The complex conductivity of partially saturated siliciclastic sediments can now be reasonably predicted through recently developed petrophysical models like the POLARIS model. However, classical crosswell-induced polarization tomography (using a galvanometric approach) is characterized by a poor sensitivity map far from the wells, and thus other methods should be analyzed for potential improvements. The presence of low-frequency ($\le 1\mathrm{kHz}$) polarization effects in earth porous materials noticeably increases the amplitude and decreases the phase of measured electromagnetic (EM) fields. As such, the quadrature conductivity (directly associated with the low-frequency polarization effect) yields a significant contribution to the EM fields. We demonstrate that these contributions can be observed in crosswell EM data in terms of signal-to-noise ratio. With a realistic amount of noise, we can recover the distribution of the in-phase and quadrature conductivities for crosswell EM tomography. We use an integral equation approach for the forward modeling and a gradient-based approach with Tikhonov regularization for the inverse problem. We also develop a new inversion algorithm to invert time-lapse frequency domain EM data using an active-time-constrain approach. This information may be used in turn to improve our ability to monitor saturation changes in enhanced oil reservoir production, the remediation of oil spills, and the exploration and production of geothermal fields.

We propose a new, simple and efficient method to image electrical resistivity between a set of wells. Our procedure consists of two steps: first, we map the interfaces between various subsurface formations using seismoelectric conversions; second, we derive the formation resistivity using image-guided cross-well electric tomography. In the first step, we focus seismic energy at a set of points located on a regular grid between wells, which enables us to map the geological formations in terms of heterogeneities in electrical, hydraulic and/or seismic properties. The density of the scanning points (i.e. the seismoelectric image resolution) is related to the wavelength of the seismic impulse used to scan the formations. Each time the seismic energy is focused at a point, the resulting electrical potential burst (equivalent to the one generated by a volumetric seismic source) is recorded remotely at a set of electrodes positioned in wells (the reference electrode can be located on the ground surface or far enough to be considered at infinity). We construct a high-resolution ‘seismoelectric’ image by assigning the electrical potential simulated at these fixed electrodes to the location of the seismic focus. In a follow-up step, the structure of this image is used in image-guided inversion to improve electrical resistivity tomography between the two wells. The structural information from the seismoelectric image is used to impose constraints on the model covariance matrix used in the inversion of the electrical resistivity data. This approach offers new perspectives in recovering fine structure of resistivity (high definition resistivity tomography) between the wells, which cannot be resolved through conventional cross-well resistivity or from seismic tomography alone.

A 2D seismic time-lapse inversion approach can image the evolution of seismic velocities over time and space. The forward modeling is based on solving the eikonal equation using a second-order fast-marching method. Wave paths are represented by Fresnel volumes rather than by conventional rays. This approach accounts for complex velocity models and has the advantage of considering the effects of wave frequency on velocity resolution. The aim of time-lapse inversion is to find changes in velocities of each cell in the model as a function of time. Each model along the time axis is called a reference space model. This approach can be simplified into an inverse problem that seeks the optimum of several reference space models taken together, using the approximation that the change in seismic velocity varies linearly in time between two subsequent reference models. The method is demonstrated on a synthetic example that includes regularization in time in the cost function and reduces inversion artifacts associated with noise in the data by comparison with independent inversions at each time.