Developing computing architectures for improving the simulation of fog


   Department of Computer Science

   Applications accepted all year round  Self-Funded PhD Students Only

About the Project

To numerically simulate the formation of fog, a horizontal spatial resolution approaching 1 metre or less is highly desirable to improve scientific understanding about fog formation to the level that could be applied to improve simulations on the kilometre scale, where the inability to accurately simulate fog remains a costly outstanding problem. The Met Office NERC convection (MONC) model was developed to simulate clouds and turbulent atmospheric motion on a very fine scale, while harnessing improvements in model parallelization on high-performance computers to make the simulation process more efficient. Still, any metre-scale simulations on domains of sufficient size to generate informative statistics presents a significant challenge in terms of both computational cost as well as data storage volumes. We aim to develop methods to accelerate MONC and its components to produce simulations much more efficiently. Preliminary work, conducted as part of The European Centre of Excellence for Engineering Applications (EXCELLERAT), involving researchers at the University of Edinburgh's Centre of Excellence, EPCC, involves redesigning MONC for GPU acceleration, the Knights Landing many-core processors, and customisable field-programmable gate array (FPGA). The most promising and practically applicable of these will be developed further to attain greater efficiency through application to the models various components (microphysics, radiation, etc.), as doing so can greatly increase the productivity of large-eddy models like MONC. With this in place, we can use what we learn from simulations enabled by these developments to improve fog forecasts at the kilometre scale in, for instance, the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS). Gains at this scale could be realised through implementation of machine learning techniques, improved representations of boundary layer processes, or enhanced parameterizations.  

Aspects of the project are likely to additionally include exploration of cloud-based computing and stroage solutions, collaborations with researchers from multiple domains and institutions, and the development of novel algorithms that enable efficient fog simulation at high resolution.

Eligibility requirement: MSc in Computer Science, Meteorology, or closely related field.

Further Enquiries contact: Dr Todd Jones ()


Computer Science (8) Environmental Sciences (13)

Register your interest for this project



Where will I study?