Abstract

In this talk I will begin by discussing the sources of uncertainty in simulations made from computer models and using statistics to help understand the effects of uncertainty on computer model predictions. I will then present the work we have been doing in Leeds applying statistical techniques to runs of a global aerosol model in order to understand and quantify the effect of parametric uncertainty on model predictions of aerosol particles important for cloud formation and model predictions of the aerosol indirect radiative forcing. In particular, I will talk about the use of statistical emulation and sensitivity analysis. I will show global maps and time series of the regional and seasonal sources of uncertainty in global aerosol modelling and their global significance. The results show where more effort is needed to reduce uncertainty in global aerosol modelling and its effect on the climate as well as helping us to understand the model itself. The work has also helped us to identify irreducible model errors which can now be quantified. I will also introduce the next stage of the work in which we will attempt model calibration taking into account the model parametric uncertainty.