In this work, we propose an approach to estimating the amount of light wasted by being sent towards the upper hemisphere from urban areas. This is a source of light pollution. The approach is based on a predictive model that provides the fraction of light directed skywards in terms of a small set of identified explanatory variables that characterise the urban landscape and its light sources. The model, built via the statistical analysis of a wide sample of basic urban scenarios to compute accurately the amount of light wasted at each of them, establishes an optimal linear regression function that relates the fraction of wasted flux to relevant variables like the kind of luminaires, the street fill factor, the street width, the building and luminaire heights and the walls and pavement reflectances. We applied this model to evaluate the changes in emissions produced at two urban nuclei in the Deltebre municipality of Catalonia. The results agree reasonably well with those deduced from the radiance measurements made with the VIIRS instrument onboard the Suomi-NPP Earth orbiting satellite.