Some remarks concerning inverse modeling and data assimilation for slow-fast atmospheric chemical kinetics

Adjoint modeling is a powerful tool that is now widely used in meteorology for computing gradients in the context of data assimilation. We present in this article the use of an automatic differentiation tool developed at INRIA, O∂yssee, for the case of Air Pollution Modeling. We simulate a reduced kinetic scheme (16 species, 12 reactions) describing the evolution of tropospheric ozone. We focus on some particular points such as the influence of the slow-fast behaviour of the chemical models.