Pattern scaling - An examination of the accuracy of the technique for describing future climates

A fully probabilistic, or risk, assessment of future regional climate change and its impacts involves more scenarios of radiative forcing than can be simulated by a general (GCM) or regional (RCM) circulation model. Additional scenarios may be created by scaling a spatial response pattern from a GCM by a global warming projection from a simple climate model. I examine this technique, known as pattern scaling, using a particular GCM (HadCM2). The critical assumption is that there is a linear relationship between the scaler ( annual global-mean temperature) and the response pattern. Previous studies have found this assumption to be broadly valid for annual temperature; I extend this conclusion to precipitation and seasonal (JJA) climate. However, slight non-linearities arise from the dependence of the climatic response on the rate, not just the amount, of change in the scaler. These non-linearities introduce some significant errors into the estimates made by pattern scaling, but nonetheless the estimates accurately represent the modelled changes. A response pattern may be made more robust by lengthening the period from which it is obtained, by anomalising it relative to the control simulation, and by using least squares regression to obtain it. The errors arising from pattern scaling may be minimised by interpolating from a stronger to a weaker forcing scenario.