Digital technologies for crop water stress detection, prediction and water-management decision-making in smallholder rainfed systems: a review

Abstract Drought remains the single largest climatic threat to crop production in rainfed and smallholder farming systems, where irrigation infrastructure is limited and crop failure translates directly into food insecurity. Over the past decade, a rapid expansion of digital technologies has transformed our capacity to detect and anticipate crop water stress. This review examines how these technologies are being connected, or fail to be connected, to water-management decisions and agronomic interventions that improve outcomes for farmers. We reviewed relevant evidence across five dimensions: (i) sensing and prediction technologies; (ii) water-stress indicators; (iii) digital decision-support and advisory systems; (iv) intervention and adoption pathways in smallholder contexts; and (v) measured effects on yield, water-use efficiency (WUE) and drought resilience. Key findings include: detection accuracy has outpaced decision integration, with machine-learning models routinely achieving R² of 0.75–0.90 under controlled conditions but rarely feeding operational decisions; a marked lab-to-field transfer gap, with accuracies falling by up to ~ 22% points in farmer fields; a systematic WUE-yield trade-off under deficit irrigation; and consistent evidence that frugal tools outperform sophisticated systems in smallholder cost-benefit terms. Digital tools deliver their largest gains when they amplify proven agronomy (deficit and supplemental irrigation, rainwater harvesting, drought-tolerant cultivars) and when they are institutionally embedded in extension and water-governance systems. This evidence was integrated into a framework that identifies the “missing middle” between sensing and action as the field’s critical bottleneck, and derive research priorities for transferable, decision-centred, smallholder-appropriate digital water management.