Department Systems Analysis, Integrated Assessment and Modelling

SPUX - Scalable high performance uncertainty quantification for stochastic models in environmental data sciences


The objective of this project is to advance high performance scientific computing in water research by developing a parallel uncertainty quantification framework for complex environmental dynamical systems with intrinsic epistemic uncertainties, often simulated using stochastic models, such as individual based models and stochastic differential equations. We aim to develop parallel supercomputer enabled Bayesian inference framework using Particle Markov Chain Monte Carlo, capable of efficiently estimating complex marginal parameter likelihoods for non-linear stochastic models and sampling the resulting parameter posterior distribution. The newly developed prototype framework SPUX will be extended using adaptive computational load balancing and hybrid parallelization techniques, modern ensemble and multi-level Markov Chain Monte Carlo samplers, and ported to energy efficient multi-core accelerators. SPUX will be applied within scientific inter-departmental collaborations in ecology, hydrological catchment modeling, subsurface ground water flows and urban flood risk assessment. SPUX can be found at and

Team Member

Dr. Marco Bacci

Dr. Jonas Sukys