Identification of Suitable Hydrologic Response Unit Thresholds for Soil and Water Assessment Tool Streamflow Modelling

Use of a non-zero hydrologic response unit(HRU) threshold is an effective way of reducing unmanageable HRU numbers and simplifying computational cost in the Soil and Water Assessment Tool(SWAT) hydrologic modelling. However, being less representative of watershed heterogeneity and increasing the level of model output uncertainty are inevitable when minor HRU combinations are disproportionately eliminated. This study examined 20 scenarios by running the model with various HRU threshold settings to understand the mechanism of HRU threshold effects on watershed representation as well as streamflow predictions and identify the appropriate HRU thresholds. Findings show that HRU numbers decrease sharply with increasing HRU thresholds. Among different HRU threshold scenarios, the composition of land-use, soil, and slope all contribute to notable variations which are directly related to the model input parameters and consequently affect the streamflow predictions. Results indicate that saturated hydraulic conductivity, average slope of the HRU, and curve number are the three key factors affecting stream discharge when changing the HRU thresholds. It is also found that HRU thresholds have little effect on monthly model performance, while evaluation statistics for daily discharges are more sensitive than monthly results. For daily streamflow predictions, thresholds of 5%/5%/5%(land-use/soil/slope) are the optimum HRU threshold level for the watershed to allow full consideration of model accuracy and efficiency in the present work. Besides, the results provide strategies for selecting appropriate HRU thresholds based on the modelling goal.

Trend detection and stochastic simulation prediction of streamflow at Yingluoxia hydrological station, Heihe River Basin, China

Investigating long-term variation and prediction of streamflow are critical to regional water resource management and planning. Under the continuous influence of climate change and human activity, the trends of hydrologic time series are nonstationary, and consequently the established methods for hydrological frequency analysis are no longer applicable. Five methods,including the linear regression, nonlinear regression,change point analysis, wavelet analysis and HilbertHuang transformation, were first selected to detect and identify the deterministic and stochastic components of streamflow. The results indicated there was a significant long-term increasing trend. To test the applicability of these five methods, a comprehensive weighted index was then used to assess their performance. This index showed that the linear regression was the best method. Secondly,using the normality test for stochastic components separated by the linear regression method, a normal distribution requirement was satisfied. Next, the Monte Carlo stochastic simulation technique was used to simulate these stochastic components with normal distribution, and thus a new ensemble hydrological time series was obtained by combining the corresponding deterministic components. Finally, according to these outcomes, the streamflow at different frequencies in 2020 was predicted.

大尺度跨境流域径流预测的迁移学习框架--敏感性分析及在数据稀缺流域的适用性

The imbalance in global streamflow gauge distribution and regional data scarcity,especially in large transboundary basins,challenge regional water resource management.Effectively utilizing these limited data to construct reliable models is of crucial practical importance.This study employs a transfer learning(TL)framework to simulate daily streamflow in the Dulong-lrrawaddy River Basin(DIRB),a less-studied transboundary basin shared by Myanmar,China,and India.Our results show that TL significantly improves streamflow predictions:the optimal TL model achieves an average Nash-Sutcliffe efficiency of 0.872,showing a marked improvement in the Hkamti sub-basin.Despite data scarcity,TL achieves a mean NSE of 0.817,surpassing the 0.655 of the process-based model MIKE SHE.Additionally,our study reveals the importance of source model selection in TL,as different parts of the flow are affected by the diversity and similarity of data in the source model.Deep learning models,particularly TL,exhibit complex sensitivities to meteorological inputs,more accurately capturing non-linear relationships among multiple variables than the process-based model.Integrated gradients(IG)analysis furtherillustrates TL's ability to capture spatial het-erogeneity in upstream and downstream sub-basins and its adeptness in characterizing different flow regimes.This study underscores the potential of TL in enhancing the understanding of hydrological processes in large-scale catchments and highlights its value for water resource management in transboundary basins under data scarcity.