Enhanced removal of very low frequency (VLF) and low frequency (LF) radio noise from transient electromagnetic (TEM) data with modeling and adaptive filtering

Abstract. Interference from very low frequency (VLF, 3–30 kHz) and low frequency (LF, 30–300 kHz) radio stations is a ubiquitous and challenging noise source in transient electromagnetic (TEM) data. It can be difficult to suppress interfering radio signals with the commonly applied methods of gating and stacking. However, the characteristics of VLF and LF radio signals encoded with minimum-shift keying methods allow for a better solution where the noise is modeled and subtracted. This approach has previously been shown to give good results for continuous streams of TEM data. Recently proposed new use cases for TEM instrumentation, such as time-lapse measurements of fluctuating groundwater levels and dynamic groundwater-saltwater interfaces produce discontinuous streams of TEM data with regular gaps between individual transients. We show that under mild constraints of data availability, radio signals can still be modeled in this case. We further show that the addition of an adaptive filter can fine-tune the radio model and improve the signal-to-noise ratio. The performance is analyzed on a synthetic noise data set and on a real field noise data set. For this field noise data set, we find that the standard errors of early time TEM data are improved by about a factor of two.