Regional Frequency Analysis of Extreme Rainfall Using L-moments and Partial L-moments in Haryana
Loading...
Date
2021-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
CCSHAU, Hisar
Abstract
Regional frequency analysis (RFA) is of great importance for planning and designing hydraulic structures by policymakers and structural engineers. In this study, we focus on regional frequency analysis of daily and monthly extreme rainfall from 1970–2017 at 27 rain gauge stations of Haryana (India) using L-moments and PL-moments. Based on mean monthly rainfall, these 27 rain gauge stations were grouped into three homogeneous regions (Region-I, Region-II, and Region-III) using Ward‟s method of cluster analysis and homogeneity of each region was tested using heterogeneity measures (H). The best fit regional distribution was selected for each region from the five candidate distributions i.e. GEV, GNO, GLO, GPA, and PE3 using the -statistic and L-and PL-moments ratio diagrams. For maximum monthly rainfall, using L-moments method, it was found that GNO was best-fitted for Region-I and Region-II while PE3 for Region-III. For maximum daily rainfall, for Region-I, Region-II and Region-III; PE3, GEV, and GLO was the best-fitted distribution, respectively. Using PL-moments method, for Region-I, for maximum monthly rainfall, GNO was best fitted. For Region-II, GEV was best fitted and PE3 for Region-III. Quantiles for various return periods were estimated using these best-fitted distributions for each region. The performance of both methods i.e. L-moments and PL-moments in quantiles estimation were studied by Monte Carlo simulations. From these simulations, accuracy measures such as relative RMSE and absolute relative bias were calculated and it was observed that these accuracy measures were smaller in the case of PL-moments as compared to L-moments. Also, quantiles were estimated using the regional and at-site base approach. The performance of regional and at-site based rainfall quantiles was studied in terms of relative RMSE. It was observed that regional analysis provided better estimates of the quantiles compared to the at-site based estimation.