Constrained inversion of RMT data using GPR sections versus their joint interpretation in investigation of an aquifer

Radio magnetotelluric (RMT) and ground penetrating radar (GPR) are known as the near-surface geophysical methods in groundwater investigations. The RMT method provides information about the variation of the electrical resistivity of 50 m of the uppermost part of the ground. High-resolution structural information can be extracted from the GPR processed sections of the very shallow ground. Combining the obtained data using these two methods lead to valuable results on the identification of near-surface layers and structures. In this study, we propose a new constraint for the two dimensional (2D) inversion of the RMT data. We have investigated a known aquifer located in Heby, Sweden, to assess the constrained inversion results using a joint interpretation approach. RMT and GPR surveys have been carried out along two survey lines having the lengths of 870 m and 550 m, respectively. The results show that thick saturated zones are distinguished quite well either in the joint interpretation results or when using the constrained inversion approach. In such cases, the main problem is to locate the water table in the inverted RMT sections. Imposing smooth regularization in the inversion results turns rather sharp boundaries into the gradual transition zone in the final resistivity models. Thus, using the GPR common-offset (CO) reflections as constraints in the inversion of the RMT can recover the water table as a sharp interface in the RMT inverted model. Thin saturated zone has not been recognized in the RMT sections, due to low resolution of the RMT method. For verification of the results, we have evaluated a synthetic model with similar physical properties to the study area. In such circumstances, the results need to be improved either in the joint interpretation or the constrained inversion approach using CO sections. Hence, harder constraints through our proposed scheme have been incorporated into the inversion routine to detect a thin aquifer and achieve a more realistic model.

The RMT and GPR methods are among the most useful non-invasive methods, which can provide continuous data for groundwater exploration. The RMT method due to its limited range of frequencies (10-250 KHz) has low resolution, especially at very shallow depth, and the GPR method itself suffers from its limited penetration depth. Hence, it seems that combining the modeling results of these two methods leads to a more accurate anomaly definition. Reflection (seismic or GPR) data are usually used as constraints in electromagnetic data inversion. Although all reflectors in seismic and GPR sections are not attributed to the distinct resistivity contrasts, in GPR they are mainly related to the dielectric contrast or may occur due to the thin layers embedded in homogenous geological formations. Thus, we propose an alternative scheme to incorporate interfaces with distinct resistivity contrast in the RMT data inversion.

Using all GPR reflections as constraints in the RMT data inversion may cause some artifacts in the final inverted model. In low clay content formations, such as clean sand and gravel formations, dielectric constant and resistivity are mainly related to the volumetric water content. Therefore, we propose a new structural constraint based on the assumption that the resistivity and water content contrasts occur at the same boundaries. To establish this constraint, we have used common mid-point (CMP) velocity analysis as well as the combination of Topp’s and Archie’s relationships. As a result, an initial resistivity model has been deduced from the CMP velocity analysis that can be used as a priori information in the RMT data inversion.

Thick saturated zones (having thicknesses of more than 10 m) have been distinguished quite well by applying smooth constraint inversion of the RMT data as the joint interpretation of The RMT and GPR data leads to a reasonable outcome in this regard. Although sharp boundaries are mapped as gradual interfaces in the inverted resistivity section of the RMT data, such interfaces are recovered well by incorporating the GPR result as a priori information in the constrained inversion of the RMT data. The water table at a depth of 10 to 20 m, and consequently, the saturated zone is resolved well in this constrained inversion method. It correlates to the borehole log information. On the other hand, thin saturated layers could not be distinguished in the RMT sections due to its low resolution. It means that the water table at a depth of 10 to 15 m is not mainly detected when only the determinant mode data are used. In such areas, the constrained inversion of the RMT data using the water table location deduced from the CO GPR data also fails. However, we have incorporated harder constraints through the model covariance matrix and prior information in our proposed constrained inversion routine. Using this approach, a local thin aquifer has been recognized well. Furthermore, our proposed technique can be used in the inversion of other electric and electromagnetic data.