Projections of future climate changes are critical for adaptation planning and for informing negotiations on mitigation. Studies of global mean temperature are common and provide a wealth of information for quantitative assessments of global change e.g. as in IPCC AR5. However, such assessments provide only low-confidence statements about spatial patterns and regional climate change. Despite ever-more sophisticated climate models, current practice often infers future climate by simply taking the mean and spread of available model simulations. This ‘model-world’ approach takes no account of model biases or our understanding of the robustness of the physical processes responsible for change. This presentation will report on work undertaken in an on-going project to improve special projections of climate change. Three topics will be discussed; the ratio between land and sea warming, tropical rainfall shifts and changes in extratropical storms.