The aim of this work is to test the ability of the Penn State/NCAR MM5 mesoscale model to reproduce correctly the dynamics in a coastal area of complex terrain in Israel. The model was configured using 4 nested domains with 1 way nesting interaction, with horizontal resolutions of 54, 18, 6 and 2 km, and vertical resolution of 26 levels with 11 levels between sigma 1.0 and 0.8. Forty-eight hour simulations initialized at 0 UTC were run for July 1994. During the summer months the atmospheric mesoscale dynamics in the area of interest is dominated by the sea-land breeze mechanism and by mountain-valley circulation. Observations collected at 8 ground stations show the influence of the topography and of the distance from the coast on the local dynamics. The evaluation of the model is done by comparison of simulated wind direction distributions, average speeds and temperatures and average wind hodographs to observations. The comparison shows that the model reproduces most of the major features that characterize the atmospheric dynamics in the area: 1. The model is able to reproduce correctly the wind direction distribution: the most frequent sectors are the same in the simulations and in the observations. 2. Average simulated wind speeds show correctly the observed diurnal cycle in most of the stations, but the model fails to reproduce the enhanced speeds observed at high terrain elevation stations. 3. Average simulated temperatures are in agreement with the diurnal cycle at most of the stations, thereby showing the effect of topography and distance from the coast. 4. Simulated wind hodographs show a similar rotation pattern to the observed ones, however the 24 hour simulated average wind vector is rotated clockwise by about 45° in most stations. Reasons for the discrepancy between the model and the observations will be discussed and ways to improve the model results will be proposed.