The Study of Depth-Dependent Centroid Frequency Shifts and an Improved Robust Method for Loss Tangent Inversion With Chang’e 4 Ground Penetrating Radar Data

Ground-penetrating radar (GPR) enables subsurface structure detection on the Moon and Mars, where the centroid frequency of radar signals is employed to invert the loss tangent of media and characterize underground structures. Observations from the Chang’E-4 (CE-4) and Tianwen-1 (TW-1) GPR systems reveal a decaying trend in both central and centroid frequencies with increasing propagation depth. Ultrawideband (UWB) attenuation model and an analysis of the correlation coefficient reveal that the reduction in centroid frequency with depth in the shallow subsurface arises from UWB attenuation, whereas its apparent increase beyond a critical depth is attributed to noise dominance. Furthermore, theoretical analysis reveals that existing centroid frequency shift methods for loss tangent inversion of the CE-4 data neglect scatterer distribution and noise, which may cause errors. Moreover, previous work used the Gaussian wave to approximate the Ricker wavelet, assuming a linear depth–centroid frequency attenuation model for inversion, which deviates from the Ricker wavelet model and caused an overestimation in the loss tangent. This study mitigates inversion errors from scatterer distribution and noise via local region selection. We analyze the spectrum difference between Gaussian and Ricker wavelets and achieve automatic region selection through wavelet transform (WT). In addition, a Ricker numerical inversion framework is proposed to eliminate approximation-induced errors. This work provides critical insights for advancing the CE-4 loss tangent inversion and subsurface interpretation in planetary radar sounding.