Abstract
The conventional statistical approach to flood-frequency analysis attempts to create “something out of nothing”, ignoring the essentially physical nature of surface runoff. The statistical methods of analysis, based on considerable extrapolation, claim to furnish one-to-one relation functions of flood frequencies. These methods make use of but few data, pushing aside most of the available data and ignoring their physical nature. They are therefore criticized. In this situation it is suggested that flood-frequency analysis be based on regional rainfall analysis.