Advanced processing of cross-hole radar-tomographic data: inversion of partial data sets and error analysis

The detection of discrete anomalies, such as cavities and tunnels, is an important application of crosshole radar tomography. However, tomographic inversion results are frequently ambiguous showing smearing effects and artifacts. This leads to uncertainties during interpretation and, hence, the size and shape of discrete anomalies can be interpreted only with limited accuracy and reliability. In this study, we present an adapted inversion strategy for the detection of discrete anomalies with crosshole tomography. For tomographic inversion, we use various partial data sets of specified angular aperture. The resulting tomograms contain different information with respect to the vertical and horizontal resolution of discrete anomalies. Ambiguities, such as smearing and artifacts, can be recognized and considered during interpretation. From this, an adapted starting model is derived combining all additional information. Although the tomographic inversion results for different starting models differ significantly regarding the resolution characteristics of anomalies, the rms residuals are equivalent. Therefore, we additionally investigate the angular contribution of the residuals to the rms values, and propose another optimization criterion, the relative data misfit. It is shown, that the angular contribution of the residuals reflects the resolution characteristics of the tomograms.