Gradient descent fusion for gravity and magnetic data

Abstract

Subsurface characterization is a crucial aspect of geophysical exploration, enabling the identification and understanding of valuable geological bodies and resources. In this context, joint inversion of gravity and magnetic data has emerged as a powerful geophysical exploration technique, allowing for a more coherent and consistent interpretation of subsurface structures. The study focuses on understanding residual gravity and magnetic anomalies by employing the gradient descent-based joint inversion approach. A MATLAB program was developed to determine the inverse gravitational and magnetic anomalies using the gradient descent approach. We explored the potential of 2D rectangular prisms as a popular geometry to represent mineralized bodies and oil and gas structures. To overcome the non-uniqueness issues, we designed code for joint inversion of gravity and magnetic data. Synthetic data was inverted using the gradient descent technique and compared with the least-squares approach. Numerical simulations and real data application successfully reconstructed the geometry of the prisms. An illustrative example of a prism fault was used for further evaluation. Real data from the Oka complex in Quebec, Canada, was collected from the literature and subjected to joint and individual gravity and magnetic modelling. The results highlighted the influence of heterogeneous mass distribution on matching forward anomalies. The high gravity anomaly in the Oka complex was attributed to carbonatite and silicate rocks. The presence of two intrusive centres within the complex caused the magnetic high. This work demonstrates the effectiveness of the gradient descent approach as it consistently outperformed the conventional method, offering a robust solution for subsurface characterization in geophysical exploration.