Native lake artwork: shorter isolated highlights

WOG-A-009280 is the current preferred lake study at the inspected fields/lake site. Its shorter, restrained highlights reduce the broad similar bright bands seen in the previous wind-patch study, while fine painted strokes remain visible. The original WOG-A-003575, retained lake source WOG-A-008307 and previous study WOG-A-009279 remain unchanged. This is a scoped artistic preference, not a full game-ready approval or source admission.

The selected native PNG ../../../../tile-library/atlas-v2/terrain/lake/lake-short-wavelets-native-v2.png remains 1536×1024, fully opaque, copied from the built-in tool without resampling or recompression. The desktop context-colour/lake-before-after-1440.png and phone context-colour/lake-before-after-390.png boards show the original on the left and selected study on the right, as native screenshot crops. Across the initial and corrected studies, 24 actual private Region/Hex frames preserve camera, source scale, natural geography and normal controls. Four frames show the selected corrected input. No world commands are issued.

The first study removed prominent highlights but was too green and flat: mean RGB about 36,88,76 and one-pixel native luminance contrast 0.621 of the original. Colour/microcontrast correction yields about 46,84,74 versus original 52,88,75, with channel standard deviations about 7.1,6.2,5.0 versus 7.0,6.7,4.8. The corrected one-pixel contrast is 0.854 of the original. Direct inspection supports retained fine marks and a calmer appearance; it does not claim identical palette/contrast or universal visual suitability. See the native detail diagnostics native-detail-validation.json.

Five real-GPU allocation comparisons preserve all 12,793,743 tested pixels at native source aspect, including fractional extent. This technical parity does not approve artistic quality. Wider lake-domain recurrence, other shore/biome contexts, Hex shore contact, real pan-return/reconnect, CPU fallback, relevant performance and source admission still need verification. The previously rejected wider sampler is not used. The public renderer, source permissions, saves and geography stay unchanged.

After exact selected-output and original preservation checks, five temporary files were retired: the low-contrast private input and its generation copy, the redundant corrected input copy, and two superseded preview boards. The retirement receipt temporary-retirement.json records 10,171,495 reclaimed bytes. Complete compact observations and source hashes remain; two final native context boards are retained.

See the visual decision visual-review.json, initial edit prompt/provenance generation.json, final colour-edit prompt/provenance generation-colour.json, selected context observations context-colour/results.json, native GPU parity pixel-validation.json, Atlas verification atlas-validation.json, and registration registration.json.

The Atlas record becomes available with this batch’s publication; the study remains private and unapproved until its recorded gates are resolved.