A Comparison of the Model Parameter Estimations from Self-Potential Anomalies by Levenberg-Marquardt (LM), Differential Evolution (DE) and Particle Swarm Optimization (PSO) Algorithms: An Example from Tamış-Çanakkale, Turkey

Abstract

In geophysics, it is particularly important to choose an adequate optimization algorithm for parameter estimation. In this study, the success of Levenberg-Marquardt (LM), Differential Evolution (DE) and Particle Swarm Optimization (PSO) inversion algorithms has been tested by applying to the synthetic and field self-potential (SP) anomalies. Even though it is not preferred to compare derivative-based algorithms with metaheuristics, thanks to a LM-based limitation procedure first proposed in this study, a comparison could be realized. First, a synthetic SP data have been inverted by LM, DE and PSO algorithms. Then, SP field data set collected from Tamis-Canakkale, Turkey was evaluated by the same algorithms. The estimated model parameters by these algorithms were compared with each other. We also inverted vertical electrical sounding (VES) data set collected from the same region, and an earth model was constructed by using both SP and VES methods. The results from each geophysical method point out the same location for a fault. Based on these studies, it can be concluded that DE, PSO, and LM algorithms may be confidently used in SP modelling studies.