The aim of this project is to co-estimate models of the core and ionosphere magnetic fields, with the longer-term view of building a "comprehensive" model of the Earth's magnetic field. In this first step we take advantage of the progresses made in the understanding of the ionosphere by global M-I-T modeling to better separate the core and ionospheric signals in satellite data. The magnetic signal generated in the ionosphere is particularly difficult to handle, because satellite data provide only information on a very narrow local time window at a time. To get around this difficulty, we are applying a technique derived from assimilation methods and that has been already successfully applied in outer-core flow studies. The technique relies on a theoretical model of the ionosphere such as the Upper Atmosphere Model (UAM), where statistics on the deviations from a simple background model are estimated. The derived statistics provided in a covariance matrix format, which can be used directly in the magnetic data inversion process to obtain the expected core and ionospheric models. We are applying these technique on the German CHAMP satellite data selected for magnetically quiet times. As an output we obtain a model of the ionospheric magnetic variation field, tailored for the selected data and a core-lithosphere field model where possible leakage from ionospheric signals are avoided, or at least reduced. The technique can in theory be easily extended to handle the large-scale field generated in the magnetosphere.