Several recent studies have shown that a remarkably wide variety of natural systems display fluctuations that may be characterized by long-range power-law correlations. Such correlations hint toward fractal geometry of the underlying dynamical system. Existence and determination of power-law correlations would help to quantify the underlying process dynamics. Traditional time series analysis of rainfall, for example using spectral analysis, often produces spurious results due to the highly nonstationary nature of rainfall signals. Here we present some new results using detrended fluctuation analysis (DFA) to quantify the correlation properties of rainfall time series. The DFA allows the detection of long-range correlations embedded in an apparently nonstationary time series and has been successfully applied to analyze highly nonstrationary DNA sequences, heartbeat time series, and temperature fluctuations. Our analysis of rainfall data ranges from hourly to annual timescales for geographically and meteorologically distinct regions. Results show a common scaling behavior for all stations analyzed. Two distinct scaling regions are identified, one corresponding to short-range and the other to long-range correlations, on either side of a breakpoint. Additional synthetic experiments demonstrate that what controls the correlation structure is not the actual rainfall values but the pattern of alternating wet and dry spells. The application of DFA on streamflow reflects the fact that streamflow is characterized by smooth fluctuations compared to its driving process, rainfall. Our results highlight the dampening effect that the soil has on the rainfall correlations. ¿ 2000 American Geophysical Union