Mixed Zero-Inflation Method and Probability Distribution in Fitting Daily Rainfall Data
In hydrological processes, rainfall is one of the important components of water supply for human life. We considered how well the statistical distribution simulates rainfall intensity. We propose an asymmetric statistical probability distribution joined by zero-inflated to fit the daily continuous record of rainfall data in Thailand. The candidate statistical probabilities are General Pareto, Exponential, Beta, Gamma, Generalize extreme value, Extreme Value, Normal, Lognormal, Weibull and Rayleigh distribution. The daily data from 123 rain gauges in Thailand collected and removed in a given year, using the null value. The statistical distributions estimated on the statistical coefficient, using the maximum likelihood estimation (MLE) method and resulted in a cumulative density function (CDF). The CDF compared to the CDF of observed data that estimated, using Kaplan-Meier algorithm. The comparisons were evaluated by Goodness of fit (GOF) in 3 null hypothesis tests (Kolmogorov-Smirnov, Anderson-Darling and Chi-Square test). The best fit distribution was identified by minimum residual (R) index and maximum correlation (Cor) index based on difference value between the estimated and observed data. The Weibull distribution matched to the 118 rain gauges while 5 rain gauges were best fitted by the Gamma distribution.