Desert areas are good candidates for the assessment of multitemporal, multiband, or multiangular calibration of optical satellite sensors. This article describes a selection procedure of desert sites in North Africa and Saudi Arabia, of size 100 x 100 km(2), using a criterion of spatial uniformity in a series of Meteosat-4 visible data. Twenty such sites are selected with a spatial uniformity better than 3% in relative value in a multitemporal series of cloud-free images. These sites are among the driest sites in the world. Their meteorological properties are here described in terms of cloud cover with ISCCP data and precipitation using data from a network of meteorological stations. Most of the selected sites are large sand seas, the geomorphology of which can be characterized with Spot data. The temporal stability of the spatially averaged reflectance of each selected site is investigated at seasonal and hourly time scales with multitemporal series of Metcosat-4 data. It is found that the temporal variations, of typical peak-to-peak amplitude 8-15% in relative value, are mostly controlled by directional effects. Once the directional effects are removed, the residual rms variations, representative of random temporal variability, are on the order of 1-2% in relative value. The suitability of use of these selected sites in routine operational calibration procedures is briefly discussed.
