The systematic modes of spatial and temporal variation in 372-month records of streamflow in Turkey are defined using the principal components (PCs) analysis. The PCs were computed from a matrix of monthly records of 78 stations during the period 1964-1994. The analysis revealed not only basic streamflow anomaly patterns but described also the variation of those patterns through time. For this purpose, the map patterns and time series of each PC scores were obtained, but only the first five significant components were selected for further analysis. They accounted for more than 70% of the total variance of the original streamflow data set. Temporal variations of the five PC scores appeared to be fairly synchronous with regional precipitation anomaly patterns documented earlier. Seasonal correlations between the PC scores and the indices of large-scale atmospheric circulation (i.e. the North Atlantic Oscillation (NAO) and the Southern Oscillation (SO)) were computed to infer the physical meaning of the components. For example, using annual and winter average series, the maximum correlation coefficients between the first PC and the NAO and SO indices were found as -0.491 and -0.506 (significant at the 99% level), respectively. This implies the sensibility of the Turkish streamflow climatology to the variations of the NAO. The fifth PC was also shown to be in a significant relation with the SO. Similar analysis was seasonally carried out for the series of precipitation-streamflow PC scores. As a result, the relationships between the series of precipitation-streamflow and precipitation-first PC were fairly strong with a correlation value of 0.657 and 0.596 at the 99.9% significance level, respectively. (c) 2005 Elsevier B.V. All rights reserved.