Previous studies using fallout radionuclides (137Cs and 210Pbex) to determine the relative contributions of surface-soil and channel erosion (including gullies and channel banks) to stream sediments have used a relatively small number of composite samples (<25) to characterize the source end members, and concentrations in each of the source end members have, through the use of means and standard errors, been assumed to be normally distributed. Here, we examine 137Cs activity concentrations to determine the erosion processes supplying sediment in seven water supply catchments in South-east Queensland. First, we test some of the underlying assumptions in the method using concentrations of 137Cs and 210Pbex in composite samples collected from 109 surface-soil sites and 39 channel-bank sites. Paired composite samples, each consisting of 20 subsamples, from five sites are used to demonstrate that this sampling approach was sufficient to average out any local variations in surface-soil 210Pbex and 137Cs concentrations across the 300 m2 sampled. We derive probability distributions for 137Cs and 210Pbex concentrations in the group of samples from each of the end members and show that only the distribution of 210Pbex in samples from surface soils is normally distributed. We use the probability distributions for 137Cs, which provides the greatest discrimination between sources, to show that the 137Cs concentrations on the river sediment samples are consistent with channel erosion being the dominant source. Conservation works aimed at reducing the supply of sediments in these catchments should therefore focus on rehabilitation of the channel network and decreasing the runoff to the channel network. These findings are consistent with other similar studies on tropical Australian rivers. Copyright © 2012 John Wiley & Sons, Ltd.