Improving Soil Heavy Metal Lead Inversion Through Combined Band Selection Methods: A Case Study in Gejiu City, China

Area of Science:

• Environmental Science
• Remote Sensing
• Analytical Chemistry

Background:

• Hyperspectral technology is vital for soil heavy metal monitoring.
• Existing band selection methods often have limitations, such as redundancy and reliance on single algorithms.
• Multi-algorithm band selection approaches are underutilized for soil heavy metal assessment.

Purpose of the Study:

• To develop and evaluate a multi-algorithm band selection method for rapid prediction of soil lead (Pb) contamination.
• To address band redundancy in hyperspectral data for improved heavy metal monitoring.
• To enhance the accuracy of soil Pb content prediction using optimized hyperspectral band combinations.

Main Methods:

• Applied single algorithms: Competitive Adaptive Reweighted Sampling (CARS), Genetic Algorithm (GA), Mutual Information (MI), Successive Projections Algorithm (SPA), and Whale Optimization Algorithm (WOA).
• Developed combined WOA-based band selection methods (e.g., WOA-CARS, WOA-GA, WOA-MI, WOA-SPA).
• Implemented multi-level band optimization using MI (e.g., WOA-GA-MI, WOA-CARS-MI, WOA-SPA-MI) for refined Pb prediction.

Main Results:

• The Whale Optimization Algorithm (WOA) initially showed the highest modeling accuracy.
• The WOA-GA-MI model achieved optimal performance with an average R² of 0.75.
• This model demonstrated significant accuracy improvements over full-spectrum and simpler models.
• Identified 22 essential common bands for Pb content inversion through spectral response analysis.

Conclusions:

• The proposed multi-level combined band selection method significantly enhances soil Pb prediction accuracy.
• This approach offers valuable insights into optimizing hyperspectral band selection for environmental monitoring.
• Provides a strong scientific foundation for assessing soil heavy metal contamination using hyperspectral data.