comparison of approaches for estimation of Depth Duration Frequency (DDF) Curves

Depth-duration-frequency (DDF) curve estimation is one of the important steps in the fields of hydrological design, flood risk management, and urban drainage planning. This gives a statistical relationship between rainfall depth, duration, and frequency, which is the basic insight required for managing extreme rainfall events. Traditional estimation of DDF includes two-parameter and three-parameter formulations based on the theoretical distribution (i.e., Gumbel distribution) for the frequency analysis of rainfall data. This approach may result in an accurate estimation of rainfall depths, usually requires vast high-resolution data, and several parameter calculation. On the other hand, scaling-invariance based methods are computationally efficient but normally of poor accuracy due to the assumption of constant values for some rainfall statistics for all the durations , which can results in a strong hypothesis especially for regions with very complicated rainfall patterns. The present thesis compares several method for the DDF curves definition and proposes a novel approach, based on the simple scaling-invariance assumption combined with the three parameters DDF formulation in the attempt to overcome the above mentioned shortcomings. The introduction of a split criteria for defining short and long durations events allows the proposed approach to take into consideration the variability of rainfall patterns for different groups of durations, making the DDF curve estimation more accurate without increasing the number of parameters to be estimated/calculated. The methodology was tested on three European rainfall stations: Helsingborg from Sweden, Pavia from Italy, and Frauenwald from Germany. It emerged that the proposed approach outperforms traditional and simple scaling methods for the DDF definition and open to new considerations about accurate design rainfall estimation.