The thesis deals with the broadly recognized problem of soiling of the surface of solar collectors, since it severely affects the energy yield of solar power plants. The specific goal is focused on Concentrated Solar Power (CSP) tower plant, which is one of the most promising technology for solar radiation harvesting and power production. Through a Matlab model developed during an exchange year at the Queensland University of Technology (QUT), Brisbane, QLD, Australia, the reflectivity of the surface of the heliostats composing the solar field is evaluated along a whole season given as inputs dust characteristics and weather conditions of a specific location. Finally, taken as input the direct normal irradiance (DNI) of the site, an estimation of the actual radiant power reflected towards the central receiver is given. The model explores many different phenomena that are involved in the whole soiling process since it has to simulate the deposition of dust onto the surfaces, the competing forces acting on deposited particles that tend either to keep them on the surface (adhesion) or to detach them from the surface (removal). The novelty of this work resides mostly in the comprehensive and detailed approach of the study that considers all the main steps of the soiling process through a deep physical insight, analysing the factors and the parameters that characterised each phenomenon, using physical laws derived from many and various field of research. The main outcomes of the model are the trend of reflectivity in time for each sector of the solar field and the related radiant power losses. The former is proved to be strongly dependent on the airborne dust concentration (that increases the detrimental effects of soiling) and on the frequency of rain (that restores the cleanliness of the heliostats). For low values of airborne dust concentration (as it is in Collinsville, QLD, Australia, where measurements were taken) the average soiling rate, expressed in percentage reflectivity loss per day, is 0.061%/day while measurements reported in literature for Saudi Arabia gives higher values that depend on the higher dust concentration in air for those areas, which corresponds to the predictions of the model, given the related data inputs, as analysed in the performed sensitivity analysis.