Abstract: The atomization of a liquid jet injected into supersonic crossflow involves complex gas–liquid interactions. Efficient and accurate numerical simulations are of great significance for analyzing flow structures and predicting spray characteristics. However, full-process simulations based on Eulerian–Eulerian frameworks usually require extremely fine computational grids to resolve small-scale liquid structures, leading to very high computational costs. In contrast, conventional Eulerian–Lagrangian methods are computationally efficient for dispersed droplets but have difficulty accurately capturing the evolution of the continuous liquid column and the interaction between shockwaves and gas–liquid interfaces near the nozzle. To address these issues, a coupled MGFM-LTDC method integrating Eulerian–Eulerian and Eulerian–Lagrangian frameworks is proposed in this work. In the continuous region, the modified Ghost Fluid Method (MGFM) using Level Set Method is employed to capture the evolution of gas–liquid interfaces with high fidelity. In the atomization region where a large number of droplets are formed, the Lagrangian Tracking of Droplet Clusters (LTDC) method is adopted to simulate the dynamics, transport, and secondary breakup of dispersed droplets, enabling efficient multiscale computation. To achieve the coupling of the two regions, droplet identification and volume fraction conversion are employed. In addition, to address the presence of incompletely fragmented liquid structures, physical information statistics incorporating a liquid filaments segmentation algorithm is developed. Numerical experiments under four supersonic conditions verify the effectiveness of the proposed method. The results demonstrate that the coupled method can efficiently and accurately reproduce the bending, breakup, and secondary atomization characteristics of the liquid jet in supersonic crossflow. Moreover, the simulated distance of detached shockwaves and penetration depth of the spray show good agreement with experimental correlations. The proposed method significantly reduces the computational cost while maintaining simulation accuracy, providing an efficient and reliable approach for the numerical simulation of liquid jet atomization in supersonic crossflow.