We present a simple method to make multi-year surface temperature forecasts using the climate change simulations of the CMIP3 database prepared for the IPCC AR4 report. By calibrating the multi-model ensemble mean with current observations, we are able to make skillful interannual forecasts of mean temperatures. The method is validated using extensive hindcast experiments and is shown to perform favorably compared to a recent forecast method based on a global circulation model with assimilated initial conditions. Five year forecasts for the global mean temperature, the Northern Hemispheric mean temperature and the summer sea surface temperatures (SSTs) in the main development region for hurricanes (MDR) are presented. Copyright 2008 by the American Geophysical Union.