Pre-Symptomatic Crop Intelligence: A Closed-Loop Framework for Anticipatory, Confidence-Aware Decision-Making in Site-Specific Crop Protection

Abstract

Abstract Purpose Symptom-triggered crop protection acts only after damage is committed, and the intervention window has narrowed. This review reframes pre-symptomatic sensing from a detection problem into a closed decision loop, establishing the physiological lead time of a signal, set against its detection confidence and the latency of the response it can trigger, as the organizing principle for anticipatory, site-specific decision-making. Methods A systematic-narrative synthesis was conducted across major bibliographic databases through June 2026. Studies reporting pre-symptomatic capability under field or realistic conditions were retained and coded onto a coupled lead-time × confidence × actionability framework spanning sensing, inference, and actuation. Results Optical modalities were found to dominate the evidence base, while electrophysiological and volatile signals extended achievable lead time. Single modalities were insufficient to separate biotic from abiotic stress, motivating heterogeneous fusion. Edge inference and temporal onset forecasting remained immature, detection confidence was rarely quantified, and the sensing-to-actuation loop was seldom closed. Reported performance degraded sharply from laboratory to field, particularly in perennial and smallholder systems. Conclusions A unifying Pre-Symptomatic Crop Intelligence framework is proposed, governed by the principle that system value is bounded by the weakest of lead time, detection confidence, and response latency; priorities identified include lead-time-labeled benchmarks, uncertainty-aware inference, field-robust fusion, and economic evaluation for perennial crops.

Plant phenotyping relevance

植物の病害・ストレス状態を早期に推定するセンシング手法を体系的に整理し、検出リードタイム、信頼度、融合推論、ベンチマークを評価するレビューであり、植物状態の取得・推定方法が中心である。

Code and data availability

This is a systematic review of pre-symptomatic crop protection. The authors state no new datasets were generated or analysed, and no public phenotype datasets, images, analysis code, or trained models specific to this paper are disclosed. The only reproducibility item is an OSF-registered review protocol, mentioned byt

No evidence-backed public reproduction asset is currently recorded.