Improving the Calibration of Low-Cost Sensors Using Data Assimilation

Area of Science:

• Agricultural Engineering
• Environmental Science
• Data Science

Background:

• Accurate soil moisture monitoring is vital for smart agriculture, optimizing irrigation and crop yields.
• Low-cost capacitive soil moisture sensors offer affordability but suffer from accuracy issues, leading to inefficient practices.

Purpose of the Study:

• To present a data assimilation method for calibrating capacitive soil moisture sensors.
• To enhance the accuracy of low-cost soil moisture sensors for precision agriculture applications.

Main Methods:

• Integrated the Hydrus 1D model with particle filter (PF) and Iterative Ensemble Smoother (IES) for sensor calibration.
• Implemented physical constraints to ensure updated parameters remain within plausible ranges.
• Validated the method using data from a drip-irrigated farm, comparing PF and IES against high-precision reference sensors.

Main Results:

• Data assimilation significantly improved sensor reading precision, aligning them with reference measurements and model simulations.
• The particle filter (PF) method achieved an 84.8% accuracy improvement over raw sensor readings.
• The Iterative Ensemble Smoother (IES) method provided a 68% accuracy improvement, though outperformed by PF.

Conclusions:

• Data assimilation is a robust and practical approach for large-scale implementation in precision agriculture.
• The PF method effectively mitigates observation noise and sensor biases in low-cost capacitive soil moisture sensors.
• This calibration technique enhances water use efficiency and crop yield potential.