A controlled model experiment for the global hydrological model WaterGAP3: Understanding recent and new advances in the model structure

The global hydrological model WaterGAP3 has a long tradition that involves increased complexity. For the first time, this study conducts a model experiment to test whether this increased complexity leads to better process representation in three individual model parts. First, the findings of our large-sample analysis reveal that the recently implemented volume-dependent river routing does not produce better process representation in about 2/3 of the basins, lacking justification. Second, the reservoir algorithms demonstrate the benefit of using reservoir-specific data. However, the approaches show individual weaknesses, indicating the need for future development. Third, implementing snow processes on wetlands is promising. However, future work is needed to enable sublimation and to address overestimated discharges during winter periods. Overall, this study demonstrates that revisiting past decisions is essential for scientific progress and shows the efficacy of model falsification through large-sample analysis. • Revisiting past decisions is beneficial for models with a long development history. • Hypothesis testing enables model falsification, supporting development decisions. • Increased accuracy does not necessarily equate to higher adequacy. • Volume-dependent river routing lacks justification and needs revision.