Recovering 3D Basement Relief Using Gravity Data Through Convolutional Neural Networks.

Gravity surveys in regional geophysical research can be used to estimate the depth of the sediment‐basement interface. In this study, we investigate a novel method using the convolutional neural network (CNN) for depth‐to‐basement inversion directly from gravity data. Based on the Random‐Midpoint‐Displacement method (RMD) and the features of the observed gravity data, we can generate a large set of realistic sediment‐basement interface models. This new method for model generation can significantly reduce the size of the training data sets which is usually considerably large to train a pervasive network. The application on synthetic models shows that the developed CNN inversion is able to capture the detailed features of the sediment‐basement interface for the complex geological model. However, so far, the training set obtained from the proposed method is still continuous and the CNN inversion still cannot effectively recover the models such as abrupt faults. We also successfully applied the developed method and workflow to a field study. The proposed approach opens a new window for estimating the physical contrast interfaces using potential field. Plain Language Summary: The sedimentary basin is related with the dynamic process of the earth and it plays an important role in natural resources exploration. As a result, it is of great importance to know the shape and depth of the basin. Generally, the sediments and the underlaid basement rocks have different densities. The depth of the basin varies at different places and such variation can cause the change of the gravity field (gravity anomaly) on the surface. Use the technique called inversion, we can transform the gravity anomaly to the depth of the basin. However, such transformation is non‐unique. In this study, we introduce the machine learning (ML) method to this field. Use tons of models and data, we train the computer to learn the relationship between the gravity anomaly and the basin depth. When the machine obtains such ability, it can automatically convert the gravity anomaly to the basin depth. This method is validated by several numerical models. In addition, we applied this method to the gravity anomaly collected at Big Bear Lake area, California. Using the ML method, we obtain a reliable basin model which fits well with the known geology and can explain the geodynamic process. Key Points: Machine learning based on the convolutional neural network is used to recover the depth‐to‐basement using gravity dataWe propose a novel model generation method for training the network to avoid adding smoothing constraints to the loss functionThe developed method can effectively recover the sediment‐basement interface and it has been successfully applied to field study [ABSTRACT FROM AUTHOR]