From Orchard Mapping to Bloom and Irrigation Events

An orchard map tells a team where trees are; a time series can begin to tell the team what phase those trees are passing through. That shift matters because many useful agricultural decisions are event decisions. A manager may need to know when bloom is spreading across a valley, whether a block changed after irrigation, or whether a seasonal decline arrived earlier than expected. Those questions are more concrete than a generic “health score,” and they can often be checked against cameras, weather records, irrigation logs, or a short field visit.

The 2026 almond-bloom study illustrates the pattern. Researchers used more than 4,000 Sentinel-2 scenes from 2019 to 2022, extracted EBI and NDVI signals for roughly 30,000 orchards, interpolated missing observations, smoothed the curves, and detected the bloom peak. The satellite-derived date had a mean absolute error of 1.9 days against time-lapse camera observations; when four temperature-based dormancy models were combined with a Random Forest ensemble, the error fell to 1.48 days. That is not a promise that every crop can be monitored with the same accuracy, but it shows why a well-defined event, a suitable index, and an independent validation source can be more useful than a vague map of “vigour.” See Lauterman et al..

For field teams, the output should remain operational: an estimated date, the confidence around it, the blocks that differ from their neighbours, and the evidence that should be checked next. Irrigation detection needs the same discipline, because a vegetation change after water application can be real without proving that irrigation was the cause of every downstream difference. The strongest system connects a satellite event to a field record and allows the grower to confirm, reject, or refine the interpretation.