2023 CSDMS meeting-097
Power, noise and scaling of Earth's surface: Disentangling causes of landscape evolution
Gareth Roberts, Imperial College London London , United Kingdom. email@example.com
Understanding how landscapes acquire their form is complicated by evolution across a large range of spatial and temporal scales. Disentangling causes of landscape evolution should carefully consider scale and scalings, but how best to do so? I summarise the work we have been doing to make use of observations, spectral analyses and forward and inverse modelling to address the following questions. Where and at what scales do fluvial landscapes acquire their physical and chemical properties? How can we use the geometries of landscapes and their compositions to recover information about driving and responsive processes? Demonstrations of how observations of landscape form and chemical concentrations can be combined with simple theory to identify where and how landscapes acquire their geometries and material provenance are given. Spectral analyses of landscape geometries are used to identify scales at which they acquire their form and scaling regimes. Consequently, it is possible to assess whether it is reasonable to ‘stitch together’ observations or theory used to understand geomorphic processes operating at small scales to determine how landscapes acquire their form. In short, that proposition is highly unlikely to be successful because of the existence of erosional ‘shockwaves’ and stochasticity. However, despite local (spatio-temporal) complexity, fluvial landscapes appear to possess emergent, deterministic, simplicity at scales > 100 km, such that processes operating at these scales (e.g. dynamic topography) are manifest in drainage networks with simple, self-similar, scalings (e.g. Brownian noise). The use of upscaling of simple physical models to generate appropriate scaling regimes and statistical insights into how fluvial landscapes acquire their form is explored. Finally, a demonstration of how statistical measures based on Optimal Transport theory can be used to identify optimal landscape evolution models is presented. Wasserstein distances are shown to have significant benefits over more widely used Euclidean measures of misfit, especially when local noise is prevalent.