Reproducibility is a credibility multiplier and increasingly a funder/journal requirement. AEC's reliance on simulations, bespoke datasets and proprietary tools makes this especially important — and often neglected.
1. FAIR data principles
- Findable — deposit with a persistent identifier (DOI) and rich metadata; index it.
- Accessible — retrievable via a standard protocol; state access conditions (open vs controlled).
- Interoperable — use standard formats/vocabularies (open formats, documented schemas; for built-env data consider IFC, CityGML, standard CSV + data dictionary).
- Reusable — clear license, provenance, and enough documentation to reuse correctly.
FAIR ≠ "open": sensitive/proprietary data (site, personal, commercial) can be FAIR yet access-controlled.
2. Data repositories
- General: Zenodo (CERN; DOIs, GitHub integration), Figshare, Dryad, OSF, Harvard Dataverse.
- Institutional repositories; domain repositories where they exist.
- Choose one that issues a DOI, supports your license, and offers versioning. Avoid sharing data only as "available on request" — it usually isn't.
3. Data Management Plan (DMP)
Plan early (many funders require it at proposal stage). Cover: what data are collected/generated; formats & volume; metadata & documentation standards; storage, backup & security during the project; legal/ethical constraints (consent, GDPR, IP, confidentiality); what will be shared, when, where, and under which license; long-term preservation; and roles/responsibilities. Tools: DMPonline / DMPTool / Argos.
4. Computational reproducibility (code & analysis)
- Version control everything (Git); tag the exact commit/release used for the paper.
- Capture the environment:
requirements.txt/environment.yml(conda),renv(R), lockfiles; or a Docker/Apptainer container or Binder for one-click reproduction. - Literate analysis: Jupyter, Quarto/R Markdown so results, figures and tables regenerate from data.
- Pin versions of tools, libraries, solvers and (for energy work) weather files — results shift between versions.
- Provide a README with exact run instructions and a mapping from scripts → figures/tables in the paper.
- Archive a release of the code with a DOI (Zenodo–GitHub integration) and cite it.
5. Reproducibility for simulation/modeling studies (AEC-critical)
- Share model input files (e.g., EnergyPlus IDF + EPW, OpenSees/FEM input, Grasshopper definitions) where IP permits; document all assumptions, boundary conditions, and solver settings.
- Report calibration/validation against measured data and uncertainty/sensitivity analysis — a model
that can't be reproduced or validated is not credible (see
research-methods).
6. Preprints, preregistration & registered reports
- Preprints: engrXiv, arXiv (some EESS/CS), SSRN, OSF Preprints — establish priority and open access; check the target journal's preprint policy first.
- Preregistration — register hypotheses, design and analysis plan before data collection (OSF, AsPredicted) to curb HARKing/p-hacking for confirmatory work.
- Registered Reports — peer review of the protocol before results; in-principle acceptance reduces publication bias. Growing in engineering venues.
7. Open access (see aec-journals)
- Gold/Green/Diamond/Hybrid OA; deposit accepted manuscripts in a repository (Green) even when publishing closed. Comply with funder/Plan S mandates; use available transformative agreements.
8. Data & code availability statements (now standard)
Write an explicit statement, e.g.:
- "The data and analysis code supporting this study are openly available in Zenodo at , under ."
- "Restricted data: the building-monitoring dataset cannot be shared publicly due to a confidentiality agreement with the building owner; aggregated data are available from the corresponding author on reasonable request." State the truth — don't promise sharing you won't honor.
9. Credit & licensing
- License data (e.g., CC BY) and code (e.g., MIT/Apache-2.0/BSD, or GPL if required) explicitly.
- Add a CITATION.cff so others cite your software/data correctly; cite the datasets and tools you used
(
citation-formats,aec-datasets-tools). - Disclose AI-tool use per venue policy (
research-ethics-integrity).
10. Integrity link
Reproducibility is an integrity practice: it lets others verify, not just trust. Never present a workflow as reproducible if it isn't, and never share data in a way that breaches consent, GDPR, IP or confidentiality.