Comparison of Ln2 and Wb2 probability distributions for low-flow frequency analysis

Estimation of low-flows for a desired duration (‘d’ in days) and return period (‘T’ in years) is utmost importance-for the assessment of water resources for many direct and indirect uses viz., municipal, irrigation, hydropower, public water supply, etc. This can be obtained by Low-flow Frequency Analysis (LFA) involving fitting probability distribution to the series of Annual Minimum d-day Average Flows (AMdAF) for different values of ‘d’ such as 7-, 10-, 14- and 30-days, that is derived from the daily stream flow data series. This paper presents a study on LFA for Haora river at Haora gauging site adopting 2-parameter Log Normal (LN2) and Weibull distributions. Parameters of the distributions are determined by method of moments, Maximum Likelihood Method (MLM) and L-Moments, and are used for estimation of low-flows. The adequacy of fitting LN2 and WB2 distributions to the AMdAF series is evaluated by non-parametric Goodness-of-Fit (GoF) test, say, Kolmogorov-Smirnov. Model-performance analysis using Model Performance Indicators (MPIs) viz., correlation coefficient and root mean squared error are applied for the selection of best suitable distribution for estimation of low-flows. The GoF test-results and MPIs values presents the LN2 (using MLM) distribution is better suited for LFA for Haora gauging site.