Fourier and wavelet analyses were used to reveal the dominant trends and coherence of a more than one-century-long time series of precipitation and discharge in several watersheds in Sweden, two of which were subjected to hydropower and intensive agriculture. During the 20th century, there was a gradual, significant drift of the dominant discharge periodicity in agricultural watersheds. This study shows that the steepness of the Fourier spectrum of runoff from the May to October period each year increased gradually during the century, which suggests a more predictable intra-annual runoff pattern (more apart from white-noise). In the agricultural watershed, the coherence spectrum of precipitation and runoff is generally high with a consistent white-noise relationship for precipitation during the 20th century, indicating that precipitation is not controlling the drift of the discharge spectrum. In the hydropower regulated watershed, there was a sudden decrease of the discharge spectrum slope when regulation commenced in the 1920s. This study develops a new theory in which the runoff spectrum is related to the hydraulic and hydro-morphological characteristics of the watershed. Using this theory, we explain the changes in runoff spectra in the two watersheds by the anthropogenic change in surface water volume and, hence, changes in kinematic wave celerity and water transit times. The reduced water volume in the agricultural watershed would also contribute to decreasing evaporation, which could explain a slightly increasing mean discharge during the 20th century despite the fact that precipitation was statistically constant in the area. Copyright © 2010 John Wiley & Sons, Ltd.