Performance of a multi-parameter distribution in the estimation of extreme rainfall in tropical monsoon climate conditions

Abstract

The changes of rainfall extremes have been recognized all over the world including the tropical monsoon regions like Bangladesh in recent years. Randomized changes from sudden low-flow year to flood year are quite common in many places of Bangladesh. Modeling extreme rainfall is thus a challenge in order to assess hydrologic risk, such as the flood risk. To appropriately account the inter-annual variability of rainfall extremes, the use of a multi-parameter distribution, namely the four-parameter kappa distribution, is examined in Bangladesh condition. The distribution is assessed against the commonly used extreme value distributions using the standard goodness-of-fit tests such as the Chi-square ${(\chi }^{2})$ and the Anderson–Darling test. The evaluation of parameters and the appraisal of estimated quantiles including its’ spatial pattern are also carried out. Annual maximum daily rainfall data from 34 gauging stations were used for the assessment. The kappa distribution is found superior to the currently practiced extreme value distributions. The quantile estimates by the kappa are increased significantly when compared to design estimates from the extreme value I distribution. The expected values of the shape parameters are also indicated for a wide use of the distribution. The effective application of the kappa distribution is expected to pave an alternate way of estimating extreme rainfall for countries where the inter-annual variation of extremes is quite high.