A layer stripping approach for monitoring resistivity variations using surface magnetotelluric responses

Abstract

The resolution of surface-acquired magnetotelluric data is typically not sufficiently high enough in monitoring surveys to detect and quantify small resistivity variations produced within an anomalous structure at a given depth within the subsurface. To address this deficiency we present an approach, called "layer stripping", based on the analytical solution of the one-dimensional magnetotelluric problem to enhance the sensitivity of surface magnetotelluric responses to such subtle subsurface temporal variations in resistivity within e.g. reservoirs. Given a well-known geoelectrical baseline model of a reservoir site, the layer stripping approach aims to remove the effect of the upper, unchanging structures in order to simulate the time-varying magnetotelluric responses at depth. This methodology is suggested for monitoring all kinds of reservoirs, e.g. hydrocarbons, gas, geothermal, compress air storage, etc., but here we focus on CO2 geological storage. We study one-dimensional and three-dimensional resistivity variations in the reservoir layer and the feasibility of the method is appraised by evaluating the error of the approach and defining different detectability parameters. The geoelectrical baseline model of the Hontomín site (Spain) for CO2 geological storage in a deep saline aquifer is taken as our exemplar for studying the validity of the 1D assumption in a real scenario. We conclude that layer stripping could help detect resistivity variations and locate them in the space, showing potential to also sense unforeseen resistivity variations at all depths. The proposed approach constitutes an innovative contribution to take greater advantage of surface magnetotelluric data and to use the method as a cost-effective permanent monitoring technique in suitable geoelectrical scenarios.

Time-lapse inversion of one-dimensional magnetotelluric data

“…Recent work includes parameterising resistivity changes as a three-dimensional (3D) plume structure and inverting using Markov Chain Monte Carlo (Rosas-Carbajal et al 2015). Another approach has been to use 1D layer-stripping, where the effect of overlying structures is removed to model the time-varying magnetotelluric responses at depth (Ogaya et al 2016). Standard MT modelling tools such as inversion can also be adapted, for example Rees et al (2016) use cascading two-dimensional (2D) inversions to model a coal-seam gas depressurisation, where the results from the preinjection inversion are used as a prior-model for the postinjection.…”

Particle Swarm Optimization (PSO) for Magnetotelluric (MT) 1D Inversion Modeling

“…However, non-linearity and ill-posedness of the problem remain to be an interesting subject for research, especially for the application of global search approach e.g. [8,9,10]. The PSO algorithm has been applied to non-linear inversion of magnetotelluric (MT) and Vertical Electrical Sounding (VES) data using 1-D model with satisfactory results [11,12].…”

Improved Parallel PSO Inversion for the MT Sounding Data

“…A knowledge of the resistivity is important since rock types important to hydrocarbon exploration can be differentiated on the basis of resistivity value [1,2]. It has some advantages such as low cost, large probing depth, not affected by high-resistivity shielding and high resolution to the low-resistivity layer.…”

“…To overcome the flaws of the classical linear inversion method, some nonlinear inversion methods, such as Monte Carlo, simulated annealing and genetic algorithm are employed to the MT inversion [4]. However, they have some disadvantages [2].…”

“…In recent years it has also been introduced into the geophysical inversion [7][8][9][10][11]. Although the disadvantage, such as local optimum, of geophysical inversion based on PSO has been improved by employing the adaptive inertial weight, it still has the problem of heavy computation time when the forward modeling becomes more complex [2]. Parallel implementation can reduce the computation time of inversion obviously.…”