Drought forcasting in eastern and central arid and semi-arid regions of Iran using time series and Markov chainmodels

In this research, the target area were regionlized into few distinctive homoginious sub-regions by applying principal component alalysis to the SPI time series at 3-, 6- and 12-months time scales and the resultant PC scores were considered as the regional SPI time series for drought forecasting using time series modellingineach identified sub-region. The probability of occurences of dry, normal and wet events were also predicted for all the considered stations using Markov chain model and the results were spatially mapped and analysed. The expected drught numebr and drught length of the prediceted drought events were also estimated and mapped to spatially display their results in order to ease their spatial variability comparrison. Furthermore, different time series models were fitted to the Regional SPI series (PC scores) to identify the best fitted model for each region in order to use for drought forcasting. The result shows that the ARMA is the best fitted model for SPI time series at 3- and 6-months time scales while for the 12-months time scales the SARIMA model is the best fitted model. Using the identified models the magnitude of the SPI was forcasted for the leading times. The result shows that the time series models can favorably forcast SPI values for three months ahead, wherease the predicted results for more than three months ahead is not reasonably accurate.