EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Estimation of erosion rate parameters from neural network inverse modeling of river profile and thermo-geochronology data

Thomas Bernard, Christoph Glotzbach, Daniel Peifer, Al Neely, and Todd Ehlers
Thomas Bernard et al.
  • University of Tübingen, GUZ, Department of Geosciences, Tübingen, Germany (

The Earth's surface topography reflects the long-term competition between tectonic and climate-driven surface processes. River erosion is a fundamental process that sets the base level for hillslope processes and drives landscape evolution. River profiles reflect external processes, such as tectonic uplift and climate, as well as intrinsic properties of the landscape, such as lithologic variations. River profiles respond to perturbations in these parameters through local changes in channel gradient, which are transmitted upstream of the river channel. River networks affected by these processes may eventually suffer drastic river captures and important drainage reorganization. As a result, river profiles can be used to extract the uplift histories of landscapes. Geochemical data with sensitivities to different time scales, such as thermochronological ages and cosmogenic nuclide concentrations, can be combined in numerical models with river profile analyses to identify governing relationships response for a landscape history. However, the estimation of a complete denudation record through time remains challenging, especially in landscapes where river capture and drainage reorganization have strongly perturbated the river system.

            In this study, we perform inverse modeling of river profiles and thermo- and geochronology data (i.e., low-temperature thermochronology and cosmogenic nuclides) to infer erosive parameters and the topographic history of different settings. The numerical model allows the prediction of river profiles, thermochronological ages (e.g., zircon fission tracks, apatite fission tracks and apatite helium ages), cosmogenic nuclide concentrations, and simplistic river captures. Variability in both rock uplift history and erodibility of different lithologies are accounted for. The model algorithm utilizes an efficient inverse modeling scheme "Simulation-Based Inference" to resolve unknown parameters such as uplift or erodibility of the different lithology. Results are presented from the Neckar catchment located in southwest Germany, which shows evidence for major river captures and drainage reorganization over the last ~10 Ma. Model results allow to reproduce the river profile and thermo-geochronological data of the Neckar catchment for specific uplift and erodibility. Moreover, early experiments indicate a better prediction of the observed data, and therefore, the parameters controlling the erosion rate, when considering river captures.

How to cite: Bernard, T., Glotzbach, C., Peifer, D., Neely, A., and Ehlers, T.: Estimation of erosion rate parameters from neural network inverse modeling of river profile and thermo-geochronology data, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-4903,, 2023.