期刊名称：Pakistan Journal of Statistics and Operation Research
印刷版ISSN：2220-5810
出版年度：2013
卷号：9
期号：1
页码：111-136
DOI：10.1234/pjsor.v9i1.461
语种：English
出版社：College of Statistical and Actuarial Sciences
摘要：The estimation of magnitude and frequency of extreme rainfall has immense importance to make decisions about hydraulic structures like spillways, dikes and dams etc The main objective of this study is to get the best fit distributions for annual maximum rainfall data on regional basis in order to estimate the extreme rainfall events (quantiles) for various return periods. This study is carried out using index flood method using L- moments by Hosking and wallis (1997). The study is based on 23 sites of rainfall which are divided into three homogeneous regions. The collective results of L-moment ratio diagram, Z-statistic and AWD values show the GLO, GEV and GNO to be best fit for all three regions and in addition PE3 for region 3. On the basis of relative RMSE, for region 1 and region 2, GLO, GEV and GNO are producing approximately the same relative RMSE for return periods upto 100. While GNO is producing less relative RMSE for large return periods of 500 and 1000. So for large return periods GNO could be best distribution. For region 3 GLO, GEV, GNO and PE3 are having approximately the same relative RMSE for return periods upto 100. While for large return periods of 500 and 1000 PE3 could be best on basis of less relative RMSE.
关键词：L-moments;Regional rainfall frequency analysis;Discordancy measure;Regionalization;Goodness-of-fit measure;Relative RMSE;GNO
