Application of Satellite Gravimetry and AI to Map the Density Distributions of Mars’s Upper Layers

Satellite gravimetry data from Mars offers a unique glimpse into the planet's interior structure. Combined with topography data of the planet's surface, measurements of the gravitational field can be used to probe the lateral density variations in the planet's upper layers. Due to inherent degeneracies between the effects of density anomalies in the mantle and the crust on gravity, and incomplete isostasy models, previous efforts for global gravity inversion to decouple the two planetary layers were unsuccessful. This study aims to aid these inversion efforts by providing constraining information about the scale and magnitude of the lateral density fluctuations. In this simulation-based approach, a two-layer planetary model is applied and the Matérn covariance function is used to simulate physically viable density distributions. The simulations are used as an input to an inference method applying Normalising Flow neural networks to infer which Matérn parameters closest align with real observations. The results can provide constraints for future inversion attempts and inform us about the sensitivity of gravimetry data to the subsurface densities.