The cube.

§1 What Bharat Strata is

Bharat Strata is a parquet-backed knowledge graph of Indian locality. Each cell is a reconciled record of what 20 witness tiers reported about a place, what 9 reconciliation passes inferred, which of 23 falsifier modules passed, and which of 10 named floor keys apply when the substrate cannot honestly answer. Every cluster carries its own SHA-256 + schema fingerprint and the doctrine version that produced it.

§2 What's inside a cell

A cluster is a polygon (typically H3-r9 at MMR / MH scope, H3-r6 at pan-India). Each cell persists a header (identity + lineage + provenance) and one row per column with sibling disagreement + witness count.

§3 Witness tiers (19)

Every column is sourced from at least one tier. Triangulated columns are sourced from three or more. The list is doctrine-pinned; tiers are not added without a closure audit.

§4 Reconciliation passes (9)

Every claim survives at least one pass. Most claims survive all nine. The order is fixed and the artefacts are emitted at R8 (closure quintet) and R9 (pristine fingerprint seal).

§5 Falsifier modules (23)

Every cluster is held against all 23 falsifiers at the publish gate. Any FAIL blocks the cluster from the cube; floor key FK8 is emitted in its place and the row is redacted, not silently dropped.

Live PASS/FAIL state per scope is at /docs/falsifier-register.

§6 Floor keys (10) — what we cannot know

A floor key is the cube's way of naming what it cannot honestly answer. A null without a floor key is forbidden by R7. There are exactly ten; the registry is doctrine-pinned.

§7 Coverage

Three scopes ship today. Numbers below are read from public/coverage.json, regenerated nightly by scripts/publish-coverage.py against the canonical S3 prefix.

§8 An example locality record

A real cluster from admin_code=482/mmr5_unified_v4.parquet. One row, ten columns shown, every column carrying witness count, disagreement and agreement, one column floor-keyed for honesty.

{
  "cluster_id": "MMR5-482-008-0014",
  "h3_r9": "892a1072b3fffff",
  "admin_code": "482",
  "geom_wkt": "POLYGON((...))",

  "built_fraction":        { "value": 0.71,  "witness_count": 4, "disagreement": 0.04, "agreement_score": 0.92 },
  "building_count":        { "value": 1238,  "witness_count": 2, "disagreement": 0.07, "agreement_score": 0.86 },
  "building_height_p50":   { "value": 18.4,  "witness_count": 2, "disagreement": 0.05, "agreement_score": 0.90 },
  "road_length_km":        { "value": 12.3,  "witness_count": 2, "disagreement": 0.03, "agreement_score": 0.94 },
  "pop_density_per_km2":   { "value": 28400, "witness_count": 1, "disagreement": null, "agreement_score": null,
                             "floor_key": "FK1" },
  "ndvi_p50":              { "value": 0.34,  "witness_count": 3, "disagreement": 0.02, "agreement_score": 0.96 },
  "lulc_class_primary":    { "value": "urban","witness_count": 3, "disagreement": 0.00, "agreement_score": 1.00 },
  "elev_m":                { "value": 14,    "witness_count": 2, "disagreement": 0.00, "agreement_score": 1.00 },

  "lineage": {
    "tier_contributions":             ["A","B","C","E","I","J","K","O","P","Q","R"],
    "reconciliation_passes_applied":  ["R1","R2","R3","R4","R5","R6","R7","R8","R9"],
    "falsifiers_passed":              23
  },

  "doctrine_version":   "v1.0.24",
  "build_id":           "mmr5_unified_v4",
  "parquet_sha256":     "9a4f...8c1b",
  "schema_fingerprint": "73340df4"
}

§9 Read further