Using a Data-driven Bayesian Inversion of Fluvial Topography and Marine Terraces to Simultaneously Constrain Surface Process and Tectonic Models

Surface process models offer a mechanistic framework to predict the landscape response to external forcing, which some argue makes them useful for determining tectonic signals from topography. However, geologic and geomorphic data must be used to calibrate surface process models before meaningful information on tectonics can be gleaned. Here we present an approach to constrain a surface process model and a tectonic model simultaneously through joint inversion of longitudinal river profiles and dated marine terraces using a Bayesian Markov chain Monte Carlo (MCMC). We apply this technique to the well-studied footwall of the Corinth Rift, Greece, where a wealth of available data allows for independent evaluation of the inversion results. Most rivers channels in the rift footwall are bedrock, so we use detachment-limited stream power incision model to simulate river profile evolution. Consistent with previous studies, we model vertical deformation as flexure of a broken elastic plate along the bounding normal fault. Using uniform prior probability distributions for model parameters, we apply the MCMC to define the posterior probability distributions for the stream power parameters, lithospheric rigidity, timing of fault initiation, and initial and final rock uplift rates. The recovered stream power parameters are reasonable, and the inferred tectonic outputs are consistent with independent datasets at the rift margin scale, giving confidence in the inversion results. The stream power slope exponent, n, is > 1, which likely reflects channel narrowing and elevated sediment flux and caliber in steepened river reaches and erosion thresholds. The plate rigidity expressed as effective elastic thickness is ~3 km. Results indicate an initial block uplift rate of 0.1 - 0.2 mm yr^-1, and fault initiation at ~700 - 800 kyrs with a final footwall uplift rate of ~1.5 mm yr^-1 along the bounding offshore fault trace that decays exponentially away from the fault. For typical normal fault uplift-to-subsidence ratios and a dip of 60 degrees, these results suggest a time-averaged fault slip rate of 6 - 8.5 mm yr^-1. These findings demonstrate that when assumptions are approximately correct, the joint inversion of fluvial topography and field data can provide meaningful insight into both earth surface and tectonic processes.