EXTREME GRADIENT BOOSTING -DRIVEN HYDROLOGICAL FORECASTING STREAMFLOW UNDER CMIP6 SCENARİOS

Accurate streamflow forecasting is essential for sustainable water resource planning under increasing climate variability. This study investigates the performance of the Extreme Gradient Boosting (XGBoost) algorithm for monthly streamflow prediction in the Ceyhan River Basin, utilizing climate projections from General Circulation Models (GCMs) participating in CMIP6. The model was tested across four Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5), using historical discharge data and climate predictors such as mean sea level pressure (mslp) and specific humidity (shum) as inputs. Evaluation metrics included RMSE, MAE, and the Nash–Sutcliffe Efficiency (NSE). The results show that the XGBoost model dynamically adapts to seasonal patterns and selects the best-performing configuration each month. Under high-emission scenarios, particularly SSP5-8.5, streamflow variability significantly increased.SHAP (SHapley Additive exPlanations) analysis enhanced model interpretability, identifying shum and mslp as key predictive features. These findings highlight the potential of XGBoost as a robust and explainable decision-support tool for climate-informed hydrological modeling.