Since the beginning of the 1990s, stable desert sites have been used for the calibration monitoring of many different sensors. Many attempts at sensor intercalibration have been also conducted using these stable desert sites. As a result, site characterization techniques and the quality of intercalibration techniques have gradually improved over the years. More recently, the Committee on Earth Observation Satellites has recommended a list of reference pseudo-invariant calibration sites for frequent image acquisition by multiple agencies. In general, intercalibration should use well-known or spectrally flat reference. The reflectance profile of desert sites, however, might not be flat or well characterized (from a fine spectral point of view). The aim of this paper is to assess the expected accuracy that can be reached when using desert sites for intercalibration. In order to have a well-mastered estimation of different errors or error sources, this study is performed with simulated data from a hyperspectral sensor. Earth Observing-1 Hyperion images are chosen to provide the simulation input data. Two different cases of intercalibration are considered, namely, Landsat 7 Enhanced Thematic Mapper Plus with Terra Moderate Resolution Imaging Spectroradiometer (MODIS) and Environmental Satellite MEdium Resolution Imaging Spectrometer (MERIS) with Aqua MODIS. The simulation results have confirmed that intercalibration accuracy of 1% to 2% can be achieved between sensors, provided there are a sufficient number of available measurements. The simulated intercalibrations allow explaining results obtained during real intercalibration exercises and to establish some recommendations for the use of desert sites for intercalibration.