Modeling Intensity‐Duration‐Frequency Curves for the Whole Range of Non‐Zero Precipitation: A Comparison of Models

Abstract

Intensity‐duration‐frequency (IDF) curves are useful in water resources engineering for the planning and design of hydrological structures. As opposed to the common use of only extreme data to build IDF curves, here, we use all the non‐zero rainfall intensities, thereby making efficient use of the available information. We use the extended generalized Pareto distribution to model the distribution of the non‐zero rainfall intensities. We consider three commonly used approaches for building IDF curves. The first approach is based on the scale‐invariance property of rainfall, the second relies on the general IDF formulation of Koutsoyiannis et al. (1998, https://doi.org/10.1016/S0022-1694(98)00097-3 ), and the last approach is purely data‐driven(Overeem et al., 2008, https://doi.org/10.1016/j.jhydrol.2007.09.044 ). Using these three approaches, and some extensions around them, we build a total of 10 models for the IDF curves, and then we compare them in a split‐sampling cross‐validation framework. We consider a total of 81 stations at 10 min resolution in Switzerland. Due to the marked seasonality of rainfall in the study area, we performed a seasonal‐based analysis. The results reveal the model based on the data‐driven approach as the best model. It is able to correctly model the observed intensities across duration while being reliable and robust. It is also able to reproduce the space and time variability of extreme rainfall across Switzerland. Intensity‐duration‐frequency (IDF) curves are used in water resources engineering for the safe design of hydraulic structures. They summarize the relationship between rainfall intensity, duration, and frequency of occurrence. IDF curves are commonly built using probability distributions that use only extreme data, which can be a drawback in terms of the efficient use of already scarce data. In this study, we use a distribution, the extended generalized Pareto distribution, that uses all the non‐zero rainfall intensities to model IDF curves. We build a total of 10 models, which are based on three commonly used approaches and some extensions around them. We compare them in terms of their reliability and robustness. The best IDF model is able to reproduce the known variability of extreme rainfall across Switzerland. We build intensity duration frequency (IDF) curves using all the non‐zero precipitation data and model the intensities with the extended generalized Pareto distribution We consider three approaches to building IDF curves: scale invariance, IDF formulation of Koutsoyiannis et al. (1998), https://doi.org/10.1016/S0022-1694(98)00097-3 and data‐driven method We compare and select the best model in a cross‐validation framework