included the Long-Term Ecological Research (LTER) program, a site-based network made up of member sites established in 1980 (Hobbie, 2003; Waide & Thomas, 2013), and the Critical Zone Observatories, established in 2007 as laboratories focusing on the processes shaping the earth’s surface (Anderson, Bales, & Duffy, 2008). In addition, the National Ecological Observatory Network started construction on a continental-scale 30-year ecological observatory platform in 2011, which focused on a network of instrumented field towers (Loescher, Kelly, & Lea, 2017; Zimmerman & Nardi, 2010). The new approaches prompted the development of new kinds of data practices and digital arrangements, specifically infrastructures for data.

Development of New Data Practices and Data Infrastructures

Ecological research has specific needs in terms of working with data. Synthesis of diverse data and knowledge is at the heart of ecology—an integrative discipline poised to address environmental challenges (Carpenter et al., 2009). Ecological research activities incorporate analysis and integration of many kinds of disciplinary data across space and time. Ecology is a field that has expanded from a small science focus to include big science. Today, ecological researchers study multicomponent, dynamic systems with associated digital data environments that are also dynamic and subject to nonlinearities and feedbacks (Harvey et al., 2017; Parsons et al., 2011). With increasing digital data and data capabilities, the concept of data infrastructure emerged.¹

Digital data activities associated with short-term funded projects and instrument use expanded in the latter part of the 20th century to include interdisciplinary collaborations, data-rich digital collections, and platform design including not only satellites but also in-situ sensors. The increase in data and growth of data work (i.e., any efforts applied to data) has led to the development of new data practices and new kinds of infrastructures for assembling, processing, and managing data collectively. In ecology, data practices and policies have recently focused on data sharing (National Science and Technology Council, 2009). Data infrastructures providing access to shared data include research-centered local repositories, resource-centered synthetic centers, and large-scale archives that aggregate collections (Baker & Yarmey, 2009). Their scopes and structures differ: Large-scale data initiatives support access to highly structured data, whereas local scale efforts vary in size and accommodate a wider range of data types, methods, and analyses (Baker & Millerand, 2010).

Bocking (2010) provides a history of collaboration in ecology that examines motivations and the social relations associated with ecologists. It was the 1970s when geographic information systems using digital field data and Landsat satellite remote sensing data became instrumental in ecological collaboration. Personal computers were available widely in the 1980s followed by use of spreadsheets and small-scale databases in the 1990s, which allowed researchers to generate and analyze their data digitally in new ways. At the same time, field instrumentation was rapidly developing, becoming more available and beginning to provide data in digital form. Collaboration increased in the 1980s as scales and interdisciplinarity of fieldwork expanded. Large projects might hold a data workshop where researchers gathered in retreat-like setting to present and integrate their data over periods of days or weeks. Large-scale field programs exposed many researchers to large-scale collaborative investigations that were often interdisciplinary.

Availability of enabling technologies both to individual laboratories and computationally intensive centers created a continuum of digital environments that range from support for local data systems (e.g., a project-specific data repository) to larger-scale data facilities (e.g., a domain-wide data repository).² Ecological scientists often work with small-scale databases using a variety of instruments and sampling methods in generating data as well as a variety of analytic methods as they work with primary data (e.g., Leonelli, 2007; Michener & Brunt, 2000). Researchers in smaller research efforts were found to “lack the tools, infrastructure, and expertise to manage the growing amounts of data generated” (Zimmerman, Bos, Olson, & Olson, 2009, p. 222). Ecological science is part of a broad, ongoing transformation in science referred to as Open Science. “Open Science” encompasses many schools of thought aimed at increasing replicability, reducing bias, and increasing the impact of science by increasing access to data (Fecher & Friesike, 2014). It is a stimulus for change as well as a subject of debate in the digital era of earth and environmental sciences (Gewin, 2016; Hampton et al., 2015; Tenopir et al., 2015). National and international reports describe Open Science as a data access challenge, an economic force, a transformational opportunity, and a key to innovation (European Union, 2016; National Academies of Sciences, Engineering, and Medicine, 2018; Organisation for Economic Co-operation and Development [OECD], 2015; Royal Society, 2012).

A turn to long-term research in ecology brought attention to the temporal dimension of data issues including time-series data and long-term data management. In ecology, long-term research prioritized time-series data and data integration over a variety of time scales stretching from a short-term focus of hours, days, and months to longer periods of years, decades, centuries, and millennia (Likens, 1987; Magnuson, 1989, 1990). For example, the U.S. LTER Network focused on periods that varied in length from days to hundreds of years, thereby supporting long-term research that extended beyond most ecology studies of the time but not such long periods as the millennia of paleoecology. Managing these time-series data foregrounded data management and lifted the temporal horizon of data managers from the lifetime trajectories of individual researchers and laboratories to the notion of preservation of data and its use over longer periods. Social studies of science document data practices and scientific endeavors (e.g., Cragin & Shankar, 2006; Hine, 2006; Jackson, Ribes, Buyuktur, & Bowker, 2011; Zimmerman et al., 2009), including tensions introduced by plans that must accommodate both short-term, knowledge-driven projects and longer-term, preservation-oriented data work (Baker & Millerand, 2010; Bowker et al., 2010; Karasti & Baker, 2004).

The U.S. mandate for data sharing (Holdren, 2013) and the European Research Infrastructure Consortium (2009) are two milestones that launched widespread discussions of data management and data infrastructure and spurred changes in data practices. Recognizing and supporting data as a first-class research object that can be cited as a research product alongside research publications (Kratz & Strasser, 2014), prompted new practices. Persistent identifiers, such as digital object identifiers (DOIs), play an important role in this recognition process.³ DOIs document data-related entities including instruments, platforms, people, and constructed digital objects called “artifacts.” Research funders and organizations have responded to this call for data sharing by focusing on data management planning. New institutional efforts are developing not only within existing organizations but also across traditional organizational and domain boundaries to support research data as a public good as well as to articulate data principles (Allison & Gurney, 2015; Mons et al., 2017). In sharing research data and tending to data infrastructure (Parker et al., 2010; Ribes & Finholt, 2009), new forums have developed to consider the issues involved (Inter-university Consortium for Political and Social Research, 2013; NRC, 2014; OECD, 2017b), including funding, sustainability, and data arrangements (Maron & Loy, 2011; OECD, 2017a).

Current Challenges with Data Management and Data Infrastructures

Data management and data infrastructures present specific challenges in ecology (Hampton et al., 2013; Porter, 2017). With only a fraction of ecological data readily discoverable and accessible (Reichman, Jones, & Schildhauer, 2011), direct access to data remains the exception rather than the norm in ecology. Supporting ecological data access involves both technical and social challenges. As a field-intensive, data-rich research domain, ecology works with heterogeneous data. Technical challenges involve data issues such as harmonization (combining data from various sources that use different formats and naming conventions into a coordinated data set) and provenance (historical record of the origins and what happens to the data thereafter). Ecological data sets are often dispersed across thousands of researchers even with large, ongoing data collection efforts such as the Group on Earth Observation Biodiversity Observation Network (Scholes et al., 2008), Knowledge Network for Biocomplexity (Andelman, Bowles, Willig, & Waide, 2004), and federating initiatives such as the Data Observation Network for Earth (Michener, Vieglais et al., 2011). Ecological data remain highly diverse given the wide range of topics studied, methods used, and production contexts (Birnholtz & Bietz, 2003; Vertesi, 2014). Data provenance is fundamental to data synthesis, but remains difficult to trace particularly in light of the loss of information that occurs over time (Michener, 2000). In ecology, definition of biological units is a central concern. Further, the quality of data is intimately linked to local knowledge and data collection details that strongly constrain standardization efforts, such as with the Ecological Metadata Language, and thereby the circulation of reusable data sets (Karasti et al., 2010; Millerand & Bowker, 2009; Zimmerman, 2008).

As noted by Reichman et al. (2011), technological issues may be challenging, but social challenges are even more daunting. Disciplines have different histories and specific data-sharing practices (Birnholtz & Bietz, 2003). The view that access to primary data collected by oneself is fundamental was initially an integral part of the training of the ecological scientist (Hampton et al., 2017; Zimmerman, 2008), but a growing number of ecologists are in general agreement about the new research opportunities offered by open data.⁴ Open data practices enable the reuse of data. The reuse of data, however, requires research, development, and new policies. On one hand, the reward system linked to the production of data sets is not yet fully in place (National Academy of Sciences, National Academy of Engineering, & Institute of Medicine, 2009; Pryor & Donnelly, 2009) for ecology or for other disciplines. On the other hand, data infrastructures are developing that involve a complex coordination of often layered and intersecting efforts at a variety of levels, such as project, program, community, domain, state, national, and international levels.

The new work of data management, description, assembly, packaging, and curation struggles with embedded assumptions within the culture of science as well as in work with data (Cutcher-Gershenfeld, 2018). STS in general and Infrastructure Studies in particular contribute to the identification and explication of ubiquitous but limited or outmoded beliefs, including the following:

As the ecological sciences mature, there is wider understanding of the complexity and the interdependence of the elements of an ecosystem. There is greater recognition of the coupling of human and natural ecosystems with investigations expanding from “how to understand an ecological biome” to “how to understand a socio-ecological biome” from which emerges the question of “how to support a research community’s data needs inclusive of research computing, data management, and data infrastructure.”

Data infrastructure is an important research object in STS and Infrastructure Studies. The following section provides a brief overview of recent studies of research infrastructure and major findings from the studies.

Definition and Characteristics of Infrastructures

In a seminal study of one of the early scientific community’s digital networking efforts, Star and Ruhleder (1996) provided a now classic definition of infrastructure and a list of its key characteristics. An initial view of infrastructure presents the concept as “something upon which something else ‘runs’ or ‘operates,’” and as “something that is built and maintained, and which then sinks into an invisible background,” such as electric power grids and the internet (Star & Ruhleder, 1996, p. 112). According to these authors, however, the initial view is not only inaccurate but also is of little use to understand what infrastructures really are and how they work. First, infrastructures are inherently social and technical, and should be considered as sociotechnical entities. Second, they constitute relevant research objects in themselves for understanding the relationship between work and technology (Star & Ruhleder, 1996), for example, to examine the uses of a new infrastructure in relation to new patterns of collaboration that are emerging in a research domain.

Infrastructure Studies theories emphasize the socially constructed aspect of infrastructure. An infrastructure such as an ecological database is, of course, a physical entity (e.g., sets of digital data sets hosted on servers), but it also implies a set of actors and institutions that contribute to its development, implementation, maintenance, and use (e.g., programmers, ecologists, research institutions), as well as socio-institutional arrangements (e.g., data standards, usage agreements). So, in order to design an infrastructure, one must take the technology, the actors and stakeholders, and the relationships and contexts into account (Star & Bowker, 2002). This theoretical perspective provides an understanding of infrastructure as a contextualized “relation” rather than a “thing.” What is recognized as infrastructure differs. For a database developer, a research lab’s local data system connected to an ecological data portal is a target object, whereas it may be just one element of support in an ecological scientist’s research program. An infrastructure becomes one in practice, for someone, and in relation to some organized practices in specific contexts (Star & Ruhleder, 1996). This conceptualization is now widely recognized in the field of Infrastructure Studies (Karasti et al., 2016), although key characteristics continue to be examined and explored, enriching the language and stimulating the thinking of those working with data and infrastructure (Karasti & Blomberg, 2018; Mongili & Giuseppina, 2014).

The following have been identified as key characteristics of infrastructure (Star & Ruhleder, 1996, p. 113): (1) embeddedness in other sociotechnical structures (any infrastructure is laid upon another and interacts with others); (2) transparency (it is transparent to use and it supports tasks invisibly); (3) spatial and temporal reach or scope (it reaches beyond a single event or one-site practice); (4) learned as part of membership in a community (new members need to familiarize with it so that it becomes taken for granted); (5) shaped by conventions of practice (while also contributing to the shaping of these conventions); (6) plugged into other infrastructures and tools in a standardized fashion (maybe after being adjusted to possibly conflicting local conventions); (7) based on an installed base (infrastructures do not grow de novo but wrestle with the inertia of an installed base and inherit strengths and limitations from that base); and (8) normally invisible infrastructures become visible upon breakdown (a server failure, for instance).

Open in new tab

Figure 1. Eight key characteristics of infrastructure distributed across technical-social and local-global axes. The distribution of characteristics will change depending on the configuration or activity to be described.

Source: Bowker et al., 2010.

These infrastructural characteristics can be ordered along two axes, one technical/social, and the other one local/global, where infrastructure building is envisioned as a set of distributed activities along these axes (Figure 1). As Bowker et al. (2010, p. 102) explain: “the question is whether we choose, for any given problem, a primarily social or a technical solution, or some combination.” For instance, metadata generation for an ecological data set can be an automated process, the responsibility of a local data manager, or an activity delegated to a national data archive. These choices imply activities that are distributed differently depending on whether the metadata production and expertise is expected to remain local or to reside at the domain level. Each infrastructure can therefore be understood as a configuration resulting from a different set of choices, the product of decisions that are not always thought through or whose implications are not always easy to anticipate.

Invisibility and Articulation Work

Invisibility is central to the infrastructural quality of a system (Star & Ruhleder, 1996), where infra means below in Latin. In Infrastructure Studies, invisibility may refer to one of three things: the invisible nature of the infrastructures themselves, the invisible work (of developing or maintaining it for instance), and the processes of making visible or invisible, deliberately or not, some work activities (Karasti et al., 2016).

A “good” infrastructure is conspicuous by its ability to blend in with its surroundings, to become transparent, invisible. As Bowker (2014) likes to remind us, we do not think about the road when we drive our cars, except when there are potholes! In the same way, we often do not think about the database or the network when we do our research, except when access fails. Infrastructures become visible when they are incomplete, break down, or malfunction. The same goes for some work tasks, often referred to as invisible work. A good example of invisible work is that of the person who edited this text to remove spelling and typographical errors. The irony is that his or her work will not be noticed until a typo appears in the text. This is the daily lot of people who may be in charge of the maintenance of systems and infrastructures or of creating metadata for data sets. These are the invisible workers of the everyday.

Invisible work often evokes a group of understudied activities, such as informal tasks, but it may specify a marginalized category of actors generally at the bottom of social and professional hierarchies such as data processors, also referred to as data cleaners (Plantin, 2018). It could also involve highly specialized or abstract work activities requiring the management of information or knowledge, such as the work of research data management in a laboratory (Millerand, 2012). As Star and Strauss (1999, p. 9) have shown in their studies of disease and hospitals: “no work is inherently visible or invisible; we ‘see’ work through a selection of indicators.” Making work activities visible or invisible may be the subject of intense negotiations; making an activity visible may lead to social recognition of the work and the worker, but it may also result in the reification of work and open the way to new forms of control (Star & Bowker, 1999; Suchman, 1995).

With the work of infrastructure and its maintenance often carried out by undervalued or invisible workers, researchers have suggested the notion of human infrastructure (Denis & Pontille, 2012; Lee, Dourish, & Mark, 2006; Shapin, 1989; Star, 1991). Mayernik, Wallis, and Borgman (2013) refer to the human-in-the-loop to spotlight the human side of infrastructural projects, thus raising awareness of the distributed collective processes that require skill and attention within research communities. In a study based on a long-term ecological research community, Millerand (2012) has shown that, in the constitution of a network-wide database, different processes work to make information managers and their work visible. Visibility results when new roles and responsibilities are acknowledged; invisibility results when a large part of the work ensures production of a routine, expected service. More recently, Plantin (2018) has investigated the work of data processors in a large social science data archive. He showed that a combined configuration of visibility and invisibility defined the social value within the archive of these workers preparing data for sharing while leaving no traces of their own work on the data set for those outside the archive.

A large part of infrastructure development and maintenance falls under the broad categories of articulation work and articulation process, that is, all the activities that bind together a group or series of tasks, actors, or environments (Baker & Millerand, 2007; Millerand, 2012). The articulating of work is what makes it possible to accomplish it (Strauss, 1988). Concretely, it essentially rests on “(largely invisible) activities of planning, organization, evaluation, mediation and task coordination which enable the ‘articulation’ of elements by a larger group (tasks, individuals, technologies, environments, etc.)” (Millerand, 2012, p. 168). Articulation work often depends on recognized mediators or liaisons to negotiate, coordinate, and manage collective planning and activities. Communication can play a central role in facilitating or constraining interactions. The medium, the format, or the chosen language can be more or less adapted to the situations at hand and especially to the nature of the relations between the persons involved.

Methods for making the articulation work visible have been developed in STS and in Infrastructure Studies. They include observing infrastructures during moments of breakdown (Star, 1999) or as they are being designed, used, or repaired (Jackson, 2014; Karasti et al., 2010; Star & Bowker, 2002). These moments make visible elements that are otherwise hard to uncover (Karasti et al., 2016). Another method consists in doing an “infrastructural inversion” (Bowker, 1994), that is, a figure/ground reversal that consists in going backstage (Star, 1999) and bringing to the forefront what has traditionally been taken for granted. And finally, the notion of creating and maintaining a category of “inclusion” in discussions of participants, designers, workers, and stakeholders of all kinds is a method that brings awareness of diversity and provides a prompt to recognize those whose work is easily forgotten in an information society.

Metaphors of Design

Developing data infrastructures involves important design issues that occur in a variety of circumstances. STS work carried out since the 1980s by historians, sociologists, and information scientists on infrastructures with the aim of understanding how infrastructures form, evolve, work and fail, has brought new metaphors and concepts for thinking about the design of infrastructures beyond just the technical issues (Edward, Jackson, Bowker, & Knobel, 2007). Words matter, and metaphors are powerful rhetorical resources that help define visions of the world (Hamilton, 2000), including design issues and ways of approaching infrastructure. Three metaphors relating to imagining data and infrastructure making are “growing” infrastructure (rather than “building” it), the process of “infrastructuring” (rather than a completed project), and a “data ecosystem” (rather than a data pipeline process).

Drawing from empirical research on infrastructure development, largely in the form of case studies, STS scholars have argued against the idea that infrastructures could be built in the sense of “deliberately designed and constructed to a plan” (Edwards et al., 2009, p. 369), in “a highly conscious, carefully controlled, and fully directed sort of way” (Jackson, Edwards, Bowker, & Knobel, 2007, p. 4). Jackson et al. (2007) found that, “a careful and historically-informed study of infrastructural dynamics and tensions” weighs against this common vision. Together with other scholars, they promote instead a metaphor of “growing” an infrastructure, “to capture the sense of an organic unfolding within an existing (and changing) environment” (Edwards et al., 2009, p. 369). Infrastructures are complex systems that can be better understood, as they say, as emergent phenomena whose properties appear as they develop—sometimes, if not always—in unexpected ways. Most projects fall into disuse at the prototype phase, while the successful ones appear at the end, less “built” than “nurtured” or “grown,” having undergone constant adaptations (Edwards et al., 2009, p. 369).

Not only is there no recipe to guarantee, for instance, use of large-scale infrastructure for data sharing (Cutcher-Gershenfeld et al., 2016), but designers often work with a limited conception of tasks, requirements, and expected uses (Edwards et al., 2009). STS scholars have suggested the notion of “infrastructuring” as an alternative to a build-and-serve model for infrastructure (Star & Bowker, 2002; Star & Ruhleder, 1996). Infrastructuring has been used to refer to “a relational and processual (in-the-making) perspective and/or design-oriented interest towards Information Infrastructures,” that is, as a “longitudinal, unfolding process” (Pipek, Karasti, & Bowker, 2017, pp. 1–2). Karasti and Blomberg (2018) describe the process of infrastructuring as having the characteristics of relational, connected, and invisible, highlighting the process as emerging and accruing as well as having associated intentions and interventions. In turning the word infrastructure into a transitive verb, the aim is to account for the complexity of infrastructure design and development, and also to emphasize the idea of an active, ongoing process.

Thinking about the design of infrastructure as a process rather than a task emphasizes its dynamic and unpredictable growth: “Infrastructures subtend complex ecologies: their design process should always be tentative, flexible and open” (Star & Bowker, 2002, p. 160). Research studies have used the notion of “infrastructuring” in many ways: Describing the work of ecological information management and environmental monitoring (Karasti & Baker, 2004; Karasti et al., 2006; Parmiggiani, 2015), supporting new kinds of data production and data curation (Baker & Millerand, 2010), enabling redesign and social innovation over extended periods of time (Bjorgvinsson, Ehn, & Hillgren, 2012; Karasti et al., 2010), encompassing many aspects of design activities (Pendleton-Julian and Brown, 2018; Pipek et al., 2017), and in conjunction with a participatory design approach (Karasti, 2014; Karasti & Blomberg, 2018; Simonsen & Robertson, 2013). Design thinking frequently opens up multidirectional interactions that support shared decision making, narratives from multiple perspectives, and negotiated digital arrangements that broaden the understanding of constraints and rationales for choices to be made. Various kinds of learning have been identified as significant in digital design activities, including incremental, continuing, mutual, and community learning.

Finally, with growing interest in making data discoverable, open, linked, useful, and safe, Parsons et al. (2011) suggest the value of conceptualizing a “data ecosystem.” The idea of a data ecosystem captures the nonlinear dynamics of data arrangements (Parsons et al., 2011; Pollock, 2011). This metaphor of the data ecosystem is found in practice in data management. Like environmental ecosystems, a data ecosystem is made up of interdependent elements in a system of subsystems. The relations of subsystems within and across systems foregrounds the need to consider feedbacks and nonlinearities in an ecosystem, whether it is an environmental system or a data system.

Interdependence of Data Arrangements

Collective Data Management and Models

Concerns about data have changed as data infrastructure and data management practices expand and models develop. Initially, data work involved collecting observations and measurements for analysis as evidence in support of the scholarly production of knowledge. Important differences in data and data practices within disciplines and communities as well as between them are still being explored (e.g., Borgman, 2015), and there is growing recognition that data as a first-class research object gives rise to a new process, that of data production (Baker & Millerand, 2010). Understanding changes in data practices involves paying attention to new geographic scales, temporal extents, institutional arrangements, documentation issues, and distributions of data work. STS work has provided understandings of the impact of these changes on data practices, particularly with respect to long-term research and collective data management that has led to the emergence of new data roles.

Collective data management activities involving the aggregation and organization of data from many community members brought recognition of the need for data systems, repositories, and platforms (e.g., Palmer, 2001). Site-based projects and networks in ecology provide early examples of collaborative work and approaches to data integration and data interoperability (Baker et al., 2000; Benson, Hanson, Chipman, & Bowser, 2006; Ingersoll, Seastedt, & Hartman, 1997; Michener, Porter et al., 2011). In these cases, data and science insights were presented and compared at data workshops that informed research narratives and knowledge making while plans were being made for coauthored papers and further field work.

An information environment supports shared resources and promotes dialogue critical to collective endeavors (Leonelli, 2016). Shared resources such as databases, data systems, project websites, and mailing lists are elements of information environments. Similar to the common information space discussed by Schmidt and Bannon (1992), an information environment provides a place for negotiation of agreed-upon interpretations of data and data procedures where differences are identified and accommodated and sometimes resolved. Such environments both enable sharing interpretations of data and facilitate cooperative decision making.

Figure 2 illustrates changes in data work spurred by collective data management and the growth of information environments. The dyad model is shown in the lower part of the diagram: a researcher working with an assistant helping with field work, instrumentation, or data analysis. As collaborative activities grow and data are aggregated within laboratories and other groups, a data management model develops to support group use of data via a designated data manager. When the development of longer-term perspectives and longer-term funding arrangements coincide in the digital age, an information management model develops to support community data use, a community-specific information environment, and a local data system supporting online data access. Strategic leadership and data vision in some cases leads to stability and funding for an informatics team model where the team is able to optimize for both immediate data use locally and for data delivery to external facilities, thereby creating the opportunity for data reuse over time. Data packaged in accordance with the requirements of other data facilities enables the data to travel upward from any of the lower levels (Leonelli, 2009, 2016). Each level in the figure represents added infrastructure supporting work with digital data.

Open in new tab

Figure 2. Data work models: Transitioning from an individual model to collective models.

Adapted from Baker and Millerand, 2010.

The increase in kinds and amounts of data as well as the advent of collective data management have led to the emergence of new kinds of analysis, expertise, and liaison work. Titles such as data manager, metadata specialist, data scientist, and software engineer designate specialists who work with data. Data or information management have been identified as dealing with a “trajectory” comprising many interdependent arenas of action that require continuous attention to managing data, supporting science, and designing technology (Karasti & Baker, 2004). Data specialists often perform mediation or liaison work. Data roles associated with data management are described in reports and case studies (Borgman, Wallis, & Enyedy, 2006; Karasti and Baker, 2008; Karasti et al., 2006; Meyer, 2009; NRC, 2015; Thompson, 2015), while the growth and distribution of data work remain topics of study critical to establishing and maintaining data quality.

A Multiplicity of Infrastructure Configurations

The term “data landscape” describes the larger context within which data resides and includes support elements such as data work arenas, repositories, centers, archives, and partners that may be positioned locally or remotely. Any arrangement of one or more of these elements may be designated a “data infrastructure” where data work occurs and across which data moves, ultimately arriving at one or more destinations. Potential partnering organizations and communities in academic, government, commercial, and public sectors may develop and support some of the elements. In practice, elements must align both internally and laterally to enable the movement of data from one location to another. The variety of configurations of infrastructure within the data landscape is large and only now beginning to be understood in terms of management and sustainability in addition to fostering invention and innovation. One example of infrastructure within an organization involves a researcher having support available to initiate submissions of a conference paper or poster to their institutional repository; another example with infrastructure that crosses institutional boundaries is the case of a researcher who annotates a data set with metadata and sends this data package to an ecological repository external to their local institution. The many data infrastructure arrangements, each made up of multiple, interacting elements, aim to support the dynamics, complexity, and interdependencies of data in support of research efforts.

Expanding from the two-axis infrastructure (Figure 1) to four axes makes visible more of the work and decision making associated with configuring a data infrastructure. The two panels in Figure 3 describe two different cases by indicating some of the choices made—although they are typically expected to change over time. The first case illustrates data use (case A) where the data work targets support for a set of defined science-driven data practices for a small project that has little existing infrastructure available. These researchers are positioned to analyze and use the data they have generated in a field project but do not have a data management plan. This configuration shows minimal activity in terms of the larger data landscape and data reuse. In the second panel illustrating data use and reuse (case B), there is a formal data management plan and a locally grown infrastructure with capabilities and functionality developed for their local environment—conventions, procedures, and systems—that facilitate local data activities and inform practices of local researchers resulting in less time dedicated to support of individual needs. The local, collective infrastructure still requires maintenance and attention to update and redesign, but if it is well planned and carried out in partnership with others it may free up resources. The project may find new, more cost-effective approaches to internal services or identify external partners that can provide resources and expertise such as for web services. With less time dedicated to individual needs, more resources are available at a collective level as well as for long-term, global efforts where extra-institutional infrastructure provides support and guidance. Whereas case A is centered around local knowledge production, case B is concerned with the production of both knowledge and data.

Open in new tab

Figure 3. Two of the many possible configurations of data infrastructure illustrated by showing their differing distribution of choices pertaining to data effort across eight characteristics on four axes: (A) case of data use with ad hoc plans for data work and (B) case of data use and reuse with a formal data management plan that includes data preservation and access.

A few examples illustrate the scope of each axis in Figure 3:

The shaded polygon for case B is more centered than for case A. In case A, data generation is locally oriented to serve needs of individual researchers, and data work remains informal with no long-tern vision for data reuse. In case B, the added demands on time, budgets, and coordination that involve assembling, packaging, partnering, preserving, and disseminating data for reuse are evident.

Sustainability of Infrastructures

Questions about approaches and costs for the sustainability of infrastructures are topics of active discussion. Although print archives and museums have focused on preservation of materials for some time, the need for long-term preservation and subsequent provision of public access to digital data sets and to research project collections are new concerns. A mistaken assumption about infrastructure’s role in supporting data work is that it can be considered a permanent solution to a problem. Rather, the existence of infrastructure requires, in the long view, watchfulness (Orr, 1996) and care for its fragility and continuing maintenance (Mol, 2008; Puig de la Bellacasa 2011). Edwards (2003) describes infrastructure as an “artificial environment” that, unlike geophysical systems that change slowly, fail because their development assumes design and maintenance are orderly, dependable, and separate from human, technical, and environmental environments.

Sustainability planning requires the development of shared ontologies (Geels, 2010; NRC, 2014) as well as shared understandings of the various characteristics and configurations of infrastructure. For instance, by the 1980s the need to coordinate satellite imagery and remote sensing efforts became evident, spurring development of agreements and vocabularies as the concepts of information systems and archives were brought together by the international satellite community (Albani, Molch, Maggio, & Cosac, 2015; CCSDS, 2012). In commenting on sustainability of digital access to data and information, the Blue Ribbon Task Force (2010) report identified risks and presented recommendations on multiple dimensions: organizational, technical, public policy, and education/outreach. To address contemporary worldwide data efforts, the United Nations (UN, 2014) reported on the necessity of “mobilising the Data Revolution for Sustainable Development” to address environmental change. Support for this aim includes coordination of data efforts at individual and community levels as well as at international levels. Some global-scale efforts include the Belmont Forum supported by international science councils and funding (Allison & Gurney, 2015), the nation-based International Council for Science Committee on Data (2014), and the multistakeholder-based Research Data Alliance (Berman, 2014).

In planning for long-term infrastructure development, whether for data generation, production, or preservation, the design notion of iterative development emerges as one key to sustainability. The iterative nature of design work occurring over time highlighted by Greenbaum and Kyng (1991) is being explored via the concept of infrastructuring (e.g., Karasti & Baker, 2004; Karasti et al., 2018; Pipek & Wulf, 2009), with emphasis on the processual nature of infrastructure making.

The Politics of Infrastructure

An investigation of the politics of infrastructure expands on the concern for sustainability to consider the ethics associated with the support and use of data and data infrastructure. Amid the politics of design (Dourish, 2010; Parmiggiani, 2017), gateways (Edwards et al., 2007; Zimmerman & Finholt, 2007), and data care (Baker & Karasti, 2018), swirl questions in the sciences about who gets to learn by designing, what gets reinforced via technology (Kraemer & King, 2006), when balances are struck, and how infrastructure can accommodate change. Forms of intervention and engagement are the subject of ongoing discussion in STS research (e.g., Ribes & Baker, 2007), including activist approaches that foreground the politics of neglected things (Baker & Karasti, 2018). Social science researchers with qualitative methodologies can raise awareness and share knowledge about the politics of infrastructure, including the impact of categories, standards, and information systems as well as the processes of design and use within a community (e.g., Karasti et al., 2018; Pendleton-Julian & Brown, 2018).

Larkin (2013, p. 330) sees infrastructure holistically as something with an ontology involving power, culture, and imaginative practices. He declares the concept of infrastructure “unruly,” pointing out that “the act of defining an infrastructure is a categorizing moment. Taken thoughtfully, it comprises a cultural analytic that highlights the epistemological and political commitments involved in selecting what one sees as infrastructural (and thus causal) and what one leaves out.” Envisioning a data infrastructure as only technical elements, or as merely ecological science and computer-IT people, excludes other key participants such as data managers and data service specialists. Trigg and Ishimaru (2012, p. 219) detail the challenges of gaining organizational support for creating new “structures of participation: informal and formal liaisons, formal committees, working groups formed around specific projects, and our own incorporation [as social science researchers] into a department of Information Management.” Finally, Jensen and Morita (2015) see today’s infrastructures as experiments with an outcome of relating to politics and power as well as an experience base from which to imagine new possibilities.

Well-Being of Natural Systems, Social Systems, and Information Systems

In responding to grand challenges and creating data infrastructures during the anthropocene, there is great value in considering the “well-being” of a trio of interdependent systems. Not long ago, the twin concepts of ecosystem services (e.g., Daily et al., 1997) and of coupled human-natural systems (e.g., Brymer et al., 2020; Liu et al., 2007) opened up research on the earth’s environment. The recognition of the well-being of natural or earth systems together with human needs and desires as an interplay of the earth and of our superimposed social systems, represents a profound conceptual move. Here, “social systems” refers to human, economic, and political developments. The addition of “information systems” as a third component in an overarching socio-info-natural system establishes the role of information as a primary agent and captures the interplay of the three systems in the digital era. The term “information systems” is used broadly, inclusive of digital technology, data and information infrastructures, and sociotechnical factors. By making information systems explicit rather than subsuming them within “social systems,” a richer context is created within which decisions must be made and balances established to ensure the well-being of natural systems, human populations, and the data that undergird society’s knowledge.

Open in new tab

Figure 4. Conceptualizing a trio of interdependent systems.

Conceptualizing this trio of interdependent systems—natural systems, social systems, information systems—ties together the confluence of forces that impact data work for which policies may be developed (Figure 4). Within information systems writ large, data is now associated with emergent information services and data products of central importance to scientific research. Further, there is a new sense of urgency about increasing efficiency while maintaining the effectiveness of data infrastructures because the data undergird scientific research that is being called upon to support societal policy making at the scale required for a managed earth. One aim of data infrastructure is to transform the perceptions of data from that of a deluge into a readily findable and accessible set of resources.

Data Infrastructure Literacy and Partnering

In developing the concept of “data infrastructure literacy”, the aim is “to make space for inquiry, experimentation, imagination and intervention.” Gray, Gerlitz, and Bounegru (2018, p. 1). These authors remind us that “critical engagement with data infrastructures has been central to various interdisciplinary perspectives from the past several decades, including infrastructure studies, data studies, science and technology studies, the history and philosophy of science, human-computer interaction, computer supported cooperative work, ethnomethodology, the history and sociology of quantification, software studies, platform studies, new media studies, critical design studies and associated fields” Gray, Gerlitz, and Bounegru (2018, p. 3).

Despite this long list, there are more items to add, such as sociotechnical systems, information sciences, change management science, and institutional science in addition to ecology, which has together with technology and library efforts developed project-, program-, and discipline-oriented research and data infrastructures. Many fields are acquiring experience and developing ontologies aiming to capture the variety of data arrangements that can enrich growth of infrastructure if mingled rather than retained separately. All are striving to understand the dynamic interplay of data management and data infrastructures in a data landscape that encompasses local scale efforts, midsize collectives, and larger-scale endeavors.

An STS perspective questions whether there is wide understanding about the aggregation and evolution of data management, the data roles emerging in response to the data work distributed across data work arenas, or how data flows across the heterogenous data landscape in the interdisciplinary field of ecology. Amid differing disciplinary perspectives such as the aims of elegant universal solutions by computer science or of diverse epistemic cultures as well as local data practices recognized by STS research, there is an ongoing emergence of insights. It behooves ecologists to develop a critical eye for infrastructure design and formation because, as interest grows in supporting data infrastructures, larger-scale enterprises develop that enable normative forces such as monitoring, auditing, and governing within layers of infrastructure management where administrative values and bureaucratic narratives flourish. Data sharing and data infrastructures create new responsibilities that require ecologists to engage in opportunities to plan, experiment, learn, and reshape data infrastructures to ensure that their science and its inquiry-driven, innovative core are served appropriately in the 21st century. Ecologists, data specialists, and social scientists such as STS researchers will benefit from active partnerships to ensure data infrastructures effectively support scientific investigative processes.