Abstract

In the atmospheric sciences, inverse modelling has mainly been applied to global source distributions of greenhouse gases. On the other hand, "trajectory statistics" has been developed as a tool to evaluate measurements of ambient concentrations of atmospheric trace substances (especially pollutants) with respect to (potential) sources on the regional scale.

In this talk, I will show how inverse modelling techniques can be used as a quantitative and straightforward tool to derive information on sources from measurements. I shall give an overview of different inverse modelling techniques and explain more in detail the technique I use, namely deriving the source-receptor matrix (SRM) with Lagrangian models, and then inverting the resulting linear system of equations using additional constraints ("a-priori information", "regularisation"). The Lagrangian model can be back trajectories in the simplest case, or a Lagrangian particle dispersion model (FLEXPART is used here). In prinicple, any model can be used in a forward or a backward (adjoint) mode for the determination of the SRM. Advantages and drawbacks of these modes will be discussed, and the novel approach of using a LPDM in backward mode (without the need for an explicit adjoint!) will be presented.

The techniques will be illustrated with a 1-year sulfate concentration data set from the European EMEP network in combination with simple back trajectories, and the first release of the European Tracer Experiment ETEX in combination with backward runs of FLEXPART.

See also my inverse modelling web page at http://www.boku.ac.at/imp/envmet/invmod.html.