Vine Loop: one vineyard, one season, one closed loop

Precision vineyard tools usually stop at a map. The satellite sees the block but not the row, the drone sees the row but lives in its own app, disease risk sits in a weather portal, and the scout in the field writes on paper.

None of it adds up to a decision a grower can act on, and nothing checks afterwards whether the treatment worked.

We built a working demo on public data from a real block in Penedès, Spain. A Python pipeline joins 1,036 vineyard parcels from SIGPAC 2024 with a full 2024 Sentinel-2 season, two drone flights published by the EU ICAERUS project, a canopy height model and hourly ERA5 weather.

The page runs the loop end to end: detect weak vines in 1 m canopy cells, rate disease risk with published mildew rules, send scouts out with an offline field app, cluster the canopy into spray zones in the browser and export a prescription file, replay a logged spray flight, then compare the second flight zone by zone.

The scouting, the spray flight and the prescription files are simulated and labelled as simulated on the page. Everything else is real data.