The alluvial deposits of braided rivers consist of channel-form bounding surfaces due to channel cutting and migration. It is unknown, however, to what extent these bounding surfaces result either from more gradual channel processes or from large but infrequent flood events. Field measurements of morphodynamically formed bounding surfaces associated with floods are not currently available. Flume experiments are useful here since concurrent observations of channel migration and the formation of bounding surfaces can be made whilst conditions are controlled. We report on two flume experiments with the objective to determine the effects of floods on i) channel cutting and migration and ii) alluvial architecture of braided gravel-bed rivers. One of the experiments has a constant bankfull discharge, and the other is run with a schematized long-duration low flow and short-duration high flow but otherwise similar conditions. High-resolution digital elevation models (DEMs) are used to monitor channel migration and to create the three-dimensional braid-belt architecture. The data demonstrate that the flood events result in longer bars and more frequent chute cutoffs. The DEMs show neither deeper channels nor bar aggradation during multiple flood events compared to a steady bankfull discharge. Consequently, the braid-belt architecture that forms during these large floods does not significantly differ from the stratigraphic architecture that forms during more frequent bankfull flows. This implies that these events cannot be differentiated based on stratigraphy. Furthermore, we find that between 10% and 40% of the mean channel depth is preserved. This can be used to characterize braid-belt architecture for purposes of reservoir engineering and provides error bounds to reconstruct paleochannel dimensions from stratigraphy.