Inverse Dispersion Modelling Based on Trajectory-Derived Source-Receptor Relationships

Inverse dispersion modelling means to derive information such as source strengths of emissions from measured concentration and/or deposition data of trace constituents, using a dispersion model. If source-receptor relationships are linear, they can be determined from a single model run, and the inversion corresponds to solving a linear system of equations. Examples for this approach are Enting and Newsam (1990), Brown (1993), and Hein et al. (1996). Non-linear source-receptor relationships require an iterative solution using an adjoint model. The calculation of large source-receptor matrices is very time consuming. Therefore, simple trajectories have been used widely to derive information on sources from pollution measurements. So-called potential source contribution functions have been calculated, e. g., by Zeng and Hopke (1989); similar techniques were employed Seibert and Jost (1994) and Stohl (1996). These are statistical methods, not based on dispersion models, and not exploiting the full information contained in the data sets. However, trajectories can be viewed as primitive Lagrangian dispersion models, and source-receptor matrices can be derived from them for use in formal inverse modelling. One has to be aware, however, that systematic shortcomings (e. g., neglection of precipitation scavenging or vertical exchange) will lead to systematic errors in the results, a problem shared with the statistical approach.