Data-driven spatio-temporal estimation of soil moisture and temperature based on Lipschitz interpolation

Area of Science:

• Agricultural Science
• Machine Learning
• Geospatial Analysis

Background:

• Accurate estimation of agricultural soil variables is crucial for precision farming.
• Existing methods like Gaussian processes and neural networks can be computationally intensive.
• Spatio-temporal modeling of soil data presents significant challenges.

Purpose of the Study:

• To adapt and validate a non-parametric machine learning technique for estimating agricultural soil variables.
• To explore the application of Lipschitz interpolation for learning spatio-temporal dynamics.
• To provide a computationally efficient alternative for soil data analysis.

Main Methods:

• Utilized a non-parametric machine learning approach based on Lipschitz interpolation.
• Adapted the method to handle two-dimensional spatial and one-dimensional temporal inputs separately.
• Validated the estimator using real-world agricultural datasets with inherent measurement noise and quantization.

Main Results:

• The Lipschitz interpolation method successfully estimated agricultural soil variables.
• The technique demonstrated effectiveness in capturing spatio-temporal soil dynamics.
• The experimental setup involved an edge layer for data acquisition and a cloud layer for processing.

Conclusions:

• Lipschitz interpolation offers a viable and simpler alternative to Gaussian processes and neural networks for soil variable estimation.
• The adapted method shows promise for real-time agricultural data analysis.
• Further research can explore its scalability and application in diverse agricultural settings.