Evaluation of Time Series Models in Simulating Different Monthly Scales of Drought Index for Improving Their Forecast Accuracy

Drought is regarded as one of the most intangible and creeping natural disasters, which occurs in almost all climates, its characteristics vary from region to region. The present study aims investigate effect differentiation operations on improving static modeling accuracy drought index time series after selecting best selected model, evaluate severity duration, well predict future behavior, Semnan city. During this process, different monthly scales was analyzed, differencing approach stationarity improvement prediction models. First, data related a one-month investigated. By using seasonal, non-seasonal, hybrid differencing, new are created examine these through analyzing ACF diagram generalized Dickey–Fuller test. Based results, indicates degree stability. Then, type number states required models determined, finally, model by applying assessment criteria. In following, same stages analyzed for derived 6-month rainfall data. results reveal that SARIMA (2,0,2) (1,1,1) 6 with calibration criteria MAE = 0.510, RMSE 0.752, R 0.218 seasonal series. addition identifying introducing six-month (SARIMA (3,0,5) 0.430, 0.588, 0.812), highlight increased 4 times correlation coefficient section 8 validation section, respectively, relative state. After comparing between reality event, duration were also examined, indicated high agreement. Finally predicted SPI next 24 months created.