Improving hydrological modeling in the Tigris–Euphrates River Basin through water-use adjustments and representation of water transfers

Human intervention directly affects the terrestrial water cycle by altering river flow regimes through water abstractions, artificial reservoirs, and water transfers. Incorporating human impacts into hydrological modeling is not straightforward. In recent decades, several global hydrological models (GHMs), including WaterGAP (Water Global Assessment and Prognosis), PCR-GLOBWB (PCRaster Global Water Balance), and H08, have incorporated representations of human interventions such as water use and reservoir operations. In addition, WaterGAP accounts for water transfers between adjacent grid cells, while H08 represents long-distance aqueduct water transfers at 55 locations worldwide, none of which are located in the Tigris–Euphrates River Basin (TERB). Despite these advances, model performance remains limited in heavily modified basins such as the TERB. This limitation is mainly due to the lack of high-quality water-use estimates and incomplete representation of alterations to the natural system, in particular, the location and the flow rates of artificial water transfers. In this study with the GHM WaterGAP, we assess the importance of explicitly representing such interventions, focusing on the upstream Euphrates River, primarily within Syria. A comparison of observed mean annual streamflow at Atatürk Dam, 823.7 m³/s for the period 1992-2011, and the Syrian-Iraqi border, 535.90 m³/s for the period 2015-2020, reveals a water loss of 287.8 m³/s between the two streamflow gauging stations. In contrast, WaterGAP simulates a smaller water loss of 82.93 m³/s with a mean annual streamflow of 638.23 m³/s at the Atatürk Dam for the period 1992-2011 and 555.3 m³/s at the Syrian-Iraqi border for the period 2015-2020. Although the upstream-downstream water loss is partially represented by WaterGAP, the remaining discrepancies between the observed and simulated streamflow losses cannot be corrected through a basin-wide uniform calibration of model parameters alone but require an adjustment of simulated water abstractions. To address this, we used the FAO (Food and Agriculture Organization) water use dataset, which suggests that water abstractions in Syria are nearly twice as high as those represented in WaterGAP; accordingly, we doubled the water abstractions in Syria. Moreover, analysis of Google Earth imagery revealed a water transfer from the Assad reservoir, located on the Euphrates River in Syria, to areas outside TERB. Based on this observation, we identified the corresponding grid cells using Google Earth imagery and implemented a demand-based water transfer from the Assad reservoir to these external areas. We assumed that whenever water demand occurred in these grid cells, it was supplied by the Assad reservoir. By doubling the water abstractions in Syria and incorporating the demand-based water transfer from the Assad reservoir to adjacent areas, WaterGAP successfully simulates a mean annual water loss of 242.33 m³/s, which is close to the observed value. These findings highlight the necessity of adjusting the simulation of human impacts in heavily modified basins as a prerequisite for a meaningful calibration of model parameters