Abstract Agriculture forms the backbone of the global economy, facing mounting pressure from populationgrowth and resource constraints. The sector increasingly relies on data-driven technologies to enhanceproductivity while reducing environmental impact. Agriculture is being revolutionized by Artificial Intelligence(AI), which is enhancing pesticide application, weed control, and irrigation management. DeepLearning techniques that have demonstrated predictive power include Generative Adversarial Networks,Recurrent Neural Networks, and Convolutional Neural Networks. Their opacity and intricacy, however,make practical use difficult. In agricultural settings, Explainable AI (XAI) enables informed decisions byproviding transparency without compromising performance. This comprehensive review analyzes peer-reviewedpublications from 2020 onwards, categorizing XAI techniques and their applications in agriculture.The starting point of 2020 was deliberately chosen to capture the most recent advancements, asthis period marks a phase of rapid growth and wider adoption of XAI within agricultural AI applications,making it particularly relevant for reflecting state-of-the-art developments. This review identifies significantchallenges, current research trends, methodological approaches, and evaluate the efficacy of variousexplainability methods, including LIME, SHAP, Grad-CAM, and rule-based models. The analysis examineskey domains including crop-weed discrimination, plant disease detection, precision farming techniques,yield forecasting, and soil quality assessment. The integration of XAI methodologies in precision agriculturepresents promising opportunities to address pressing challenges related to resource optimization,climate adaptation, and global food security. This review also provides a structured framework for futureresearch directions and practical implementation guidelines to enhance the interpretability, trustworthiness,and adoption of AI-powered agricultural systems among farmers, agronomists, and policymakers.
Pai et al. (2025) studied this question.