Known artifacts
Vertical seams at 72 / 78 / 84 / 90°E are UTM zone boundaries. Source tiles
overrun their zone and pad the overrun with NoData; our dedup let the pad win, so ~40% of
cells in those 0.25° strips read valid = false (9.7% elsewhere).
Sampler is fixed — needs a re-run to clear. CHIRPS is unaffected.
The box is a rectangle, not India. 68–97.5°E, 6.5–37.5°N also catches Pakistan,
Nepal, Tibet, Bangladesh and Myanmar. No country mask applied yet.
Bright Himalaya on drift is mostly snow-cover variability, not land-use change.