We present a statistical method to reconstruct continuous atmospheric fields on various pressure levels, given a constraint at the surface. The method is based on analogues of circulation and taking into consideration the time sequence of the analogues. The method is tested on the 2001-2011 period, with an emphasis on the year 2010. We base the atmospheric reconstruction on reanalysis data from 1948 to 2000, and use a constraint of sea-level pressure for 2001-2011. The pattern correlation scores appear to be significant most of the time, although score flaws are occasionally detected. Those flaws are mainly due to the time continuity constraint that is imposed on the reconstruction, which lowers the possibility of finding matching analogues. This method offers an ensemble of atmospheric reconstructions, and presents a computationally cheap alternative to data assimilation for a climate model, although with lower scores. © 2012 Springer-Verlag Berlin Heidelberg.