General Circulation Models (GCMs) are considered to provide the best basis for the construction of future climate change scenarios. GCM output cannot, however, be widely or directly used in many impact assessments because of their relatively coarse spatial scale. In order to overcome the problems of mismatch in scale and of reliability, techniques for downscaling GCM output have been developed. A number of different downscaling methods have been proposed and can be divided into two general categories: model-based and empirical. Two examples of empirical downscaling are presented. First, scenarios for the Mediterranean region as a whole, constructed using a Geographical Information System to interpolate GCM output to a 1 km by 1 km grid based on information such as height above sea level, distance to the sea, and latitude/longitude. Second, scenarios at the river-basin scale for study regions in Spain and Italy, based on changes in atmospheric circulation patterns.