Importance of Estimating Measurement Noise in the Resistivity Imaging Inversion

Since measurement noise is a substantial source of uncertainty in tomographic inversion, this study estimates measurement noise and evaluates its propagation on resistivity data inversion and model results. The observed and numerically simulated resistivity datasets are analyzed regarding noise distributions. The resistivity data are numerically generated by perturbing a dike conceptual model with 5% Gaussian noises, a considered measurement noise. However, the data are inverted by estimating 1%, 5%, and 15% noises to examine the noise estimation effect on model inversion. In addition, we evaluate observed data error using reciprocal measurements, showing a 5.3% error. Based on numerical and field data analysis, inversion for underestimated noise fail to achieve a good data fit and produces artifacts and rough image. On the other hand, inversion for overestimated noise displays a noticeable smooth model and reduces resolution. Inverting resistivity data for adequately estimated noise can reduce overfitting and underfitting data and gives a good quality model. Thus, estimating measurement noise in tomographic inversion can determine model accuracy.