Two noise reduction techniques are proposed for the removal of sferics noise from airborne transient electromagnetic data. Both techniques use multi-resolution analysis via a stationary wavelet transform of the data. The analysed signal is divided into several successive lower resolution components. The transient character of the sferics can be seen as high amplitudes of the wavelet detail coefficients close to the time of the sferics event. The first noise reduction strategy, named the wavelet extraction technique, identifies sferics in the first detail coefficients using an energy detector. The corresponding detail coefficients are set to zero, and the electromagnetic signal is reconstructed by inverse transform. This technique is very robust and successful both for on-time and off-time data and even in the case where several sferics are present. However, when sferics occur near the switch on or the switch off times of the airborne electromagnetic transmitter signal, or if the low frequency components of the spheric are very high, this technique becomes less effective. To overcome this problem, the second strategy, named the wavelet stacking technique, uses the shift invariance and linearity of the stationary wavelet transform to perform data stacking in the wavelet domain. Tests on synthetic data results show that the wavelet stacking technique performs better than the mean and median stacking techniques. The wavelet extraction and median stacking present equivalent performance. On very noisy real data, the wavelet stacking technique makes the detection of weak anomalies more straightforward. After additional smoothing by filtering, wavelet extraction and median stacking can produce similar results to wavelet stacking. However, the amplitude and temporal decay of anomalies can be affected by high residual sferics noise. The wavelet extraction technique has the advantage that it can be used to extract sferics for an audio frequency magnetic-like method to map subsurface conductivity changes. When a large number of sferics are observed, the current practice is to stop data acquisition; these techniques allow data collection to continue.
The flexibility of geostatistical inversions in geophysics is limited by the use of stationary covariances, which, implicitly and mostly for mathematical convenience, assumes statistical homogeneity of the studied field. For fields showing sharp contrasts due, for example, to faults or folds, an approach based on the use of nonstationary covariances for cokriging inversion was developed. The approach was tested on two synthetic cases and one real data set. Inversion results based on the nonstationary covariance were compared to the results from the stationary covariance for two synthetic models. The nonstationary covariance better recovered the known synthetic models. With the real data set, the nonstationary assumption resulted in a better match with the known surface geology.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.