IoT and AI Integration for Climate‑Smart Farming: A Predictive and Adaptive System for Smallholder Farmers

Abstract

Climate variability increasingly threatens the stability and productivity of smallholder farming systems. Traditional decision-making approaches cannot reliably address rapid shifts in rainfall, soil moisture, pest pressure, and crop stress. This research presents an integrated Internet of Things (IoT) and Artificial Intelligence (AI) based climate-smart farming system that continuously monitors environmental conditions, predicts crop responses, and generates adaptive management recommendations. IoT nodes collect soil moisture, temperature, humidity, rainfall, and nutrient-level data and transmit them to a cloud-based analytics engine. Machine learning models including Random Forest for irrigation prediction, Long Short-Term Memory (LSTM) networks for yield forecasting, and Gradient Boosting for disease-risk estimation form the predictive core of the system. A rule-based adaptive module converts model outputs into actionable recommendations. Experiments using 11,200 sensor-hours, 240 field observations, and 90 climate reports demonstrate irrigation prediction accuracy of 96.2%, disease-risk detection accuracy of 93.7%, and a yield-prediction RMSE of 0.18. Field deployment results indicate 27% water savings and 12-18% productivity gains. The findings show that combining IoT sensing with AI-driven analytics significantly enhances decision-making, reduces resource waste, and supports climate-smart agriculture for smallholder farmers.