Faster Emergency Response:

NAME model development and parallelisation

[with Matthew Hort, David Thomson (Met Office), Graham Riley, Rupert Ford (Manchester)]

As a scientist in the Met Office Atmospheric Dispersion group I contributed to the development of the NAME dispersion model, which is used for research applications and operational prediction of atmospheric pollutants such as volcanic ash. In particular I was responsible for the OpenMP parallelisation of the Model and my work lead to a significant reduction of the model runtime [Müller et al. (2013a)]. The figure on the right is taken from this publication and demonstrates the excellent strong scaling of an operational volcanic ash run on up to 12 cores.

I also contributed significantly to the development of a new emergency response system for volcanic ash events. In particular I developed a novel Java webinterface which allowed the easy and reliable use of the system by forecasters.

Strong scaling of NAME operational volcanic ash run

Multilevel methods in atmospheric dispersion modelling

[with Sarah Cook, Rob Scheichl, Tony Shardlow (Bath), Ben Devenish, David Thomson (Met Office)]

Together with Prof. Rob Scheichl I co-supervised the MPhil project of Sarah Cook on the application of multilevel Monte-Carlo methods in atmospheric dispersion modelling (Sep 2012 - Sep 2013). Based on the ideas in [M. Giles, Operations Research, 56(3):607-617, 2008], multilevel methods can lead to a significant reduction of computational cost when solving the stochastic differential equations that are currently used by the NAME model to predict the atmospheric transport of pollutation. This work is carried out in collaboration with the Atmospheric Dispersion Group at the Met Office. I am also investigating the performance of different numerical time stepping methods for SDEs of the Langevin-type together with Dr. Tony Shardlow.