Winter ground coverage for the snow-pass stage
The actual 19:17 Alps03 driving view showed snowy trees above largely beige/green ground. User references 05–09 were reviewed alongside it: snow forms connected banks and forest-floor cover, while trunks, understory openings, steep rock and cleared asphalt remain distinguishable. Reference 09 retains a bare rock face beside snowy trees; reference 07 shows interrupted snow on steeper banks. Their target is a coherent winter landscape, not a fully white image.
The old mask starts snow only at local height 1170 m, becoming fully altitude-enabled at 1425 m. The third stage starts near 809 m. At the actual view0 motorcycle position, race time 7.2652879 s, the existing snow weight was zero. Thus the soil discrepancy had a concrete material-weight cause before any broader tree/lighting review.
Scalar coverage change
Tools/Art/build_alpine_winter_landcover.py generates only B in the separate stage-03 SchneebergSnowLayers.png. R forest, G geology and A base stay byte-identical. The existing native HDRP snow layer receives the new weight; no source albedo, tree texture, normal map or material color is repainted.
- Lower woodland receives broken coverage through a local-height transition from 560 to 820 m, varied by a deterministic ±45 m snowline.
- Coverage grows more connected between 780 and 1040 m; high gentle surfaces use approximately 0.90–0.975 snow weight.
- The unchanged 20 m DEM supplies slope. Near-vertical faces shed snow; existing geology G further suppresses it on steep exposed limestone. Flat summit rock can still accumulate snow.
- Deterministic 650/120/45 m scalar fields vary the snowline and patches, using seeds 91531/91532/91533. These are material weights, not new terrain detail or finer elevation data.
The full-map draft remains zero over 78.29% of its pixels. Below local 500 m the generated weight is zero. Mean weight on DEM slopes steeper than 65 degrees is effectively zero. The maximum encoded B is 248/255. These figures check scope, not artistic quality.
Probes tied to the native view
The motorcycle trace position is (-1880.82251, 838.48547, 4427.84180). Side/forward probes use its horizontal velocity direction; they are geographic ground samples, not camera raycasts. The road uses its own material, so sampling terrain B at the motorcycle X/Z does not add snow to the road.
| Sample | World X / Z | DEM height | Old B | New B |
|---|---|---|---|---|
| Native view0 position | −1880.823 / 4427.842 | 842.94 m | 0.000 | 0.650 |
| 40 m right | −1865.988 / 4464.989 | 797.72 m | 0.000 | 0.516 |
| 40 m left | −1895.657 / 4390.694 | 873.76 m | 0.000 | 0.699 |
| 160 m forward, steeper ground | −2029.412 / 4487.181 | 834.64 m | 0.000 | 0.229 |
The old/new scalar-map review and full B draft were personally inspected before publication. The source comparison shows more connected high-ground coverage along the actual route while exposed bands remain open. It is labeled as a technical map, not a gameplay render. It does not replace checking the result from the motorcycle camera.
Reproduction and boundaries
Immutable input: ArtSource/Alps/WinterGround/SnowGeologyBaseline.png, SHA256 ce576c652446466f12718966ec2adb82a052b184388c7c56ef7c5b0b5ffdb53a. Draft/published SHA256: 21ceed5911dd1fd6a5dd646dea094de435cbf5fa5febe05da8d790f7639b3791. ALPINE-WINTER-GROUND-ASSETS.json mirrors ArtSource/Alps/WinterGround/winter-ground-build.json.
Run the tool without arguments to regenerate the draft; use --publish after reviewing it and releasing the build freeze. It refuses a changed baseline or unexpected live revision. This is a later pipeline step after geology mask preparation; it never reruns the broader summer producer. The second run reproduced the identical PNG hash before publication.
The published file retains its GUID and linear, clamped, 2048-pixel import settings. Channel equality and SHA256 checks confirm unchanged R/G/A, elevation bytes, AlpineRoadData.cs and summer SchneebergLayers.png. No runtime C#, model, source photograph, collider, checkpoint, RidingSurface, road geometry, physics or stage-01/02 material changes. DEM attribution remains in THIRD-PARTY-NOTICES.md.
Native visual acceptance is pending a build containing this mask. The 20 m DEM cannot describe every exact road-cut normal; review the left rock face and both shoulders in view0, later snow detail and the terrain/tree relationship. Winter forest grouping and overly bright trees remain separate work. This ground correction does not establish AAA quality.