oa A thorough synthetic study on IP effects in AEM data from different systems

IP effects in AEM data are subject of current research around the world, due to the recent recognition of their significance for exploration and general (hydro)geological mapping. There have been success stories and it is now practical to model AIP on thousands of line kms of AEM data. However there still is a need to study more accurately the boundaries of the effect and of its relevance, beyond common past acceptance. In this paper we present a systematic, extended analysis of AIP effect in different AEM (TEM) systems’ data, , based on synthetic modelling of different pseudo geologies. Its goal is to provide a clear overview of possible AIP effects in the data space, without imposing simplistic assumptions (e.g., fixing some parameters to arbitrary values or limited boundaries). We produce 1D forward responses with dispersive resistivity for hundreds of thousands of combinations of Cole-Cole model parameters and AEM system transfer functions. The results are analyzed using various metrics (e.g., sum of negative voltages, exponential fitting) that capture different AIP signatures in the transients. Experiments include half spaces, 2 and 3 layer models, combined with different waveforms, Rx types (dB/dt and B), Tx-Rx geometries, flying heights, transients’ binning, base frequencies. The results, presented as 4D hyperspaces, each with 10⁴transients obtained from the combinations of 4 x 10 different Cole-Cole parameters, allow a clear assessment of the AIP effects over a wide range of geophysical situations. Some of the main observations are: AIP effects are increased most often by the presence of a resistive bedrock, often using slow turn-off of the waveform, are generally better observed recording the B field instead of its derivative, and in any most cases lowering base frequencies to 12.5 Hz. In general, they are more pervasive than previously thought and should be carefully considered in virtually any AEM survey. If present, they can often be sensitive to model parameters down to depth.