[Application of wavelet packet analysis in estimating soil parameters based on NIR spectra]

Area of Science:

• Agricultural Science
• Analytical Chemistry
• Spectroscopy

Context:

• Accurate estimation of soil parameters like total nitrogen (TN) and soil organic matter (SOM) is crucial for sustainable agriculture.
• Traditional methods for soil analysis are often time-consuming and labor-intensive.
• Near-Infrared (NIR) spectroscopy offers a rapid alternative but requires robust modeling techniques.

Purpose:

• To enhance the accuracy of Partial Least Squares (PLS) models for estimating soil TN and SOM using NIR spectra.
• To investigate the effectiveness of wavelet analysis in preprocessing NIR spectra for improved soil parameter prediction.
• To develop a practical and efficient method for real-time soil parameter monitoring.

Summary:

• Soil samples were analyzed using NIR spectroscopy, and initial PLS models were established.
• Wavelet packet decomposition was applied to NIR spectra to denoise and extract characteristic features.
• Reconstructed spectra were used to build new PLS models, significantly improving prediction accuracy for TN and SOM.

Impact:

• The developed wavelet-assisted PLS models achieved high calibration and validation coefficients for both TN (0.960/0.920) and SOM (0.922/0.883).
• Wavelet analysis effectively reduced spectral noise and interference, enabling more accurate linear modeling.
• This research demonstrates a feasible and potential method for the real-time estimation of soil parameters, benefiting agricultural management.