Method
Where the data comes from, what this site does and — above all — what it does not do.
Where the references come from
Each study is added from its DOI. A script queries OpenAlex, then Crossref as a fallback, and stores the metadata exactly as returned: title, authors, journal, year, abstract where one is released, and open-access status.
These fields are never typed or edited by hand. The retrieval date is shown on every record: beyond it, the metadata may have changed at the source.
What is an editorial decision
Assigning a study to a theme, including it in a topic brief, and any contextual note are human decisions made when the study is added. They are fallible and open to challenge.
The selection of indexed studies does not follow a systematic protocol: this is not a literature review. It favours peer-reviewed work, reports from scientific institutions, and widely cited syntheses.
How a study is selected
Four criteria, in this order. The publication must have been peer-reviewed, or come from a scientific institution whose review procedure is public — the IPCC, space agencies, national weather services. A preprint is indexed only if it carries authority in its field, and its status is then displayed.
Priority goes to syntheses and meta-analyses over isolated primary studies: a literature review states what a field has established, a single result states what one team found. This ordering avoids giving exploratory work the same weight as a consensus.
Citation counts serve as a signal, not a criterion. They identify the structuring work of a field, but mechanically favour older publications and popular subjects: they are corrected by hand so that whole parts of the corpus are not left in the dark.
Finally, thematic coverage is watched. An under-represented theme is filled in deliberately, which the radar on the Themes page makes visible. A directory whose balance depended only on the accidents of research would give a distorted picture of the literature.
Abstracts and numbers
Abstracts are the publishers' own, reproduced without rewording. Some publishers release no reusable abstract; those records link straight to the DOI instead.
Citation counts come from OpenAlex and are labelled as such. They differ between databases and sometimes badly understate the real figure: treat them as orders of magnitude, not measurements.
Publication type also comes from OpenAlex, which often gets it wrong: a literature review is frequently tagged as an article. Where the classification is plainly mistaken it is corrected by hand and the record says so. Without that, the directory's facet would contradict the selection criterion stated above.
Charts
Numerical series are versioned in the repository along with their source and retrieval date, both displayed under the chart. A chart whose provenance cannot be shown is not published.
What this site does not do
It publishes no original research and organises no peer review. The work indexed here was reviewed — or not — by its own publishers, not by this site.
It does not rank studies by quality or reliability, and offers no voting mechanism. An approval counter says nothing about the soundness of a piece of research.
It is not affiliated with any institution, journal, NGO or public body, and does not display any organisation's branding.
Reporting an error
Mistakes in theme assignment, translation or context are possible. Content and code are public: corrections can be proposed directly on the repository.