Research Fellows Directory
Professor Tim Palmer FRS
University of Oxford
Unlike other areas of science, we cannot test “What If…” ideas about Earth’s climate by doing laboratory experiments. Hence, to answer questions like: “What if we continue to emit greenhouse gases into the atmosphere”, “What if we spray sulphate aerosols in the stratosphere to try to offset global warming”, we must resort to solving the basic mathematical equations of climate. However, climate is a complex system with chaotic quasi-turbulent components (the atmosphere and oceans), and solving these equations accurately requires use of the most powerful computers available. Crucially, any predictions we make using computational representations of the underlying equations of climate have to be qualified with estimates of uncertainty. This is not straightforward. My research at Oxford University is based around the concept that the representation of unresolved processes in comprehensive climate models should use stochastic rather than deterministic closure schemes. (Traditional climate models represent unresolved processes using deterministic formulae that mimic molecular diffusion and viscosity. A stochastic closure scheme is one that includes some random-number processes.) We have made substantial progress showing not only that climate models with stochastic closures provide not only more reliable predictions, but also reduced systematic errors – the latter being related to the mathematical notion of noise-induced rectification in nonlinear dynamics. Leading on from this, we have been making some excellent progress on what could be a major new initiative in climate modelling. If the sub-grid closure schemes should be formulated stochastically, then it may be possible to increase the accuracy of our predictions by enhancing the resolution of the models but where the smaller-scale variables are represented using reduced precision and with reduced levels of bit reproducibility. Such models will use computer resources much more (energy) efficiently.
Interests and expertise (Subject groups)