AbstractPresent lithospheric field models, like the MF4 and CM4, are produced by least squares estimation using spherical harmonic basis functions with global support. Accounting for the different properties of magnetic data at low and high latitudes, a method that can take regional differences into account is proposed. Using four years of CHAMP satellite data, a detailed lithospheric magnetic field snapshot is obtained at 400 km altitude over the entire sphere by stitching together a dense coverage of regional models. The individual forward models computed on a quasi-regular grid over the Earth are then transformed to spherical harmonics by direct integration. Despite the stitching procedure, the long wavelength lithospheric features are correctly reproduced and small scale features are well resolved. Without regularization, the resulting model is stable to spherical harmonic degree 56. In addition to accounting for regionally varying noise levels, the proposed technique is also well suited to deal with incomplete data coverage issues when combining satellite with near surface data. The method could therefore make an important contribution to one of the main goals of the Swarm mission: to close the spectral gap between satellite and near-surface magnetic surveys.