Tectonics from topography: Embracing noise and uncertainty in inverse modelling of landscapes

Disentangling the roles of climatic or lithologic processes from tectonic ones in shaping landscapes remains an important goal within the geosciences. However, this endeavour is complicated by the existence of ‘geomorphic noise’---the shaping of topography by surface processes operating on shorter timescales than large scale tectonic uplift.  We present a probabilistic inverse modelling framework that can recover histories of uplift from noisy topography. It is shown that noise added to landscape simulations generates variability in their resultant geomorphic properties, but that ensemble-based approaches to landscape evolution modelling are well-suited to quantifying this uncertainty. Unlike Euclidean approaches (e.g., root mean square), optimal transport-based techniques for comparing observed and theoretical topographies are able to ‘see through’ local complexity. Crucially, this precludes the need for precise knowledge about initial conditions, reproducing facsimiles of observed topography, and assumptions of topographic steady state, all of which are likely to be unrealistic expectations in recovering tectonics from topography. We show that recovering spatially- and temporally variable uplift histories with this modelling framework is applicable to large portions of Earth’s surface and may have valuable implications for identifying mineral resources.