To explore the feasibility of introducing remote sensing soil moisture data to assist in distributed hydrological model parameter calibration in semi-arid small watersheds, a joint parameter calibration method that integrated hard data (flow rate) and soft data (remote sensing soil moisture) was proposed. CLDAS satellite remote sensing soil moisture data were applied to the Grid-Multi-GA model. A multi-objective optimization framework was adopted, and the Nash efficiency coefficient of flow rate simulation and the Spearman correlation coefficient of soil moisture's spatiotemporal distribution were used as dual evaluation indicators. By adjusting the weights of the system to dynamically balance two types of indicators, the optimal weight and its corresponding runoff generation and flow routing parameter combination were ultimately determined. To validate the feasibility of the joint calibration method, three model scenarios were established, namely uncalibrated parameters, calibration using only discharge, and joint calibration of discharge and soil moisture. The case study results in the Ningxia Yuanzhou Watershed demonstrate that the Grid-Multi-GA model jointly calibrated by flow rate and soil moisture achieves a Nash efficiency coefficient greater than 0.7 and a Spearman correlation coefficient of 0.84 in flood simulation for small watersheds, significantly outperforming both the uncalibrated model and the model calibrated only with flow rate.

In order to accurately reconstruct the process characteristics of “23·7” catastrophic flood and provide theoretical support for the prevention and early warning of catastrophic floods in the Yongding River Basin, the grid-based Xin’anjiang (Grid-XAJ) model and Grid-XAJ based on saturation-infiltration double excess (Grid-XAJ-SIDE) model were used to simulate and analyze five key control sections in the Yongding River Basin: Zhaitang Reservoir, Qingbaikou, Yanchi, Sanjiadian, and Lugou Bridge. Typical flood events from 2000 to 2022 were selected for parameter calibration, and the “23·7” catastrophic flood event was used for validation. The results show that both the Grid-XAJ and Grid-XAJ-SIDE models can accurately simulate the entire process of the “23·7” catastrophic flood. In the Qingbaikou section, the determination coefficients of both models achieve Class A accuracy; at the Yanchi, Sanjiadian, and Lugou Bridge control sections, the model accuracy reaches Class B, indicating excellent simulation performance. The study demonstrates that the Grid-XAJ and Grid-XAJ-SIDE models perform well in simulating flood volume, peak flow, and peak time at the five control sections, making them effective for simulating catastrophic floods in semi-arid and semi-humid regions.

This study aims to construct the estimation method of river network regression coefficient C_s that does not depend on the rainfall-runoff data. Four typical small and medium-sized river basins located in different hydrometeorological divisions are selected in this study. River networks are extracted from the DEM information, simplification schemes for the river network is constructed based on Strahler’s rule, the differential equation for the storage based on the self-similar river network method is improved, and the influence of different river length ratios on the C_s calculation results is analyzed. The results show that: the improved method proposed in this paper can effectively reduce the difference of C_s between the original river network and the simplified scheme to improve the stability; the increase of river network complexity in the basin usually leads to the increase of C_s calculation results; the river length ratio affects the applicability of simplified scheme by influencing the information covered by the river network. For small and medium-sized basins with the 3-level structure, when the length ratio of the level 1 river network is less than 70%, the simplification solution of river network retained to level 2 can be used as an alternative input.

To improve the simulation accuracy in the Chinese semi-humid watershed, this study took the Chenhe Catchment in the semi-humid area as the study area and selected 6 parameters including ZSOILFAC, GWEXP, REFKDT, RETDEPRTFAC, OVROUGHRTFAC and MANNFAC for the sensitivity analysis of WRF-Hydro model parameters. The sensitivity of these parameters was qualitatively analyzed by studying the influence of different parameter value in the reasonable range on the flood simulation and each evaluation index, and the modified Morris screening method was used to quantitatively determine the sensitivity of each parameter. The results show that ZSOILFAC, REFKDT and MANNFAC are very sensitive, and it is suggested to take the value within 0.1 to 1.0, 0.0 to 1.0 and 0.5 to 1.5 respectively.

From the perspective of physical mechanism of rainfall-runoff behaviour, three small watersheds in different hydrometeorological regions were selected and the design storms were calculated based on the attenuation formula of rainstorm in this study. According to the characteristics of the underlying surface of the basin, the model parameters were deduced, and the hydrological model was constructed to realize the design flood calculation. The results of using the hydrological model to estimate design floods were compared with those of the reasoning formula method. The results show: the design flood obtained by the hydrological model method is similar to that obtained by the inference formula method, with relative errors of flood peak and flood volume not exceeding 30%; the hydrological model method takes into account the variation of soil infiltration capacity during rainfall processes and the uncertainty of the catchment process at various points within the watershed; the uneven spatiotemporal distribution of hydrological model parameters is more in line with the actual runoff generation and concentration laws of the basin, improving the reliability of design flood calculation in the complex situation.

To support the decision-making and forecasting work for the catastrophic flood in the Haihe River Basin, the Zijingguan and Manshuihe sub-basins, located in the Daqing River where a catastrophic flood occurred in July 2023, were selected for the flood simulation and prediction with the Xin’anjiang-Haihe model based on measured rainfall and discharge data. For the Zijingguan sub-basin, 5 floods from 1996 to 2020 were chosen for the parameter calibration, while for the Manshuihe sub-basin, 8 floods from 1953 to 2016 were used. The simulation was validated against the Haihe “23· 7” basin-wide extreme flood event. The results showed that the relative errors of peak discharge and flood volume in both sub-basins were within 20%, with a peak timing error of 0 hours. The high accuracy of the hydrological model constructed in this study, along with the good agreement between simulation and measured results, demonstrates its ability to reflect the actual flood process. A comparison between different models revealed that the Xin’anjiang-Haihe model performed better in the Manshuihe sub-basin, which is more affected by human activities, compared to the Zijingguan sub-basin with lesser human influence.