A meteorology approach to assess mantle flow induced dynamic topography using object-based image processing methods

Abstract

The construction and assessment of model trajectories that link multiple mantle states is essential to constrain poorly known mantle convection parameters. Previously, volumetric approaches have been applied to assess the quality of constructed mantle flow trajectories. However, there is a need to assess these trajectories based on their dynamic topography predictions because mantle convection cannot be directly observed and may be inferred via its surface geological expressions. Typical metrics for assessing dynamic topography suffer from the double penalty problem — a prediction that is correct in intensity, size, and timing, but incorrect in location, results in large root-mean-square errors when compared to an observation. Here, we introduce metrics, gleaned from meteorology, that decompose any number of dynamic topography fields into their distinct objects after which the similarity between objects is compared. We find that this object-based approach overcomes double penalty and assesses models in a robust manner by providing the ability to assess separately the quality of match between subsidence and uplift areas. Additionally, the approach allows independent quality assessment of multiple aspects of a dynamic topography field, including amplitude and location of dynamic topography.