Advancing nonlinear density interface inversion from gravity, gravity-gradient, and gravity-curvature data: Application to Moho recovery

Abstract

Density interfaces are discontinuous boundaries characterized by significant density contrasts within Earth's interior, such as the seafloor, sedimentary basement, and the Moho. Gravitational observations can be used to infer the geometry of these undulating interfaces. Current and future satellite missions provide global high-quality gravity, gravity-gradient, and gravity-curvature data, enabling local- to global-scale studies and mitigating the limitations of terrestrial measurements. Density interface inversion is intrinsically non-linear, and conventional linearization inevitably introduces bias, particularly in regions with pronounced interface relief. To address this limitation, non-linear forward and inverse models under a spherical approximation are considered, and a higher?order regularization framework is developed to approximate the nonlinear solution. Synthetic tests based on theoretical models, together with applications to Moho recovery beneath the Tibetan Plateau, demonstrate superior performance relative to conventional Tikhonov regularization, indicating that higher-order contributions are non-negligible when the interface undulations are moderate. The application also reveals limitations of gravity-curvature Moho inversion, as the refined gravity-curvature disturbances are strongly influenced by short-wavelength signals associated with imperfect stripping reductions, resulting in weak sensitivity to Moho geometry. Overall, the proposed higher-order regularization provides a robust and effective framework for nonlinear density interface inversion and offers improved algorithmic support for further applications.