[1] The objective of this paper is to describe a statistical downscaling model based on established links between large scale variables from a general circulation model (GCM) and regional rainfall to project climate change over East Asia in the 21st century. We develop a SVD (singular value decomposition)-based regression model to downscale the GCM data. In order to establish a link between monthly regional precipitation and large scale variable (mean sea level pressure, 10 m wind speed, or 2 m temperature), we used the IS92a scenario data of ECHAM4/OPCY3 and the data from the Climate Research Unit from 1901 to 1990. In the 20th century, their connections have remained unchanged over time for all seasons. The continuity of these relationships to the next century is tested by checking the similarity in the main SVD modes of predictors and the empirical orthogonal functions of predictors throughout the 20th and 21st century. Furthermore, comparing the SVD-downscaled and ECHAM4-projected changes of precipitation in the 2090s with respect to the 1990s reveals similar patterns over most areas with much more detailed regional characteristics, although the magnitude of change is not compatible. In all, the annual rainfall over East Asia is projected to increase over northern China and the Indochina peninsula and to decrease over the coast of southern China by the end of the 21st century.