Natural Field Electromagnetics Using a Partially Known Source: Improvements to Signal to Noise Ratios

Abstract

SUMMARY We aim to provide a novel approach to processing and interpretation of natural fields electromagnetic (EM) data through automated interpretation of sferic source parameters provided by the World Wide Lightning Location Network (WWLLN). Accurate sferic time stamps obtained from WWLLN are hypothesised to improve signal to noise ratios (SNR) through precisely controlled extraction of sferics with amplitudes both above and below observed noise levels in time series audio-magnetotelluric (AMT) measurements. Averaging of extracted data of equal source moment increases signal in proportion to the square root of the number of averages whilst decreasing noise since data between sferic events is inconsequential and can be discarded. Knowledge of source characteristics allows further improvements to data quality to be achieved through discrimination of sferic sources that do not meet the required assumptions of AMT. Since sferic propagation occurs primarily along great circle paths, antipodal sferics propagating around the reverse side of the earth can in principle be extracted for inclusion as additional signal. These increases in SNR afford reduced measurement time and an increase in data quality, leading to more cost effective exploration. Use of source information is unique amongst existing approaches to time domain AMT. The existing approaches are often limited in application to signals with amplitudes well above the noise level. ARMIT sensors developed at RMIT are sensitive to natural EM fields generated by world wide lightning activity and can measure up to 40 sferics per second. We collected AMT data using one such sensor in order to carry out an initial feasibility study on our hypothesis. Preliminary results are encouraging and demonstrate that WWLLN data are accurate and efficient enough to predict useful sferic arrival times. Future efforts will be applied to characterizing waveform similarity and investigating the relationship between stacking techniques and data reliability. This may improve SNR and hence the prediction of subsurface geological structure.