Enforcing multiple constraints on the interior structure of Ganymede: A machine learning approach

Determining the internal composition of planetary bodies remains a challenging problem due to observational degeneracies. In the 2030s, ESA’s Juice mission will orbit Ganymede and provide key constraints on its interior structure, including estimates of the polar moment of inertia, Love numbers, libration amplitudes, and the amplitude and phase of the induced magnetic field due to a subsurface ocean. To impose these constraints in the most effective way, a joint inversion of all available parameters would be ideal. In this work, we applied a machine learning approach to predict the thicknesses and densities of Ganymede’s internal layers, the viscosity of the icy shell, and the ocean conductivity from these observables. We generated a synthetic dataset of plausible internal structures via Monte Carlo sampling and computed the corresponding observables using existing physical models. A neural network was then trained to learn the intricate relationships between them. The trained model retrieves the internal structure parameters with varying degrees of accuracy across different layers. It performs well in predicting the thicknesses and densities of the icy shell and ocean, with mean absolute errors on the order of 10 km and 10 kg m −3 , respectively. These errors increase to about 40 km and 20 kg m −3 for the high-pressure ice layer beneath the ocean. The trained model also estimates the shell viscosity with a mean absolute error of 0.05 log 10 Pa s, and the ocean conductivity with an error of 0.1 S m −1 . However, the neural network performs poorly in the task of inferring the thickness and density of the deeper interior, suggesting limited sensitivity of these parameters to the chosen set of observables. The Monte Carlo dropout method was utilized to estimate the uncertainties in the predicted parameters. These results highlight the potential of machine learning as a fast, preliminary tool for detecting families of internal structures compatible with the observed parameters.