Sources & Methodology

This page states the evidentiary standard the rest of the site is held to. It exists because the public record on this event is unusually polluted: decades of official silence, followed by a mix of foreign press speculation, advocacy writing, and viral secondhand retellings, have left the topic with very few claims that are both specific and verifiable. This project would rather publish less, slowly, than repeat a number or a story it cannot trace to a source.

How sources are ranked

1. Primary engineering and technical records — dam design and construction documents, hydrological survey data, post-failure engineering investigations, and reservoir operation records.
2. Contemporaneous meteorological and hydrological data — rainfall and river-flow records from the period, and later peer-reviewed hydrological analyses that reprocess that raw data.
3. First-hand testimony — interviews with survivors, engineers, and former local officials, collected directly for this project under the consent and anonymization protocol described below.
4. Government and administrative archives — internal records, correspondence, and reports from the relevant period and its aftermath.
5. Peer-reviewed academic literature — used where it engages directly with (1)–(4) and shows its own sourcing; treated more cautiously where it does not.
6. Contemporary and later press accounts — used only where corroborated by one of the above, and always flagged as press reporting rather than primary fact.

Claims that exist only in secondary retellings — with no traceable origin in (1)–(4) — are not presented on this site as established fact. Where such a claim is widely repeated but unverifiable, that will be noted explicitly rather than left unaddressed, particularly on the Data & Statistics page, where casualty and damage figures are the most contested part of the record.

Provenance on every item

Every document, photograph, data point, and interview excerpt published here will carry:

• Source — where it came from (repository, collection, or interview subject).
• Date — of the original material, not of its publication here.
• Language and translation notice — most primary material is in Chinese; translations are the researcher’s own unless stated otherwise.
• Confidence notes — anything uncertain about authenticity, transcription, or interpretation.

Interview ethics

Interview subjects, and in some cases their family members, may face personal or professional risk from being publicly associated with this topic. Subjects are given the choice to be identified, partially identified, or fully anonymized (referred to by a coded identifier rather than a name), and identifying details are withheld whenever a subject requests it. Consent is sought explicitly for any material published on this site, separately from consent to be interviewed at all.

Rights & reuse

Much of this project’s underlying material consists of scanned published books, gazetteers, and government-affiliated compilations. Publishing full scans of copyrighted works on a public website is not something this project does, regardless of research purpose. Material is instead handled by tier:

• Reproduced in full — only material that isn’t copyrightable to begin with: Chinese laws, regulations, and official orders (excluded from copyright under PRC Copyright Law Art. 5), and factual data extracted from a source and re-presented in this project’s own tables (facts themselves are never copyrightable, only their original expression).
• Quoted — short, specific excerpts needed to support a specific claim, always with a full citation, under fair-use/reasonable-quotation principles — not extended to whole chapters or documents.
• Cited only — a full bibliographic citation and, where one exists, a link to a legitimate access point (library catalog, publisher, or lending service), with no file hosted here.
• Restricted — material still under copyright that isn’t reducible to the categories above stays out of public distribution. It may be described and cited, and made available to other researchers on request via Contact, but is not published as a downloadable file.

Every entry in the Archive carries a rights status noting which of these applies, so a reader always knows why something is — or isn’t — available to download here.

Corrections

This is a working archive maintained alongside an active research project, and it will contain gaps and, at times, errors. If you can point to a misattribution, a mistranslation, or a claim that needs a better source, see Contact. Corrections are tracked and, where they change a published claim, noted on the relevant page.