Estimation of Denudation Parameters and River Capture Events From Neural Network Inverse Modeling of River Profiles and Thermo- and Geochronology Data

Abstract

Earth's topography represents the cumulative effects of tectonics and surface processes modulated by climate and lithology. These factors shape landscapes through time. River profiles can be inverted to estimate the rock uplift histories or lithology-specific erodibilities. However, river systems are dynamic and evolve in response to spatial and temporal internal dynamics, such as river capture events. Here, we present a modeling framework to infer denudation rates from the inversion of river profiles and thermo- and geochronology data. We achieve this by coupling a landscape evolution model and an efficient inverse modeling scheme to infer poorly resolved erosional and tectonic parameters. An application of the approach is presented for the Neckar catchment, southwest Germany, characterized by stark lateral variation in bedrock erodibility and rock uplift, and that have demonstrably undergone multiple river capture events. Different end-member scenarios are explored in the simulations. First, we test uniform and spatial variability in rock uplift rate and bedrock erodibility, and second, temporal variations in rock uplift rate and base level. Finally, we simulate river capture events by adding upstream sections (drainage area) at specific times and locations within the fluvial network. We find that spatial variation in rock uplift rate is necessary to reproduce the Neckar's river profile while honoring analytical observations. Simulations integrating river captures allow improved river profile predictions of specific tributaries of the Neckar catchment, leading to potentially more realistic erodibility and rock uplift history estimates. The time and location of the capture events determined from the modeling agree with previous estimations from geological evidence.