The applicability of meteorological general circulation models (GCMs) is li
mited by their spatial resolution. In this paper, a method is developed for
improving the resolution of GCM-generated rainfall fields, using ideas fro
m Bayesian image analysis to improve the resolution of the binary wet-dry i
mage. This approach incorporates both the spatial and temporal memory of th
e rainfall held and can be adapted to utilize any available physical inform
ation. The method is illustrated using data from a network of weather radar
stations in Arkansas, and some informal diagnostic procedures are develope
d for assessing the adequacy of the underlying model.