Application of air quality models for determining the source distribution is an emerging science. Both forward- and backward-modeling methods are currently being developed. Forward-modeling methods use forward running transport and dispersion models or CFD codes, which are run many times and the resulting dispersion field is compared against the actual data from multiple sensors. Such methods are based on Bayesian updating/inference, Kalman filtering, and Markov chain Monte-Carlo techniques. Utilizing sensor and meteorological data, inverse modeling methods solve the adjoint of a transport and dispersion model to estimate the source distribution, and reverse diffusion methods consider diffusion along backward trajectories to estimate the upwind area of influence for each sensor measurement. Lagrangian stochastic dispersion models, which calculate the ensemble-mean dispersion quantities from the trajectories of a large number of fluid particles representing the pollutant mass, are well suited for both forward and backward modeling. The need for characterizing the uncertainties in source estimation using atmospheric dispersion models is emphasized.