European Journal of Sustainable Agroecosystems

Background: Inefficient fertilizer use remains a persistent constraint on global food security and environmental sustainability, with nitrogen use efficiency in cereal systems commonly reported below 50% worldwide. Conventional blanket-recommendation approaches to nutrient management are increasingly being supplemented, and in some contexts replaced, by predictive analytics, machine learning, remote sensing, and Internet of Things (IoT)-enabled sensing, forming the emerging paradigm of precision agronomy.

Objective: To synthesize peer-reviewed literature published between 2015 and 2025 on predictive analytics and decision-support technologies for fertilizer recommendation and resource use efficiency, and to critically evaluate methodological trends, areas of consensus and disagreement, and unresolved research gaps.

Method: A structured literature search was conducted across Scopus, Web of Science, PubMed, ScienceDirect, SpringerLink, IEEE Xplore, and Google Scholar. The search was supplemented with reports from the Food and Agriculture Organization (FAO), the World Bank, and CGIAR, following a PRISMA-informed study selection process.

Results: Machine learning models, particularly ensemble and deep learning approaches, consistently demonstrated higher predictive accuracy than traditional statistical and rule-based fertilizer recommendation methods. Remote sensing and IoT-based sensing platforms enabled spatially and temporally resolved nutrient diagnosis. Across the reviewed studies, predictive analytics and decision-support interventions were associated with yield increases of approximately 10–20% and fertilizer-input reductions of 15–40%, although outcomes varied considerably depending on crop type, region, and farm scale. Key research gaps included limited model generalizability across agro-ecological zones, insufficient explainability of black-box algorithms, a lack of long-term multi-season validation, underrepresentation of smallholder and Global South farming systems, and weak integration of agronomic expertise with data-driven models.

Conclusion: Predictive analytics has significant potential to enhance fertilizer-use efficiency and support sustainable precision agronomy. However, achieving widespread, reliable, and equitable farmer-level adoption will require interdisciplinary validation, improved model explainability, long-term field evaluation, and stronger policy and agricultural extension support.