The geomagnetic field, as observed at the Earth's surface or satellite altitudes, is the combination of signals generated by several sources: internal sources due to the liquid outer core flow and magnetized rocks in the lithosphere, and external sources due to electric currents in the ionosphere and magnetosphere. We describe an approach for modeling the magnetospheric contribution to the geomagnetic field, based on magnetic observatory data. Magnetic survey satellite data provide a good spatial resolution for fields generated in the Earth, but their spatiotemporal resolution is limited for rapidly varying external magnetic fields. Modern magnetic observatories, located over a large range of latitudes and longitudes, record minute-mean (or second-mean) vector magnetic data. Therefore, magnetic observatories provide together a data set with a better temporal resolution particularly well suited to characterizing magnetospheric signals. The objective of this work is to produce a reliable description of theobserved magnetospheric components up to spherical harmonic degree 6 with a time resolution of an hour or better. The approach followed is based on a Kalman filter and correlation-based modeling. In order to properly separate fields of internal and external origins in magnetic observatory data, the data are firstly treated by removing the core contribution as described by MCM core model, and then selected for night time (from 23:00 to 05:00 UT) and geomagnetically quiet time (Dst between -30 nT and 30 nT). The first results are presented and compared with anapproach where the day-side data are used and the magnetospheric model co-estimated with the ionospheric field contribution in the observatory components. We anticipate that using day-side data is a necessary step to derive robust magnetospheric field models.