# Mars Arena — provenance (rebuilt 2026-08-06) This replaces the 2026-08-04 lift, which wrapped the same catalogue in a leaderboard. That page described a scoring UI over 1,062 rows; this one is an instrument over two populations with live orbital imagery, and the numbers below are the ones it actually reads. Source tree: `LEGIONFARM-3.0/mars-arena/`. Built with `tools/build_static.py`, which refuses to publish a payload that references the dev server, is missing a script it loads, lacks a canonical or description, or whose data disagrees with itself. ## Data sources | What | Source | Retrieved | Notes | |---|---|---|---| | 1,062 cave candidates | NASA PDS4 Mars Global Cave Candidate Catalog (MGC³), Cushing 2016 · `urn:nasa:pds:mars_mro_odyssey_multi_cavecatalog_cushing_2016` | archive copy, verified against its PDS4 label | Parsed positionally against the label, not by the CSV header — see "column order" below | | Corpus graph: 56 nodes, 52 edges, 72 claims | `05_graph/` in this repo, extracted from 136 canonical episodes | 2026-08-06 | Read only by this lane; the graph itself is maintained elsewhere | | 9 plotted anomaly coordinates | IAU Gazetteer of Planetary Nomenclature; HiRISE / CTX / MOC product labels; CIA Reading Room (Stargate) | 2026-08-06 | Every one carries the URL it came from, shown in the card | | Global basemap | NASA Trek, Viking MDIM21 colour mosaic, 232 m/px, equirectangular | live, per session | Fetched by the browser directly; `access-control-allow-origin: *` | | Close crops | USGS Astrogeology WMS, `planetarymaps.usgs.gov` | live, per candidate | THEMIS controlled mosaic by default, with a fallback chain for places it does not cover | ## Four corrections to the previous build 1. **Targeting priority was inverted.** The archive's scale is a targeting queue, not a confidence score, and it does not run low-to-high with the number: 1 is the highest priority, 3 is the most ambiguous ("80–90% will probably not show cave-like characteristics"), and 0 means HiRISE has already imaged it — off the scale, not at the bottom of it. The previous build treated 3 as strongest, in both the marker sizes and the science term of the score. 2. **Six of fifteen type-code expansions were invented** rather than read from the archive description. `srp` — 218 records — was labelled "sinuous rille pit"; the authors define a small, shallow, circular rimless collapse pit. `pol` was labelled "polygonal feature" and means polar ice region, which the data confirms: all eight `pol` records sit at 84–85° N or −71.9° S. Two codes (`con`, `smrp`) appear in the data and in no table in the archive; they now say so. 3. **The count is 1,062.** `data/Mars_Cave_Catalog.csv` is 1,063 lines including its header, and the PDS4 label declares `1062` for the table. The figure 1,079 appears nowhere in the archive. 4. **Column order.** `data/Mars_Cave_Catalog.csv` declares `longitude,latitude,Label,…` and its first data row is `APC001,240.4585,-0.94678,…` — the label is first and the header is shifted by one. The shapefile CSV in the same archive declares the same first two columns and means them. Anchor on APC001 = lon 240.4585, lat −0.94678. ## The anomaly layer, and what it refuses to do 14 places from the corpus were researched. **9 are drawn. 5 are not**, and the file records why for each: - `AA-OBJ-CYD-004` The City / City Square — no cited coordinate; the research returned `missing-evidence` rather than a guess. - `AA-OBJ-CYD-005` The Fort — same. - `AA-OBJ-OTH-002` Phobos monolith — on Phobos. - `AA-OBJ-OTH-005` ALH 84001 — a meteorite recovered in Antarctica. - `AA-OBJ-OTH-007` Mars — the planet itself, the base node for planet-wide claims. The build script asserts, and refuses to write the file otherwise, that every drawn object has both a cited coordinate source and an established scientific reading. An object rendered with the claim and nothing beside it is the failure this layer exists to prevent. Coordinates are planetocentric, east longitude 0–360. Confidence is one of `exact`, `approximate` or `region-only`, and `region-only` is shown to the reader as "treat the position as a label, not a location" — two of these objects are named regions ~3,500 km across, and the imagery tab says so when the crop is smaller than the region. ## The one number this project measured itself The Syrtis–Isidis mound (object 9) has no published size in the literature. Measured here on the HiRISE `ESP_077288_1890` RED browse image at 3.596 m/px, derived from the product label's `MAP_SCALE` and footprint: **3.1 km** across the ridge crests, **4.85 km** including the bright apron, quoted as **3.4 km** for the bright edifice at a mid threshold. The threshold is a real judgement and the sensitivity to it is published rather than hidden. The viral post's "about 3 miles corner to corner" = 4.83 km sits at the top of that range. Method: `mars-arena/research/out/pyramid.md`. ## Score weights (unchanged, and still a hypothesis) `safety 0.4 · location 0.3 · logistics 0.2 · science 0.1`, renormalised by their sum if a reader changes them. Every term, its value, the rule that produced it and whether a missing measurement was substituted are shown in the inspector. Archive priority feeds exactly one term — scientific interest — and that is stated in the rule column rather than folded silently into the total. ## Not claimed - No NASA, USGS or JPL validation of any ranking, score or ordering on this page. - The habitability score is not a peer-reviewed habitability model. - The anomaly layer endorses neither the claims nor, where they conflict, the established readings; it reports both with their sources. - `APC_Diameter` and `APC_Depth` are strings by schema. Non-scalar values such as `135x195` are preserved verbatim and left unparsed. An unmeasured feature is not a zero-metre feature, and the previous import's zeros came from pretending otherwise.