feat: terrain mesh from the DEM, and correct the blockout frame

Tools/terrain_from_dem.py builds a 191x106 grid at 2 m over the block, Z being
NAVD88 minus the origin ground. Building footprints from the OSM extract are cut
out so rooftops never become terrain (72 footprints, 6310 cells). Spot checks
reproduce the measured profile: origin 0.00, Orange Dr +3.39, south gutter
+0.93, crown +1.38, north sidewalk +1.43.

Blosm projects with a sphere, not an ellipsoid, so its frame was ~0.4% oversized
and rotated by meridian convergence: about 2 m at the theatre. The blockout now
fits a six-parameter affine into EPSG:26911 on five control points (0.0 cm
residual) and transforms root objects only, since children inherit and double
transforming them read as a 2 m placement error.

Art/blender/README.md

@@ -29,18 +29,50 @@ Run time is about 36 s for this extent. Two headless-only workarounds are baked
 1. Blosm refuses to import until an asset directory is set; the paid asset pack path is passed explicitly and is **not** committed.
 2. In background mode the GUI never fills Blosm's default level table, and the add-on then takes a code path with a broken relative import. The script seeds one entry.
 
+## Frame correction
+
+Blosm projects with a **sphere** of radius 6 378 137 m about the origin, not an ellipsoid. Against EPSG:26911 that is conformal but about 0.4 % oversized and rotated by the meridian convergence, which put the theatre ~2 m off. `Tools/blosm_blockout.py --origin` therefore fits a six-parameter affine from Blosm's frame into UTM on five spread control points and applies it to the imported roots. Fit residual: **0.0 cm**. After it, the blockout and the terrain share one metric frame.
+
+Two traps, both already handled in the script: transform only **root** objects (children inherit, and doing both applies the transform twice — that looked like a 2 m placement error), and fit an affine, not a similarity (a similarity leaves ~30 cm).
+
+## Terrain
+
+`terrain/terrain-2026-11-06.blend` — regular grid from the committed 3DEP DEM, built by `Tools/terrain_from_dem.py`.
+
+```bash
+B=/var/lib/flatpak/app/org.blender.Blender/x86_64/stable/active/files
+LD_LIBRARY_PATH="$B/blender/lib" "$B/blender/blender" -b --factory-startup   --python Tools/terrain_from_dem.py --   --dem Geo/raw/dem-3dep-1m-26911-2026-11-06.tif   --osm Geo/raw/osm-api-map-2026-11-06.osm   --out Art/blender/terrain/terrain-2026-11-06.blend
+```
+
+- 191 × 106 grid at 2 m over X −330…+50, Y −100…+110: 20 246 vertices, 13 640 faces.
+- Z is NAVD88 minus 115.911 m, the DEM value at the origin, so the origin ground is Z 0.
+- Building footprints from the OSM extract are **cut out** (6 310 cells), so rooftops never become terrain. 72 footprints masked.
+- Z range −2.80…+8.19 m. The high values are the ground rising north of the boulevard, not rooftops: the maximum sits at the northern edge, Y +110.
+
+Spot checks against the measured cross-sections:
+
+| Point | Terrain Z |
+| --- | ---: |
+| Origin, Hollywood × Highland | 0.00 |
+| Orange Drive end of the block | +3.39 |
+| North sidewalk mid-block | +1.43 |
+| South gutter, mid-block | +0.93 |
+| Road centre, mid-block | +1.38 |
+
+That reproduces both the 3.4 m eastward fall and the ~2 % crown, so the DEM, the cross-sections, and this mesh agree.
+
 ## Scale check (blockout, 2026-11-06)
 
 Blender works in metres, and the imported meshes match the OSM geometry, so the blockout is 1:1 before any Unreal export:
 
-| Building | Measured in Blender | OSM geometry | Height source |
+| Building | Blender dimensions | OSM way outline | Height source |
 | --- | --- | --- | --- |
-| El Capitan Entertainment Centre | 25.68 × 54.88 × 14.50 m | 25.75 × 54.88 m | `height` tag 14.5 |
-| El Capitan Theatre | 38.59 × 54.87 × 20.62 m | 38.59 × 54.87 m | `building:levels=6` × 3.2 m, plus roof |
-| Madame Tussaud's | 42.41 × 57.20 × 15.90 m | 42.0 m frontage | `height` tag 15.9 |
-| TCL Chinese Theatre | 43.75 × 76.19 × 18.10 m | 44.0 m frontage | `height` tag 18.1 |
+| El Capitan Entertainment Centre | 25.72 × 54.47 × 14.50 m | 25.75 × 54.88 m | `height` tag 14.5 |
+| El Capitan Theatre | 38.66 × 54.46 × 20.96 m | 38.59 × 54.87 m | `building:levels=6` × 3.2 m, plus roof |
+| Madame Tussaud's | 42.48 × 56.77 × 15.90 m | 42.0 m frontage | `height` tag 15.9 |
+| TCL Chinese Theatre | 43.78 × 75.90 × 18.10 m | 44.0 m frontage | `height` tag 18.1 |
 
-The 0.07 m width difference on the hero is Blosm's own projection of the outline, not a units problem. Anything larger than a few centimetres would be a flag.
+Widths agree to a few centimetres; depths differ by up to 0.4 m because Blosm assembles **multipolygon relations**, not the raw way I measured. Placement after the frame correction agrees with the UTM control points to 0.1–0.5 m per building — the residual is relation assembly and Blosm's edge straightening, not projection.
 
 Buildings **without** a `height` tag get levels × 3.2 m, which is a placeholder, not evidence. Every such height is a modelling decision to be replaced from the reference photography.