Comparative Analysis of Methods for Recovering Gravity Anomalies by Combining Satellite Gravity and Topographic Potential Models

Objectives: Currently, extensive terrestrial areas remain either unmeasured or sparsely covered by gravity data, and a commonly used approach is to use topographic data to densify and fill in the gravity field. Considering the complementary spectral advantages of satellite gravity field models and topographic potential models, we compare and analyze three methods for recovering gravity anomalies by combining satellite gravity field model and topographic potential model.Methods: Four satellite gravity field models are combined with the dV_ELL_Earth2014_5480 topographic potential model to conduct comparative experiments in five regions of the United States with different terrain types. We analyze and compare three methods for recovering regional gravity anomalies: the potential coefficient direct splicing (DS) method, the potential coefficient weighted combination method, and the normal equation (NEQ) combination method, along with the parameter settings for each approach.Results: The optimal degree for the DS method varies in different regions, with higher optimal degrees in mountainous areas. The optimal transition zone for the weighted combination method also differs, with the models performing best when the transition zone is between degrees 230–300. Among the weight functions of the weighted combination method, the Hanning function demonstrates superior performance. Compared to the DS method, the NEQ method provides smoother spectral transitions and better accuracy. The standard deviations of gravity anomaly differences for the three methods are 8–9 mGal in plains and 15 mGal in mountainous areas. Discrepancies in north-central United States highlight the limitations of current models in detecting certain gravity signals.Conclusions: Overall, the NEQ method and the potential coefficient weighted combination method each have their strengths and weaknesses. The NEQ method is more complex as it requires combination of normal equations, but it is applicable to any region globally. The spectral domain weighted method is simpler but relies on selecting optimal weighting strategies for different regions, making it more empirical.