Efficient and systematic evaluation of the global hydrological model WaterGAP against multiple types of observation data

Global hydrological models (GHMs) play a crucial role in understanding Earth's water resources. To evaluate the strengths and limitations of these models, and how their performance changes with parameter modification or calibration, model outputs are compared against observational data. Traditionally, hydrological models have been evaluated against in-situ streamflow observations. However, this can lead to incomplete assessments of models because models might simulate streamflow well while failing in other aspects, such as overall terrestrial water storage anomaly (TWSA) or the dynamics of specific storage compartments. Nowadays, geodetic and remote sensing data (time series) have become available and are suitable for model evaluation in addition to streamflow. This study takes a comprehensive look at the GHM WaterGAP2.2e by extending the evaluation beyond streamflow observations by integrating different GRAC TWSA products, snow cover fraction, and the dynamics of lake and reservoir surface areas, and the corresponding storage anomalies. The multi-variable evaluation approach is particularly valuable in identifying areas where the model might need improvement. As an example, by comparing the model against GRACE TWSA and streamflow observation, we can test the effect of increasing water storage capacity in soils or decreasing the groundwater discharge coefficient. These parameters govern the flow from groundwater to surface water bodies, offering viable options to address, for example, underestimation or overestimation of the temporal variability of GRACE TWSA when using models like WaterGAP. Evaluating the snow cover fraction model output against observed data improves the model’s ability to simulate snowpack dynamics, a crucial element for estimating seasonal water supply in areas that depend on snowfall. Furthermore, comparing the model output with observed surface areas and storage anomalies of lakes and artificial reservoirs helps to improve, for example, the estimation of surface water use and the simulation of reservoir management. Ultimately, the multi-variable evaluation approach could pave the way for creating models better suited to address the complex questions in global water research.