Publication Name: Australasian Exploration Geoscience Conference 2019
Authors: Rohitash Chandra, Danial Azam*, R. Dietmar Müller
Date Published: September 2019
Number of Pages: 5
Abstract:
Badlands is a basin and landscape evolution forward model for simulating the evolution of surface topography, sediment transport and sedimentation at a large range of spatial and time scales. Here we use the Bayesian paradigm to find the best-fit parameters driving basin evolution models using Badlands. Inference in a Bayesian framework is obtained via the modelled distribution of the unknown parameters. We implement parallel tempering Markov chain Monte Carlo (PTMCMC) using high-performance computing to accelerate parameter space exploration of the computationally expensive Badlands model. Our results show that traditional implementations of single chain MCMCs rarely converge and lead to misleading inference. In contrast, PT-MCMC not only reduces the computation time, but also provides a means to improve the sampling for multi-modal posterior distributions. This motivates its usage in regional basin and landscape evolution models, allowing us to determine the relative importance of different parameters driving basin stratigraphic evolution. Parameters that can be explored include time-dependent tectonic and dynamic topography, precipitation, rock erodibility, flexural rigidity of the lithosphere and relative sea level fluctuations.