# Findings from the supplied NYC files

Scope: the supplied local collection, inspected 2026-10-18. Measured findings, sampled failures, and proposed design applications are intentionally separated.

## Confirmed collection properties

- 49 readable native files, totaling 41,878,112,224 bytes; 44/59 district files plus five special-area files.
- Missing district filenames: BK03, BK07, BK12, BK15, BK16, BK17, QN04, QN05, QN06, QN08, QN09, QN10, QN13, QN14, SI03.
- Native unit systems: {'Millimeters': 18, 'Feet': 31}. Unit-normalized lengths and areas are the comparison basis.
- 19,320,233 object records; type totals: {'PolylineCurve': 11039206, 'Brep': 8249164, 'TextEntity': 3002, 'PolyCurve': 28250, 'LineCurve': 5, 'NurbsCurve': 606}.
- Normalized building-role record counts: {'context_or_other': 10212216, 'footprint_curve': 859810, 'facade_surface': 6667461, 'roof_surface': 968054, 'footprint_surface': 612692}. Counts refer to components, not buildings.
- 0 object bounds failed the reader's bounding-box validity check.
- 23 files have unreadable creation dates; filenames are listed in `evidence/summary.json`. They remain readable for geometry.

## Footprint evidence

- Eligibility exclusions: {'nonhorizontal_over_1mm': 211734, 'open_curve': 4}; another 14 non-curve records occur in nominal footprint layers. Most nonhorizontal exclusions occur in QN01; the follow-up sample identifies facade-like upright outlines. See [diagnostic checks](DIAGNOSTICS.md) before interpreting these exclusions as terrain or failed buildings.
- 648,058 eligible footprint entries, with 648,058 unique source object GUIDs. 0 GUIDs recur across file records. Neither figure is a deduplicated building count.
- Pooled footprint area: median 92.27 m²; p10–p90 31.45–226.05 m². Pooled values are entry-weighted, include possible duplicates, and inherit the collection's geographic gaps.
- Median longest-edge-aligned fill: 1.0000; median aligned aspect ratio: 2.05. These are shape descriptors, not lot coverage, use labels, or proof of a rectangle.
- 326,882 footprints have a retained height proxy. Coverage is incomplete and spatial matching is heuristic.

## District contrasts: measured examples

![Selected district footprint and height-proxy distributions](evidence/district-distributions.png)

The figure uses source-file quantiles, not a geographic population estimate. Its height panel is conditional on incomplete geometric matching; the retained counts differ from the footprint counts.

- [NYC_3DModel_MN05.3dm](files/NYC_3DModel_MN05.md): footprint area n=3,150; median 364.91 m²; p10–p90 139.51–1,572.42 m²; range 3.68–16,714.39 m²; height proxies n=2,121; median 50.03 m; p10–p90 14.08–116.19 m; range 2.26–443.67 m. These compare supplied file labels and eligible entries, not independently checked zoning districts.
- [NYC_3DModel_BK04.3dm](files/NYC_3DModel_BK04.md): footprint area n=11,421; median 104.67 m²; p10–p90 68.17–186.29 m²; range 9.16–18,354.62 m²; height proxies n=5,588; median 10.42 m; p10–p90 3.50–14.59 m; range 1.94–53.55 m. These compare supplied file labels and eligible entries, not independently checked zoning districts.
- [NYC_3DModel_BX01.3dm](files/NYC_3DModel_BX01.md): footprint area n=4,282; median 109.62 m²; p10–p90 57.26–836.08 m²; range 5.51–22,298.48 m²; height proxies n=1,236; median 16.47 m; p10–p90 3.48–29.56 m; range 1.68–80.81 m. These compare supplied file labels and eligible entries, not independently checked zoning districts.
- [NYC_3DModel_QN01.3dm](files/NYC_3DModel_QN01.md): footprint area n=24,685; median 92.28 m²; p10–p90 35.95–268.63 m²; range 3.83–30,268.03 m²; height proxies n=9,441; median 4.73 m; p10–p90 3.01–17.54 m; range 0.97–76.74 m. These compare supplied file labels and eligible entries, not independently checked zoning districts.
- [NYC_3DModel_SI01.3dm](files/NYC_3DModel_SI01.md): footprint area n=48,295; median 84.03 m²; p10–p90 32.72–155.05 m²; range 5.56–22,682.45 m²; height proxies n=20,347; median 4.36 m; p10–p90 3.28–9.48 m; range 0.24–68.94 m. These compare supplied file labels and eligible entries, not independently checked zoning districts.

## Confirmed sampled geometry failures

Across 60,627 sampled object observations, 49 failed Rhino validity checks. Reservoir samples are per layer; do not turn this pooled count into a citywide failure rate.

- `NYC_3DModel_QN_Parks_LGA_JFK.3dm` / `Buildings::Surface_FootPrint`: 20/64 invalid sampled objects. Example `4389ebe6-49c7-4fea-a407-59ad113a86ca`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_QN_Parks_LGA_JFK.3dm` / `Buildings::Surface_RoofTop`: 4/64 invalid sampled objects. Example `292a1201-bab4-4597-8810-38e712057214`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_QN_Parks_LGA_JFK.3dm` / `Buildings::Surface_Facade`: 8/64 invalid sampled objects. Example `2331d49c-1cd3-48f6-a927-6932e9d7d6e0`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BX01.3dm` / `Linework::Water`: 2/8 invalid sampled objects. Example `6fd301ef-8b9e-4227-b183-dc3d79f4014b`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BK13.3dm` / `Linework::Water`: 1/15 invalid sampled objects. Example `92c098ab-118c-4759-9e6e-5f7db720dfe0`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BX08.3dm` / `Linework::Parks`: 3/64 invalid sampled objects. Example `e9da25d7-c46a-4be0-8ab9-1d493518aff1`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BK04.3dm` / `Linework::Open Space`: 1/64 invalid sampled objects. Example `5f99fccd-e7e7-4c52-aac0-29b2cb3a4c6d`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BX09.3dm` / `Linework::Water`: 2/11 invalid sampled objects. Example `13d41277-87a8-4982-9478-c0e3518894d6`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BX12.3dm` / `Linework::Bridges & Tunnels`: 1/55 invalid sampled objects. Example `c3246f2b-0d8b-4f97-9939-444d9ea3bfbb`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BK01.3dm` / `Linework::Water`: 2/10 invalid sampled objects. Example `cda46d52-687d-4d04-9747-f6751895598a`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BK10.3dm` / `Linework::Water`: 1/7 invalid sampled objects. Example `05585cee-9a6c-4405-a697-91de511af8a0`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_BK18.3dm` / `Linework::Water`: 1/64 invalid sampled objects. Example `8d5654f9-74ff-4376-8668-f1bf136399f6`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_SI01.3dm` / `Linework::Park`: 1/64 invalid sampled objects. Example `2c3aef53-89a4-4eb9-bcac-0bd48899cad7`; see that file's note and JSON for Rhino diagnostics.
- `NYC_3DModel_SI01.3dm` / `Linework::Water`: 2/64 invalid sampled objects. Example `d51fa8ae-52ee-479c-baa8-a329e259f6b9`; see that file's note and JSON for Rhino diagnostics.

## Interpretation safeguards

- Open facade and rooftop Breps are expected component surfaces. A valid open patch is not a closed solid, and its naked edges alone are not evidence of a broken source model.
- Multiple naming conventions exist for semantically similar building surfaces. Literal layer-name equality is insufficient for cross-file extraction.
- A file includes broader context and labels; full-file bounds are not a clean district boundary. Repeated context GUID signatures, when present, are listed in the aggregate evidence. No geometry-based citywide deduplication is claimed.
- Large source coordinates, small CAD tolerances, and fine decimal display do not establish survey accuracy. No CRS/datum certification is performed.
- Geometric outlines and roof patches do not establish floor counts, internal layouts, facade materials, occupancy, structural systems, current condition, or design intent.
- Design applications are proposed separately in [Design principles](DESIGN_PRINCIPLES.md).
