# Preliminary evidence package: lunar boulder-track candidate

Candidate `e824bf3ee775b25b10ab721b` was surfaced by the Phase 15 unsupervised ranking in the 2026 LROC product `M1528367567RC`. This package asks a narrower question: **is the apparent boulder/track morphology real and persistent in independent observations?** It does not claim that the movement occurred in 2026.

## Preliminary finding

The strongest historical frame, `M143053461RC` acquired 2010-10-30, predates the discovery frame by more than 15 years. On the terrain-projected north-up image it shows a continuous, slightly sinuous trace that crosses the model target and terminates at a bright block with a cast shadow. The terminal block remains independently visible under substantially different illumination in multiple later frames through 2026-03-17.

This supports the interpretation **genuine transported boulder with a persistent track**. It rules out an explanation based solely on 2026 detector patterning or illumination. It does **not** establish the fall date: the track may be geologically old. The upper trace is continuous, while the late segment contains repeated-looking marks compatible with bouncing, irregular tumbling, or shape-driven intermittent contact. That distinction remains a preliminary visual interpretation.

- Exact terrain-projected model target: `-43.790978195235°, 240.480250549430°E`.
- Preliminary terminal-boulder centre: `-43.79438976°, 240.48135610°E`, about 106.2 m southeast of the 8 m model target.
- Preliminary hand trace: about 228 m; treat this as a 190–230 m visual estimate until a specialist trace is accepted.
- Common display grid: 0.5 m/px, north up, 640 m square, Kaguya terrain reference.

## Historical observations

| Acquisition | LROC CDR | oblique m/px | incidence ° | phase ° | evidence weight |
|---|---:|---:|---:|---:|---|
| 2010-10-30 | `M143053461RC` | 0.556 | 49.83 | 48.79 | usable |
| 2011-07-30 | `M166637697RC` | 0.567 | 66.41 | 65.60 | usable |
| 2016-10-19 | `M1231525802LC` | 0.844 | 80.96 | 82.69 | caution |
| 2021-06-29 | `M1379606327RC` | 1.291 | 72.59 | 71.66 | usable |
| 2024-11-29 | `M1487481035LC` | 1.031 | 53.83 | 54.90 | usable |
| 2026-03-17 | `M1528367567RC` | 1.174 | 58.78 | 57.61 | usable |

Two initially acquired products are excluded from the review and all evidence counts: `M179585542RE` (human review determined image is noise and scientifically invalid); `M1439367760LE` (human review determined image is noise and scientifically invalid). Their source identity remains recorded for provenance, but they are not weak evidence; they are invalid inputs.

## Published inventory cross-check

Bickel_et_al_2020_i80_06.csv: no entry within 1 km; nearest is 2.157 km away (M1224464562LC_pyr). Bickel_et_al_2020_i60_06_2021-rerun.csv: no entry within 1 km; nearest is 2.157 km away (M1224464562LC_pyr).

The comparison is against both public `CT ≥ 0.6` catalogs from Bickel et al.: the 136,610-entry ±80° catalog used for the 2020 paper and the 92,842-entry ±60° 2021 rerun. Absence is not proof of novelty: the catalog is a CNN detection product with stated limitations, not an exhaustive nomenclature registry. Dataset: [Bickel catalog, DOI 10.17617/3.OG927P](https://doi.org/10.17617/3.OG927P), CC BY 4.0. Paper: [Bickel et al. 2020](https://doi.org/10.1038/s41467-020-16653-3).

The paper reports recall of about 0.44–0.69 at the selected threshold and explicitly lists missed detections, coverage gaps, illumination and noisy images among the limitations. A real feature being absent from these catalogs is therefore plausible. The same paper summarizes estimated track survival around 1.55–35 million years, with a proposed upper bound around 150–300 million years. The historical sequence here only proves “already present by 2010”; it cannot date the fall within that much larger geological interval.

## Reproducibility and uncertainty

1. The source centre was transformed from native pixel `line 20419.5, sample 4339.5` in `M1528367567RE` using ISIS 10, web SPICE and the Kaguya LALT terrain model. That produces the target coordinate above; it differs from the earlier catalog-coordinate approximation.
2. Eight NAC observations whose catalog footprints cover that terrain coordinate were downloaded as native EDRs, SPICE-initialized and projected onto one fixed equirectangular grid. Human review rejected two as noise, leaving six valid evidence observations.
3. The viewer keeps the six valid observations on that grid and lets the reviewer change date without changing viewpoint. It also retains exact LROC observation links and native line/sample positions.
4. Remaining registration error from spacecraft pointing and the coarse terrain model can be several to tens of metres on steep local relief. Use stable crater geometry, not pixel-perfect opacity, when comparing years.
5. The thin terminal circle and track polyline are preliminary markup and are **off by default**. They are not measurements accepted by a lunar geomorphologist.

## What is established vs pending

**Established locally:** the feature exists in public native LROC data; the clearest track is already present in 2010; the terminal block persists across independent observations; the candidate is not within 1 km of an entry in either public Bickel `CT ≥ 0.6` catalog.

**Pending external verification:** specialist confirmation of the transported-boulder interpretation; refined photogrammetric registration; a defensible block diameter/height and slope measurement; an authoritative search of site-specific publications and informal feature names. Official IAU naming is a separate question and is not implied by this package.