ANALYSIS OF EXTREME RAINFALL EVENT IN THE METEOROLOGICAL SUBDIVISION OF UTTAR PRADESH USING THREE-PARAMETER EXTREME VALUE DISTRIBUTIONS

Abstract

Extreme rainfall is a global phenomenon occurring in almost every major country of the world that cause significant damage such as floods and erosion that can destroy infrastructure, human and animal life, disruptive economic activities, and related development. The forecasts of heavy rainfall help to implement strategies, and measures before they occur. In this study, we used statistics strategies to create models that could work to predict maximum rainfall in Uttar Pradesh. For this purpose, the annual maximum rainfall from 1979-2018 applies in the subdivision of Uttar Pradesh. Extreme value distribution GEV, GLO, GPA, and UEV considered analyzing extreme events. The parameters of the 1st three distribution are determined using the method of moment (MOM) and probability-weighted L- moment (PWM-L). The last distribution parameters are determined using the simple objective method (SO). Two methods used to analyze the best fit distribution among four distribution, i.e., graphical method (coefficient of determination, R2 ), and goodness of fit test (GOF). Five different GOF tests apply in this study, i.e., RRMSE, RMSE, MAE, MADI, and PPCC. Basic time series analysis such as outliers test, normality test, homogeneity test, and stationarity tests performed to ensure that the information used is adequate and appropriate. The results obtained indicate that the GEV (PWM-L) was an appropriate method for the distribution of the annual maximum rainfall series in the west Uttar Pradesh subdivision and the GLO (PWML) which was an appropriate distribution method to analyze the series of the East Uttar Pradesh subdivision. The coefficient of determination (R2 ) for the observed versus predicted rainfall based on the best fit model observed to be 0.9899 and 0.9865, respectively, West Uttar Pradesh subdivision and East Uttar Pradesh subdivision. Finally, the most appropriate distribution at each site applies to predict maximum rainfall at different return periods.