Earth and environmental data repositories

In general, data repositories archive, curate and preserve information for the long term and make it available for use by a variety of audiences (CCSDS 650.0-B-1 2012). The kinds of data a repository handles can vary widely depending on the repository’s mission and the needs of its users. While some repositories – such as Figshare and Zenodo – are generalist, others are limited to data from specific missions or projects sponsored by their funding agency, certain topics, scientific domains or regions, or the data from specific institutions. A repository may fall under several of these categories, e.g., the holdings of a domain-specific institutional repository may be limited both by its scientific scope and to researchers associated with that institution. Staffing varies widely in repositories as well – from small groups of a few, perhaps part-time employees, to large repositories with perhaps 100+ FTE (Baker and Duerr 2017).

Repository operations vary widely, and funding sources and contractual obligations determine many aspects of their operation, such as the level of curation provided for the data. In most cases, the content, quality and completeness of a dataset’s metadata depend strongly on the researchers or agencies that provided it (Baker and Duerr 2017), and consequently, the adage ‘garbage in – garbage out’ often applies. Repositories’ audiences and services provided also vary widely and evolve over time as recognition of the utility of their data changes. This often leads to a general broadening of the audiences served and more recently, increased recognition of the rights and needs of communities and individuals from which the data originate (Baker, Duerr, and Parsons 2015). This may result in plans for expanded services if support can be found.

Several organizations provide repository certification that cover the basic activities and responsibilities of data repositories (Downs 2021), in line with the TRUST Principles for Digital Repositories that recommend applying Transparency, Responsibility, User focus, Sustainability, and Technology to manage data (CoreTrustSeal Standards and Certification Board 2022; Lin et al. 2020). Many Earth and Environmental repositories have or are in the process of undergoing certification by one or more of these organizations. However, such certification processes omit repository responsibilities that address the CARE Principles, although some of the recommendations here are related to these certification processes. Readers should already be familiar with the basic requirements of certification boards, as these often can form a basis to support specific activities related to CARE Principles.

Every repository has one or more designated communities that it was designed to engage both in terms of data acceptance and access support (CCSDS 2012). For the repositories represented in ESIP, historically this has primarily been environmental or Earth scientists. In some cases, data consumers (including the public) are the intended users – that is, the repository designs specific data products or visualizations (e.g., Socioeconomic Data and Applications Center, (CIESIN 2024)). In other cases, a significant part of the community may be data depositors, e.g., researchers individually archiving ad hoc data (e.g., Environmental Data Initiative (EDI, n.d.), Biological & Chemical Oceanography Data Management Office (BCO-DMO, n.d.)). Many repositories do both (e.g., National Snow and Ice Data Center (NSIDC 2023; EDI, n.d.)).

Repositories are rarely involved in the entire data lifecycle (Ball 2012), typically playing a role during the archival phases (Figure 1, lower). The repository typically assists the scientist in describing the data with metadata, becoming responsible for longer term integrity and preservation. The repository makes the data available for discovery and integration into other research efforts, acting as a mediator between data providers (researchers and/or agencies) and all types of reuse. While not optimal, researchers typically submit data to repositories when the project is nearing completion and activities during the planning and collection stages (Figure 1, top right) are often overseen by universities and research funding agencies. In cases where repositories are not involved from project conception, repositories have no control over what happens during data collection; but they can record information about that process during ingestion and can make that information available when data are reused.

CARE Principles

The CARE Principles for Indigenous Data Governance state that the application of Indigenous data should result in tangible benefits for Indigenous communities (GIDA 2023a). These benefits can be derived through an in-depth understanding of the community goals and by centering the values-based relationships as defined by community-specific culture and knowledge systems. This push towards inclusively developed outcomes strives to uphold Indigenous communities’ governance, sovereignty, and self-determination. Thus, implementation of the CARE Principles should be required to ensure that the use of data aligns with Indigenous Peoples’ rights (Hudson et al. 2023; Davis 2016).

The four CARE Principles each have three subprinciples (Figure 2). The Collective benefit (‘C’) triad states that data ecosystems must function in ways that enable Indigenous Peoples to derive benefit from the data. Authority to control (‘A’) details how Indigenous Peoples must have the authority and ability to control how aspects of data are represented and identified, e.g., geographical indicators, Indigenous knowledge, or material about Indigenous lands, territories and resources. Responsibility (‘R’) underscores that when working with Indigenous data, there is a responsibility to share how the data are being used and how that supports Indigenous Peoples’ self-determination and collective benefit. Finally, Ethics (‘E’) places the rights and wellbeing of Indigenous Peoples as the focus of concern throughout the data lifecycle and in all data systems.

Local Contexts

Local Contexts is a ‘global initiative that supports Indigenous communities with tools that can reassert cultural authority in heritage collections and data’ (Local Contexts 2024). It introduces the concepts of Metadata Labels and Notices. Labels are applied by Indigenous communities to develop and assert their conditions for using and sharing knowledge and data. Labels imply that a community has created a rule, which is conveyed publicly through metadata (e.g., data or image). Notices are applied by institutions internally or directly on data, metadata, etc. to recognize rights, and are often used in cases where details are yet to be worked out with the appropriate nations or communities. Traditional Knowledge (TK) Labels (or Notices) cover cultural material that has community-specific conditions for access and use. Biocultural (BC) Labels and Notices cover genetic resources and biological and genomic data. Both Labels and Notices are examples of the broader concept of badges, which in this context involves imagery associated with metadata (Anderson et al. 2013; Foundation of Digital Badges and Micro-Credentials, n.d.). For example, data from the recent International Polar Year utilize Polar Information Commons (PIC) badges to 1) indicate the norms and expected behavior of both data users and providers; 2) make the rights of users and providers explicit and 3) to make the data as close as legally possible to open domain (Carlson 2012; Kanao, Okada, and Kadokura 2014).

Recommended repository activities

The 47 individual activities in each category are listed below, each with a brief description and its relationship to specific CARE Principles and sub-principles indicated by initial and sub-principle number (Carroll et al. 2020). During discussions, we noted that in some cases, the association between an activity and the CARE Principles would depend on the type of data being managed or other factors related to the activity. For example, the activity ‘include funds for specific targeted tools’ (R10) would be associated with A3 if the intent for building the tool was to facilitate Indigenous data governance (e.g., a tool to allow Indigenous control over data access), and R2 if the tools were to build capacity in Indigenous communities (e.g., tools for using or displaying the data). In general, we did not deeply consider these alternate factors and tried to associate activities to the CARE Principles that would apply no matter what additional factors were present.

Activities were separated into categories for the purposes of this paper, but it should be clear that an activity in one category will often trigger activities in others, forming a chained response. Duerr (2024) provides a tabular example of chained activities that might be triggered after a repository determines that it holds Indigenous data. The examples in that dataset were extracted from the activities described below.

Observations

First, we note that this work was not an empirical survey, but is the result of discussions among approximately 60 people over several years. Consequently, it is important to temper interpretations of our results; as noted above, this work is meant to initiate further conversations rather than provide authoritative guidance.

In the results, we noted nearly all repository activities were related to more than one CARE Principle, which is consistent with other work linking CARE to data stewardship practices (Belarde-Lewis et al. 2024) and the interdependence of the CARE Principles themselves. The most associations made were to activities in the Technology category, which reflects the interdependence of repository activities within the ESIP community and that technical infrastructure supports the other three activity categories. We noted that there were fewer associations with principles concerned with local (e.g. tribal) governance (A2, C2), capacity building (R2), and ameliorating past injustice (C3, E2), although all these principles had associations to some (five–eight) activities. These trends are logical, since repositories are focused on data preservation, access and support of data’s future use (Lin et al. 2020), and generally speaking, recognition of rights and allocation of benefits can be communicated through dataset metadata. The CARE Principles that have lower numbers of alignments also might reflect the fact that those needs are likely to have more complex solutions than can be accomplished within a data repository.

Activities that represented multiple steps of a process were broken down (e.g., as described in Methods for ‘attach TK Labels to data’). However, in an effort to simplify, we also combined what initially appeared to be some very similar activities. For example, an early iteration contained separate activities pertaining to needed metadata fields (for uses, permits limitations, badges, other documentation). For simplicity, these were combined into one, T13, (‘identify metadata fields for…’). However, those needs (permits, badges, etc.) are somewhat disparate, and each is related to a different CARE principle. Therefore, T13 has many associations to CARE, whereas simpler alignments might have been achieved if they remained separated. Likewise, a multistep activity like ‘apply TK badges to data’ alone would probably be aligned primarily with E3 (ethics of future use), but its components steps are aligned to many principles as they cover communication, intended use, and technology. In a project management context, using the multistep examples of the Duerr (2024) dataset may result in fewer and more streamlined associations.

Access and control

The definition of ‘Indigenous data’ used here is quite broad, and could be argued to apply to any data, anywhere. For example, a concern arose in the ESIP community that, with a definition this broad, even something like the daily weather would count as ‘Indigenous’. Certainly these data should be included – especially as climate change causes more extreme weather – making related data available to Indigenous communities in ways that are easily accessible and understandable can be literally a matter of life or death.

The issues of handling (and solutions) will differ for different types of data. For example, for climate data or general biodiversity data, access may be the major concern. Whereas, for data from or explicitly about Indigenous people, the main issue is likely to be one of control. In this discussion, the question of control by Indigenous communities is considered separately from access, since access is a primary function of many repositories, and some issues of control may be beyond the scope of some repositories to resolve (see below).

Access to data in ways that are meaningful and useful is a fundamental requirement for any user. For some, simple permission settings might be enough. However, products tailored to community-specific needs may be important for some Indigenous Peoples, and logically, must be designed with their input (e.g., as described in Christen 2012). As noted above, granular, well-structured, -annotated (meta)data and derived products will support user interface/user experience developers in creating such tailored views. Vis a vis the weather data mentioned above, solutions might include tailoring warnings, forecasts, and other weather products for users with low bandwidth, or to provide products in Indigenous languages.

Sovereignty and control become relevant when data are explicitly about Indigenous Peoples or their territories, whether generated by them or by others (Carroll et al. 2020). In these cases, data exposure or reuse may need to be controlled by the Indigenous community or its designate. Data categories could be developed to help prioritize repository work, ideally in conjunction with the Indigenous communities involved. For example, a high priority might be to identify data collected manually on Indigenous land or about Indigenous Peoples or artifacts, where sovereignty is clear, permits were required, and communication is relatively straightforward (e.g., tribal elders can be identified). The Smithsonian instituted Directive 609 (Smithsonian Institution 2019) to recognize the complexity within their digital assets and metadata, which identifies specific restriction categories for material that might need additional management, with general guidelines for documenting the nature of the restriction in collection systems. The directive also outlines which roles within the institution are responsible for aspects of improved stewardship and potential restrictions.

However, it should be noted that once data is publicly available, knowing how it is being used becomes practically impossible. Two possible remedies for that would be to impose access protocols that include legally binding contractual or usage constraints or to add information (i.e., metadata) to the data that describe the permissions, protocols and usage constraints that should be used. In either case, the problem becomes one of creating tracking and verification mechanisms as well as determining what could be done if violations are detected. All of this is beyond the scope of the work done here.

Lastly, as mentioned in the recommendations, in some cases returning the data to the Indigenous community may be necessary (e.g., upon loss of repository funding or when data repatriation is needed). It needs to be recognized that undertaking such a transition is likely to be quite complex, involving at the very least the transfer of the data and its metadata to the community, deleting instances of the data from all servers and backups under the control of the repository, replacing the targets of identifiers for the data and its metadata with notices recording this action (unless requested not to), and supporting propagation of these changes to all data and metadata aggregators, such as DataCite (Vierkant, n.d.), that maintain records about the data.

Calls for open data

The CARE Principles are designed to complement FAIR, TRUST and other mainstream data frameworks (Wilkinson et al. 2016) (Lin et al. 2020). FAIR is focused mainly on data clarification and reusability plus simplified retrieval, while TRUST stresses repository operational issues like transparency and sustainability – tenets designed to promote global data openness and awareness. CARE extends awareness to promote equitable participation and outcomes from said data access, use, reuse, or attribution, and emphasizes that the level of openness and accessibility be aligned with Indigenous rights. Resolving the tension between protecting Indigenous individual and collective rights and interests while encouraging open data in a global research environment can only be accomplished by strong ties between the repositories entrusted with that data and the Indigenous Peoples and communities that relate to that data, in particular, by increased communication and adoption of Indigenous standards. Yet, as noted below that communication is itself fraught.

A fundamental purpose of data repositories is to curate data and information for future generations and to make that data available to the repository’s designated communities now and into the future. With current increasing calls for open data, funders and repositories may be assuming that all data can be ‘… freely used, re-used and redistributed by anyone – subject only, at most, to the requirement to attribute and sharealike’ (OKFN 2016). In reality, even the full definition from the Open Knowledge Foundation recognized the need to preserve provenance (version 2.1, ibid), and further, that early conversations around data openness and sharing were incomplete (Avila 2023). In fact, most open data calls discuss the concept of ethically open data, because unrestricted public access or ‘open by default’ might be incompatible with the need for authority, ownership and control. Data that align with the CARE Principles is ethically open data (Kukutai 2023).

Implementing CARE: Repositories cannot do it alone

Repositories have an important role to play in implementing the CARE Principles, but they cannot resolve all of the issues or tensions that exist. Most of the issues below will require discussions among not just repositories but governments, international bodies and funding agencies, in collaboration with consortia of Indigenous Peoples who have specific interests in the outcome.

First, are issues of scale. Controlled access may work well for isolated data about particular people or places that are rarely studied, but controlled access on a granular level becomes almost impossible for data collected at large scale, e.g., global data from satellite remote sensing. Worldwide, Indigenous Peoples represent at least 5,000 distinct groups (United Nations, n.d.-a), thus it is unrealistic for a single repository or even group of repositories to consult and collaborate with all. Similarly, it is unlikely that any Indigenous Peoples would have the resources to consult with all of the many thousand repositories that exist worldwide. Solutions must be found that protect and promote Indigenous rights and relationships with data at scale.

Second, repositories recognize that their role in the application of CARE is limited to particular parts of the data lifecycle (Figure 1). Repositories steward data and create enabling environments for preservation and sharing in a responsible manner. Should they be caught between competing interests, for example where an Indigenous community would like to restrict access to openly accessible third-party data (e.g., a government agency, a scientific project), it is not necessary for the repository to be the arbiter of the dispute. However, the repository may put the parties in contact with each other so that they can decide on an acceptable, actionable solution, while ensuring data preservation in the interim. In an instance where an Indigenous community feels that the use of data from a repository has been inappropriate and caused harm, there may be little a repository can do beyond recording the communication, directing the community towards the user and submitter of the data, and providing a basic assurance that the data were accessed in a manner consistent with the legal and ethical provisions associated with it.

Third, the CARE Principles apply to the entire data lifecycle, but repositories are rarely involved in all stages (Figure 1). This fundamental mismatch makes it difficult for many repositories to ensure that all of their data align with CARE. To increase the likelihood of CARE-alignment, data management planning should begin early, the CARE Principles should be integrated into data management plan (DMP) templates, and DMPs also should cover plans for data archive in an appropriate repository. If funding agencies require and enforce collaboration with Indigenous Communities prior to funding research some of the burden of tracking and verifying researcher compliance could be simplified (e.g. NSF PAPPG (NSF 2024)). However, this currently is only feasible for data about a community as opposed to data of potential interest to many communities, unless all communities are in agreement.

Fourth, the subject of metadata standards here is complex, especially if we are to avoid imposing ‘cognitive imperialism’ (Montenegro 2019). Indigenous data governance and standards organizations need to be established so that they can provide guidance to the Earth Science Repository community in implementing metadata standards that address the needs of Indigenous communities. For effective implementation, these organizations will need to create well defined specifications for all of the metadata fields, semantics, and rules implied – for example, by the recently published Indigenous Metadata Bundle (Taitingfong et al. 2023) as well as the content of any TK or BC Labels. Some existing metadata standards could be promising points of departure e.g., PREMIS (Library of Congress 2023) or ODRL (W3C 2018) but it is not yet clear whether they are fully adaptable. Clarifying data’s particular cultural setting, plus potential reciprocity and obligations incurred when accessing it may be complex, depending on the granularity of metadata elements (Christen 2012) and the intended meaning imposed by the affected Indigenous communities. Furthermore, concepts of ownership, property rights, copyright law as it relates to misappropriation and metadata categories may conflict with TK and its associated religious, spiritual, or worldview meanings (Scassa and Taylor 2017).

Fifth, throughout, we have stressed the importance of working with Indigenous Peoples in any implementation. This sort of engagement has already overloaded certain individuals and groups like the Collaboratory, and will continue to do so. Communication strategies that defend against this possibility and also provide for diverse representation of Native voices would be wise; for example, creating representative bodies to communicate with multiple repositories and multiple communities would improve efficiency. While repository-to-community examples are rare, it should be noted that the Canadian territories have been working on the issue of communications with researchers for years (see ‘Conducting Research in Canada’s North’ (Canada 2018)) and Tribal Nations in the US have their own laws, policies, and practices for research oversight, as well (Garba et al. 2023; Carroll et al. 2022).

Lastly, we recognize that most of the activities listed here cannot be covered by repositories within their current budgets. Many repositories have little or no budget for outreach, nor for communication beyond technical support (USGAO 2022). When they do, it is highly unlikely to cover working with all the Indigenous communities that a repository’s holdings may potentially be connected to. Foundations, agencies, universities, and other institutions that fund the creation and maintenance of repositories must appropriately support the transformation of repositories to implement CARE.

Where should repositories start?

Although repositories may desire to implement sophisticated documentation of the CARE Principles, a limited workforce or limited funding and shortfalls in metadata specifications may make it difficult to do so in a timely manner. A goal is for a repository to apply the CARE Principles to their data service and management in a manner that respects their development and maintenance resources.

The first course of action should be to determine if you already have Indigenous data in your repository. A simple first step would be to determine if the geographic coverage of existing datasets overlaps with culturally relevant or known Indigenous areas, for instance by using resources such as Native Lands Digital (https://native-land.ca/). Parallel (or later) steps could involve making public disclosures about the presence of Indigenous data within your repository with tools such as the Local Contexts ‘Open to Collaborate’ Notice, or engaging in targeted outreach to communities who are now actively asserting access and governance in their ancestral homelands, particularly around data and plant specimens (Mc Cartney et al 2022). Similarly, identifying already-deposited datasets with related collection permits or other documentation and re-curating them can help the repository community plan mechanisms for linking these resources widely. It’s likely that most repositories hold datasets that meet these criteria, but have not yet characterized their holdings this way.

The next steps may depend on the type of data held. If your data are global satellite remote sensing data, your activities will be very different than for a repository holding data or collected from a particular Indigenous community’s land. In the former case, the next step would be to address the technical and protocol elements to ensure the repositories’ responsiveness to both Indigenous data and information requests from Indigenous communities. Following that, active engagement could determine if any global groups with both Indigenous and repository representatives have developed a set of general guidelines for this situation (at the time of this writing, the answer was “no”). Without that in place, the implementation of CARE within your repository is likely to remain a work in progress.

In the latter, an assessment of your repository’s existing technical capabilities should help ascertain whether you already have systems that can handle all of the technical activities described above. For example, do you already have the ability to embargo data, apply TK Labels, attach documents and other information to data, etc.? What about your communication strategies and capabilities? Is there coordination among linked repositories or data aggregators to handle the transfer of new metadata? Do you have an advisory board and if so what communities are represented? Do you have funding for engaging with new communities? Do you have all of the needed policies and the procedures that implement them or do these need to be developed or updated to be more responsive to the CARE Principles (see Repository Protocols above)? In all likelihood you will need to prioritize the work needed to address any gaps, though we advise against investing in expensive upgrades until a community asks for them.