This study proposes an algorithm to identify streamflow source areas through the interpretation of the predictive uncertainty analysis of streamflow generation. The method interprets model-free variance-based sensitivity indices and response surfaces to track the spatiotemporal evolution of streamflow source areas that define the components of a single rainfall-runoff event-driven hydrograph. Under a Monte Carlo framework, a distributed model was used to investigate (at daily resolution) a mountainous catchment with simple terrain in Japan and (at monthly resolution) a mountainous Andean watershed with complex terrain. Results suggest that for simple terrains, integrative river discharge observations at the outlet describe, to a great extent, most states of the catchment runoff response. However, for the time steps during which rainfall strikes, the validity of the observations at the outlet is restricted to the main stream. In the watershed with simple terrain, permanently saturated areas were found in the low lands and the riparian areas (slope gradients from 1% to 10%). The storage zones are on the headwaters (slope gradients up to 23%). Transitional zones cover the remaining areas. In complex terrains, the role of the topographic slope gradients is more relevant than the topology for the streamflow source areas delineation. In the Andean watershed, the dominating processes are in the headwaters, where the slope gradients are gentle (from 3% to 20%), and wetlands act as transient water reservoirs. In conclusion, the transferability of information at different spatiotemporal scales between catchments with heterogeneous topography and complex terrain seems unlikely. Copyright © 2011 John Wiley & Sons, Ltd.