Local noise prediction filtering (LNPF) is a method of noise reduction designed for use with the in‐loop transient electromagnetic (TEM) method of exploration. LNPF is based on simultaneous measurements of three orthogonal magnetic‐field components and the calculation of a time‐domain filter which predicts the vertical component from the two horizontal components. For in‐loop, or central induction, sounding over a horizontally layered earth, the TEM signal is contained wholly within the vertical component. Thus the predicted time series obtained with LNPF is the predicted electromagnetic noise which can be subtracted from the measured vertical component in subsequent processing. For low‐frequency noise, the method results in improvements in the signal‐to‐noise ratio of about a factor of five using a three‐point prediction filter. Longer filters give superior performance but are computationally intensive. The depth of exploration should increase by 40 percent, due to the five‐fold increase in signal‐to‐noise ratio, a figure which could be obtained otherwise only by increasing the transmitter power or the acquisition time by a factor of 25. In areas with lateral variations in conductivity the TEM response has both horizontal and vertical components. Applying LNPF in nonlayered areas will bias the vertical‐component TEM signal by a proportion of any horizontal TEM signal present. This bias error will generally be less than 5 to 10 percent, and usually will be insignificant in geologic environments where EM soundings are carried out. A new parameter, termed TEM tipper by analogy with magnetotellurics, is defined as the ratio of horizontal‐to‐vertical TEM response and is used to estimate the maximum LNPF bias error. The TEM tipper is useful for interpretation of TEM in‐loop data in areas of complex geology.
