I propose to join the Department of Statistics and the Centre for the Analysis of Time Series (CATS) at the London School of Economics and Political Science (LSE) for one year in order to contribute to the assessment of the quality of climate predictions. The objective of the project is to develop and to apply a statistical methodology for quantifying the uncertainty of climate model outputs in respect to direct observations. Although this kind of quality check is of fundamental importance for the justification of the use and faith in climate models and its predictions, it has not yet been done and the analysis of uncertainty in climate forecasts remains rudimentary. Just as uncertainty in the initial conditions limits the utility of a single forecast, so model error due to imperfectness in model formulations severely bounds attempts to obtain the 'true' forecast probability density distribution. It is thus suggested to use multi-model ensembles to define a region of model-state space which will contain the verification (i.e. observations) at a certain level of confidence. The volume describing the certainty range is called a bounding box. An expected result of the proposal is a better understanding of model's uncertainties as an important aspect of successful model evaluation. Hence this project will contribute to the huge potential Casino-21 has to raise awareness of the issue of uncertainty in climate predictions in society.