The NAO, defined by the first EOF of the geopotential height in the North Atlantic area, is well reproduced by a long-term run of the model forced by ERA40 data. Similar properties appear in the observed NAO and simulated NAO, in particular, the relationship between the nature of the waves breaking and the NAO. Furthermore, a large-scale low-frequency ridge anomaly is identified in the Northeast Pacific in the days prior to the maximum of the positive NAO phase both in the reanalyses and model. A large-scale Northeast Pacific trough anomaly is observed during the negative NAO phase but does not systematically precede it. In order to understand how the large-scale ridge anomaly can act as a precursor of the positive NAO phase, short-term linear and nonlinear simulations were performed. The ridge acts in two ways. First, it tends to prevent the downstream propagation of short synoptic waves compared to long synoptic waves. Second, it deflects the propagation of the synoptic wave trains in such a way that they mainly propagate equatorward in the Atlantic. The two modes of action favor the anticyclonic wave breaking and so the positive NAO phase. With the trough, the wave train propagation is more zonal, disturbances are more meridionally elongated and cyclonic wave breaking is more frequent in the Atlantic than in the ridge case.