The Observed Isotopic Errors in Extreme Precipitation Leads to Overestimation of Long-Term Soil-Streamflow Hydrological Connectivity

Abstract

The tracer-aided modelling has gained widespread attention in recent years as a crucial approach for investigating watershed hydrological functions. However, errors in model inputs, such as precipitation, evapotranspiration (ET) and isotopes in precipitation, can lead to uncertainty in physically meaningful model parameters, which, in turn, affects the accurate depiction of watershed hydrological functions. In this study, we focused on the Hemuqiao watershed, a typical humid mountainous region in southeast China, equipped with intensive isotopic and hydrological monitoring. The Two Reservoirs StorAge Selection (TRSAS) model was adopted to explore the impact of input data in understanding watershed hydrological connectivity and preferential flow. The results show that observation errors in precipitation and ET data do not significantly reduce model performance (with the optimal NSE value decreasing by up to 0.05). However, they do increase the uncertainty of simulation parameters, primarily due to errors associated with large precipitation and evapotranspiration events. In contrast, observed isotopic errors in precipitation, particularly during extreme precipitation events, reduce model performance and meanwhile lead to significant differences in some simulation parameters compared to no error data. Although the proportion of young water fraction in the streamflow does not show a noticeable difference, the proportions of lateral subsurface flow and young water fraction in lateral subsurface flow tend to be overestimated (i.e., by approximately 0.14 and 0.08, respectively, on average over the long term). This results in an overestimation of lateral preferential flow and hydrological connectivity between soil and streamflow. These findings suggest that in tracer-aided models, improving the observation accuracy of isotopes in extreme precipitation is more critical for accurately understanding watershed hydrological processes than enhancing spatial observations of precipitation and ET.