Statistical Analysis and Modelling of Annual Mean Temperature in Gorgan Station (1956-2005)

Statistical methods are useful tools for understanding the behavior of climatic variables. In this study, statistical techniques such as identification of trends, oscillations and fluctuations of temperature are used in order to identify any change in the behavior of Gorgan station during 1956 to 2006. Run test (P<0.5) has shown a random trend to the annual mean temperature. The results of conducted tests such as the Pearson, Spearman and Mann - Kendall and linear regression have shown no significant trend in the time series of Gorgan. To predict the duration and sequence of years with higher temperatures than the annual mean temperatures (temperature anomalies) Markov chain model was used. According to test on transition matrix of annual mean temperature, no dual-or multi state were accepted. Moreover, a spectral analysis techniques was conducted in order to identify the fluctuations in temperature, in which no significant trend was observed. Furthermore, the 95% confidence level in spectral analysis technique showed a significant cycle of 20 with short-term returns period of 2.5 years. Studies carried out by many researchers indicate that the mentioned cycle is in relation with oscillation periods of the quasi-biennial oscillation (QBO) in other parts of the world. ARIMA models were used to predict the behavior of the annual mean temperature in Gorgan. In this method, a number of three patterns were processed as the initial patterns. AIC test has shown that the third ARIMA model, M3 (0,1,1) is a better choice for temperature prediction. Based on this predictions of temperature for next 10 years (2015- 2006) with 95% confidence level were performed.