RAINFALL TIME SERIES ANALYSIS IN SEMIARID REGION OF DUHOK GOVERNORATE, IRAQI KURDISTAN REGION

This study was aimed to investigate data homogeneity, detect trend and changes in historical rainfall and forecast future rainfall pattern. Rainfall historical records taken from 22 gauging and non-gauging stations distributed all over Duhok governorate. The dataset includes long time series of monthly and annual rainfall totals (mm) from 1998 to 2020. To detect data variability, four different homogeneity tests were used. To analyze trends and sequential shifts in historical data, parametric and non-parametric tests were applied. Theil-Sen's slope estimator test was applied to calculate the magnitude of change over time. Rainfall forecasting was based on the Box-Jenkins methodology. The homogeneity test results revealed that the majority of the monthly and annual rainfall series were labeled as useful. Furthermore, the annual rainfall at most of the study stations presented positive trend with sen’s slope ranged between 0.545 -43.03 mm yr-1. At five stations tended to beyond the upper limit of the CUSUM chart during 2019 and 2020. Conversely, it tended to be below the lower limit at 3 stations during the period from 2000 t0 2002. ARIMA (0, 1,1) was the best model for predicting yearly rainfall for 75% of the stations, while the model (1,1,1)(1, 0, 1)12 was the best suited model for predicting and forecasting monthly rainfall for more than 83% of the selected stations.