APPLICATION OF POISSON PROCESS TO DROUGHT PREDICTION – THE CASE STUDY OF YUCHENG CITY

Abstract. Open-source R language can implement quantitative research using the flexibility, adaptability and simplicity of bayesian inference models. Counts of drought which are regarded as the realizations of drought “event” in a Poisson process follows an Gamma distribution (prior distribution) in the case study of Yucheng city with the support of R language. That is, the annual drought counts in Yucheng City during the 10 years from 1974 to 1983 is regarded as the 10 realizations of the Poisson process, and the drought counts in Yucheng city during the 10 years is described by the Poisson distribution. The Gamma (1,1) of theta is used as the prior distribution, and its posterior distribution is calculated to be Gamma (60,11). The sample value of theta is obtained by random sampling from the posterior distribution of theta, and a Poisson sample is randomly generated for each value as the predicted value. The posterior mean value of theta sampling results, the mean and quantiles of the predicted value, and the probability estimation of theta and the predicted value are calculated to obtain the highest density region of theta. Studies have shown that Poisson model of drought prediction in Yucheng city has more flexibility, adaptability and simplicity and can provide posterior mean, highest-posterior density of the data, and the drought prediction of Yucheng city is a feasible method on basis of the stochastic Poisson process, thereby providing better reference value for drought reduction.