GreenAI: An Explainable AI and IoT-Enabled Offline Framework for Precision Agriculture

Precision agriculture requires intelligent, interpretable and deployment-ready systems capable of operating under native farming conditions and limited-connectivity environments. This article presents GreenAI which is an Explainable Artificial Intelligence (XAI) and Internet of Things (IoT)-enabled offline-first framework for precision agriculture tailored to Pakistan. The proposed framework integrates crop recommendation, plant disease diagnosis, environmental monitoring and edge-assisted decision support within a unified multilingual ecosystem. A Punjab-specific crop recommendation dataset comprising 9,500 balanced samples across 19 major crops was developed and evaluated using multiple machine learning algorithms. Among these evaluated models XGBoost achieved the best performance with a test accuracy of 97.40% and a macro F1-score of 0.9741. For plant disease diagnosis an EfficientNetV2-S model trained on a unified 113-class disease dataset achieved a validation accuracy of 99.03%. To improve transparency and user trust, SHAP-based feature attribution, partial dependence analysis and Grad-CAM visualizations were incorporated into the decision-making pipeline. The framework further integrates an ESP32-based IoT sensing platform and lightweight edge inference to support low-latency operation in resource-constrained environments. Results of this experimental demonstrate that the effectiveness of the proposed framework in delivering accurate, interpretable and scalable agricultural decision support. GreenAI provides a practical pathway toward sustainable, data-driven and farmer-centric precision agriculture.