f Fast denoising of AEM data using Singular Value Decomposition

Airborne Time-Domain ElectroMagnetic (TDEM) surveys are increasingly carried out in anthropized areas as part of environmental studies. In such areas, noise arises from either natural sources, such as spherics, or cultural sources, such as couplings with man-made installations. TDEM data may therefore be affected by many distortions, such as spikes, oscillations and shifts, which make the EM noise spectrum complex and may lead to erroneous inversion and subsequent misinterpretations. In such noisy environments, thresholding and stacking standard techniques, commonly used to filter TDEM data, are hardly efficient. Time-consuming and subjective manual cleaning of the data is therefore required. We propose an alternative fast and efficient user-assisted filtering approach. We adapted the Singular Value Decomposition (SVD) to denoise TDEM data. The SVD method uses the principal component analysis. It allows separating noise from geological signal extracting into components the dominant shapes from a series of raw input curves. The signal components are then used to reconstruct the EM decays without the noise. The SVD procedure was implemented in the denoising of several EM datasets acquired over anthropized areas in various contexts. The comparison between each reconstructed decay and its corresponding measured decay allows efficiently detecting noisy gates and rejecting mainly spikes and oscillations. Moreover, an ad hoc analysis of the map of weights of the components explaining noise showed high correlation with man-made installations. Thus, the SVD also provides a tool to reject most likely soundings biased by coupling noises, which may result in artefacts on the inverted models. However, some distorted decays can only be localized based on the analysis of specific SVD components. It was also shown that the maps of the weights of the components explaining the geological signal could be useful as a first rapid view of the contrasts that exist in subsurface. This established SVD based procedure is fast and provides accurate denoising tools; it makes, at least, the manual cleaning less time consuming and less subjective.