Rainfall frequency analysis and isopluvial mapping for Bangladesh using best fitted probability distribution function

Abstract

Frequency analysis is an important tool to find the most suitable method that can be used to anticipate the magnitude of extreme events of some climatic phenomena such as rainfall, floods, etc. Rainfall frequency analyses are used extensively for designing engineering works that control storm runoff including agricultural drainage system, highway and railway culverts and most importantly municipal storm sewer systems. Establishing a good probability distribution method for a particular region that provides a good estimate to daily rainfall magnitude is an important study and is imperative for different projects related to water resources. A reliable estimate of probability distribution and trend analysis is a valuable guide for practitioners in determining return period and risk analysis of flood, drought, cyclones and other extreme climate events for the design of infrastructures. In this study, the best-fitted probability distribution function for annual maximum daily rainfall of Bangladesh have been determined using rainfall data of 35 stations from 1948-2015. Rainfall data have been checked for trend, homogeneity and randomness first. In the Mann-Kendall trend test, it is found that 5 of the total 35 stations have significant trends which have been removed from these stations for further analysis. The rainfall data of Jessore is found to be non-homogeneous in both the Standard Normal Homogeneity test and the Pettitt test and therefore, it has not been considered for frequency analysis. The rainfall data of all the stations are found to be random in randomness test. Seven commonly used probability distribution functions have been used and four goodness of fit tests have been applied to find the best fitted probability distribution function (PDF). The lowest sum of the ranks from the four goodness of fit tests has been taken as the best fitted PDF for a station. The PDF with the lowest rank for a maximum number of stations is considered as the best fitted PDF for Bangladesh. The Log-Normal 3-parameter distribution has been found to be the best fitted PDF for annual maximum rainfall of Bangladesh in this study. This distribution function has been used for calculating the amount of rainfall corresponding to 2.33-year, 5-year, 10-year, 25-year, 50-year and 100-year return periods for the 34 stations. Isopluvial maps have been generated using these rainfall estimates for Bangladesh. These maps can be used for designing engineering works that control storm runoff. The best fitted PDF has been assessed for the different hydrologic zones of Bangladesh as well. In this analysis, Log-Normal 3-parameter distribution has been found as the best fitted for the North-West (NW), North-Central (NC), South-West (SW), South-East (SE) and the Eastern Hills (EH) zones, where Log Pearson type III distribution was found as the best for the North-East (NE) and the River & Estuary (R&E) region. Pearson type III distribution was found as the best fitted for the South-Central (SC) zone.