Amid concern over the decline of salmonid and other fish populations, the traditional basis of much stream restoration has been identification of degraded river sections and localised attempts to restore them. Research has demonstrated that fine sediment, solutes and organic matter also influence instream aquatic ecosystems and these may be influenced by upstream watershed land use. However, results from modelling the role played by land use impacts are contradictory. This is not surprising for three reasons. First, measurements of river ecology taken at any one point will be influenced by processes operating at scales ranging from the local (e.g. presence of suitable spawning habitat) through the reach-scale to the tributary and watershed scales. Second, upstream land use is only important if it can transmit or deliver its signal (e.g. a fine sediment source) to the river network, as moderated by the extent (frequency, duration) to which it connects with the river network. Third, once a signal is transmitted to a river, its importance can only be judged with respect to other signals, as a result of dilution and/or accumulation effects. These latter two issues are often overlooked by studies of land use impacts upon aquatic ecology. In this paper we bring together two critical research developments: system-scale semi-quantitative electrofishing of salmonids; and risk based modeling of hydrological connectivity; for a 2300 km2 catchment. We support these developments with quantification of both local- and reach-scale influences on salmonid habitat. We use multivariate inference to show that an index of delivery based upon hydrological connection is a first order control upon the presence/absence and the abundance of juvenile salmonid fry: the topographic control of watershed hydrological response exerts a fundamental filtering effect upon the spatial structure of salmonid fry when evaluated at the catchment-scale. We also show that the nature of this relationship is scale dependent, varying between and within sub-watersheds. We conclude that if topographic control mediates the watershed to stream linkage, land use impacts can only be understood with respect to their position in the landscape. The spatial organisation of landscape elements becomes crucial to understanding the ecological impacts of particular management activities, with hydrological flow paths providing the functional linkage. This is of practical importance as locations of high connectivity should be a primary objective in targeting watershed restoration measures to where they will deliver most instream benefits.