Using a System of Models of Various Complexity in Inverse Modeling of Pollutant Transport and Transformation in the Atmosphere

The problem of using integrated models to assess and predict air quality require the development of data assimilation algorithms. The use of computationally complex models in inverse modeling algorithms, which often require multiple solutions to direct and adjoint problems, may be difficult from a practical point of view. An algorithm involving two models (the detailed model and the simplified one) is considered. The detailed model is used in the direct modeling mode, and the simplified model serves to recover information about key parameters for modeling (for example, emission sources) based on observational data. This raises a problem of adjusting (training) the simplified model to reproduce the behavior of the detailed one. The algorithm has been tested on the regional air quality assessment scenario for the Siberian region.