Every number on this page is read from a committed record and reconciled against the other records before the page is written. Nothing here is fetched at runtime: no network, no API key, no computation on load.
Every window the decisions were made on, with the metadata needed to check that both sides of a comparison cover the same ground. Footprint is what was compared; support is the cell size; the effective floor is the coarsest support the source can honestly carry.
| regime | reference | body | footprint | support | cells | posting | floor | valid |
|---|---|---|---|---|---|---|---|---|
| lunar_panel_v2 | LUNAR_Site01 | moon | 9.0 km | 1000 m | 9×9 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site04 | moon | 9.0 km | 1000 m | 9×9 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site07 | moon | 9.0 km | 1000 m | 9×9 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site11 | moon | 9.0 km | 1000 m | 9×9 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site20 | moon | 9.0 km | 1000 m | 9×9 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site23 | moon | 9.0 km | 1000 m | 9×9 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site01 | moon | 9.0 km | 500 m | 18×18 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site04 | moon | 9.0 km | 500 m | 18×18 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site07 | moon | 9.0 km | 500 m | 18×18 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site11 | moon | 9.0 km | 500 m | 18×18 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site20 | moon | 9.0 km | 500 m | 18×18 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site23 | moon | 9.0 km | 500 m | 18×18 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site01 | moon | 9.0 km | 750 m | 12×12 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site04 | moon | 9.0 km | 750 m | 12×12 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site07 | moon | 9.0 km | 750 m | 12×12 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site11 | moon | 9.0 km | 750 m | 12×12 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site20 | moon | 9.0 km | 750 m | 12×12 | 5 m | 15 m | 1.000 |
| lunar_panel_v2 | LUNAR_Site23 | moon | 9.0 km | 750 m | 12×12 | 5 m | 15 m | 1.000 |
| mars_panel_v3 | mars_daedalia | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_daedalia | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_daedalia | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_elysium | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_elysium | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_elysium | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_amazonis | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_amazonis | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_amazonis | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_syria | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_syria | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_syria | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_hesperia | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_hesperia | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_hesperia | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_noachis | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_noachis | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_noachis | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_coprates | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_coprates | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_coprates | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_arcadia | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_arcadia | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_arcadia | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_cerberus | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_cerberus | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_cerberus | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_solis | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_solis | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_solis | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_terra_sabaea | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_terra_sabaea | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_terra_sabaea | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_terra_sirenum | mars | 60.0 km | 1500 m | 40×40 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_terra_sirenum | mars | 60.0 km | 2000 m | 30×30 | 463 m | 1389 m | — |
| mars_panel_v3 | mars_terra_sirenum | mars | 60.0 km | 3000 m | 20×20 | 463 m | 1389 m | — |
Each row is one comparable window. Ranking asks which candidate is closest among those tested. Adequacy asks whether any of them is inside the reference's own morphological variability. Sufficiency asks whether the evidence is enough to answer either. They are filled in independently and they disagree, which is the point of showing them side by side.
| cell | ranking | unit | worst axis | adequacy | sufficiency |
|---|---|---|---|---|---|
| Site01@1000m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.12 | NOT_ESTABLISHED | marginal |
| Site01@500m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.21 | NOT_ESTABLISHED | marginal |
| Site01@750m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.12 | NOT_ESTABLISHED | marginal |
| Site04@1000m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.71 | NOT_ESTABLISHED | marginal |
| Site04@500m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.70 | NOT_ESTABLISHED | marginal |
| Site04@750m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.70 | NOT_ESTABLISHED | marginal |
| Site07@1000m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.11 | NOT_ESTABLISHED | marginal |
| Site07@500m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.16 | NOT_ESTABLISHED | marginal |
| Site07@750m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.12 | NOT_ESTABLISHED | marginal |
| Site11@1000m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.36 | NOT_ESTABLISHED | marginal |
| Site11@500m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.25 | NOT_ESTABLISHED | marginal |
| Site11@750m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.32 | NOT_ESTABLISHED | marginal |
| Site20@1000m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.17 | NOT_ESTABLISHED | marginal |
| Site20@500m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.17 | NOT_ESTABLISHED | marginal |
| Site20@750m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.17 | NOT_ESTABLISHED | marginal |
| Site23@1000m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.36 | NOT_ESTABLISHED | marginal |
| Site23@500m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.24 | NOT_ESTABLISHED | marginal |
| Site23@750m | RECOMMEND_FOLLOW_UP provisional | earth_nevada_basin | 0.31 | NOT_ESTABLISHED | marginal |
30 perturbations were declared before the stress test ran, with the pass rule fixed in advance. 28 of 30 left the ordering untouched. The grid below is every cell against every counted perturbation — the two reversals are the falsifiable part and they are as visible as the passes.
| +400m_e | -400m_e | +400m_n | -400m_n | 500-to-4 | 500-to-6 | 750-to-9 | 1000-to- | 6000m | 3000m | 10500m@5 | 9750m@75 | 10000m@1 | bilinear | nearest | border_r | tail10pc | noise3pc | tilt0.5d | smooth3x | mean_slo | mean_rel | elev_std | earth_nu | earth_ll | earth_de | earth_sn | earth_ha | earth_od | earth_an | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Site01 @ 500 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | R |
| Site01 @ 750 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | R |
| Site01 @ 1000 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | R |
| Site04 @ 500 m | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site04 @ 750 m | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site04 @ 1000 m | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site07 @ 500 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | R |
| Site07 @ 750 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site07 @ 1000 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site11 @ 500 m | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site11 @ 750 m | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site11 @ 1000 m | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site20 @ 500 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | R |
| Site20 @ 750 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | R |
| Site20 @ 1000 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | R |
| Site23 @ 500 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site23 @ 750 m | · | · | · | · | · | · | · | · | · | R | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
| Site23 @ 1000 m | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · | · |
footprint:3000m — 11 of 18 cells change. footprintadd_unit:earth_andean_volcano — 7 of 18 cells change. pool_changeThe lunar panel decides a winner by Pareto dominance over three declared axes. This section asks whether that winner is a property of the terrain or of the rule — and answers it against a control, because a Pareto frontier over eight units in three dimensions is expected to be fragile when a fourth coordinate arrives, any fourth coordinate.
| measurement | value | what it means |
|---|---|---|
| candidate axes decision-tested | 141 | every admissible scalar in the engine’s own descriptor bank was added as a fourth axis in turn and the layer-1 set recomputed at all 18 cells |
| cells whose winner set changed | 0 to 18 of 18 | 42 of the 141 axes changed nothing at all; adding a coordinate can only ever promote a co-winner, never remove one |
| an arbitrary coordinate | 15.90 of 18 cells on average | over 300 random draws, median 18, leaving the frontier intact in 12% of them — this is the baseline the numbers above must beat |
| axes beating that control | 0 | none. Every change above is what adding a dimension does, not evidence about a candidate |
| most independent descriptor | /variogram/nugget, adjusted R² -0.511 | the three declared axes predict it no better than three random numbers would (chance level R² 0.4286) — and it changes the decision at 0 of 18 cells, which is how independence and decision-relevance come apart |
| what the axes cannot see | 48 of 180 bank scalars | the three declared axes are exactly blind to a quarter-turn rotation of a planar surface (verified to 1e-13); the scalars listed here are the ones that are not, and they are the only ones this symmetry argument can even point at |
EXPLORATORY SENSITIVITY PROBE.| diagnostic | value | what it means |
|---|---|---|
| descriptor dependence | effective rank 1 of 3 | mean slope and 3×3 relief are the same gradient up to a direction factor; the “three axes” are one measurement, so this is a one-dimensional screen |
| the relation, verified | worst relief error 4.4e-16 m | checked against the engine's own slope and relief functions on synthetic planes, and ρ observed inside the provable [1.0000, 1.4142] |
| planar fit on the panel | ρ = 1.319, R² = 0.9989 | a one-parameter fit to real terrain recovers the direction factor the algebra predicts |
| error ensemble | 6 of 100 published realisations | the archive's own perturbed DEM clones, not independent per-pixel noise, which would average away under aggregation and manufacture stability |
| adequacy bands | 0 of 18 usable | a calibrated band that admits known-different terrain is not a yardstick, so it is demoted rather than reported as a positive |
| resolution limits | Mars floor 1389 m > every lunar support | the comparability gate refuses a candidate below its source's effective resolution; Mars is excluded from the lunar regime by that arithmetic, not by preference |
| the nearest unit | a sibling lunar site in 18 of 24 site × descriptor statistics | Nevada is the closest Earth candidate everywhere, and not the closest unit: the six nearby sites are more like each other than any Earth terrain is |
The atlas is assembled from six committed records and refuses to publish if any two of them disagree. Each check below names both records it holds together, so a reader can go and read them.
| check | the claim it holds | the records |
|---|---|---|
| documented_hash:v3_bundle | the v3 bundle hash printed in the v3 protocol equals the file on disk | docs/BENCHMARK-PROTOCOL-v3.md <-> artifacts/mf_panel_v3/matched_panel_v3_real.json |
| documented_hash:v3_config | the v3 config hash printed in RESULTS-v3 equals the config on disk | docs/RESULTS-v3.md <-> config/matched_footprint_v3.yml |
| lunar_headline:panel | all 18 lunar cells share one single-member non-dominated set, and it is Nevada | artifacts/lunar_sites_v2/lunar_sites_v2.json |
| lunar_headline:stress_baseline | the stress test re-derived the panel's headline rather than assuming it | artifacts/lunar_adversarial/lunar_adversarial.json |
| ranking:adequacy_matches_panel | every cell the adequacy report ranks recommends the panel's survivor | artifacts/lunar_adequacy/lunar_adequacy.json <-> artifacts/lunar_sites_v2 |
| dependence:agrees_with_stress | the audit says one axis and dropping any single axis changes nothing | artifacts/dependence <-> artifacts/lunar_adversarial |
| stress:summary_matches_rows | stable + reversing perturbations equal the counted total | artifacts/lunar_adversarial/lunar_adversarial.json |
| adequacy:summary_matches_cells | the summary's sibling-nearest count is what the cells say | artifacts/lunar_adequacy/lunar_adequacy.json |
| regimes:disjoint_supports | the two regimes share no support, so nothing may be pooled across them | the two declared regimes |
| v3:hypothesis_counts_sum | the R-rule change counts sum to the number of cells they classify | artifacts/mf_panel_v3/matched_panel_v3_real.json |
| axis:probe_reproduces_panel | the probe re-derived the panel's single-member frontier at all 18 cells | artifacts/lunar_axis_probe <-> artifacts/lunar_sites_v2 |
| axis:no_pareto_removals | every added axis promoted co-winners only; dominance makes a removal impossible | artifacts/lunar_axis_probe/lunar_axis_probe.json |
| axis:control_is_present | the probe's frontier numbers travel with the control that bounds their reading | artifacts/lunar_axis_probe/lunar_axis_probe.json |
artifacts/discovery_atlas/discovery_atlas.json (sha256 f50ff74ea8bf693f…) and artifacts/lunar_axis_probe/lunar_axis_probe.json. Regenerate with python scripts/build_atlas_page.py. Scope: map-scale terrain-morphology screening only — no traversability, geologic-equivalence, site-access or mission-readiness claim, and no composite analog score anywhere on this page.