Ephemeral rivers commonly occur in regions with a shortage of water resources, and their channel configuration tends to change substantially owing to cultivation, tree planting and sand extraction. There is an urgent need to restore degraded river ecosystems. During short-term water conveyance, water storage in sand pits and leakage in dry riverbeds retards the flow of water, which is detrimental for ecological restoration of the riparian zone. A coupled dynamic leakage loss and flood routing model was established to predict the flow processes in the complex river channel of the Yongding River in China. The model mainly included three sub-models of flow dynamics, dynamic leakage loss, and water balance along multiple cross sections of the river channel. The complex section is reflected in the different infiltration properties for each section, and the existence of sand pits. The water head was dominated by flow velocity and the overflow from sand pits. Owing to the difference in landforms and the deposited sediment size of the riverbed bottom, the river channel was divided into 11 cross sections and a sand pit to ascertain the respective infiltration or leakage loss processes. The input parameters of the model came from field surveys of sand pits, river geometry and hydrogeology. The model was also calibrated and validated using monitoring data from ecological water releases into the Yongding River in 2019 and 2020. This coupled model can predict the water leakage loss and flow process of the water head and also provide important guidance for river reconstruction and ecological restoration.

Particle selectivity plays an important role in clarifying sediment transport processes in vegetative filter strips (VFS). 10-m long grass strips at slopes of 5○ and 15○were subjected to a series of silt-laden inflows experiments with different particle sizes to investigate the sediment transport and its response to overland flow hydraulics. The inflow sediments came from local soil, river-bed sand, and mixed, with median particle size d50 of 39.9, 207.9, and 77.4 μm, respectively. Three independent repeated experiments were carried for each treatment. The results show that when the sediment trapping lasted for a certain length of time, the re-entrainment of some small-sized particles was greater than the deposition; that is, negative deposition occurred, which was not erosion of the original soil. Negative deposition of particles is mainly determined by the particle diameter. The coarser the inflow sediment particles and/or the steeper the slope, the coarser the particles can be negatively deposited. Deposited sediment causes the VFS bed surface to become smooth and hydraulic resistance decrease exponentially. Stream power P is more suitable than shear stress τ of overland flow to be used to describe the process of sediment particle transport in VFS. The relationship between P and d50 of outflow sediment is very consistent with the form of power function with a constant term. These results are helpful to understand the physical process of sediment transport on vegetation hillslopes.