2.1 WGI—physical science of climate change

Working Group I (WGI) implemented the FAIR Guidelines for the full report. It took a more detailed approach for the provenance documentation compared to WGII and WGIII, including the development of the innovative WGI Interactive Atlas [18]. The implementation in the WGI began in early 2021 when AR6 was already underway. Initially, workflows were designed to spread responsibilities among authors, DDC, and the TSU and communicated to authors. The data curation process was not made mandatory for chapters but recommended, emphasizing the benefits it would have for IPCC products, as well as for authors, in terms of recognition and broadening the uptake of their work. Supplementary materials and Virtual Workspaces were provided for the authors [19]. For final data, a dedicated server, known as the Figure Manager, was custom-made by the TSU and then used to assist authors in meeting the minimum information requirements deemed necessary for proper data storage. This information, along with the data itself, was quality-checked by the TSU and then sent to the DDC for long-term preservation.

In the case of software code, although minimal information requirements were requested, the verification of information quality was directly entrusted to the provider. For intermediate data, the workflow was simplified and customized for specific cases. This involved direct exchange with the authors without the use of automation tools.

For datasets related to international model intercomparison projects, e.g. CMIP6 input datasets, detailed information on dataset usage was gathered from the authors and provided to the DDC Partner. The DDC quality-checked the information, identified the datasets, gathered data and information (including usage of individual datasets in figures), and long-term archived the data. Authors coordinated directly with the DDC Partner for products related to the Atlas.

WGI made final data and code available for over 200 figures, which represents about a third of the figures in the main report. Additionally, all plotted data for the SPM and the data behind about 20 figures from the TS were provided. Furthermore, selected intermediate datasets resulting from the assessment were curated for some key indicators used in the report (e.g., projections of global surface temperature and sea level rise, where uncertainties were constrained through the assessment of multiple lines of evidence). Moreover, an innovative digital product was produced to support and expand the access to the datasets and assessment of regional information conducted in the WGI AR6, synthesizing key findings for climatic impact drivers: the Interactive Atlas (see subsection 2.4).

2.2 WGII—impacts, adaptation and vulnerability

Working Group II (WGII) assesses the vulnerability of socio-economic and natural systems to climate change, the negative and positive consequences of climate change and options for adapting to the observed and projected impacts of climate change. During AR6, WGII followed a careful curation process often involving personal communications with the authors of each data source, requesting access to data and/or permissions for their inclusion in the DDC catalog. Overall, the FAIR principles were implemented in the final data curation for the SPM and TS figures and tables, and the integrated database of observed and projected climate change impacts, as well as the risk assessment database across Working Groups. The critical input data used across WGs, such as INFORM Global Risk Index were also curated. In total, 42 input and final datasets covering most of the figures and tables from the WGII TS and SPM were successfully cataloged.

2.3 WGIII—mitigation of climate change

Working Group (WGIII) assesses the mitigation options to reduce and remove green-house gas emissions. It utilizes various datasets for AR6, including a synthetic historical emissions dataset (intermediate data) based on the EDGAR database [20] and scenarios data from the Scenarios Explorer hosted by the International Institute for Applied Systems Analysis (IIASA) on behalf of the IAMC [9]. However, due to licensing restrictions, input data, like that from the IEA, could not be curated by WGIII (it was made available to IPCC WGIII authors through a Memorandum of Understanding, granting access to IPCC authors and reviewers, and giving IPCC the right to share numbers derived from IEA datasets on which figures and tables were based). The process of data curation for WGIII’s contribution to AR6 was adapted from the TG-Data guidelines [21], and includes providing a clean final dataset in a standard tabular format (covering data class, element, and type), structural metadata explaining the dataset (location in the report, authors’ details, data source and underlying research papers, and any post-processing undertaken), and minting a DOI.

2.4 New interactive products supporting the IPCC assessment

The representation of findings (figures, tables, etc.) in a traditional report is limited due to report length restrictions and prevents users from exploring the digital information in more detail such as specific regions, scenarios or variables of interest. This motivated the development of new interactive products with flexible visualization options, supporting and extending the information provided in the report and its figures. The IPCC WGI Interactive Atlas (https://interactive-atlas.ipcc.ch/) was developed as part of the WGI AR6. In addition, the IPCC NASA Sea Level Projection Tool (https://sealevel.nasa.gov/ipcc-ar6-sea-level-projection-tool) further extended the accessibility of sea level projections, which were developed in collaboration and endorsed by the IPCC. WGIII AR6 included a call for mitigation scenarios (https://data.ece.iiasa.ac.at/ar6) which was issued jointly by the IAMC and the IIASA, and supported by IPCC WGIII under a Collaboration Agreement [22,23].

The main challenge for the Interactive Atlas was to meet the requirements of the formal IPCC review process since interactive products cannot be reviewed in full detail due to the many combinations of choices. The implementation of FAIR principles was a key element for this, since it allowed for an indirect review, focusing on the underlying data (and rich end-to-end metadata) and software [18]. Besides the formal reviews, the Atlas GitHub repository was open for scrutiny, aiming to collect feedback and suggestions, resulting in notable enhancements.

Overall, the repository significantly advanced reproducibility by consistently leveraging existing technical FAIR enablers within the digital infrastructure providing climate data, e.g. for the production of provenance information attached to all final products visualized in the Atlas [24]. This contribution is instrumental in fostering a Community of Practice dedicated to effectively disseminating open and reliable climate change assessment tools and products, facilitating users’ active engagement. For further details and implemented solutions, refer to [18].

3.1 Coordination and harmonization

FAIR principles focus on the user experience. In the IPCC context, however, great care has to be taken to minimize the additional burden for IPCC authors and TSU staff. This proved difficult in practice, especially because data and code collection protocols were developed whilst the AR6 assessment process was already underway. TSUs and DDC Partners had to quickly harmonize their practices to exchange data and metadata records and avoid duplicating work. A common schema (https://github.com/MetadataWorks/Schemata) describing metadata requirements was agreed across all WGs and DDC Partners, which allowed the creation of a joint DDC data catalog. The schema describes 40 different metadata elements, of which 22 are derived from DCAT (Data Catalog Vocabulary: https://www.w3.org/TR/vocab-dcat-2/), while most of the remainder are derived from Schema.org. The metadata elements are grouped into 8 categories as follows: Summary, Documentation, Coverage, Provenance, Accessibility, Enrichment and Linkage, Structural Metadata and Data Status. These groupings allow researchers to find and evaluate datasets quickly and easily before accessing and reusing the data directly. It is worth highlighting that such close collaboration between DDCs and TSUs was unprecedented in the history of the IPCC.

One noteworthy achievement stemming from these new collaborations is the organization of joint training activities like the WGI Training on Data and Software Development [28], and the delivery of international outreach activities, notably regional workshops on the WGI Interactive Atlas [29] hosted by different regional partners, and on the AR6 scenario database [30] co-hosted by IIASA and local governmental and academic organizations within each region. Multiple regional webinars were held with the support of local organizers, including live sessions held in different languages.

3.2 Licensing

Some of the data assessed by the IPCC are published with licensing restrictions. For example, the IEA shared datasets with IPCC authors, but these could not be redistributed publicly by the DDC. Another example relates to CMIP6 climate projections, initially published by climate modeling groups under a “sharealike” license (CC BY-SA 4.0). The “sharealike” clause requires data users to distribute derived products, such as figures, under the same license as the original data. Given that this was at odds with IPCC’s own license (CC BY-NC-ND 4.0), TG-Data and the DDC managers reached out to the World Climate Research Programme Working Group on Coupled Modelling (WGCM) and made recommendations that were adopted by the individual CMIP6 modeling centers to share model outputs under more open licenses (CC BY 4.0 and CC0).

Another licensing challenge arises from input datasets published in various scientific journals, each having its own license and conditions for sharing published data. Authors and TSU staff had to consider each license and, in some cases, make inquiries to journals to be granted permission to redistribute data under IPCC’s terms. This has proven to be extremely time consuming and solutions will need to be explored to streamline the process for AR7.

3.3 Data and software curation

The curation of the majority of data and scripts started after the SPMs were approved. This led to the overall challenge of decreasing author availability and TSU staff capacity during the archival and limited the ability to archive report data. This was especially challenging to finalize the CMIP6 input data archival.

The implemented flexible and inclusive curation approach adopted did not require authors to use specific software to conduct their analysis. This heterogeneity made reviews of software, final and intermediate datasets considerably more difficult and often required time consuming additional solicitation of input from authors to ensure consistency and uniformity to the documentation requirements. Another time-consuming task was to create a record of all the CMIP6 datasets used in the AR. The DDC focuses on datasets (variables and versions) used in the report to generate figures or tables. It turned out, however, to be very difficult to track which datasets and which versions were used by different authors over multiple chapters and hundreds of figures, especially at a time when author teams and TSUs were gradually winding down their activities. Despite considerable efforts, gaps remain in the preservation of those input datasets, and tools and training will have to be developed to ease the process in AR7.

In retrospect, it is remarkable that over 300 final datasets (Fig 1), a handful of intermediate datasets and ca. 65,000 CMIP6 input datasets and a couple of further input datasets have been archived and issued DOIs when needed. Around 130 codes have been created by WGI authors and included in WGI’s official GitHub repository (https://github.com/IPCC-WG1) by WGI TSU. Codes for individual figures used the GitHub-Zenodo link to become citable with a DOI. WGI’s GitHub repository additionally includes the scripts used to produce analyses and figures. Together, these efforts increase the transparency of the IPCC process and give credit to scientists who work behind the scenes to produce the report’s content. A visible outcome of these efforts are the links to the dataset that now accompany these report figures on the IPCC website.

3.4 Implementation of the IPCC error protocol

One of the motivations to implement FAIR principles is to support the implementation of the IPCC Error Protocol [14] in case errors are identified in the reports’ contents, and datasets or code underpinning report figures or calculations used to underpin estimates in the assessment. The Error Protocol is part of the principles governing IPCC work. It serves to document and correct errors of fact or accuracy that could have been avoided in the context of the information available at the time the report was written. Once a potential error is identified, the Protocol foresees an exchange managed by TSUs with authors of the report to check, confirm, and provide a correction if needed. The outcome of this process is appropriately documented and is made available publicly on the IPCC website.

The Error Protocol has so far not been invoked for data or software produced during the AR6. Should this arise, TSU will need to collaborate with DDC managers to put in place a new process to curate and document any updated, corrected versions of datasets. Corrections made to code that is archived in TSU GitHub repositories will also need to be documented alongside new versions. The implications for the published report (corrected figures or estimates in tables or text) will need to be documented following the standard Error Protocol approach alongside the relevant information on corrections made to the underlying data and/or code.

4.1 External collaborations on input data

The World Climate Research Programme (WCRP) coordinates a wide range of climate research activities, including the CMIP and CORDEX model intercomparison projects. Their results have always been one of the most important sources of information and data for the IPCC assessments. For AR7, high-level consultations between representatives of the IPCC and WCRP are ongoing with the aim of aligning WCRP’s research priorities with the scope of IPCC’s AR7 in order to meet the tight AR7 timeline. Based on this consultation and further community engagement, WCRP CMIP7 (https://wcrp-cmip.org/cmip7) redesigned CMIP by identifying and grouping a set of experiments relevant to AR7 in the so-called "CMIP7 AR7 Fast track". This subset of CMIP7 experiments must conform to a higher level of standardization and adhere to the AR7 schedule.

Requirements for data and information from the IPCC data guidelines are brought into the CMIP infrastructure development and standardization process by TG-Data representatives involved directly in the WGCM Infrastructure Panel (WIP), CMIP Panel and relevant CMIP7 Task Teams. CMIP is strengthening errata, data citation and model documentation services (ES-DOC: https://es-doc.org/) with stricter requirements for minimum viable information that accompany model submissions. This work is coordinated by the WIP and implemented in collaboration with the Earth System Grid Federation (ESGF: https://esgf.llnl.gov/) and further CMIP data infrastructure providers. ESGF itself is moving to a new architecture meant to improve data discoverability, provenance tracking and the communication/integration of further services like the errata service, among others. Plans also include a messaging service that can be utilized to inform scientists of data updates and errata. We highlight the need for a federated data citation service to ensure homogeneous data citation approaches across ESGF data nodes.

For WGII and WGIII, the International Committee on New Integrated Climate Change Assessment Scenarios (ICONICS: https://depts.washington.edu/iconics/) and the IAMC (https://www.iamconsortium.org/) coordinate important activities related to scenario development, in particular when it comes to the Shared Socio-economic Pathways (SSPs) and curation of related data products (e.g., SSP Extensions Explorer: https://ssp-extensions.apps.ece.iiasa.ac.at/, SSP Update: https://data.ece.iiasa.ac.at/ssp/). Up to now, scenario collection and vetting were conducted under the tight deadlines set by the assessment timeline. In the IPCC workshop on scenarios [32], it was generally agreed that the IAMC implements an ongoing community-coordinated process, irrespective of IPCC timelines. This would allow the IPCC authors to take a snapshot of the data whenever needed, and hence optimize the time to focus on the scientific assessment. In addition, a separate call for collecting national scenarios [33] following a harmonized scenario protocol has recently been opened.

4.2 Development of tools supporting the IPCC process

The IPCC has initiated the implementation of a centralized Figure Manager; the internal server enables the consistent management of the figures included in its reports. Its functionalities are inspired by similar software developed by AR6 WGI TSU that in turn built on a resource developed for the US National Climate Assessment (NCA). AR7 WGIII has decided to use the US NCA Figure Manager. The Figure Managers aim to provide TSU with an overview of figures status, collect metadata and figure datasets, facilitate communication with authors, record versions and link to external assets (input data and software) that have contributed to the generation of a particular figure version. All this information characterizes the provenance of a particular figure.

We foresee that collecting standardized provenance records for the figures’ generation through the Figure Manager together with a set of internal analysis tools would enable the systematic citation of input data and the traceability of the reports’ figures. Furthermore, the information contained in the provenance records is essential for the identification of input datasets for long-term archival in the DDC and for establishing relations between report (full report, chapter, figure), all data types (input, intermediate, final) and applied software, making these relations accessible to users and to infrastructures such as Knowledge Graphs.

Ideally, collecting figure data and metadata should require low levels of effort from authors. In AR6, some chapter writing teams and the WGI Interactive Atlas [24] have adopted analysis tools that automatically track provenance information. In cases where those tools are not applicable, clear instructions will have to be provided for authors to record the relevant information and submit it to the figure manager database.

4.3 External collaboration on data management practices

There are multiple initiatives where the community actively encourages and works towards the transparent sharing of climate-related data, fostering collaboration among scientists, data providers and, eventually, policymakers. In Europe, the RODEO project is creating services for the retrieval of meteorological and observational climate datasets identified as High Value Datasets (HVD). The distribution process aligns with the FAIR principles, trying to homogenize and accommodate metadata models for discovery and detailed description of the informative content, adhering to an open license. IPCC participation in networking initiatives aiming at raising global awareness of the importance of achieving FAIRness and equity of the information provided by future climate services will be fundamental to steer priorities and support. It will also impact the creation of traceable climate scenarios at a regional scale by national meteorological services, as the IPCC depends on data produced and managed by international organizations. This highlights the significance of these collaborations aimed at concrete implementations of a FAIR climate data infrastructure.

Ultimately, an effective strategy towards the implementation of FAIR can benefit from the emerging FAIR evaluation frameworks, which can be used for the compilation of guidelines, as well as their incremental implementation. This allows advisors and implementers to refer to a common and detailed model which disambiguates terminology and fosters an incremental application of the principles. For instance, the Research Data Alliance published the FAIR Data Maturity Model that organizations can use to assess their progress in FAIR coverage, thanks to a collection of detailed indicators. This could be used in combination with the FAIR implementation profile, delivered by GOFAIR organization, which aims at gathering and disseminating FAIR implementation choices made by a Community of Practice. These are made openly available to others for reuse: https://www.go-fair.org/how-to-go-fair/fair-implementation-profile/.

Further collaborations with groups such as the Research Data Alliance (RDA) and the Open Geospatial Consortium (OGC) on technical solutions for open issues related to the implementation of the AR7 recommendations are ongoing. For example, TG-Data members are engaged in the RDA Complex Citations WG, aiming to provide recommendations for the citation and credit assignment of many datasets, partly subsets, stored across multiple repositories using a single complex citation object. The documentation of the figure creation process in the figure captions has been one of the three guiding use cases for WG [34].

Other RDA groups with case statements relevant for the IPCC guidelines include the emerging ‘RDA Coordinating Earth, Space, and Environmental Science Data Preservation and Scholarly Publication Processes Events WG’, the ‘RDA & ReSA: Policies in Research Organisations for Research Software WG’ or the FAIR Digital Object Fabric IG. OGC’s Climate Resilience Domain WG provides an open forum for the discussion and presentation of OGC Standard usage in the context of cross-sector climate actions, where TG-Data can contribute its use case to the discussion.

Informal exchanges with other assessment bodies such as the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) and national climate assessments have been initiated to share tools, practices and jointly improve data and code handling in assessments.

4.4 Funding and sustainability

The DDC Partners provided in-house data curation expertise and data storage infrastructures to the IPCC as a national or institutional in-kind contribution up to 2024. With the introduction of the IPCC data and software guidelines, the workload for DDC Partners increased substantially. At the same time, the in-kind contributions of some members were undergoing substantial reductions, jeopardizing the future operation of key DDC services. Given the growing importance of data curation activities within the IPCC, TG-Data recommended the IPCC Panel establish a stable and predictable funding mechanism for DDC activities. At the 60th Session of the IPCC [35], a three year budget conditional on the approval of the TG-Data programmer of work for the Seventh cycle was adopted, paving the way for the planning of DDC activities covering the full AR cycle. These funds support DDC activities and data for the new cycle. The long-term preservation of the data and the maintenance of the digital products from previous cycles still rely on in-kind contributions or alternative funding sources of the DDC Partners.

4.5 Ideas for future work

Beyond the immediate operational requirements for AR7, TG-Data is also thinking more broadly about future developments. One avenue could be to generate a machine-readable version of IPCC Assessment Reports, similar to what has been done by the IPBES. IPBES has made three reports available in the Linked Open Data format (https://ipbes-data.github.io/IPBES_LOD/) [36], with the objective of facilitating its ingestion into databases and machine learning training datasets. We are also investigating the use of complex citations to give credit to all data providers contributing to the figures included in IPCC reports and in support of provenance documentation of the figures. Many figures aggregate results from dozens of different modeling groups, and at the moment they cannot be credited directly.