Predicting Soil Salinity with Vis-NIR Spectra after Removing the Effects of Soil Moisture Using External Parameter...

Area of Science:

• Agricultural Science
• Soil Science
• Spectroscopy

Background:

• Accurate soil salinity prediction is crucial for agriculture.
• Existing visible and near-infrared (vis-NIR) spectroscopy models struggle with soil moisture variations.
• Soil moisture significantly impacts spectral reflectance, hindering salinity assessment.

Purpose of the Study:

• To develop robust soil salinity prediction models using vis-NIR spectroscopy.
• To mitigate the influence of soil moisture on spectral data.
• To enhance the accuracy and reliability of soil salt content (SSC) estimations.

Main Methods:

• Collected 570 soil spectra (380-2400 nm) across varied soil types, salt concentrations, and moisture levels.
• Applied External Parameter Orthogonalization (EPO) to spectra to remove soil moisture variations.
• Compared Partial Least Squares Regression (PLSR) models with and without EPO correction.

Main Results:

• EPO effectively removed soil moisture's influence on spectral reflectance.
• Salinity prediction models built with EPO-corrected spectra showed significantly improved accuracy and robustness.
• The enhanced models performed better under diverse soil moisture conditions.

Conclusions:

• EPO is a valuable technique for correcting soil moisture effects in vis-NIR spectral analysis.
• This method enhances the reliability of soil salinity quantification in agricultural fields.
• The findings support improved soil salinity monitoring and management strategies.