Statistical Modeling And Return Levels Of Extreme Rainfall Events In Flood Prone Area Of Bangladesh

ABSTRACT: The connection between lives and water is well-known where rain is the important source of water for the plants and animals on earth. However, excessive rainfall raises water level and sometimes results serious floods, flash floods, river/sea flooding etc. In this regard, efficient planning considering characteristics of heavy rainfall is essential to protect lives. This article intended to determine the best-fit probability distributions of annual maximum rainfall from eight flood prone areas in Bangladesh. Consequently, the Lognormal, Generalized extreme value distribution and Inverse-Gaussian distribution were considered and maximum likelihood methods was utilizing to estimate model parameters. The power of the fitted model assessed through some goodness of fit test such as KolmogorovSmirnov statistic, Cramér–von Mises criterion and Anderson-Darling statistic. Furthermore, some of the graphical measures such as QQ plot, PP plot and density plots examined the competence of fitted model. The lognormal and Inverse-Gaussian distribution had similar inference, resulted best fitted distribution for almost all station. The return levels of 10-year, 20-year, 50-year and 100-year return period estimated from the best fitted models of this study were essential measures in applied point of view. Hence, protective measures for higher return levels areas may save from the adverse loss of agriculture, ecosystems, bio diversities and livelihood. 

Keywords: Return level, probability distribution, GEV, trend, flood