Abstract: Data assimilation is a process where an improved prediction is obtained from a weighted combination between experimental measurements and mathematical model data. In the present work this procedure is applied to pollutant atmospheric dispersion by using a Kalman filter (KF). This is interesting aprroach, because the KF gives an output in which the balance between the data from the diffusion model and the experimental data is done automaticaly, through the Kalman gain. In addition, the Kalman filter computes the propagation error.