Modeling of Daily Rainfall Extremes, Using a Semi-Parametric Pareto Tail Approach

Abstract

Various approaches have been widely proposed for simulation of the occurrence and the amount of daily rainfall. In this study, a piecewise distribution approach has been developed to improve extreme event amount simulation. The gamma-generalized Pareto (GGP) approach employs a combination of the generalized Pareto (GP) and the gamma density estimation method to model the daily rainfall distribution. Furthermore, three stochastic rainfall time series generations have been developed and compared to simulate the daily rainfall occurrence based on the first-order Markov chain (MC1), the second-order Markov chain (MC2) and the third-order Markov chain (MC3) approaches. 30 years daily datasets from 5 synoptic stations have been used in the semi-arid extra cold in Iran. The performance of different approaches has been compared using Akaike information climate criterion (AIC) and root mean square error (RMSE). Results show that MC1 performs relatively better than MC2 and MC3 for daily rainfall occurrence modeling. Results also show that the GGP probability density performs better to reproduce extreme daily rainfall compared with gamma, GP and exponential distributions.