LAUSR

Drought forecasting and monitoring play a significant role in reducing the negative effects of global meteorological droughts caused by different intensities at different temporal and spatial scales in different regions, especially in regions with high dependency on rainwater. The present study tries to compare the accuracy of stationary time series (ST) models including autoregressive moving average (ARMA), moving average (MA) and autoregressive (AR) and cyclostationary time series (CT) models including periodic autoregressive moving average (PARMA), periodic moving average (PMA) and periodic autoregressive (PAR) to predict drought index (i.e. monthly reconnaissance drought index (RDI)) in periodic data series considering that CT models are more powerful and efficient than ST models by using data series of 8 synoptic stations with different climate conditions in Iran from 1967 to 2017. According to the results the monthly RDI was significantly periodic in all selected stations. The PAR (25) model was the best fitted CT model in data series at all stations and on the other hand, the following models were the best-fitted ST models in data series: the AR models at Babolsar and Rasht AR (25) and at Gorgan AR (24) and ARMA models at Tehran ARMA (2, 3), at Zahedan and Shiraz ARMA (2, 4) and at Esfahan and Shahre Kord ARMA (2, 5). Based on the best fitted CT and ST models, the results showed that the correlation coefficients (R) between observed and simulated RDI vary from 0.882 to 0.946 and from 0.693 to 0.874, respectively from January 1967 to December 2017. According to the best fitted CT and ST models, the validation test of the best fitted models indicated that the R between observed and simulated RDI vary from 0.634 to 0.883 and 0.585 to 0.847, respectively from January 2012 to December 2017. In total, it can be concluded that that the accuracy and capability of CT models in predicting the RDI were more than those of the ST models at all stations and the hypothesis of the study was confirmed.