IoT based soil moisture measurement and type prediction using advanced regression and machine learning models.

Area of Science:

• Environmental Science
• Geotechnical Engineering
• Agricultural Science

Background:

• Accurate soil moisture measurement is crucial for resource efficiency in agriculture and environmental management.
• Current methods can be time-consuming or lack real-time capabilities.
• Developing advanced monitoring systems is essential for sustainable practices.

Purpose of the Study:

• To present a novel Internet of Things (IoT)-based method for real-time soil type and moisture prediction.
• To evaluate the accuracy and effectiveness of the proposed system compared to traditional methods.
• To demonstrate the potential for scalable applications in precision agriculture and environmental monitoring.

Main Methods:

• Utilized capacitance sensors to collect soil moisture data, creating a custom lab dataset.
• Employed traditional oven-dry methods for ground-truth water content verification.
• Developed machine learning models, including polynomial regression and Random Forest classification, for prediction and classification.

Main Results:

• Achieved 96.49% accuracy in water content prediction using logarithmic regression, outperforming linear regression.
• A polynomial regression model demonstrated a strong correlation (R2=0.79, MAE=1.71%) between capacitance and water content across different sand types.
• Random Forest classifier accurately identified soil types with approximately 97.77% accuracy.

Conclusions:

• The IoT-based system provides an effective, non-destructive method for analyzing soil properties in real-time.
• The methodology offers significant potential for environmental management, soil monitoring, and precision agriculture.
• Further research will focus on dataset enrichment and incorporating environmental factors to enhance predictive accuracy.