Inverse modeling of atmospheric chemistry with a differential evolution solver: Inverse problem and Data assimilation

Abstract

Abstract In this paper, we numerically compare an inverse problem for an atmospheric chemistry model, when all the measurement data is available a priory, to a corresponding data assimilation problem when the data are obtained in some portions during the simulation. In both cases, we reconstruct the unobservable parts of the model state function by the observable ones (i.e., solve a continuation problem). This is done by identifying the model reaction rate parameters with the available measurement data by solving an optimization problem with a derivative-free differential evolution solver. In our numerical experiments, the inverse problem formulation of the continuation problem has not necessarily provided better results than the data assimilation formulation despite the latter being more limited in the measurement data.