A study on sferic removal methods from actual airborne transient electromagnetic data

In airborne transient electromagnetic (ATEM) surveys, sferic significantly degrades data quality due to its randomness and uncertain energy amplitude. This study explores three methods for removing sferic: wavelet decomposition and reconstruction, α-trimmed mean filter, and Hampel filter. Their respective advantages and application scenarios in signal processing are compared. Wavelet decomposition and reconstruction offer multiscale and multiresolution signal analysis, allowing simultaneous consideration of time and frequency characteristics. The choice of basis functions affects global properties and phase shifts, while the number of decomposition levels influences data smoothing, and threshold settings impact noise removal effectiveness. α-trimmed mean filtering removes noise by averaging the remaining data after excluding the maximum and minimum values within a window, thus demonstrating adaptability. The Hampel filter detects outliers using median and median absolute deviation (MAD), effectively targeting noise removal, particularly suitable for scenarios requiring signal detail preservation. This study enhances the Hampel filter, reducing computational overhead while ensuring accurate outlier detection. Simulation experiments on random signals and field-collected ATEM data validate the denoising capabilities of the three methods, revealing that the Hampel filter achieves optimal sferic reduction.