Processing and Inversion of Airborne Gravity Gradient Data
In this thesis, a data-driven method for determining and reducing noise in AGG data will be presented first. The new noise reduction method is based on the idea of iteratively projecting survey data onto a lower level, upward continuing the data back to the original survey height, and then subtracting the upward continued data from the survey data. This method is successfully applied to the AGG data over Karasjok, Norway. The results show that the new noise reduction method can detect and reduce some high-frequency noise. Next, a fast equivalent source approach based on Landweber iteration and Gauss-FFT is developed. By applying the method to a synthetic dataset, the method shows great efficiency. Subsequently, two applications to real data over Karasjok are presented. The first is to jointly denoise the data with carefully selected parameters. The results are comparable to the routinely processed data which represents the industry standard. The second is to estimate densities of the topography in Karasjok with the data. The results show that the estimation method is a fast way to acquire an overview of densities of topography when only sparse petrophysical samples are available. At last, to obtain detailed density distributions of the survey area and evaluate the possibilities for mineralization, a stochastic inversion constrained by a prior lithology model and petrophysical data is applied to the AGG data. By inverting various combinations of AGG components, the results suggest that noise reduction prior to inversion is not necessary when the existing noise level is low and behaves like zero-mean Gaussian noise. The results also indicate that the constructed and the measured components both can be used for inversion and the inclusion of more than four components in the inversion does not provide additional information. From the acquired density models, insights into potential mineralization in the Karasjok area are provided.