AI-IOT Enabled Soil and Weather Monitoring System for Smart Agricultural Decision Support

Abstract

Background: The agricultural sector increasingly adopts digital technologies to combat environmental volatility, yet conventional smallholder farming still relies heavily on intuition and experience-based decision-making. Existing systems frequently monitor soil or weather parameters in isolation, creating fragmented insights that lead to inefficient resource utilization. This study aimed to design, develop, implement, and validate an integrated AI-IoT Enabled Soil and Weather Monitoring System to provide real-time, predictive decision support for optimized farm management.

Using a developmental-descriptive research design and Agile Scrum methodology, a functional telemetry prototype was engineered using an ESP32 microcontroller connected to an industrial 7-in-1 soil probe, a three-cup anemometer, a light sensor, and a rain detector. The hardware was integrated with a FastAPI cloud backend, a PostgreSQL database, and machine learning models trained via Scikit-Learn. Pilot testing and user acceptability evaluations were conducted with 36 purposively sampled farmers and agricultural technicians.

Standard machine learning regression metrics revealed that the system achieved an accuracy rate exceeding 99% for irrigation prediction and approximately 98% for disease-risk classification. Ordinary Least Squares (OLS) linear regression for crop growth modeling demonstrated high statistical significance (P < 0.001, R² = 0.994). Empirical data from 36 evaluation respondents demonstrated strong performance across all technical indicators, with data update latency scoring the highest mean value of 4.50. The system successfully bridges the gap between raw data acquisition and low-cognitive-load operational directives, offering a highly reliable, energy-efficient, and cost-effective precision agriculture framework that enhances smallholder confidence and resource optimization.