Abstract Simulations are carded out to verify a mesoscale model in order to perform sensitivity tests of satellite response to atmospheric dust content. The model chosen is the mesoscale model of Colorado State University with a modified radiation parameterization in order to take atmospheric dust content into account. Downward and upward longwave irradiances are estimated using a 25-interval model. The shortwave pan of the spectrum is processed by a very fast, highly parameterized, single-interval code. Tests using experimental data gathered during the Etude de la Couche Limite Atmospherique Tropicale Seche (ECLATS) experiment performed during the 1980 dry season near Niamey (Niger, West Africa) prove that dust content is satisfactorily handled. Three 24-h simulations performed under various meteorological and turbidity conditions show that ground surface energy exchanges are satisfactorily described, so that surface temperature is predicted with a standard deviation of about 1°C. Vertical profiles of co...