Uncertainty quantification of multi-scale and multi-physics computer models

Developing new tools to investigate and quantify uncertainties in computer models, with applications to climate, earthquake and tsunami models


What possible tsunamis will hit the west coast of India? What are the uncertainties in future predictions of global warming over Europe? 'Uncertainty Quantification' (UQ) can help answer these questions by analysing the propagation of uncertainties in complex simulators, such as climate or tsunami computer models that run on supercomputers and require multiple scales and physical processes to be combined.

Explaining the science

Typically UQ makes use of surrogate models that are much faster to run than simulators, in order to sample uncertainties efficiently. These are often 'Gaussian processes' that need to be fitted using a smart design of computer experiments. Building a UQ workflow that integrates heterogeneous models (both in scale and in nature) is a challenge, and the corresponding designs also have to be investigated.

Another scientific endeavour is identifying the distributions of parameters that will allow the computer simulations to match best observations, using Bayesian calibration based upon surrogate modelling. New methods, as well as hardware and software solutions will enable scientific progress in these areas.

Project aims

This project will allow a joint multidisciplinary effort on four research strands:

  1. Computational workflow integrating highly heterogeneous high-performance computing and novel statistical design of computer experiments. At a high level, the Research Engineering team at Turing will create a UQ workflow that integrates heterogeneous models (both in scale and in nature).
  2. Parallelisation and coupling of multi-scale and multi-physics models on GPUs
  3. Scaling UQ on next generation accelerator architectures (FPGAs)
  4. Applications to climate and risk assessments - Climate model (calibration, future uncertainties) and tsunami, storm surges and earthquake risk


UQ for specific combinations of computer models (e.g. earthquakes and tsunami models) will allow the project team to carry out novel risk assessments in shipping ports where currents may damage vessels and facilities. This will help build resilient and robust infrastructure and protect populations.

Other applications include climate predictions with more precise uncertainty estimates than currently. A wider group of applications to other science and engineering problems will be reached through the new UQ workflow platform developed by the Turing Research Engineering team, both within Data-centric Engineering (e.g. energy models) and beyond. Another application would be to early warning systems where uncertainties are currently lacking.

Recent updates

September 2020

The ExCALIBUR workshop on 'Data Assimilation and Uncertainty Quantification at the exascale'

July 2020

Our new paper published on numerical dispersion and dissipation in tsunami modelling on HPC

April 2020

Serge Guillas becomes Met Office Joint Chair in Data Sciences for Weather and Climate

Preprint on emulating systems of computer models (multi-physics or even multi-disciplinary).

August 2019

Workshop on 'Effective and efficient Gaussian processes'.

April 2019

Paper published on landslide-generated tsunami risk assessment for the Indian Ocean.

January 2019

Project kick-off meeting with co-investigators, the Research Engineering team, and researchers involved in the project.

Members of the project team



Researchers and collaborators

Contact info

[email protected]

PhD students

Deyu Ming, UCL
Mariya Mamajiwala, UCL
Ayao Ehara, UCL
Theodoros Mathikolonis, UCL
Ryuichi Kanai, UCL