Quasi-2D Resistivity Model from Inversion of Vertical Electrical Sounding (VES) Data using Guided Random Search Algorithm

Abstract

Abstract. Vertical electrical sounding (VES) data are usually interpreted in terms of a 1D resistivity model using linearized inversion. The local approach of a non-linear inverse problem has fundamental limitations, i.e. the necessity of a starting model close to the solution and possible convergence to a local rather than a global minimum solution. We studied the application of a global search approach for non-linear inversion using the guided random search method to model VES data. A quasi-2D resistivity model can be created by stitching 1D models obtained from VES data along a profile. Both vertical and lateral resistivity variations are minimized to incorporate a 2D smoothness constraint. The proposed method was applied to invert synthetic VES data as well as field data from a sedimentary environment. Both synthetic and field data inversions resulted in models that correlated well with the known synthetic model and with the geology of the study area, respectively.

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

“…This approach can effectively overcome difficulties of local search or linearized approach of strongly non-linear inverse problems. The MonteCarlo based algorithms such as Simulated Annealing (SA), Genetic Algorithm (GA), Markov Chain Monte Carlo (MCMC) are among the most popular non-linear inverse modeling algorithms that have been applied to geophysical inverse problems [3,4,5]. One of global approach algorithms which is gaining interest for geophysical inverse modeling is the Particle Swarm Optimization (PSO) proposed by Kennedy and Eberhart [6,7].…”