TESTING THE PERFORMANCE OF PROBABILITY FUNCTIONS WITH LONG TERM DATA OF EXTREME RAINFALL

Shreyas, H.C, Mysooru R. Yadupathi Putty

Abstract: Frequency analysis of data on flood and extreme rainfall, for estimation of the peak flood magnitude, is among the most commonly adopted design procedures in Hydrology. The best fitting among the different available Probability functions, selected based on analysis of recorded data of 30 – 40 years, is used in estimation of design values of different return periods like 50, 75 and 100 years. The present study has been carried out to test the performance of the ‘best fitting’ functions in estimating design values of extreme daily, 2- daily and 4-daily rainfall in the region of Western Ghats in Karnataka, for which rainfall records of over a century is available. The work has been carried out by analyzing 40 years’ data from four stations, and by adopting the Normal, Log-Normal, the Gumbel and the Log-Gumbel distributions. Commonly used tests of goodness of fit have been used to select the best fitting function. Peak values of 50, 75 and 100 years are estimated and compared with the observed values. The results are discussed and inferences drawn.

Keywords: Extreme events, Probability functions, Peak rainfall estimation, Reliability of prediction

DOI: https://doi.org/10.15623/ijret.2017.0614010