• Exploration Geophysics
• Electromagnetism
• Induced Polarization

Theoretical and empirical studies have shown that the Time Domain Airborne Electromagnetic (TDEM) methodology is sensitive to the response of chargeable bodies, i.e. to the Induced Polarization (IP) effects.
The Airborne EM data are acquired using an airborne system (typically a helicopter or a fixed wing platform) that induces electric currents in the halfspace (Grant & West 1965). The induced currents generate in the ground a time varying magnetic field whose decaying (∂B/∂t) is measured and represented as a discrete time decaying curve. Inverting these data is possible to map the halfspace’s resistivity and, in some cases, its chargeability. In fact, in correspondence of a capacitive halfspace, the chargeable body will behave like a capacitor able to complete a charge and discharge cycle and to generate an additional electric current (the polarization current). As a consequence, during the charging phase, the polarization current has the same sign as the induction one and the total resulting current causes an early time pull-up of the recorded signal. Whilst the induction currents proceed downwards, the capacitor begins its discharging phase and the polarization current changes its sign. Therefore, the total current gives a faster signal late time decay which can culminate in its change of sign if the halfspace is sufficiently chargeable.
This complex physical phenomenology cannot be explained with a non-dispersive resistivity model and makes the AEM data modelling highly time consuming and complex. At the same time, this computational effort is necessary since failing to recognize and properly model IP effects can lead to recover false structures with incorrect conductivity-thickness parameters.
In order to improve the AIP data modelling, the current thesis is focused on the possibility to elaborate an efficient method able to detect the presence of AIP effects in the raw dataspace. The study was developed both on helicopter and on fixed wing platform. There is a disparity between the two systems in knowledge and treatment of the airborne IP phenomenology: while for the TDEM concentric loop helicopter systems is nowadays commonly possible to model kms of AIP acquired lines, for fixed wing systems is still necessary to verify with a systematic study its capability in AIP detection.
For the fixed-wing system, it was decided to study a great number of forward responses relative to several thousand synthetic models. The intuition was to compare the forward responses between a purely resistive halfspace (i.e. defined by resistivity and layer thickness) and its equivalent, in terms of electrical stratigraphy and layering, capacitive halfspace (characterized by the Cole & Cole parameters). It was assumed that, in this condition, recording different signals guarantees a system sensitivity to AIP effects. To quantify these differences, we define a vectorial metric, called the “IP Ratio”, calculated as the ratio between the gates voltage with IP and the gates voltage without IP.
We summarize each IP Ratio into a scalar metric, called “IP Datum”, given by the number of gates with a ratio values higher than 1.1 and lower than 0.9 (considering the additive noise that affects the data).
The IP Datum is then displayed in a 3D plot in order to map the results in function of the Cole & Cole parameters. For each calculated scalar metric its respective 3-dimensional hyperspace is defined, with the displayed point that represents the Datum value and, its coordinates, the model parameters from which is calculated. Is thus possible to define for which parameter domains the Induced Polarization effects produce visible distortions.
The fixed wing data resulted highly sensitive to capacitive conditions, presenting diffuse IP distortions and negatives in several parameters domains. These effects monotonically increase with the resistivity in a homogeneous halfspace and become even more marked when a purely resistive basement is located under a shallower capacitive layer. In a three layers configuration with a deep conductive layer, the Datum is highly dependent to the second resistive layer thickness and to the first strata chargeability. All these results demonstrate how the IP phenomenology is detectable and diffusively presents in a great variety of parameters conditions distorting the fixed wing EM data.
In the light of these conclusions and of those obtained by Viezzoli & Manca (2019) on the AIP phenomenology for helicopter TDEM systems, it is demonstrated that the IP effects in AEM datasets are highly diffuse and pervasive. Following these results, a novel robust tool (the “AIP Scanner”) was developed, able to detect and spatially quantify the AIP effects in airborne EM datasets.
This tool performs an extended dataspace and a partial model space analysis. For the dataspace analysis different scalar metrics are calculate over the acquired dataset in order to define, for each sounding, its main AIP signature content. These metrics are calculated on the negative voltages (sum of negatives metrics, first negative gate metrics ecc...) and on transient decay rates.
Following the basic assumption for which inverting a capacitive halfspace without modelling the IP will produce high misfit values (Viezzoli & Kaminski 2017), few dataset lines containing both IP and non-IP effects are selected (by the metrics results) and inverted with a non-dispersive resistivity model. An inversion misfit for each sounding is thus obtained and will constitute the sub-dataset for which the dataspace metrics will be inverted. The inverse problem is formalized as:
d_obs=Misfit_i=∑(W_j*M_ij)
Where the observed data is the inversion misfit resulted from the non-dispersive resistivity modelling, W is the unknown weight-parameters and M is the metric associated to the sounding for which the misfit is calculated. We formalize the inverse problem assuming that the measured misfit for a sounding is given by the sum of the weighted metrics associated to that sounding. Solving this linear inverse problem, we obtain the weight parameters associated to each metric.
By extending the weights to the entire dataspace, a map will be produced that combines the weighted metrics to predict the AIP distribution in all the dataspace. This map allows to define different areas with variable probability of IP effects detection. This map gives back a good correlation with the geology and with the inverted chargeability cross sections. It also provides a relevant tool for the starting model choice for inversion, warning about the dataspace portions that needs a different treatment to accurately perform the inversion. The possibility of being able to define areas with different probability of detection of IP effects also allows to guide in a conscious and target-oriented way the entire data processing and modelling workflow, strongly reducing the risk of generating artifacts and false structures in the inversion output. The AIP scanner is a useful tool also to reprocess and re-model targeted portions of dataset for which, not taking into account of the polarization effects, the recovered models could contain artifacts.