Statistical methods are one of the useful tools for climatic elements and their behaviors analysis، as well as، modeling and predicting them subtly. In this paper the behavior of time series of annual temperature of zanjan during 1956-2005 have been studied. So، based on autocorrelation، correlation coefficient، and Spearman and Mann - Kendall ways، the data trend was determined. The amount of the annual trend based on the least error squares have been estimated about -0.04◦C، and based on line equation of regression the direction of the modeling of the trend has been shown. Based on spectrum analysis the significant harmonics of temperature through 95 percent of Interval Confidence have been examined. The cycles of 55 year that reflect trend and the cycles of 2.3 - 2.7 years has been shown in temperature of Zanjan. To predicate annual temperature of Zanjan، ARIMA modeling has been used. Two initial models، ARIMA (0,1,1) and ARIMA (1,1,2) have been chosen. With regard to residue testing، as well as AIC criteria Model، the ARIMA (1,1,2)، was found suitable. According to this model the temperature has been predicted for next 20 years. More over Markov chain model were applied on monthly temperature of Zanjan during 1956-2005. Monthly temperature data have followed the two- state Markov chain، so the probability occurrence of a warm month at Zanjan is 0.1607 and probability occurrence of cold months is 0.8393. The return period of warm months is about 6 months and return period of cold months is about one month.