Alpine surface structure and canopy cohorts

The actual Alps01 images in QA/20260916-032128-campaign view0/view1 and Alps02 QA/20260916-022247-campaign/alps-02-pass0-view1.png were inspected. They show nearly uniform near meadow colour, separated similarly shaped dark tree silhouettes, and a very broad pale planar cutting. The earlier colour-space, UV and forest-density diagnoses remain applicable; none of these changes is presented as a repaired lighting calculation or proof of photorealism.

Surface changes

AlpineTerrainMaterial retains the existing summer base colours, all original surface photographs, the 1.4 m grass/normal mapping, geographic layer masks and snow material. Only summer layer 0 gains independently scaled native HDRP detail. A deterministic 256-square linear RGBA map describes ground-cover variation at a 32 m period, with 8 m, 4 m and 1 m features. Its normal is neutral. This supplies spatial variation after the small grass blades become their mip average without magnifying the original blades or stretching an aerial picture over the road verge.

The existing continuous road-cut opacity pass still uses two identical rock surfaces with alpha 0/1 so native vertex-R blending preserves the edge opacity. Its limestone source and 9 m mapping remain. A second deterministic linear 256-square detail map supplies irregular bedding and fissure shading/normal response at a 24 m period. Both are authored material data, not photographed or surveyed geological structure. The previous detail photograph's 2.8 m overlay is replaced; the source's base normal remains. The rock's authored colour input is (0.82, 0.84, 0.83), and mask smoothness has a 0–0.48 range to reduce the broad pale wall response. These are artistic settings requiring native review.

Both maps use a periodic lattice, trilinear mipmaps, Repeat wrapping and linear RGBA32 storage. Channels follow the pinned HDRP contract: R centred albedo, G normal Y, B neutral smoothness, A normal X. Their total raw mip allocation is about 683 KiB before engine overhead. They are constructed once and shared statically, with no per-frame generation, GPU buffer or new shader. Existing representative materials retain the native detail-map variants. No original texture bytes, geology masks or terrain collision geometry change.

Canopy changes

Summer forest height now varies with the existing density and a spatial cohort field: mature conifers reach 15–25 m while low-cover edges carry shorter trees. Far crowns also receive bounded site-stable width variation, and mirrored atlas UVs break repeated asymmetry. Near mesh-to-billboard LOD dimensions remain exact, and winter is unchanged. Billboard bounds are computed from the actual varied half-width.

The third sparse summer source tree was deliberately omitted by the previous design; it was not a UV bug. It is now used only for 18% of eligible low-cover edge firs (Density < 0.60), retaining both original source views. Closed forest keeps the full A/B crowns. No trees are added and no new random draws alter the placement sequence, shared spacing registry or later candidates. Existing road clearances and tree ground pivots remain unchanged.

Validation boundary

This author did not launch Unity, a Player, a build, or repeated render tests. Root owns the consolidated managed/native checks and the full-campaign graphics capture. Before acceptance, inspect near meadow repetition, cut brightness/strata, forest overlap and retained winter appearance in those actual captures. Wider/taller crowns increase alpha-covered pixels despite unchanged tree/triangle counts; actual GPU cost must be observed. Macro material detail cannot reconstruct a non-planar rock silhouette or replace a full 3D mid-distance forest, so the photorealism target remains open until the native result is reviewed.

Native review and one bounded correction

The RTX4060/D3D11 run QA/20260916-110529-graphics-all-levels supplied actual Alps01 views0/1/2, Alps02 views0/1 and Alps03 view0. All six images were personally inspected. Summer forest crowns overlap somewhat more, but dark repeated silhouettes remain visible; the native canopy counts remain exactly 77,545/107,855. The cutting retains its continuous toe and local grain, but still looks like a large manufactured beige plane. There is no photorealism acceptance.

The new layer-0 cover variation at strength 0.85 produces conspicuous cloudy light/dark streaks, especially on the right near meadow in Alps01 view1 and the right slope in Alps02 view0. That visual regression justifies a single correction to 0.35, retaining the exact data map and every other surface parameter. HDRP's detail albedo overlay has a squared strength response; 0.35 therefore substantially restrains the generated contrast rather than merely reducing it by a linear ratio. This correction is later source than the 11:05 capture and must not be claimed as present in those images. Root owns its final corrective build and capture.