Severe storms and associated floods have been a major natural catastrophe in a number of cities over the last decade. The risk of extreme precipitation events is usually assessed by return periods, which are evaluated from frequency analysis of point rainfall accumulations. Point return periods associated with design quantiles are traditionally used in the design of structures to control storm runoff and floods. Return periods have also acquired great meaning in hydrology when analyzing observed events and qualifying their severity. This study investigates the characterization of the return period of a rainfall event over a given area, highlighting the importance of considering its spatial dimension. Three storm events observed in the city of Marseilles, France, are studied. Severity graphs and severity diagrams are developed for each storm event. They allow us to visualize the strong variability of return periods within a storm. For a given rainfall duration, one can observe point return period estimations varying from ordinary values (less than 1 or 2 years) to rare or even exceptional values (10–100 years) when considering different rain gauges. Also, for a given rain gauge, different return periods are estimated for different rainfall durations and important differences can be found between the obtained values. The degree of severity of each storm event for the city is shown to depend also on the extent of heavy rain areas and therefore on the spatial structure of rainfall intensities. Storm severity visualization could be useful in evaluating the importance of observed rainstorms over a city and in helping decision makers in real time water management.