the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Mar 2026
A Reproducible Research Framework for Wind Energy
Abstract. With the rapid growth of wind energy installations and the corresponding increase in data-driven research outputs, establishing robust reproducible research (RR) practices has become essential to ensure reliability, transparency, and long-term scientific value. Wind energy research is characterised by high data volumes, complex computational workflows, and increasing reliance on advanced modelling and machine learning techniques, all of which amplify reproducibility challenges. This paper examines the current state of reproducible research practices within the wind energy domain and identifies key gaps that limit computational reproducibility and replicability. In response, it proposes a structured, sector-specific reproducible research workflow designed to improve transparency, reliability, and ease of replication. The proposed workflow spans three stages; Conceptualisation and Planning, Implementation and Execution, and Dissemination. These emphasise essential components such as systematic data management, code sharing, platform selection, version control, and comprehensive documentation. In addition, the paper introduces a set of Python-based tools and best practices that support reproducibility at each stage of the workflow. A practical case study in wind power forecasting using an open-access dataset and a publicly available GitHub repository is presented to demonstrate the application of the workflow and to highlight common reproducibility challenges, particularly those related to data sharing, preprocessing, and documentation. The results show that adopting structured reproducible research practices enables transparent verification, facilitates independent replication, and enhances the reusability of computational wind energy studies. Collectively, the proposed framework and case study provide actionable guidance to support more reliable, verifiable, and collaborative wind energy research.
- Preprint
(14482 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (extended)
- RC1: 'Comment on wes-2026-52', Anonymous Referee #1, 17 Apr 2026 reply
-
RC2: 'Comment on wes-2026-52', Anonymous Referee #2, 27 Apr 2026
reply
The manuscript outlines a framework for improving the reproducibility of the computational and data analysis steps within wind energy research. The authors provide a high-level overview of the elements required for reproducibility, including a diagram (Figure 2 in the manuscript) with the general workflow elements.
The topic of the study is timely and relevant and can potentially have positive impact on the practices in the wind energy research community. However, the current descriptions are high-level and lack a structured process, specifications and acceptance criteria for the different elements of the framework. I suggest the manuscript is significantly reworked to introduce concrete and structured guidance. Below are a few additional comments providing further elaboration:
General comments
- The core elements of the framework, sections 2 and 3, are vague and not fully organized – they do not provide a concrete procedure that a user can independently follow. I suggest that a concrete procedure is defined, with clear objectives, targets, constraints (e.g. minimum requirements) and recommended practices (or relevant standards) for each step. Basically, this means adding detailed specifications to each white box in Figure 2.
- I believe that reproducing experimental conditions in wind energy is an extremely difficult task since field experiments and operational data are subject to large variations in conditions and large measurement uncertainty / lack of observability. Due to that, I believe one can’t interchangeably use the term “reproducibility” when meaning computational reproducibility in particular. I suggest changing the paper title to specifically suggest that the framework targets computational reproducibility, and also make this distinction completely clear throughout the paper, and avoid using the term “reproducibility” in the parts of the paper where the meaning of the term can be confused.
- I am not sure – is the goal to reproduce a study where the researcher doesn’t have access to the full codebase, or is the goal to enable another researcher to run code as provided by the original authors? I would say the former is reproducibility, while the latter may better be described as code portability. In my view, the text in sections 1-3 leans towards a reproducibility scenario, but the example in section 4 demonstrates a mix of code portability and reproducibility aspects. Further, in line with my previous comment, consider making the distinction between reproducibility and code portability throughout the paper and in the title if necessary.
Technical comments
- Introduction: the authors first make a distinction between experimental reproducibility and computational reproducibility, later in the same section they consider computational reproducibility against replicability, and finally the mention “consistency” within the same context. Based on the explanations provided by the authors, there is an overlap between experimental reproducibility, replicability and consistency. Consider rephrasing and clearing out the terminology to avoid confusion.
- One way of standardizing data handling and sharing is using data ontologies, schemas and fixed vocabularies. Consider adding this standardization step explicitly in the conceptualization and planning step.
- Lines 129-140 and elsewhere in the text: multiple confusing citation formats (name repetitions, long reference titles, no punctuation) – please update.
- Figure 8: I am not sure this figure and the comparison used to generate it supports the reproducibility claim. I believe a more convincing case of computational reproducibility would be if two teams independently apply all analysis steps (including and least partially independent code implementation) on the same dataset and obtain identical or nearly identical results.
Citation: https://doi.org/10.5194/wes-2026-52-RC2 -
RC3: 'Comment on wes-2026-52', Anonymous Referee #3, 07 May 2026
reply
The theme of this manuscript is reproducible research where the underlying workflows and tools have a significant computational component. The authors argue that the concepts and workflows are specifically relevant to wind energy research. The discussion, proposed computational tools and techniques, and the example workflow are tied to specific, commonly used software (at least commonly used at the time of this review).
Regardless of the methods and tools used for published research, the need for reproducibility is understood and appreciated. However, this manuscript does not rise to the level required of an archival journal paper. Rather than a journal-quality manuscript with long lasting and broad impact, it is written more like a tutorial document focused on a specific workflow with a specific set of tools.
The authors' motivation is that the "gap in wind energy research practices" is that "computational reproducibility of research findings remains limited". If that is the case, it is not because of lack of knowledge or resources. The methods (e.g., modern software development best practices) and tools (e.g., git, GitHub) described here are well known and computational researchers should already be adopting them in some form. Many statements are so obvious that they are not needed, e.g., "the core principles of the algorithms and methods used in the study should be clearly described in the published paper." The following are some links to organizations and workshops showing that the topics covered in this paper are very well tread:
https://www.repro4everyone.org/
https://events.library.ucdavis.edu/event/reproducible-research-intensive-2024
https://www.science.org/doi/10.1126/science.aah6168
https://www.swissrn.org/The manuscript touches on software engineering best practices, but again these are documented in many places and are commonly used. I encourage the authors to review the extensive resources gathered by the Better Scientific Software foundation (BSSw, https://bssw.io/).
The narrative applies to any research field with a significant computational component, so the workflows described do not necessarily fit well in a wind-energy specific journal.
The paper's content is a snapshot of the current open-source software tools for software development, e.g., docker, GitHub, and the current specifics of those tools, e.g., GitHub file size limits. These specific values are likely to change, and makes any value of this manuscript too temporary.
While there is a myriad of codes, packages, languages used in wind energy research, the authors further narrow the applicability of their proposed framework given than the tools discussed are "primarily related to Python-based wind energy research."
The paper should be written in more general terms describing best practices, which would have longer term impact. Even still, those best practices should be already be known and practiced by professional computational researchers, so it is not clear that there is sufficient novelty to warrant a journal paper.
Minor Points:
- Many of the inline references are formatted incorrectly, e.g., "turbines located at the Kelmarsh wind farm Plumley (2022)"
- Spelling mistakes in Figures 1 (e.g., "Anaylsis") and Figure 2 (e.g., "Reprduction"); other spelling mistakes peppered throughout paper.
- "Github" -> "GitHub"
- pg 8: "Github is the platform that compliments Git" -> "GitHub is a platform that compliments Git"; there are hosting option, e.g., GitLab, bitbucketCitation: https://doi.org/10.5194/wes-2026-52-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 277 | 205 | 20 | 502 | 16 | 24 |
- HTML: 277
- PDF: 205
- XML: 20
- Total: 502
- BibTeX: 16
- EndNote: 24
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This paper emphasizes the importance of adopting a reproducible research framework in wind energy research and proposes recommendations for what such a framework should entail. While the underlying motivation is both timely and important, the proposed framework is presented in a way that limits its practical applicability for many researchers in the wind energy community. In the following, I outline several key shortcomings that, in my view, prevent the paper’s recommendations from being readily actionable.
The paper basically describes a software process. Granted, many wind energy related evaluations or computations require resource intense calculations or modeling. For those, such a process as described in this paper is justified. There are, however, many other research avenues in wind energy: observational analyses, the challenge of how to match observations to model output, how to perform wind resource assessments, choosing thresholds for various analyses, dealing with observations, how to use AI in wind energy research, etc. Section 4.2.1 is getting close to this and what I would expect this whole paper to be about. I would argue this paper is rather narrow in scope and more about “A reproducible research framework for complex computational simulations: Applications to wind energy forecasting”. I would further argue that you could substitute wind energy in this paper for any other discipline. What would make the paper wind energy specific? Proposing a reproducible framework to tackle the research avenues I mentioned above would make the paper wind energy specific.
It is not clear to me how a researcher in the wind energy industry would apply what is proposed. The authors state that the framework can be “readily adopted”, but probably only if you have a software engineering background. The framework is very specific to highly computational workflows, and while wind energy does contain them, the research done by the community also contains other workflows. I can see this framework adopted by academia rather than industry. What it lacking is a plan for adoption, a plan for how to convince others to put in the extra effort to follow this approach. Researchers are sometimes hesitant to learn a new coding language or change the way they have been working, so I think equally important to suggesting a framework is to suggest how to adopt a rigorous software process across an entire industry.
An important question that needs to be tackled is also: How does AI play into this? A lot of what you mention in section 3.2 can be done by an AI agent.
The authors state that “The results show that adopting structured reproducible research practices enables transparent verification, facilitates independent replication, and enhances the reusability of computational wind energy studies. Collectively, the proposed framework and case study provide actionable guidance to support more reliable, verifiable, and collaborative wind energy research.” In my opinion, the guidance is not “actionable” without describing how to convince the (international?) wind energy community to adopt such a framework.
Minor comments:
Reference are wrongly formatted throughout the paper.
Line 84: “is presented”
Line 205: “Furthermore”
I’d encourage you to think more about the framing of the paper and what you would like to accomplish. What would make this idea being adopted, adopted in what way? You might want to include more case studies to show how relevant this is across the wind energy community, or narrow the scope of the paper.