Characterizing Martian Lithospheric Density Distribution Through Bayesian Gravity Analysis

Abstract

One of the best sources of information for probing the inner structure of Mars is gravimetry data from satellites. By combining these measurements with topography data and isostatic theory, the aim of this study is to infer information about lateral density variations in the Martian lithosphere. Previous efforts at direct gravity inversion had limited success due to inherent degeneracies between the gravitational effects of different planetary layers. This study takes a different approach, where the aim is not to obtain spatial density maps, but rather to gain information about the spatial distribution of density values. The density distributions are simulated via multivariate Gaussian sampling using a Matérn covariance function. Constraints on the Matérn parameters for Mars are obtained through Bayesian inference, which can inform about the sensitivity of gravity data to different aspects of the density distribution and can be translated into physically meaningful insights about the lithospheric structure. The inference is performed with a machine-learning technique, namely Normalizing Flows. The results of this study are consistent with previous findings and indicate that the upper mantle exhibits significantly smoother density variations than the crust, with both layers containing large, connected regions with similar densities.