Most of the commonly used flood frequency distributions, when extrapolated toward exceedingly large recurrence intervals, yield floods of unlimited magnitude. However, it is well recognized that there must be upper limits to local accumulations of precipitable moisture. This paper describes a simple procedure by which to reconcile these upper limits with conventional methods of flood frequency analysis.
An interesting analogy is developed between the probability distribution of maximum floods and the time distribution of a logistically growing population. This analogy yields a suitably bounded "logistic-normal" frequency distribution. A graphical procedure for presenting the logistic growth curve as a straight line is then useful for linearizing a logistic-normal distribution using ordinary log-probability graph paper.
Sample applications of this new technique to some existing sets of flood data indicate that, for most rivers, a simple logistic transformation tends to yield frequency distribution curves that are straighter, and therefore more confidently extrapolated, than conventional log-probability plots.