Almost 65 years after the first design storms were proposed for urban hydrology (Keifer and Chu, 1957), the concept of using a synthetic rainfall pattern to generate runoff parameters with computer modeling is well integrated into water engineering practice. Even if many pitfalls were recognized early on and that continuous simulations were perceived as more appropriate for different reasons, applying design storms for a variety of engineering analyses has now become mainstream and is somewhat taken for granted. Despite all the caveats, practitioners have indeed continued using synthetic design storms, as they provided a simple and apparently appropriate tool for routine designs. In many cases, the origins and limitations of different design storms are however lost in the historical technical literature and are now being derived with the same simplified approaches, without recognizing their fundamental limitations. These could typically be revealed only when the results of modeling are questioned following for example specific events that have produced more important flooding than predicted with design storms.
After a brief historical reviews of design storms that are now routinely applied in Canada and an exposure of their limitations and assumptions on which they are based, with also a critical review of newer approaches developed internationaly, another approach is proposed to evaluate the performance of different rainfall patterns when discharges measurements are available. Using the actual hydrological response as given by the discharge measurements and statistical analysis, the shape and temporal distribution of existing design storms are adjusted so that the runoff parameters(peak discharge or runoff volume) are reproduced for a given return period. An application of the approach is provided using several models developed for different types of watersheds (both natural and urbanized) in the Quebec City area in Quebec.