f Two-dimensional interpretation of long geoelectric apparent resistivity profiles

Using up-to-date computer-controlled multi-electrode geoelectric instruments a large amounts of measured data can be acquired relatively quickly. An appropriate example of the application is the geoelectric measurement performed along flood-control embankments where the purpose of investigation is either the status survey of the embankment, e.g. delineation of fissured or mudded zones, or the detection of buried old river-beds below embankment. In both cases it is expedient to measure densely ( regularly at every meter ) sounding curves having enough points to obtain a suitable vertical resolution. Using multi-electrode system GRM-30 developed in KBFI TRIAS Ltd., Hungary two person can measure more than 1 km long profile during one working day measuring at every meter a 30 point pole-pole curve ( AM=1-30m ). To process a such large amounts of measured data very quick and efficient interpretation procedures are needed. We compare the efficiency of four inversion procedures based on different principles. These methods are as follows: 1. Method of alpha-centers; 2. Smoothness-constrained least-squares method using finite-difference based forward calculation; 3. Deconvolution method based on Born-approximation; 4. Occam-inversion using a multigrid+finite element technique in the forward modelling. In the first three methods a flat topography is assumed but in the last case an arbitrary two-dimensional topography can be used. All procedures are applied to the same simulated and measured datasets. We present here two examples to be inverted: 1. In the Figure 1 a buried old river bed can be seen. The resistivity decrease between 5900 and 5980m is caused by an old river bed having material with high clay content( old river beds can be recognized as an resistivity anomaly having or lower or higher resistivity depending on the curvature of old river). In the Figure 2 the interpreted results of measurements performed on the top of a partially corrected dam can be seen. Corrected segments and those which have to be corrected are well distinguishable. To decrease the computational time and the amount of required memory in the second and fourth case the long profiles are split up into overlapping shorter segments before inversion. After connecting cell-resistivities obtained from the inversion of different segments often artificial anomalies occur in the real resistivity-depth sections. Therefore it seems that the most reliable results can be obtained from a such method which inverts total measured profile simultaneously. Deconvolution method is suitable for this purpose but in the regularization in the solution of relatively small but unstable, illconditioned linear system of equations there are some unsolved problems yet. We investigate the possibility of segmentation in the solution of forward problem, too. If only short geometrical parameters ( comparing to the length of profile ) are used in forward modelling of geoelectrical measurements, e.g. when pole-pole configuration is used, than it is allowable to divide the two-dimensional resistivity model into segments and to build up the calculated apparent resistivity pseudosection from the values calculated above different segments.