ABSTRACTChanges of the land surface affect the water balance components over seasonal, annual and decadal time scales. This study explored the role of vegetation cover transitions on evapotranspiration in forested watersheds of the North American West. We applied empirical time‐recovery functions describing the impact of forest removal and subsequent regrowth on actual evapotranspiration (AET) or runoff. A generalized function (K‐curve) was adapted to the North American West and tested using three different datasets of observed or estimated AET in forest chronosequences: AET from flux towers equipped with eddy covariance sensors, AET estimated from the water balance in experimental paired watersheds and in a set of gauged watersheds with considerable forest cover history dating back to the 18th century. AET from the first two datasets showed a behaviour similar to the K‐curve, although the timing and the magnitude differed substantially. To reconstruct long‐term changes in AET for the gauged watersheds, we applied a transfer function approach linking the K‐curve and the reconstructed forest cover history at the watershed scale. In several watersheds, correlation coefficients between the reconstructed AET changes and the annual water balances suggest that changes in time were driven by the land cover transitions. In watersheds with low correlations, disturbance activities peaked before the 20th century, and the effects of vegetation have phased out in the period of streamflow observations. The findings of this paper suggest that trends in the observed water balance in forested watersheds can be associated to land cover disturbances well before the start of hydro‐climatic observations. Copyright © 2011 John Wiley & Sons, Ltd.