Climatic Variables Prediction Using Time Series Analysis of Zohre Watershed

Climatic variables such as discharge, rainfall, temperature and evaporation, have autocorrelation, seasonality, trend and stochastic processes. In present study of time series analyses in prediction of climatic variables in monthly and seasonality scales were used. Zohre and Kheirabad subwatersheds were selected. In the first step, nonparametric tests such as kolmogorov-smirnov and Mann-Kendall tests, for determination of normalization and trend, was applied respectively. The result of Mann Kendall test showed that discharge variable have a significant trend (P<0.05) while, variables of rainfall, temperature and evaporation, have not a significant trend (P >0.05). Also, the rainfall, discharge and evaporation variables have negative trend (Z> 0) and the temperature variable have a positive trend (Z>0). In the next step, difference operator for converting non- static to static data, were used. The techniques of model identication which are most commonly used were propounded originally by Box and Jenkins. Their basic tools were the sample autocorrelation function and the partial autocorrelation function. Minimum and maximum P and Q between 0 and 2 changed, respectively and using of Akaike criteria the most appropriate model was selected. Then using maximum likelihood method, significant parameters confirmed. Results of validation test indicated that temperature and evaporation models have the height accuracy (R=0.98) and Monthly scale is better than seasonally scale.