DEVELOPMENT OF AI-DRIVEN SENSOR PLATFORMS FOR SUSTAINABLE ENVIRONMENTAL MANAGEMENT AND CLIMATE MITIGATION

Abstract

Rural Nigeria confronts escalating climate change impacts, including unpredictable rainfall, soil degradation, and rising air pollution, which severely affect rain-fed agriculture and community livelihoods. This study develops a low-cost, AI-driven environmental monitoring platform designed for resource-constrained rural settings. The system integrates sensors measuring CO₂ (400–5000 ppm), PM₂.₅ (0–500 μg/m³), soil moisture (0–100%), and ambient temperature (−10 to 60 °C), connected to ESP32 microcontrollers with Global System for Mobile Communications (GSM) modules to enable real-time data acquisition. Field deployments in rural Adamawa State communities lasted 1–12 months, with manual cross-validation using gravimetric and portable reference methods. AI models (Long Short-Term Memory LSTM, Random Forest, and bidirectional LSTM-CNN hybrids) were applied for calibration of raw low-cost sensor outputs and predictive modeling of environmental trends, attaining prediction accuracies >90%, Root Mean Square Error (RMSE) <20, and R² ≥ 0.98 for the calibrated and predicted parameters (CO₂, PM₂.₅, soil moisture, temperature). Deployed platforms demonstrated 90–98% uptime, and >95% data transmission success, enabling near real-time data acquisition and transmission for subsequent offline AI calibration and predictive modeling. These processed insights support alerts for irrigation optimization, air quality exceedances, and heat stress mitigation. These results demonstrate the platform’s potential as a scalable tool for precision agriculture and early environmental hazard warning, contributing toward Sustainable Development Goal, SDG 2 (Zero Hunger) and SDG 13 (Climate Action) in resource-constrained regions.