A contribution to statistical depth-duration-frequency analysis

Abstract

Present methods for evaluating the depth-duration-frequency relationship to be assumed for any given location are not completely satisfactory because of the uncertainty due to the low density of recording station networks and the assumed significance of every difference between stations. A new procedure is suggested with the aim of taking full advantage of all the information at hand. The procedure is based on the grouping of stations exhibiting homogeneous characteristics and on the use of the one-day rainfall characteristics available from the much more numerous non-recording stations. Station grouping is accomplished under rigorous statistical control by means of analysis of variance theory while a convenient statistical criterion is given to relate the one-day depth distribution to the shorter duration depth characteristics. Moreover, a new method for the depth-duration-frequency analysis of point rainfall is suggested on the basis of linear regression theory. Although not completely rational from a statistical point of view, because of the lack of independence between observations, the method is believed to be superior to the ones commonly used. As an example, an application to the rain-gauge network of Sardinia is reported.