How does the mixing assumption influence the distributed tracer-aided hydrological model?

Abstract

Hydrological simulations bear large uncertainties in mountainous basins due to complex water sources and runoff pathways. Tracer-aided modelling is useful for improving hydrological simulation and reducing uncertainties. The mixing assumption is an important issue for tracer-aided hydrological models. Although numerous studies have shown that partial mixing phenomena are common in small-scale catchments and groundwater storages, no research has yet adopted the partial mixing assumption in distributed tracer-aided hydrological models, and the impact of the mixing assumption on distributed models has not been fully understood. This study developed a two-reservoir method to simulate the partial mixing processes in the groundwater storage for the tracer-aided hydrological model THREW-T. Adopting the model in a typical large mountainous basin on the Tibetan Plateau, we analyzed the influence of mixing assumption on the estimated contribution of subsurface runoff. Results showed that: (1) The model with the partial mixing assumption (THREW-PM) can effectively simulate the isotopic variation of subsurface runoff. The calibrated parameters related to partial mixing indicated a small active storage controlling hydrological response and (∼0.05) a long time required for complete mixing between active and passive storages (∼100 days). (2) The mixing assumption and isotope data had an important influence on the subsurface runoff estimated by the THREW-T model. Adopting the partial mixing assumption and calibrating the model toward the subsurface runoff isotope resulted in a 10 % difference in the estimated contribution. (3) The mixing assumption affected the models by influencing the isotope variation of subsurface runoff. As the temporal variation in the simulated isotope composition of subsurface runoff increased, the estimated contribution of subsurface runoff also increased. This study conducted the first analysis on how mixing assumptions influence distributed tracer-aided hydrological models, and highlighted the importance of tracer data from various water bodies in verifying the simulations of the tracer composition in runoff components, and, the consequently the separation among them.