Improving the inversion of helicopter-borne frequency-domain electromagnetic data with depth constrained

Generally, the measured secondary field data is inverted into resistivity using two principal models; the homogeneous half-space model and the layered half-space model. While the homogeneous half-space inversion uses single frequency data, the inversion is done individually for each of the frequencies used, the multi-layer 1D inversion is able to take the data of all frequencies available into account. The resulting parameter of the half-space inversion is the apparent resistivity which is the inverse of the apparent conductivity. It's possible that using the fast method to calculate the apparent resistivity, if the distance between the HEM sensor and the top of the half-space is known. Unfortunately, the dependency of the secondary field on the half-space resistivity is highly non-linear. Thus, the inversion is not straightforward and the apparent resistivities have to be derived by the use of look-up tables, curve fitting or iterative inversion procedures (Fraser, 1978; Siemon, 1997; Siemon, 2001).The usual technique for inversion of airborne electromagnetic data frequency domain (HEM) data is a 1D single site inversion, because of the 2D and 3D inversion of HEM data wants very powerful computer hardware. Some inversion method for electromagnetic data inversion suggested. Usually this method updated for ground electromagnetic methods. One of the methods employed in the inversion of airborne electromagnetic data frequency domain (HEM), Levenberg-Marquardt method inversion (MLI) is looking for smoothing fitted to the data in the inversion algorithm; this inversion method based on least squares criteria, seeking a modelby minimizing the residuals of an objective function. Marquardt’s inversion only pursuits the largest fitting of simulation data to original measurements, and has the characteristics of simple algorithm and fast calculation. In this procedure usually HEM data smoothed and then used in the inversion procedure, but any variation in data change results. For stability of inversion procedure, it is suggested that stitched-together 1-D models along the profile that each sounding inverted by constrained neighbor sounding and each layer of each sounding inverted by depth constrained neighbor layers. In addition used smoothing constrained in inversion procedure instead of smoothing a data like Marquardt–Levenberg inversion.
In this paper, Starting model determined for apparent resistivity with Mundry technique and for centroied depth with Weidelt technique. To using this method, the auto inversion cod written in MATLAB software environment that inputs are real and imaginary part of data with sensor altitude and output is inverted model with misfit. In the following this algorithm tested on standard synthetic data, the model chosen for the generation of synthetic data represents a layered earth structure having an inhomogeneous top layer in order to study the influence of shallow resistivity variations on the appearance of deep horizontal conductors in one-dimensional inversion results. The inversion of synthetic data results shown this technique for inversion HEM data improved the results and is much more accurate than Marquardt–Levenberg inversion. Finally the inversion algorithm used to invert a set of real DIGHEM field data from Mirgah Naqshineh area in Saqqez of Kurdistanand interpretation of results according to geology information of area.