Determination of best fit probability distribution for prediction of the rainfall of the rice-growing season in the main rice growing areas of the country

Despite the water supply, rainfall has multiple and conflicting roles during the rice cultivation period and choosing the appropriate distribution of the probability of its occurrence is an important step in planning water resources management and adjusting the planting calendar and reducing damage in rice farming. In this research, to determine the most appropriate distribution of rainfall probability during the rice-growing season, from the data of eight synoptic stations of the southern coast of the Caspian Sea, including the stations of Astara, Bandar Anzali, Rasht, Ramsar, Babolsar, Qarakhil, Nowshahr and Gorgan with a statistical period of 30 the year (1991-2020) was used. After quality control and homogenizations, Bernoulli-log-normal, Bernoulli- Weibull and Bernoulli-gamma distributions were fitted to the rainfall data in the daily time range (in windows of three days without overlap) as well as the length of the rice-growing season. Kolmogorov-Smirnov goodness of fit test (K-S) and Akaike's index (AIC) were used to identify the most suitable one. The obtained results showed that the Bernoulli-Gamma distribution is the most suitable probability distribution for estimating the rainfall of the rice-growing season in the southern shores of the Caspian Sea. After the Bernoulli-Gamma distribution, the Bernoulli-Weibull distribution showed a better fit, especially for Nowshahr station located in the central part of Mazandaran province. The findings of this research can be used to quantify the amount of expectation and risk caused by rain during the rice-growing season.