# Moon Arena — provenance

Retrieved 2026-08-05. Everything on the board traces to one of the sources below or is
recorded as absent. Nothing is imputed, estimated, or filled forward.

## Source of truth

**LROC Lunar Pit Atlas** — <https://lroc.im-ldi.com/atlases/pits> (redirected from
`https://www.lroc.asu.edu/atlases/pits/list`), maintained by the Lunar Reconnaissance
Orbiter Camera team at Arizona State University's School of Earth and Space Exploration.

Morphometric fields — latitude, longitude, funnel diameters, inner diameters, depth, and
the Overhang flag — were read from each pit's individual atlas page on the retrieval date.

Peer-reviewed context for the three classic mare pits and for the thermal and radar
evidence comes from:

- Haruyama et al. 2009 — SELENE/Kaguya discovery of the Marius Hills hole
- Robinson et al. 2012 — LROC NAC stereo confirmation of sublunarean voids
- Wagner & Robinson 2014 — pit distribution by host terrain
- Horvath, Hayne & Paige 2022 — Diviner thermal environments of lunar pits
- Carrer et al. 2024 — Mini-RF radar evidence of an accessible conduit below MTP

## Denominator

The atlas listed **278** pits on the retrieval date. This dataset captures **18** of them —
16 confirmed, 2 candidates. It is a curated seed for the board, not the catalogue, and the
board states so above the table rather than in a footnote. Wagner & Robinson 2014 reported
mare n=8, impact melt n=221, highland n=2 at that time; the atlas has grown since.

Any statement of the form "the best lunar pit is X" is therefore out of scope for this
artefact. The board ranks eighteen pits against each other and says so.

## What is derived rather than sourced

Four things on the board do not come from the atlas and are this project's own judgement:

1. **The four axes.** Access, mouth, shelter, thermal. Chosen because each maps to a field
   the atlas actually publishes for at least some pits.
2. **The weights** — 0.35 shelter, 0.30 access, 0.20 mouth, 0.15 thermal. A stated
   judgement, not a derived result. Thermal is weighted lowest because only two pits have
   the measurement, and a heavy weight on a near-empty field would let absence drive the
   ranking.
3. **The normalisation ranges.** Depth 10–120 m, inner diameter 10–160 m, both linear,
   both bounded by the extremes present in this sample. A different sample would move them.
4. **The two readings.** Measured-only divides by the weight that was measured;
   unmeasured-charged divides by the full weight. Both are in `app/app.js` in the open.

## Null handling

A missing measurement scores `null` and propagates to the display as an em dash. It is
never a zero, never an average, never a nearest-neighbour fill. Specifically:

- `depth_m` is null for two pits where the atlas publishes only a lower bound (">25 m",
  ">8 m"). A bound is not a depth.
- `overhang_or_cave` distinguishes three states: confirmed, explicitly absent, and not
  assessed. Eight pits are not assessed, and that is the largest hole in the board.
- `inner_diameter_m.max` is null for four pits the atlas lists as N/A; those maxima are not
  estimated from the minima.
- Sorting sends nulls to the bottom in both directions. A pit with no published depth is
  not the shallowest pit.

## Known retrieval problems

Carried in `data/moon_pits.json` under `errors` and rendered on the case page:

- The ResearchGate copy of Haruyama 2009 failed to scrape on every engine; the AGU DOI page
  was used instead.
- Horvath et al. 2022 full text sits behind the publisher. Values come from the ADS
  abstract and the DOI record.
- Carrer et al. 2024 is largely paywalled. The abstract and open figure captions were
  retrieved; conduit length is stated in the paper as "tens of metres" and is not given a
  precise figure here.
- Haruyama et al. 2010, which first reported the Mare Tranquillitatis and Mare Ingenii
  pits, was not separately retrieved. Discovery credit for those two is attributed through
  Robinson et al. 2012, which cites it.
- The atlas landing page has no prose description; maintainer attribution is taken from the
  page footer.

## Files

| Path | What it is |
|---|---|
| `data/moon_pits.json` | The dataset, with `denominator`, `errors` and `field_null_log` intact |
| `app/index.html` | The board. Data inlined, so it runs offline |
| `app/app.js` | Axes, weights, both scoring readings, null handling |
| `app/app.css` | Board styles, on the site's shared tokens |
| `index.html` | This case study |