A model for interpolating 2D thermal satellite data over a 3D urban surface is described, which combines the ViSC and GIS functions inherent in a modern GIS interface. Vertical surfaces are represented as spatial entities whose temperature attributes are determined from image data supplemented by field sampling in urban canyons. Thus a model representing the temperature of the complete urban surface is created, for which the viewpoint can be varied according to Sun angle and azimuth at the time of imaging and the viewing angle required by the user. Due to the close correspondence between surface and air temperature the model indicates micro-scale climatic variations due to variations in building geometry and surface materials which are not readily apparent from a twodimensional perspective. It also illustrates the magnitude of systematic error introduced by the nadir viewing angle of satellite sensors.