A resolution study of buried valleys using laterally constrained inversion of TEM data Esben Auken ⁎, Anders V. Christiansen, Lars H. Jacobsen 1 , Kurt I. Sørensen The Hydrogeophysics Group, Department of Earth Sciences, Høegh-Guldbergs Gade 2, 8000 Aarhus C, University of Aarhus, Denmark ABSTRACT ARTICLE INFO Article history: Received 11 April 2007 Accepted 10 March 2008 Keyword: LCI SkyTEM TEM Buried valleys Paleo channels Inversion In this paper we present a study where the 1D laterally constrained inversion (1D-LCI) algorithm is used to invert continuously sampled synthetic 2D TEM data sets with 3D near-surface resistivity variation. The models are intended to closely resample typical hydrogeological targets such as paleo-channels or buried valley structures. In many parts of the world, these structures carry significant groundwater resources, or they can be associated with mineral deposits. Generating synthetic responses over known targets is an efficient to quantify how well a model is recovered by a combination of the applied geophysical method and the inversion algorithm. The 1D-LCI algorithm gives quasi 2D images of the subsurface efficiently suppressing 3D effects and the effect of data noise. Furthermore, layers with little signature in the data become resolved because the LCI algorithm distributes laterally the information. Based on the inversion of the synthetic 2D data sets we have constructed a robust setup of the inversion algorithm in terms of strength of the laterally constraints and regularization. This setup is used to invert measured data sets from a SkyTEM survey carried out over a buried valley structure. © 2008 Elsevier B.V. All rights reserved. 1. Introduction The transient electromagnetic (TEM) method is an induction method by which an electric current is induced into the ground from a high-powered transmitter loop. The size of the transmitter loop for groundbased systems can be in the order of 40 × 40 m 2 up to more than 200×200 m 2 . When the transmitter has build up a steady current in the loop, the current is abruptly turned off, by which – due to Faraday induction – new currents are induced into the ground. It is the decaying magnetic field from these currents that is measured in an induction coil at the surface of the ground. As the method does not need any galvanic contact with the ground it can efficiently be applied on the ground or from airborne platforms. The TEM method has been used worldwide for hydrogeological surveys since Fitterman and Stewart (1986) undertook a theoretical study on the applicability of the method for groundwater investiga- tion. It is a fast and relatively cheap method for exploring the sub- surface. The method yields high resolution of layers with low- resistivity making it suitable for delineation of high-resistivity aquifers bound by heavy clays, or for mapping fresh water–salt water interfaces (e.g. Meju et al., 1999; Dugue et al., 2008). During the last decade new and enhanced helicopter-borne TEM systems have been developed. These systems not only increased the data volume significantly, but also gave so far unseen resolution ca- pabilities of the subsurface. In some systems the data quality is comparable to that of similar data from groundbased systems and therefore calls for application of quantitative modeling and inversion algorithms. We maintain that the speed of processing and inversion is not as important as stability and the capability to extract maximum information about the subsurface resistivity structures from the data. Preferably, the inversion algorithm should benefit from the dense spatial data distribution obtained from the helicopter systems. This is possible if a traditional inversion scheme of single-site data sets is expanded to simultaneously inverting a large number of data sets along a profile thereby creating pseudo-2D images using a 1D based forward solution. An example of such an algorithm is the 1D laterally constrained inversion (1D-LCI) by Auken et al. (2005). Santos (2004) has published a similar algorithm for EM34 data. A number of approaches for the 3D forward modeling of time domain EM response have been presented in the literature by e.g. Arnason (1995), Best et al. (1995), Alumbaugh et al. (1996) and Sugeng (1998). Full 3D inversion of TEM data is more rare. However, algorithms have been presented by e.g. Alumbaugh and Newman (2000) and Haber et al. (2004). Even though these studies show that 3D inversion of TEM data is possible, the computation power involved is significant, and presently efficient use of the algorithms requires access to small computer clusters. Multi-source data as airborne TEM data are even more challenging because the electric fields in the subsurface have to be solved for each source position. Journal of Applied Geophysics 65 (2008) 10–20 1 Currently at Orbicon Inc, Viby, Denmark. ⁎ Corresponding author. E-mail addresses: esben.auken@geo.au.dk (E. Auken), anders.vest@geo.au.dk (A.V. Christiansen). URL: http://www.hgg.au.dk (E. Auken). 0926-9851/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jappgeo.2008.03.003 Contents lists available at ScienceDirect Journal of Applied Geophysics journal homepage: www.elsevier.com/locate/jappgeo