We test the reliability of an inverse model to estimate the amount of mass released instantaneously by a source of pollutant in a turbulent boundary layer. To that purpose we use wind tunnel experiments simulating the dispersion of puffs due to the impulsive release of pollutant. For each of these emissions we recorded time dependent signals at fixed receptors. These single signals have been used as input data for an inverse dispersion model in order to evaluate its error in estimating the real emission of pollutant. The inversion has been also performed using the ensemble average of the 100 signals as input data. The comparison of both approaches allows us to discuss the reliability of an atmospheric dispersion inverse model in real situations.
Keywords: inverse modelling, accidental release, deliberate release, risk management, atmospheric dispersion