Towards integrating tomographic resolution and uncertainty information into geodynamic mantle flow reconstructions

Abstract For reconstructing mantle flow back in geologic time, geodynamic inversions require input from seismology in the form of tomographic images. However, a practical representation of image uncertainty is needed for robust inferences. Addressing the scale discrepancy between fluid dynamic predictions and seismically visible heterogeneity is crucial, since the subsequent validation of mantle flow trajectories involves surface dynamic topography predictions that are highly sensitive to the tomographic input. Here, we conduct a synthetic experiment to illustrate the challenges in quantitatively integrating tomographic and geodynamic models, using a linear tomographic framework, the subtractive optimally localized averages (SOLA) method and a mantle circulation model (MCM) as reference. We propose a possible workflow for adjoint flow reconstructions that aims to leverage the capabilities of the SOLA method. This includes the construction of tomographic averaging kernels that can be spatially optimized and which define local resolution, together with model averages and their uncertainties. For the geodynamic adjoint framework, we suggest incorporating the SOLA estimates in the cost function and testing the implementation of specific tomographic realizations in closed-loop experiments. We stress that only after accounting for the effects of resolution, an ensemble approach for uncertainty quantification can provide meaningful constraints on the flow history.