Doubling the spectrum of time-domain induced polarization: removal of non-linear self-potential drift, harmonic noise and spikes, tapered gating, and uncertainty estimation

This paper presents an advanced signal processing schemefor time-domain induced polarization full waveform data.The scheme includes several steps with an improvedinduced polarization (IP) response gating design usingconvolution with tapered windows to suppress highfrequency noise, a logarithmic gate width distribution foroptimizing IP data quality and an estimate of gatinguncertainty. Additional steps include modelling andcancelling of non-linear background drift and harmonicnoise and a technique for efficiently identifying andremoving spikes. The cancelling of non-linear backgrounddrift is based on a Cole-Cole model which effectivelyhandles current induced electrode polarization drift. Themodel-based cancelling of harmonic noise reconstructs theharmonic noise as a sum of harmonic signals with acommon fundamental frequency. After segmentation ofthe signal and determining of noise model parameters foreach segment, a full harmonic noise model is subtracted.Furthermore, the uncertainty of the background driftremoval is estimated which together with the gatinguncertainty estimate and a uniform uncertainty gives atotal, data-driven, error estimate for each IP gate. Theprocessing steps is successfully applied on full field profiledata sets. With the model-based cancelling of harmonicnoise, the first usable IP gate is moved one decade closerto time zero. Furthermore, with a Cole-Cole backgrounddrift model the shape of the response at late times isaccurately retrieved. In total, this processing schemeachieves almost four decades in time and thus doubles theavailable spectral information content of the IP responsescompared to the traditional processing.