Due to its unique geographical location and complex geological evolution processes, the South China Sea has been a focus of extensive research. Previous studies on the density structure of the South China Sea mostly focused on 2D density structures, with relatively limited research on 3D density structures. A comprehensive study is still needed to refine the expansion mechanism and tectonic evolution of the South China Sea. In this study, we utilized wavelet multi-scale analysis of gravity data to obtain a 3D density model of the South China Sea and discussed its tectonic evolution from the pattern of density anomalies. The inversion results show that (1) the expansion of the South China Sea caused the typical thin oceanic crust and parts of the continent–ocean transition zone may fracture due to the expansion; (2) the low-density anomaly in the upper mantle of Luzon Island may indicate partial melting or the upwelling of asthenosphere materials; and (3) the expansion of the South China Sea is influenced by multiple plate forces and uneven forces affect the distribution of high-density anomalies in the upper mantle.

The South China Sea (SCS) is the result of several major tectonic plates’ interaction, and a more detailed density structure is essential to study the tectonic evolution of the marginal seas. In the past decades, density studies in the SCS have focused more on profiles, and the lateral density structure has been less mentioned by scholars. Previously, gravity data have been used many times to study the SCS; this can satisfy the demands of large-scale lateral studies, but a point that cannot be ignored is that the gravity method has certain deficiencies in vertical resolution. Therefore, this paper applies a gravity multiscale analysis method to isolate the Bouguer gravity anomaly signal in the SCS, and then uses the power spectrum method to estimate the mean field source depth, which corresponds to the decomposed gravity anomaly signal. On this basis, we modeled stratified crustal and upper mantle structures in the SCS to determine the lateral density distribution at different depths. The results show the following: (1) There are two large high-density materials in the SCS. The first one is located near the Yinggehai Basin, where the depth ranges from 42.4 km to 71.2 km, with a diameter of nearly 220 km; the second one is located near the northwest sub-basin, where the depth ranges from 106.8 km to 128.8 km, which is probably part of the Paleo-Pacific remnant in the SCS. (2) The proto South China Sea subduction zone is located in the northwestern part of the Sulu Sea and has a northeast–southwest trend.

In existing regional geomagnetic field modeling, the smoothness of basic functions and the insufficient data constraints in marginal regions lead to the omission of detail features and extrapolation oscillations. To address these limitations and develop a high-precision marine regional geomagnetic field model, we develop a back propagation neural network (BPNN) method enhanced by particle swarm optimization (PSO). The PSO-BPNN method has the ability of adaptive learning and could extract local features. By combining the magnetic field data measured by ships with the previous model data, a high-precision geomagnetic field model of the northern South China Sea (SCS) is developed. The fitting error of the PSO-BPNN model is 18.05 nT, which is 16% and 20.1% lower than those of the traditional Legendre Polynomial (LP) and Taylor Polynomial (TP) models, respectively. The proposed PSO-BPNN model demonstrates superior robustness and higher accuracy, while retaining more magnetic signals of small geological bodies.

The Tibetan Plateau is a region where the Indian and Asian-European plates collide. A high-precision Moho topography model is important for the study of plate motion and internal tectonics of the Tibetan Plateau. However, previous gravity-based models failed in representing adequately the Moho topography because crustal effects and biases in the inversion parameters were not fully considered. To address these issues, we extracted the gravity effect caused by the crust using wavelet multi-scale analysis, and used an adaptive linear inversion method with available gravity and seismic data to estimate more accurate inversion parameters. With these two improvements, we inverted for a high-precision Moho topography model of the Tibetan Plateau. The results show that the majority of the Tibetan Plateau has a Moho depth of 60–70 km. The Moho depth in the center of the Qaidam Basin is about 50 km, and the Moho topography between the east and west of the basin has obvious undulations. The Moho depth of the Tarim Basin ranges from 38–50 km. While the undulation of Moho is limited in the center of the basin, obvious uplifts are shown in northern and southern basin. The plate driving force between the Tibetan Plateau and the Tarim Basin may be the primary cause of this phenomenon. By comparing the difference with the seismic data, we found that the RMS of our model is 2.8 km smaller than that of the CRUST 1.0 model, which shows that our model is more accurate.