Climate research and weather forecasting both rely heavily on simulations using numerical models. While increasing computational power has greatly improved such simulations, further optimization remains an important goal. Developing and using new computational methodologies to increase the accuracy and fidelity of climate and weather simulations is an important challenge for applied mathematics and computational science, in collaboration with climate, atmosphere and ocean science.
Our group is active in two areas. One is the development of particle filters for data assimilation, a methodology for combining numerical models and data in a suitable way. The other area is that of parameterization (or subgrid-scale modelling), where we develop both stochastic methods and computational multiscale methods to represent the influence of (for example) clouds and convection processes on the atmosphere and climate system.
The aim is always to develop new numerical algorithms and computational methods that can be used by climate and weather modellers to increase the computational speed and accuracy of their simulations.