The inversion of geoelectric data is nonunique. Therefore, electric resistivity tomography (ERT) usually results in different subsurface models that fit observed apparent resistivity values equally well. To reduce the uncertainty, constraints on the geometry and resistivity of subsurface structures can be incorporated into the ERT inversion. We test different ways of constraining ERT by applying (1) improved starting models, (2) structural constraints, and (3) structural and resistivity constraints. A priori information is needed for these approaches, which is acquired from direct push electrical conductivity (DP-EC) logs and vibracores in our study. We found that adapting high vertical resolution of DP-EC logs to coarser ERT resolution requires blocking of the logs. These blocked logs can be used for constraining the ERT inversion with resistivity and structural constraints along so-called regions around the log locations, which was found to be the best approach. In this case, ground truthing with additional DP-EC logs that were not used for constraining showed the lowest root-mean-square error between logs and ERT model.

Hyperbolic diffractions in Ground Penetrating Radar (GPR) data are caused by a variety of subsurface objects such as pipes, stones, or archaeological artifacts. Supplementary to their location, the propagation velocity of electromagnetic waves in the subsurface can be derived. In recent years, it was shown that deep learning tools can automatically detect hyperbola in radargrams using data measured over urban infrastructure, which are relatively clear. In contrast, in this study, we used an archaeological dataset with diverse underground structures. In the first step we used the deep learning network RetinaNet to detect hyperbola automatically and achieved an average precision of 0.58. In the next step, 10 different approaches for hyperbola fitting and thus velocity determination were applied. The derived information was validated with manually determined velocities and apex points. It was shown that hyperbola extraction by using a threshold and a column connection clustering (C3) algorithm followed by simple hyperbola fitting is the best method, which had a mean velocity error of 0.021 m/ns compared to manual determination. The average 1D velocity-depth distribution derived in 10 ns intervals was in shape comparable to the manually determined one, but had a systematic shift of about 0.01 m/ns towards higher velocities.

Frequency-domain electromagnetic induction (FDEMI) methods are frequently used in non-invasive, area-wise mapping of the subsurface electromagnetic soil properties. A crucial part of data analysis is the geophysical inversion of the data, resulting in either conductivity and/or magnetic susceptibility subsurface distributions. We present a novel 1D stochastic optimization approach that combines dimension-adapting reversible jump Markov chain Monte Carlo (MCMC) with artificial bee colony (ABC) optimization for geophysical inversion, with specific application to frequency-domain electromagnetic induction (FDEMI) data. Several solution models of simplified model geometry and a variable number of model knots, which are found by the inversion method, are used to create re-sampled resulting average models. We present synthetic test inversions using conductivity models based on 14 direct-push (DP) EC logs from Greece, Italy, and Germany, as well as field data applications using multi-coil FDEMI devices from three sites in Azerbaijan and Germany. These examples show that the method can effectively lead to solutions that resemble the known DP input models or image reasonable stratigraphic and archaeological features in the field data. Neighboring 1D solutions on field data examples show high coherence along profiles even though each 1D inversion is independently handled. The computational effort for one 1D inversion is less than 120,000 forward calculations, which is much less than usually needed in MCMC inversions, whereas the resulting models show more plausible solutions due to the dimension-adapting properties of the inversion method.