Multi-core parallel tempering Bayeslands for basin and landscape evolution

In recent years, Bayesian inference has become a popular methodology for the estimation and uncertainty quantification of parameters in geological and geophysical forward models. Badlands is a basin and landscape evolution forward model for simulating topography evolution at a large range of spatial and time scales. Previously, Bayesian inference has been used for parameter estimation and uncertainty quantification in Badlands, an extension known as Bayeslands. It was demonstrated that the posterior surface of these parameters could exhibit highly irregular features such as multi-modality and discontinuities making standard Markov Chain Monte Carlo (MCMC) sampling difficult. Parallel tempering (PT) is an advanced MCMC method suited for irregular and multi-modal distributions. Moreover, PT is more suitable for multi-core implementations that can speed up computationally expensive geophysical models. In this paper, we present a multi-core PT algorithm implemented in a high performance computing architecture for enhancing Bayeslands. The results show that PT in Bayeslands not only reduces the computation time over a multi-core architecture, but also provides a means to improve the sampling process in a multi-modal landscape. This motivates its usage in larger-scale problems in basin and landscape evolution models.