Flow forecasting is an essential topic for flood prevention and mitigation. This study utilizes a data-driven approach, the Long Short-Term Memory neural network (LSTM), to simulate rainfall–runoff relationships for catchments with different climate conditions. The LSTM method presented was tested in three catchments with distinct climate zones in China. The recurrent neural network (RNN) was adopted for comparison to verify the superiority of the LSTM model in terms of time series prediction problems. The results of LSTM were also compared with a widely used process-based model, the Xinanjiang model (XAJ), as a benchmark to test the applicability of this novel method. The results suggest that LSTM could provide comparable quality predictions as the XAJ model and can be considered an efficient hydrology modeling approach. A real-time forecasting approach coupled with the k-nearest neighbor (KNN) algorithm as an updating method was proposed in this study to generalize the plausibility of the LSTM method for flood forecasting in a decision support system. We compared the simulation results of the LSTM and the LSTM-KNN model, which demonstrated the effectiveness of the LSTM-KNN model in the study areas and underscored the potential of the proposed model for real-time flood forecasting.

The Weather Research and Forecasting (WRF)-Hydro model as a physical-based, fully-distributed, multi-parameterization modeling system easy to couple with numerical weather prediction model, has potential for operational flood forecasting in the small and medium catchments (SMCs). However, this model requires many input forcings, which makes it difficult to use it for the SMCs without adequate observed forcings. The Global Land Data Assimilation System (GLDAS), the WRF outputs and the ideal forcings generated by the WRF-Hydro model can provide all forcings required in the model for these SMCs. In this study, seven forcing scenarios were designed based on the products of GLDAS, WRF and ideal forcings, as well as the observed and merged rainfalls to assess the performance of the WRF-Hydro model for flood simulation. The model was applied to the Chenhe catchment, a typical SMC located in the Midwestern China. The flood prediction capability of the WRF-Hydro model was also compared to that of widely used Xinanjiang model. The results show that the three forcing scenarios, including the GLDAS forcings with observed rainfall, the WRF forcings with observed rainfall and GLDAS forcings with GLDAS-merged rainfall, are optimal input forcings for the WRF-Hydro model. Their mean root mean square errors (RMSE) are 0.18, 0.18 and 0.17 mm/h, respectively. The performance of the WRF-Hydro model driven by these three scenarios is generally comparable to that of the Xinanjiang model (RMSE = 0.17 mm/h).

The Xin’anjiang model is a conceptual hydrological model, which has an essential application in humid and semi-humid regions. In the model, the parameters estimation of runoff routing has always been a significant problem in hydrology. The quantitative relationship between parameters of the lag-and-route method and catchment characteristics has not been well studied. In addition, channels in Muskingum method of the Xin’anjiang model are assumed to be virtual channels. Therefore, its parameters need to be estimated by observed flow data. In this paper, a new routing scheme for the Xin’anjiang model is proposed, adopting isochrones method for overland flow and the grid-to-grid Muskingum–Cunge–Todini (MCT) method for channel routing, so that the routing parameters can be estimated according to the geographic information. For the new routing scheme the average overland flow velocity can be determined through the land cover and overland slope, and the channel routing parameters can be determined through channel geometric characteristic, stream order and channel gradient. The improved model was applied at a 90 m grid scale to a nested watershed located in Anhui province, China. The parent Tunxi watershed, with a drainage area of 2692 km², contains four internal points with available observed streamflow data, allowing us to evaluate the model’s ability to simulate the hydrologic processes within the watershed. Calibration and verification of the improved model were carried out for hourly time scales using hourly streamflow data from 1982 to 2005. Model performance was assessed by comparing simulated and observed flows at the watershed outlet and interior gauging stations. The performance of both original and new runoff routing schemes were tested and compared at hourly scale. Similar and satisfactory performances were achieved at the outlet both in the new runoff routing scheme using the estimated routing parameters and in the original runoff routing scheme using the calibrated routing parameters, with averaged Nash-Sutcliffe efficiency (NSE) of 0.92 and 0.93, respectively. Moreover, the new runoff routing scheme is also able to reproduce promising hydrographs at internal gauges in study catchment with the mean NSE ranging from 0.84 to 0.88. These results indicate that the parameter estimation approach is efficient and the developed model can satisfactorily simulate not only the streamflow at the parent watershed outlet, but also the flood hydrograph at the interior gauging points without model recalibration. This study can provide some guidance for the application of the Xin’anjiang model in ungauged areas.

The distributed Grid-Xin’anjiang (Grid-XAJ) model is very sensitive to the spatial and temporal distribution of data when used in humid and semi-humid small and medium catchments. We used the successive correction method to merge the gauged rainfall with rainfall forecasted by the Weather Research and Forecasting (WRF) model to enhance the spatiotemporal accuracy of rainfall distribution. And we used the Penman–Monteith equation to calculate the potential evapotranspiration (PE_PM). Then, we designed two forcing scenarios (WRF-driven rainfall (Wr) + PE_PM, WRF-merged rainfall (Wm) + PE_PM) to drive the Grid-XAJ model for flood forecasting. We found the WRF-driven Grid-XAJ model held significant potential in flood forecasting. The Grid-XAJ model provided only an approximation of flood hygrographs when driven by scenario Wr + PE_PM. The results in scenario Wm + PE_PM showed a high degree-of-fit with observed floods with mean Nash–Sutcliffe efficiency coefficient (NSE) values of 0.94 and 0.68 in two catchments. Additionally, scenario Wm + PE_PM performed better flood hygrographs than scenario Wr + PE_PM. The flood volumes and flow peaks in scenario Wm + PE_PM had an obvious improvement compare to scenario Wr + PE_PM. Finally, we observed that the model exhibited superior performance in forecasting flood hydrographs, flow peaks, and flood volumes in humid catchments compared with semi-humid catchments.

Flood simulation in small- and medium-sized catchments is hindered by data scarcity and strong hydroclimatic heterogeneity. Distributed models with pedotransfer functions offer new opportunities, yet their parameter sensitivity and regional applicability remain insufficiently understood. In this study, the wflow_sbm model was applied to two catchments: the humid Tunxi basin and the semi-humid Chenhe basin, China. Model seamless parameters, defined as spatially continuous fields derived directly from global datasets using pedotransfer functions without local calibration, were generated using the HydroMT system. The parameter sensitivity, applicability of pedotransfer function derived parameters, and model performance were systematically evaluated and benchmarked against the well-established Xin’anjiang (XAJ) model, which is a conceptual lumped hydrological model widely used for flood simulation in humid and semi-humid regions of China. Sensitivity analysis identified KsatHorFrac and InfiltCapSoil as dominant in Tunxi, and KsatHorFrac and SoilThickness in Chenhe. SoilThickness derived by HydroMT underestimated flood volumes in the Chenhe basin but was substantially improved after applying a uniform scaling factor of 0.1, resulting in an effective SoilThickness of approximately 0.2 m. The wflow_sbm model achieved performance comparable to the XAJ model. Optimal calibration achieved N_SE = 0.85 in Tunxi with good performance at internal sub-catchments (Yuetan and Chengcun, N_SE > 0.70), and generally above 0.7 in Chenhe. These findings highlight the region-dependent validity of parameterization and provide guidance for distributed flood modeling in data-scarce basins.