Thermometric measurements over Belgium provide a considerable amount of spatiotemporal data, the study of which can provide valuable information for a variety of atmospheric studies, agricultural activities, pollution control, etc. In this work, a data set that includes a year of daily temperature measurements at 121 stations and shows a complicated variation in space and time is selected and analyzed using a spatiotemporal random field model. This model has three random components and a space/time mean function, which allow considerable flexibility in analytical calculations and rigorous assessment of the spatiotemporal variation of temperature. A space/time estimation system is formulated and solved to provide detailed temperature maps over the whole country. The composite space/time analysis offers valuable physical insight that could not have been obtained on the basis of a purely spatial or a purely temporal analysis of the temperature data set. The variability between successive days at a monitoring station is considerably more important than spatial variability throughout the country at a given day, and therefore limited improvements should be expected regarding temperature forecasting and interpolation by focusing only on the spatial aspect of the temperature variation. ¿ 1997 American Geophysical Union