Improving the Findability of Digital Objects in Climate Science by adopting the FDO concept

Climate science relies heavily on the effective creation, management, sharing, and analysis of massive and diverse datasets. As the digital landscape evolves, there is a growing need to establish a framework that ensures FAIRness in handling climate science digital objects. Especially, the machine-to-machine actionability of digital objects will be a crucial step towards future AI assisted workflows. Motivated by a use case, this contribution proposes the adoption of the Fair Digital Object (FDO) concept to address the challenges associated with the emerging spread in interdisciplinary reuse scenarios of climate model simulation output.

FDOs are encapsulations of data and their metadata made accessible via persistent identifiers (PIDs) in a way that data and their context will remain a complete unit as FDO travels through cyberspace and time. They represent a paradigm shift in data management, emphasizing the machine-actionability principles of FAIRness and the requirements enabling cross-disciplinary research. The FDO concept can be applied to various digital objects, including data, documents and software within different research disciplines and industry areas.

The aim of this work is to commit to an FDO standard in climate science that enables standardized and therefore automated data analysis workflows and facilitates the extraction and analysis of relevant weather and climate data by all stakeholders involved. The current work  expands on the efforts made to enable broad reuse of CMIP6 climate model data and focuses on requirements identified to enable automated processing of climate simulation output and their possible implementation strategies. The exemplary use case of an automated, prototypical climate model data analysis workflow will showcase the obstacles occuring when analyzing currently available climate model data. In particular, the findability of digital objects required for a particular research question in climate science or a related field shows to be challenging. In order to mitigate this issue, we propose certain strategies: (1) Enriching the PID profiles of climate model data in accordance with the FDO concept and taking into account the needs of the climate science community will lead to improved findability of digital objects, especially for machines. (2) Defining a standardized, unique association between climate model variables and their meaningful long names will increase the findability of climate model data, especially for researchers in other disciplines. (3) Furthermore, combining the FDO concept with existing data management solutions, such as the intake-esm catalogs, can lead to improved data handling in line with prevailing community practices.

Eventually, implementing an FDO standard will benefit the climate science community in several ways: The reusability of the data will facilitate the cost-effective use of existing computationally expensive climate model data. Improved data citation practices will promote data sharing, and ultimately, high transparency will increase the reproducibility of research workflows and consolidate scientific results.