Modeling of daily rainfall sequence and extreme values in the east Azerbaijan province

In this study using 4 synoptic stations of semi-arid extra cold climate of east Azerbaijan province were used in order to modeling extreme values and occurrence rainfall. To this aim, a stochastic rainfall time series generation consisting of first, second and third-order Markov models and the generalized Pareto and Exponential distribution density functions were used for reproducing amount rainfall. Also, the Exponential- generalized Pareto density function was used to improve the estimation of extreme values. The proposed model essentially was a piecewise distribution approach created by parametrically modeling the tails (i.e. above a threshold) of the distribution using a generalized Pareto, , and the rest Exponential density estimation methods. The results Based on the AIC criterion indicated that the first-order Markov performs relatively better than another model for daily rainfall occurrence. The average of preference first- order Markov chain compared with second and third order was 79 and 66% for all study stations, respectively. Also, results from RMSE showed that Exponential- generalized Pareto probability density performs better to reproduce extreme daily rainfall comparing another distribution. The RMSE criterion is varying between 0.0015 to 0.0017 for Piecewise Exponential-generalized Pareto distribution to estimate extreme daily rainfall daily rainfall.