Selection of the best fit probability distribution for annual maximum daily rainfall data for northwest Bangladesh

Design rainfall is an important input to hydrological analysis. This is generally expressed in the form of a three-dimensional curve known as intensity-duration-frequency (IDF) curves. Many countries like Australia have developed national IDF portal, which allows extraction of IDF data at any arbitrary location of the country. In case of Bangladesh such national IDF database is non-existent although there have been few studies at different locations of Bangladesh to derive local IDF curves. This study focuses on finding the best fit probability distribution function for estimating design rainfall at six meteorological stations in northwest Bangladesh. The performance of eighteen probability distributions is evaluated to find the best fit distribution for annual maximum daily rainfall data. The stations analysed include Dinajpur, Saidpur, Rangpur, Bogra, Rajshahi, and Ishurdi, covering a data period from 1965 to 2022. Two goodness-of-fit tests, the Kolmogorov-Smirnov (KS) and Anderson-Darling (AD) tests, were employed to determine the most suitable probability distribution for each station. The Lognormal distribution emerged as the most appropriate probability model for most of the stations according to the KS test, while the Gamma (3P) and Fréchet distribution was identified as the best fit distribution for the most stations based on the AD test. Further research is being conducted to cover a greater number of meteorological stations across the whole of Bangladesh. The results of this analysis will be used in deriving national design rainfall for Bangladesh. The paper also presents learning methods of statistical hydrological tools utilised in this study by the first author.