Good knowledge of the environmental conditions of deserts on Earth is relevant forclimate studies. The Atacama Desert is of particular interest as it is considered tobe the driest region on Earth. We have performed simulations using the WeatherResearch and Forecasting (WRF) model over the Atacama Desert for two week-longperiods in the austral winter season coincident with surface temperature and relativehumidity in-situ observations at three sites. We found that the WRF model generallyoverestimates the daytime surface temperature, with biases of up to 11◦C, despitegiving a good simulation of the relative humidity. In order to improve the agree-ment with observed measurements, we conducted sensitivity experiments in whichthe surface albedo, soil moisture content and five tuneable parameters in the NoahLand Surface Model (namely soil porosity, soil suction, saturated soil hydraulic con-ductivity, thebparameter used in hydraulic functions and the quartz fraction) areperturbed. We concluded that an accurate simulation is not possible, most likelybecause the Noah Land Surface Model does not have a groundwater table that maybe shallow in desert regions. The WRF-predicted land surface temperature is alsoevaluated against that estimated from the Moderate Resolution Imaging Spectrora-diometer (MODIS) instrument. While at night the satellite-derived and ground-basedmeasurements are generally in agreement, during the day MODIS estimates aretypically lower by as much as 17◦C. This is attributed to the large uncertainty inthe MODIS-estimated land surface temperatures in arid and semi-arid regions. Thefindings of this work highlight the need for ground-based observational networksin remote regions such as the Atacama Desert where satellite-derived and modelproducts may not be very accurate.