A novel method for gravity data continuation from undulating surface to a horizontal plane based on deep learning

The presence of undulating terrain introduces uncertainty in the inversion and interpretation of gravity data. Therefore, the operation of continuation from undulating surface to a horizontal plane for gravity data is crucial for subsequent studies. Several methods have been proposed in the literature to deal with this problem. However, most of the existing methods yield unsatisfactory in the case of highly undulating terrain and large continuation distances. When large-scale undulating surface data is involved, the processing of continuation from undulating surface to a horizontal plane is inefficient. To address above issues, this study proposes a novel method for gravity data continuation from undulating surface to a horizontal plane based on deep learning. The proposed method utilizes the U-Net network architecture to design an end-to-end framework, followed by continuous training and parameter optimization to determine the optimal network structure. Afterward, the presented method is tested on a synthetic example. The results indicate that the proposed method is effective for gravity data continuation from an undulating surface to a horizontal plane without suffering from above-mentioned problems arisen from conventional methods.