Generalized distributions for modeling precipitation extremes based on the L moment approach for the Amman Zara Basin, Jordan

Abstract

Extreme event magnitude and frequency are prerequisites for accurately designing and managing various water infrastructure systems. This paper studies precipitation extremes in the Amman Zara Basin (AZB) using daily precipitation records from 24 weather stations during a period that exceeds 50 years. Two extreme precipitation series (the annual maximum (AM) and the peak over threshold (POT)), four generalized probability distributions (generalized extreme value (GEV), generalized Pareto (GP), generalized lognormal (GLN), and generalized logistic (GLO)) and the L moment method for distribution parameter estimation are used. A mix of increasing and decreasing trends is observed at different stations for both the AM and POT series over the study period. Since the POT series considers up to the fifth largest precipitation in some years, in contrast to the AM series, and skips the largest precipitation in the AM series in other years, the trend analysis results for the POT series differ slightly from those for the AM series. According to the goodness-of-fit (Kolmogorov-Smirnov test and L moment diagram), the probability distributions GEV, GLN, and GLO can better fit the AM series, with no unique distribution among them consistently ranking the best for all stations, while the POT series is better fitted by the GP distribution. The calculated extreme precipitation amounts of the AM series are up to 18% greater than those of the POT series for the same return period. Additionally, the AM series can describe extreme precipitation events better than the POT series based on relative error calculations. The 50- and 100-year extreme precipitation events occurred more frequently in recent years.