In the past decade, more than 300 people have died per year on average due to mountain torrents in China. Mountain torrents mostly occur in ungauged small and medium-sized catchments, so it is difficult to maintain high accuracy of flood prediction. In order to solve the problem of the low accuracy of flood simulation in the ungauged areas, this paper studies the influence of different methods on the parameter regionalization of distributed hydrological model parameters in hilly areas of Hunan Province. According to the terrain, landform, soil and land use characteristics of each catchment, we use Shortest Distance, Attribute Similarity, Support Vector Regression, Generative Adversarial Networks, Classification and Regression Tree and Random Forest methods to create parameter regionalization schemes. In total, 426 floods of 25 catchments are selected to calibrate the model parameters, and 136 floods of 8 catchments are used for verification. The results showed that the average values of the Nash–Sutcliffe coefficients of each scheme were 0.58, 0.64, 0.60, 0.66, 0.61 and 0.68, and the worst values were 0.27, 0.31, 0.25, 0.43, 0.35 and 0.59. The random forest model is the most stable solution and significantly outperforms other methods. Using the random forest model to regionalize parameters can improve the accuracy of flood simulation in ungauged areas, which is of great significance for flash flood forecasting and early warning.

Flash floods are typically associated with short, high-intensity and extreme rain-storms, and they are characterized by short response time and severely impact and damage communities in different areas in China. In order to scientifically assess the risks of flash floods, this paper takes Hubei Province as an example to carry out risk assessment. Based on Pearson correlation coefficient and principal component analysis methods, 14 factors were selected from 98 factors to establish a risk assessment model. The confidence coefficient model and multi-factor superposition method were used to determine the weight of each risk factor, and a risk map of Hubei Province was finally constructed. The results show that medium-high risk areas in Huanggang account for 47.00%, and high-risk areas account for 8.70%, with both areas adding up to more than 50%, followed by more than 40% in Shiyan, E’zhou and Xianning, and more than 30% in Huangshi, Yichang, Xiangyang, Jingmen and Suizhou. The risk level distribution is highly consistent with the location and frequency of flash flood disasters, shows high reliability, and can provide data support for flash flood disaster prevention and control. This study used a quantitative method to determine the key factors affecting flash flood disasters and provides a reference and basis for flash flood risk assessment in other provinces in China.

Short-term heavy precipitation is a crucial factor that triggers urban waterlogging and flash flood disasters, which impact human production and livelihood. Traditional short-term forecasting methods have time- and scale-based limitations. To achieve timely, location-specific, and quantitative precipitation forecasting, this study applies the precipitation spectral decomposition algorithm, along with variational echo tracking and autoregressive AR2 extrapolation techniques, to forecast three cases of heavy precipitation events during the rainy season in Hebei Province. The variational optical flow extrapolation forecasting based on precipitation spectral decomposition has a forecasting lead time of up to 3 h. However, noticeable discrepancies in forecast accuracy can be observed around 2 h, and the forecasting skill gradually weakens with longer lead times. For 3 h lead time forecasts, substantial variability occurs among different performance metrics, lacking clear comparability. The effective forecast lead time for variational optical flow forecasting based on precipitation spectral decomposition is up to 1.6 h for severe convective weather systems and up to 2.2 h for stratiform cloud weather systems. Overall, the forecast effect of this method is good in the three rainfalls—the highest CSI is up to 0.74, the highest POD is up to 0.87, and the forecast accuracy and success rate are high.

Flash floods are ferocious and destructive, making their forecasting and early warning difficult and easily causing casualties. In order to improve the accuracy of early warning, a dynamic early warning index system was established based on the distributed spatio-temporally mixed model through a case study of riverside villages in Hubei Province. Fully taking into account previous rainfall and assuming different rainfall conditions, this work developed a dynamic early warning threshold chart by determining critical rainfall thresholds at different soil moisture levels (dry, normal, wet, and saturated) through pilot calculations, to support a quick query of the critical rainfall at any soil moisture level. The research results show that of the 74 counties and districts in Hubei Province, more than 50% witnessed higher mean critical rainfall than empirical thresholds when the soil was saturated, and about 90% did so when the soil was dry. In 881 towns, a total of 456 early warnings were generated based on dynamic thresholds from 2020 to 2022, 15.2% more than those based on empirical thresholds. From the perspective of total rainfall, dynamic early warnings were generated more frequently in wet years, while empirical early warnings were more frequent in dry years, and the frequency of two warnings were roughly the same in normal years. There were more early warnings based on empirical thresholds in May each year, but more based on dynamic thresholds in June and July, and early warnings generated based on the two methods were almost equal in August and September. Spatially, after dynamic early warning thresholds were adopted, Shiyan and Xiangyang, both northwestern cities in Hubei Province, witnessed significant increases in early warnings. In terms of the early warning mechanism, dynamic early warning took into account the impact of soil moisture and analyzed the flood discharge capacity of river channels according to the flood stage of the riverside villages. On this basis, the rainfall early warning thresholds under different conditions were determined. This is a refined early warning method that could improve the accuracy of flash flood warnings in Hubei Province.

High-quality rainfall data are crucial for accurately forecasting flash floods and runoff simulations. However, traditional correction methods often overlook errors in rainfall-monitoring data. We established a screening system to identify anomalous stations using the Hampel method, Grubbs criterion, analysis of surrounding measurement stations, and radar-assisted verification. Three rainfall data-fusion methods were used to fuse rainfall station data with radar quantitative precipitation estimation data; the accuracies of the fused data products with and without anomalous data identification were compared. Validation was performed using four 2012 rainfall events in Hebei Province. The 08:00–19:00 July 3 rainfall event had the highest number of anomalous stations (11.5% of the total), while the 01:00–17:00 August 9 event had the lowest number (7.8%). By comparing stations deemed to be anomalous with stations that were actually anomalous, we determined that the accuracy of reference station determination using Hampel’s method and Grubbs’ test was 94.2%. Radar-assisted validation improved the average accuracy of anomalous station identification during the four typical rainfall events from 89.7 to 93.7%. Excluding anomalous data also significantly impacted the efficacy of rainfall-data fusion, as it improved the quality of the rainfall station data. Among the performance indicators, 95% improved after the exclusion of anomalous data for all four rainfall events.