Qualification of inversion inputs in certain engineering geophysical methods

The present paper analyzes the possibilities for input qualification necessary for inversion of certain engineering geophysical measurement data series. The work formulates at first the requirements for determination of geophysical conclusions reliability and/or accuracy. Then it reviews the quantities playing a role in qualification of geophysical data and geophysicl conclusions deduced from the data. Statistical characterization of the ek = (yk - fk (x, z)) random differences between the yk geophysical measurement data and the fk (x, z) theoretical forward problem solutions calculated for the (x, z) model parameters is indispensable in quality controlled geophysical inversion. To qualify the x̂ final result of inversion first of all the error components of observational and modelization origin of ek random differences should be determined. To give the σ 2κ variances of the ek differences characterizing the data-model relationship experimental testing should be carried out. To estimate the input errors of least squares type geophysical inversion approximate methods can be suggested beside the ideal testing procedures. Such are: the subjective, the literature-based, the partial and the feedback data-model relationship qualification. In addition to them for certain inversions using 1D and 2D forward problem solutions, self-control based model class adaptive engineering geophysical measurements can also be applied. According to the testing experiences gained with different approximate input qualifications not any theoretical speculation or sophisticated computation procedure is able to substitute the experimental measurement studies. © 2005 Akadémiai Kiadó, Budapest.