Gravity inversion of sub-ice shelf bathymetry in West Antarctica using a geostatistical Markov Chain Monte Carlo approach

Sub-ice-shelf bathymetry controls the delivery of warm water to the ice-shelf bottom in West Antarctica, making the bathymetry beneath ice shelves in the Amundsen Sea critical inputs to ice-sheet and ocean models. Previous estimates of the bathymetry have often used deterministic inversion frameworks or do not account for the non-uniqueness of the inverse problem, and ultimately lack robust uncertainty quantification. To provide more robust and reproducible bathymetry models, we implement a random walk Metropolis-Hastings Markov Chain Monte Carlo (MCMC) inversion approach, which iteratively generates model perturbations using random Gaussian fields and forward models the gravity disturbance of proposed bathymetry models. After convergence, our approach samples the posterior distribution allowing for estimation of the mean and variance of the bathymetry while providing realistic models of the sub-ice-shelf bathymetry. An ensemble of bathymetry models can then be used in ice-sheet and ocean simulations to propagate the uncertainty in bathymetry to dynamic ice processes, resulting in better uncertainty quantification of future sea-level rise. In addition to providing more robust bathymetry models, this work provides a step forward in the reproducibility of geophysical inversions by leveraging the growing open-access geoscientific computing ecosystem of Python.