Denoising of deep-tow magnetic anomalies based on discrete wavelet transform

Deep-tow magnetic surveys deploy magnetometers close to the seafloor to obtain high-spatial-resolution magnetic anomaly data, yet such datasets are inevitably contaminated by complex noise. This study adopts the discrete wavelet transform (DWT) to denoise deep-tow magnetic data and systematically evaluates the denoising performance of four wavelet families (Haar, Daubechies, Symlet, and Coiflet). The evaluation is conducted using synthetic magnetic models with varying noise intensities, together with deep-tow magnetic profiles collected from the western Pacific Ocean. The results indicate that the selection of mother wavelet exerts a crucial influence on denoising performance: Haar wavelets produce prominent staircase artifacts at high decomposition levels and severely distort the morphology of magnetic anomalies, whereas Daubechies, Symlet, and Coiflet wavelets deliver far better noise suppression and signal fidelity. The denoising effect improves with increasing decomposition levels, yet excessive decomposition leads to over-smoothing and eliminates subtle weak anomalous signals. Accordingly, sensitivity tests coupled with quantitative indicators are used to identify the optimal decomposition parameters. This study verifies that DWT serves as an efficient and robust technique for denoising deep-tow magnetic data.