Neural Posterior Estimation of Terrain Parameters from Radar Sounder Data

2026-05-05

AI Summary

Traditional radar sounding inversion methods rely on simplified assumptions and yield only point estimates that neglect parameter correlations and noise characteristics. This work introduces neural posterior estimation (NPE) to this domain for the first time, leveraging GPU-accelerated electromagnetic simulations to generate synthetic training data and train a conditional neural density estimator capable of inferring the full posterior distribution over subsurface terrain parameters. The approach enables rigorous uncertainty quantification and explicit modeling of parameter interdependencies, while systematically assessing the impact of reference surface assumptions through explicit conditioning. The model demonstrates well-calibrated performance on simulated data and successfully generalizes to real Martian radar profiles, achieving probabilistic inversion grounded in literature-derived priors.

Abstract

Radar sounders are electromagnetic instruments that can probe deep into the subsurface of Earth and other planetary bodies by processing the echo of transmitted radar waves. Conventional approaches for analyzing such data rely on approximate assumptions and often produce point estimates that ignore parameter correlations as well as galactic and measurement noise. We propose a simulation-based inference approach to terrain parameter inversion from radar sounder data, where synthetic observations from a GPU-based simulator are used to train a neural network-based density estimator for neural posterior estimation (NPE). By explicitly conditioning on reference surface assumptions, the proposed framework allows systematic evaluation of posterior robustness to reference surface variability. We demonstrate that our NPE model is well calibrated on simulated data and transferable to real Mars radar profiles, where we analyze terrain parameters using literature-informed reference values.

Problem

• radar sounder
• terrain parameters
• parameter estimation
• uncertainty quantification
• reference surface variability

Innovation

• neural posterior estimation
• simulation-based inference
• radar sounder
• terrain parameter inversion
• GPU-based simulator