3D non-linear inversion of magnetic anomalies caused by prismatic bodies using differential evolution algorithm ☆ Çağlayan Balkaya a, ⁎, Yunus Levent Ekinci b,c , Gökhan Göktürkler d , Seçil Turan d a Süleyman Demirel University, Engineering Faculty, Department of Geophysical Engineering, West Campus, TR-32260 Isparta, Turkey b Bitlis Eren University, Faculty of Arts and Sciences, Department of Archaeology, TR-13000 Bitlis, Turkey c Career Research and Application Center, Bitlis Eren University, TR-13000 Bitlis, Turkey d Dokuz Eylül University, Engineering Faculty, Department of Geophysical Engineering, TR-35160 İzmir, Turkey abstract article info Article history: Received 4 April 2016 Received in revised form 24 October 2016 Accepted 25 October 2016 Available online 5 November 2016 3D non-linear inversion of total field magnetic anomalies caused by vertical-sided prismatic bodies has been achieved by differential evolution (DE), which is one of the population-based evolutionary algorithms. We have demonstrated the efficiency of the algorithm on both synthetic and field magnetic anomalies by estimating horizontal distances from the origin in both north and east directions, depths to the top and bottom of the bodies, inclination and declination angles of the magnetization, and intensity of magnetization of the causative bodies. In the synthetic anomaly case, we have considered both noise-free and noisy data sets due to two vertical-sided prismatic bodies in a non-magnetic medium. For the field case, airborne magnetic anomalies originated from in- trusive granitoids at the eastern part of the Biga Peninsula (NW Turkey) which is composed of various kinds of sedimentary, metamorphic and igneous rocks, have been inverted and interpreted. Since the granitoids are the outcropped rocks in the field, the estimations for the top depths of two prisms representing the magnetic bodies were excluded during inversion studies. Estimated bottom depths are in good agreement with the ones obtained by a different approach based on 3D modelling of pseudogravity anomalies. Accuracy of the estimated parame- ters from both cases has been also investigated via probability density functions. Based on the tests in the present study, it can be concluded that DE is a useful tool for the parameter estimation of source bodies using magnetic anomalies. © 2016 Elsevier B.V. All rights reserved. Keywords: Differential evolution Metaheuristic Non-linear inversion Magnetic anomaly Prismatic bodies Granitoids 1. Introduction Surface and/or airborne magnetic surveying are frequently used in exploration geophysics. Surface magnetic surveying is an effective tool especially for small-scale investigations such as archaeological remains (e.g., Quesnel et al., 2011; Stampolidis and Tsokas, 2012; Büyüksaraç et al., 2014; Ekinci et al., 2014), while airborne magnetic surveys, the most common type of airborne geophysical surveys, are mainly used for large-scale explorations dealing with mineral deposits, oil and gas, boundaries between magnetically contrasting rocks and magnetic crust (e.g., Eventov, 1997; Jallouli et al., 2003; Ates et al., 2005; Airo and Mertanen, 2008; Xu et al., 2011; Ekinci and Yiğitbaş, 2012). One of the most commonly used data processing techniques in magnetic anomaly interpretation is the determination of model parameters by a linearized inversion approach. Inversion of magnetic data by both con- ventional deterministic algorithms using the gradient information (e.g., Rao and Babu, 1991; Li and Oldenburg, 1996; Fedi and Rapolla, 1999) and the stochastic algorithms displaying some randomness (e.g., Boschetti et al., 1997; Shamsipour et al., 2011) has been performed over the years to investigate subsurface materials and geology. Gradient-based algorithms widely applied for estimating the model parameters depend on the initial guess required to start the optimization process. In addition, they lack of a mechanism to avoid local minima. Thus, they may be trapped in a local minimum instead of reaching the global minimum. It is well known that a priori information for the param- eters is either weak or non-existent for many optimization problems. Contrary to conventional least-squares approaches, metaheuristics do not require good initial estimates to reach to the global minimum. Metaheuristics, often inspired by the processes in the nature, can be classified in two groups as population-based (e.g., genetic algorithm, particle swarm optimization, differential evolution algorithm, firefly al- gorithm, cuckoo search) and trajectory-based (e.g., simulated annealing algorithm, tabu search) algorithms (Yang, 2014, p.16). Even though the metaheuristic algorithms suffer from high computational cost due to Journal of Applied Geophysics 136 (2017) 372–386 ☆ Some parts of this paper were presented at the 20th International Geophysical Congress and Exhibition of Turkey, 25–27 November 2013, Antalya (Turkey) and at International Conference on Engineering and Natural Science (ICENS), 15-19 May 2015, Skopje (Macedonia). ⁎ Corresponding author. E-mail addresses: caglayanbalkaya@sdu.edu.tr (Ç. Balkaya), ylekinci@beu.edu.tr (Y.L. Ekinci), gokhan.gokturkler@deu.edu.tr (G. Göktürkler), secil.turan@deu.edu.tr (S. Turan). http://dx.doi.org/10.1016/j.jappgeo.2016.10.040 0926-9851/© 2016 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Journal of Applied Geophysics journal homepage: www.elsevier.com/locate/jappgeo