In this work models of pollutants transport in atmospheric air were described in form of stochastic differential equations. The way of their solution was shown. On the base of measurments results of average daily SO2 concentrations given during a year an ten automatic stations of monitoring set in Szczecin agglomeration the most profitable distributions (gamma, Weibull's) were set. Their parameters were defined and characteristic values were calculated: average concentration (average annual), modal value (daily average concentration the most of ten occured) and the highest in the given station concentration so called episodes. Values of arithmetical means were given as well and annual, average concentration calculated with application of present obligatory estimating quality of air system. Normalized distribution were established and was stated that obtained on their base for certain stations modal values and dispersion coefficients can give the base of monitoring station estimation. It was shown that stochastic models describe atmospheric transport in better way then deterministic models.