Predicting cadmium concentration in soils using laboratory and field reflectance spectroscopy

Area of Science:

• Environmental Science
• Soil Science
• Spectroscopy

Background:

• Heavy metal contamination poses significant environmental risks.
• Visible and Near-Infrared Spectroscopy (VNIRS) offers a rapid method for soil analysis.
• Understanding soil contamination requires accurate and efficient detection techniques.

Purpose of the Study:

• To investigate the prediction of heavy metal (Cadmium - Cd) concentration in soils using VNIRS.
• To assess the transferability of prediction models between laboratory and field conditions.
• To explore the role of soil organic matter in VNIRS-based heavy metal prediction.

Main Methods:

• Collected soil samples from mining and agricultural areas for laboratory and field spectral measurements.
• Utilized VNIRS (350-2500 nm) to obtain soil reflectance spectra.
• Employed Genetic Algorithm (GA) for band selection and Partial Least Squares Regression (PLSR) for Cd concentration prediction, focusing on spectral bands related to organic matter.

Main Results:

• VNIRS combined with GA-PLSR, focusing on organic matter-associated spectral bands, significantly improved Cd concentration prediction accuracy (higher R2 and RPD values) compared to using the entire VNIR region.
• Prediction models showed improved performance for both laboratory and field spectra in both mining and agricultural soil samples.
• The method demonstrated feasibility and potential for application in field conditions.

Conclusions:

• Predicting soil Cadmium concentration using GA-PLSR with VNIRS, specifically targeting spectral features linked to soil organic matter, is a feasible and effective approach.
• The developed prediction method shows potential for successful transfer and application to real-world field conditions for soil contamination monitoring.
• VNIRS offers a valuable, rapid, and non-destructive tool for assessing soil heavy metal contamination.