DOWNSCALING OF GLOBAL CLIMATE-CHANGE ESTIMATES TO REGIONAL SCALES - AN APPLICATION TO IBERIAN RAINFALL IN WINTERTIME

A statistical strategy to deduct regional-scale features from climate general circulation model (GCM) simulations has been designed and test ed. The main idea is to interrelate the characteristic patterns of obs erved simultaneous variations of regional climate parameters and of la rge-scale atmospheric flow using the canonical correlation technique. The large-scale North Atlantic sea level pressure (SLP) is related to the regional, variable, winter (DJF) mean Iberian Peninsula rainfall. The skill of the resulting statistical model is shown by reproducing, to a good approximation, the winter mean Iberian rainfall from 1900 to present from the observed North Atlantic mean SLP distributions. It i s shown that this observed relationship between these two variables is not well reproduced in the output of a general circulation model (GCM ). The implications for Iberian rainfall changes as the response to in creasing atmospheric greenhouse-gas concentrations simulated by two GC M experiments are examined with the proposed statistical model. In an instantaneous '' 2 CO2'' doubling experiment, using the simulated chan ge of the mean North Atlantic SLP field to predict Iberian rainfall yi elds, there is an insignificant increase of area-averaged rainfall of 1 mm/month, with maximum values of 4 mm/month in the northwest of the peninsula. In contrast, for the four GCM grid points representing the Iberian Peninsula, the change is - 10 mm/ month, with a minimum of - 1 9 mm/ month in the southwest. In the second experiment, with the IPCC scenario A (''business as usual'') increase of CO2, the statistical-mo del results partially differ from the directly simulated rainfall chan ges: in the experimental range of 100 years, the area-averaged rainfal l decreases by 7 mm/month (statistical model), and by 9 mm/month (GCM) ; at the same time the amplitude of the interdecadal variability is qu ite different.