Ensemble Method for Intervention Analysis to Predict the Water Resources of the Tigris River

Abstract

Iraq faces real challenges and great concerns regarding water resources, including the noticeable decrease in the flow of the Tigris River into Iraqi territory due to various irrigation projects being implemented on the Turkish side, the latest of which was the construction of the Ilisu Dam, which exacerbated the water crisis and placed Iraq before a serious challenge. The flow river data utilized in this paper was the annual revenue of the Tigris River, representing the amount of water entering Iraq at the Turkish border data for the period (Oct-2014–Sep-2021), which equals 84 months. To enhance the accuracy of Tigris River flow forecasting with the ARIMA models as a classical statistical approach and the nonlinear model of eXtreme Gradient Boosting (XGBoost), a Random Tree ensemble model was proposed in this study. Two distinct ARIMA models are employed to capture the linear characteristics of the Tigris River flow: SARIMA and ARIMAX. XGBoost model was utilized to capture the nonlinear characteristics of the Tigris River flow. The results reveal that the Tigris River flow prediction using the Random Tree ensemble model achieves better than the other models introduced in this paper regarding the evaluation measurements. The forecast suggests stabilizing the river flow aligning with the low average river flow level, with variations observed. These seasonal changes reflect the impact of increased river flow during the rainy season in Iraq and Turkey during peak times and reduced river flow in the summer months.