AbstractSpatial and temporal average behaviour of winter precipitation in the Duero basin (DB) were analysed for 1958–1993. Monthly observed data (observed data) from 34 weather stations and a subset of daily precipitation data from the NCEP/NCAR reanalysis project (reanalysis data) were used.The spatial variability of winter precipitation was examined using principal component analysis (PCA) for both types of data. The winter precipitation fields are well represented by the first three principal components (PC), which describe 81% and 87% of the total variance for the observed and reanalysis data, respectively. The relationship between the three leading PCs and circulation patterns was also investigated. The PC1 is related to the North Atlantic Pattern (NAO), East Atlantic Pattern (EA) and Pacific North American Pattern (PNA). The PC2 is related to the East Atlantic/West Russia Pattern (EAWR), while the PC3 depends mainly on Scandinavia Pattern (SCAND) influences.The reanalysis data were used to calculate a 10‐day period to study the intra‐annual variation (S‐mode of PCA) and the spatial distribution of those precipitation amounts during the year (T‐mode of PCA). For the intra‐annual variation of precipitation, two modes statistically significant were obtained. The first mode was characterized by a broad maximum that extended from mid October to early April, and a minimum that extended from early May to September. The second mode exhibited two maxima and a minimum. The first maximum occurred in late February and the second maximum extended from early April to June. The minimum extended from mid July to early February. For the spatial distribution of precipitation, we obtained three patterns. The ‘winter’ pattern, which showed a maximum over the western area, the ‘spring’ pattern with a maximum over the eastern area, and the ‘autumn’ with a maximum over the easternmost edge of the basin. Copyright © 2006 Royal Meteorological Society.