{
  "project": "gpu-pilot",
  "doc": "crosswalk/CROSSWALK.md",
  "title": "Block-Group \u2194 H3 Crosswalk + hex_hta",
  "oneline": "TIGER 2022 geometries \u2192 22M-cell crosswalk \u2192 hex-level CNT H+T view in the commute DB \u2014 closes the long-flagged blkgrp\u2194H3 item.",
  "tldr": [
    "3 new objects in the commute DB: tiger_blkgrp (35,559 CT/NJ/NY/PA block groups, TIGER 2022 = 2020 delineation, MultiPolygon 4326), blkgrp_h3_crosswalk (22,018,397 rows \u2014 res 9: 2.75M cells, res 10: 19.27M; PK (h3,res), centroid containment so each cell maps to exactly ONE block group), view hex_hta (h3 \u2192 ht_ami/h_ami/t_ami/transit_cost/vmt/autos/% transit + cbsa_name).",
    "Coverage validated both ways: 100.00% of the 1,310,600 nyc hex_demographics res-10 cells get a block group; 99.4\u2013100% of BGs get a res-10 cell per state (3,254 NYC-scale micro-BGs miss at res 9, 97% recover at res 10; only 98 statewide miss both).",
    "Spot checks bit-exact pg\u2194h3-py: Times Sq \u2192 360610119001 (ht 12%, t 5%, $2,736/yr transit \u2014 hyper-transit Manhattan); Piscataway \u2192 340230006031 (ht 57%, $3/yr transit \u2014 car suburbia). Household-weighted avg ht_ami over the nyc cells = 43.0, matching the CBSA row exactly.",
    "Fast + cheap: whole build \u2248 9 min, 2.55 GB (h3_postgis extension's h3_polygon_to_cells at ~60K cells/s). The 72% match on the RAW res-9 bbox grid is the grid's Atlantic rectangle \u2014 92% of unmatched are >5 km offshore.",
    "National rollout is the same loop \u00d747 states: \u2248 91M cells res-9-only (~11 GB, ~30 min) or 727M both-res (~84 GB, ~4 h single-stream); recommendation res 9 everywhere + res 10 in CBSA counties \u2248 25 GB. Alaska needs an antimeridian split. Bulk alt: IPUMS NHGIS national shapefile."
  ],
  "updated": "2026-06-12 06:25",
  "sections": [
    {
      "level": 1,
      "title": "Block-Group \u2194 H3 Crosswalk + hex-level CNT H+T (`hex_hta`)"
    },
    {
      "level": 2,
      "title": "What this gives you"
    },
    {
      "level": 2,
      "title": "Objects created (commute DB)"
    },
    {
      "level": 2,
      "title": "Sources"
    },
    {
      "level": 2,
      "title": "Method"
    },
    {
      "level": 2,
      "title": "Validation results (sql/04_validate.sql, full output in `validate_run.log`)"
    },
    {
      "level": 2,
      "title": "Runtime"
    },
    {
      "level": 2,
      "title": "How to extend nationally (the remaining 47 states)"
    },
    {
      "level": 2,
      "title": "Files"
    },
    {
      "level": 2,
      "title": "NATIONAL ROLLOUT \u2014 2026-06-12 (overnight, geocoder \u2192 gtfs3)"
    },
    {
      "level": 3,
      "title": "What changed"
    },
    {
      "level": 3,
      "title": "res-10 scope decision (documented choice)"
    },
    {
      "level": 3,
      "title": "Alaska disposition"
    },
    {
      "level": 3,
      "title": "Runtime (wall clock, 2026-06-12 UTC)"
    },
    {
      "level": 3,
      "title": "Validation (full output: `validate_national.log`)"
    },
    {
      "level": 3,
      "title": "Per-state row counts"
    },
    {
      "level": 3,
      "title": "New files from the rollout"
    }
  ],
  "stats": {
    "lines": 305,
    "words": 3157
  },
  "html": "https://board.certihomes.com/project-docs/gpu-pilot/crosswalk-CROSSWALK.html",
  "out_name": "crosswalk-CROSSWALK.md",
  "comments_api": "https://board.certihomes.com/api/doc-comments?project=gpu-pilot&doc=crosswalk/CROSSWALK.md"
}