This paper describes a process of analysis of the Gaussian model ISCST3 of air pollutants dispersion applied in a densely populated area, the Higher Polytechnic Institute José Antonio Echeverría (CUJAE) in Havana. For the numerical simulation was carried out a sources and emissions inventory of air pollutants in the area, to estimate the concentration of the conservative pollutant, in this case sulfur dioxide. The objective is to show the sensitivity of numerical simulation with variation in the input of several parameters: average albedo, Bowen ratio, surface roughness and wind speed. After obtaining the results, the influence of variations in these parameters input and contaminant concentration studied is analyzed. Concluding that the wind speed is the parameter to which the model is more sensitive and has the lowest sensitivity to changes in albedo input. The least affected receptor concentration values estimated for 24 hours is the number 2, located in the center of the screen and land in the campus CUJAE. While for hourly values, receptors 1 and 3 are less sensitive.