Historical and Contemporary Crop Yield Prediction Models: Key Lessons and Innovations

Abstract

Crop yield prediction models (CYPMs) are essential for ensuring global food security and sustainable agricultural planning. This systematic literature review compared the overview of historical and contemporary CYPMs evolution, challenges, innovations, and key lessons learned from peer-reviewed literature. The study analyzed peer-reviewed papers published between 2015 and 2025, sourced from the Scopus, Web of Science, and PubMed databases, following PRISMA guidelines. Twenty-three studies met the inclusion criteria and were evaluated for methodological quality and risk of bias. Historical empirical and mechanistic models offered valuable theoretical foundations but were limited by data scarcity and scalability. Contemporary approaches, particularly those using machine learning, deep learning, and remote sensing, demonstrated superior predictive accuracy (R² = 0.85–0.93) compared with traditional models (R² = 0.60–0.75). Key lessons emphasize the importance of data integration, contextual calibration, and expert validation. Persisting challenges include computational demands and limited applicability in data-scarce regions. The review concludes that hybrid, interpretable, and resourceefficient models are critical for improving prediction reliability and achieving sustainable, equitable food systems.