Diatoms, microscopic single-celled algae, are present in almost all habitats containing water (e.g. streams, lakes, soil and rocks). In the terrestrial environment, their diversified species distributions are mainly controlled by physiographical factors and anthropic disturbances which makes them useful tracers in catchment hydrology. In their use as a tracer, diatoms are generally sampled in streams by means of an automated sampling method; as a result, many samples must be collected to cover a whole storm run-off event. As diatom analysis is labour-intensive, a trade-off has to be made between the number of sites and the number of samples per site. In an attempt to reduce this sampling effort, we explored the potential for the Phillips sampler, a time-integrated mass-flux sampler, to provide a representative sample of the diatom assemblage of a whole storm run-off event. We addressed this by comparing the diatom community composition of the Phillips sampler to the composite community collected by automatic samplers for three events. Non-metric multidimensional scaling (NMDS) showed that, based on the species composition, (1) all three events could be separated from each other, (2) the Phillips sampler was able to sample representative communities for two events and (3) significantly different communities were only collected for the third event. These observations were generally confirmed by analysis of similarity (ANOSIM), permutational multivariate analysis of variance (PERMANOVA), and the comparison of species relative abundances and community-derived indices. However, sediment data from the third event, which was sampled with automatic samplers, showed a large amount of noise; therefore, we could not verify if the Phillips sampler sampled representative communities or not. Nevertheless, we believe that this sampler could not only be applied in hydrological tracing using terrestrial diatoms, but it might also be a useful tool in water quality assessment.