Modeling and interpretation of electrical resistivity data in anisotropic media using a generalized regression neural network: A case study on the Lokbatan mud volcano, Azerbaijan

This study explores the electrical anisotropy effects in a vertical electrical sounding dataset acquired from the Lokbatan mud volcano in Azerbaijan. Vertical electrical sounding is a fundamental geophysical technique for mapping subsurface resistivity distributions; however, electrical anisotropy is a factor often ignored, can significantly influence field measurements, especially with respect to layer thicknesses. This study investigates electrical anisotropy effects on vertical electrical sounding data using generalized regression neural networks for estimating parameters of layered subsurface. Schlumberger electrode arrays were employed for these purposes with a 22-sounding dataset. In general, a classical approach for estimating electrical anisotropy coefficient fails. Vertical and horizontal resistivity values cannot be estimated using vertical electrical sounding data. Thus, generalized regression neural networks are considered. The generalized regression neural networks for appraising parameters of electrical anisotropy can be used with keeping vertical resistivity higher than horizontal resistivity in each layer as an assumption. Further, the proposed method applied how the surrounding clay-rich geological material and layered structure influence electrical properties of the Lokbatan mud volcano in Azerbaijan. As a conclusion, without considering electrical anisotropy makes interpretation erroneous especially with respect to layer thicknesses.