Uncertainty analysis for complex chemical systems
Résumé
The validation of the models for complex reactors, such as planetary atmospheres, ionospheres or interstellar dark clouds relies on the estimation of the accuracy of their predictions. For a deterministic chemistry model, uncertainty has two major sources, the variability of the environment (temperatures, photon and electron fluxes...) and the uncertainty about the intrinsic parameters of the model (chemical kinetics parameters). We focus here on the latter source of uncertainty. Individual reactions are characterized by parameters (rate constants, branching ratios) which result from laboratory measurements or theoretical estimations. Uncertainty on these parameters has to be evaluated and propagated through the model. This basic problem becomes challenging when hundreds or thousands of parameters have to be considered. We will review different approaches developped in the recent years to deal either with uncertainty representation for the parameters, or with efficient uncertainty propagation. Application to Titan's ionospheric chemistry will be presented.