Application of observed data denoising based on variational mode decomposition in groundwater pollution source...

Area of Science:

• Environmental Science
• Data Science
• Signal Processing

Background:

• Groundwater pollution source recognition (GPSR) is crucial for remediation and risk assessment.
• Observed data in GPSR can be noisy, impacting recognition accuracy.
• Existing denoising methods struggle with complex, nonlinear, and non-stationary data, and lack comprehensive applicability analysis.

Purpose of the Study:

• To introduce Variational Mode Decomposition (VMD) for denoising in GPSR.
• To comprehensively analyze the applicability of denoising by considering both noise and observed data attributes.
• To enhance GPSR accuracy using a novel collective decision optimization algorithm.

Main Methods:

• Applied Variational Mode Decomposition (VMD) for denoising noisy observed data in GPSR.
• Investigated the influence of noise level and observed frequency on denoising effectiveness across 12 scenarios.
• Integrated a collective decision optimization algorithm to improve recognition accuracy in four representative scenarios.

Main Results:

• VMD demonstrated effective denoising across various scenarios.
• Denoising effectiveness decreased with increased noise levels and decreased observed frequencies.
• Denoising proved more effective for GPSR under high noise and multiple observed frequencies.
• The collective decision optimization algorithm exhibited good inversion accuracy and robustness.

Conclusions:

• VMD is a suitable method for denoising in GPSR, with performance influenced by data and noise characteristics.
• Comprehensive analysis of denoising applicability requires considering both noise and observed data attributes.
• The collective decision optimization algorithm offers a robust approach to enhance GPSR accuracy.