Notebooks Now! The Future of Reproducible Research

In this paper, we introduce a new ambitious project led by AGU in collaboration with Wiley and Curvenote and a team of geoscientists, computer and data scientists. The project, aptly called Notebooks Now!, aims to develop an end‐to‐end workflow for the preparation, submission, peer‐review and publication of Earth and space science papers, based on computational notebooks, specifically Jupyter Notebooks or R Markdown. The idea, first proposed by Dr. Peter Fox, Editor in Chief of Earth and Space Science from 2019 to early 2021, is to revolutionize the way data‐ and computation‐rich scientific research is performed and published. The vision is to establish the computational notebook as the centerpiece of the scientific work, around which the manuscript is then written. In this model, the notebook is the primary publication, accompanied by a static “traditional” paper as the enduring archive. Advantages of this model include: ease of editing of data and code during manuscript writing and revision; open access to the same by reviewers and readers; prompt reusability and repurposing of research outputs by other researchers; access to sophisticated computational tools by the global community; “multiplier effect” of the research. The model thus fulfills the findable, accessible, interoperable, and reusable principles of scientific publishing, and commits to Open Science. This is part of AGU's strategic initiative to accelerate openly accessible scientific contributions to sustainable development.

data, metadata and code must be easily findable and openly accessible through a unique digital identifier, the software used to process and analyze data should be non-proprietary (as much as possible) to facilitate reuse by the scientific community in a non-exclusive manner, the code and algorithms should be adequately documented and explained and ready to be applied to similar projects.From a scientific perspective, the combination of these qualities fulfills the requirement of transparency and reproducibility of published research (Gil et al., 2016).From an ethical perspective, they advance us one step closer to Open Science (https://www.agu.org/learn-about-agu/about-agu/open-science), the global effort to increase participation and access to scientific research.
Achieving FAIR and Open Science is not without challenges, however: what online repository can be "trusted" to permanently host data and methods (script, workflows, etc.) openly and accessibly?What restrictions are imposed on the dissemination of data and methods by the authors' organizations?How can we recognize the additional time and resources required by the authors to provide all the information needed to properly archive data and methods in an online repository?And overall, what are the costs associated with these efforts?There is no easy answer to these questions, and the model may not be unique, strongly depending on the type of research published.
A new way of thinking about FAIR and Open Science was promoted and actively researched by Dr. Peter Fox, Editor in Chief of Earth and Space Science from 2019 to early 2021.Throughout his career, Dr. Fox worked to challenge and advance the way scientists acquire, process, analyze and interpret data.He collaborated with NASA, NOAA, USGS, and other organizations, to explore ways in which published papers could be enhanced by direct links between the papers and the data and algorithms underpinning results and interpretations.He was therefore inspired to shift the publication paradigm, specifically to move toward a model whereby computational tools could feed directly into the preparation, submission, peer-review, publication and archiving of papers (Fox, pers. comm.).
A demonstration of this new publication paradigm is found in two seminal papers by Iacovino et al. (2021) and Elkins and Spiegelman (2021).The first paper presents Volatile Equilibria and Saturation Identification calculator, a comprehensive tool to model the solubility of water and CO 2 in magmas.Here, the authors explain the properties of the program, which is provided as python code, ready to be incorporated in other scripts.The second paper presents the theoretical underpinnings and computational tools for the forward melting modeling of uranium-series disequilibria in basalts.The static versions of both papers are published in the journal, but they also contain the link to their Jupyter notebook versions, which readers can use to run the code for their research without the need to compile a python script.Importantly, the notebooks were reviewed as part of the publication process.From these beginnings, stems the inspiration for Notebooks Now!
The current project flips the script to deliver the notebook as the version of record, while providing a static pdf as a secondary product.AGU invited experts from the computational notebook and publishing communities to join our efforts, first to identify the challenges involved in building and implementing this workflow at a November 2022 workshop, and then to begin to answer those questions and plot a path to a minimum viable product that can be pilot-tested, by joining working groups that looked at specific parts of a potential workflow: (a) Pre-submission; (b) Metadata; (c) Editorial and Peer-review; (d) Production and Post-production; and (e) Content Consumer.These efforts led to the development of a workflow that relies on a Manuscript Exchange Common Approach (MECA) bundle to package the contents of a notebook to usher it through peer-review and production systems.Currently, the project is progressing on schedule and at the stage of testing the technical workflow for peer review and publication of papers.The final stage of the project will involve soliciting new manuscripts through this journal, to run through the entire process of manuscript preparation, submission, peer-review and publication.

Concluding Statement
We are excited about the NN! project, which represents a revolutionary approach to the publication of scientific papers: primary research outputs are shared within the Notebooks Now! platform, and are directly linked to publication outputs.The principles of FAIR are applied, transparency and reproducibility are integral to the publication from the beginning of the research, data and code can be tested by peer-reviewers in the computational notebook sandbox, and, once the paper is published, data and code can be reused and repurposed directly from within the publication.Thus, the publication itself becomes a live research laboratory, from which new scientific