This research implements 2D gravity data inversion using the Levenberg-Marquardt algorithm to analyze the sensitivity of modeling results to variations in initial models. Testing was conducted on two synthetic models, namely intrusion and fault, with density contrasts of 0.5 g/cm³ and 0.7 g/cm³ respectively. The inversion results demonstrate a critical dependence on the initial model. A homogeneous initial model produced distorted density estimates concentrated near the surface. Conversely, an initial model approximating the actual synthetic model yielded faster convergence and more accurate density distribution reconstruction. Comparative analysis revealed that the fault model achieved better fit, indicating the superior lateral resolution of the gravity method compared to its vertical resolution. These findings confirm that the accuracy of gravity data inversion heavily relies on geological constraints integrated through the initial model, where an informed approach produces more reliable subsurface models.
Copyrights © 2025