# Vumarii 2019 upper diagram: Dunhuang Star Map attribution

## Finding

The Vumarii upper panel is a cleaned, re-inked, cropped rendition of the north-polar detail of the Dunhuang Star Map (British Library Or.8210/S.3326), as reproduced in the archived Wikimedia reference image. The identification is supported by the shared position and topology of multiple distinctive asterism clusters, the enclosing paths, and the southern chain; the labelled comparison marks nine non-independent visual regions.

This is a **visual-reference attribution**, not a claim that the panel independently maps the modern sky, identifies every Chinese asterism, selects a traversal, or decrypts the Vumarii text.

## Reproducible geometry support

The fixed overlay uses source crop `[200, 280, 1760, 1560]` resized to the Vumarii diagram's `830 x 592` crop. In a 500-crop bounded comparison, the original orientation's best local-darkness contrast is `3.1935`. The same search gives `2.5729` for a horizontal mirror, `2.3113` for a vertical mirror, and `2.0162` for a 180-degree rotation.

The reference was found visually before the overlay was tuned. These orientation values are therefore a transparent supporting check, not a blind-discovery probability or proof of authorial intent.

## Boundary

The source chart explains why the upper panel looks like an astronomical diagram. It does not give a source-selected mapping for the 2019 glyph body, a hidden payload, a Vumarii language, or a relationship to Page 23. It narrows the open question from ‘what picture is this?’ to ‘what, if anything, does the author intend readers to do with this historical astronomical reference?’

## Sources

- Dunhuang reference image: `archive/recon/reference-images/dunhuang-star-map.jpg` (SHA-256 `fa7a2ac9d5530e48c36d3b04af314733cf2aba3df46257ead2f5fd019488cb1a`), public source: https://commons.wikimedia.org/wiki/File:Dunhuang_star_map.jpg
- Vumarii target image: `archive/gaps/tr1-00-files/vumarii-2019-imgur-7a4FZ6Q.png` (SHA-256 `bfbdd8b5847e8ae87cbfbdb4cb1894e69fd2a40a90894ab4db875d1c410160a4`)

## Reproduce

```powershell
& '<bundled-python>' scripts\audit_vumarii_2019_dunhuang_attribution.py --fetch-reference
```