A lightweight soil moisture prediction model based on irrigation cycle segmentation and Kalman filtering

Area of Science:

• Agricultural Engineering
• Environmental Science
• Data Science

Background:

• Efficient soil moisture prediction is crucial for precision irrigation, sustainable water use, and maximizing crop yields.
• Existing models often require extensive data and computational power, hindering their use in resource-limited farming settings.

Purpose of the Study:

• To develop a lightweight, cost-effective soil moisture prediction model for precision irrigation.
• To enhance the applicability of soil moisture prediction in resource-constrained agricultural environments.

Main Methods:

• Integration of Discrete Wavelet Transform (DWT) for time series segmentation and change point identification.
• Autocorrelation analysis to extract temporal dependency features from soil volumetric water content (VWC) data.
• Kalman filtering (KF) framework for recursive prediction, incorporating a dynamic updating and rolling mechanism with a sliding window.

Main Results:

• The proposed model achieves competitive prediction accuracy compared to existing methods.
• Demonstrates substantially reduced computational cost, making it efficient.
• The model exhibits adaptability to varying field conditions through its dynamic updating mechanism.

Conclusions:

• The developed model offers a low-complexity and adaptable solution for soil moisture prediction.
• Well-suited for real-time deployment on low-cost edge devices in precision irrigation systems.
• Facilitates improved water management and crop yield enhancement in irrigated agriculture.