Through a systematic modelling analysis for focus catchments in Costa Rica (131 km(2)), Ecuador (10 km(2)), Chile (0.35 km(2)) and Argentina (12.9 km(2)), the hypothesis is tested that, as the size of the hydrological event increases, the effect of forest cover on the peak discharge becomes less important. For each focus catchment, a 1000-year synthetic rainfall time series was generated, representative of the current climate. This time series was used to run SHETRAN hydrological models for each catchment with two contrasting land use scenarios (generally with and without a forest cover). The corresponding maximum daily discharges for the contrasting scenarios were then compared to show the extent to which the two responses converged as the size of the peak discharge increased. For a given forest catchment discharge there could be a range of larger non-forest catchment discharges, depending on antecedent soil moisture content. The simulations show consistently for the rainfall dominated sites that the width of this range either remains constant or narrows as discharge increases, indicating either relative or absolute convergence of the responses. The pattern is more difficult to distinguish for a snowmelt regime but a relative convergence of response still appears possible. The results therefore support the test hypothesis. However, the pattern is complicated by factors such as catchment scale, soil depth, antecedent moisture content and land management. Forests may also still offer significant flood mitigation benefits for moderate (and more frequent) rainfall events and they protect against soil erosion and sediment transport for a wide range of events. (C) 2011 Elsevier B.V. All rights reserved.