Frequency analysis of precipitation extremes is significant for the selection of design rainfalls, which are essential inputs for the design of water infrastructure projects, especially when the climate has changed. Therefore, the objective of this study was to propose a framework for more reasonably analysing the frequency of extreme rainfalls. The proposed framework consists of a maximum likelihood estimate (MLE) method for analysing the parameter trends, a hydrological variation diagnosis system to determine abrupt change times, generalized extreme value (GEV) and generalized Pareto distribution (GPD) models for frequency analysis of precipitation extremes, and an ensemble‐methods approach for choosing the most appropriate distributions. The methodology was successfully implemented using a 52‐year time series (1963–2014) of rainfall data recorded by eight rain gauges in Chengbi River basin (south China). The results show that the rainfall series mutated in 1993 and that the entire data set could be divided into two slices (1963–1992 and 1993–2014). Climate change was found to have some impacts on the precipitation extremes: the extreme rainfall value and the parameters of GEV and GPD were variable in the context of climate change. Furthermore, the GPD distribution model outperformed the GEV distribution model.

A reasonable flood season delineation can effectively implement staged reservoir scheduling and improve water resource efficiency. Therefore, this study is aimed at analyzing the flood period segmentation and optimizing the staged flood limit water levels (FLWLs) for a multi-purpose reservoir, the Longtan Reservoir, China. The rainfall seasonality index ($S{I}_P$) and the runoff seasonality index ($S{I}_R$) are used to evaluate the feasibility and rationality of the flood period staging. The fractal method is then used to segment the flood season. Finally, the design flood is carried out to optimize the staged FLWLs. The results show that the $SI$ is an effective indicator for judging the feasibility and verifying the rationality of flood segmentation. The flood period can be segmented into the pre-flood season (12 April–29 May), the main flood season (30 May–3 September), and the post-flood season (4 September–9 November). The FLWLs in the main flood and the post-flood season can be raised by 2.05 m and 3.45 m, and the effective reservoir capacity is increased by 5.810 billion m³ and 6.337 billion m³, according to the results of the flood season division.

Frequent flood disasters have caused serious damage to karst areas with insufficient measured rainfall data, and the analysis of the applicability of satellite rainfall data in runoff simulation is helpful to the local water management. Therefore, the purpose of this study is to analyze the accuracy of IMERG satellite rainfall data and apply it to long-term runoff simulations in a karst area—the Xiajia River basin, China. First, R (correlation coefficient) and POD (probability of detection) are applied to analyze the accuracy of the IMERG data, and the SWAT model is used for runoff simulation. The results show that the accuracy of the original IMERG data is poor (R range from 0.412 to 0.884 and POD range from 47.33 to 100), and the simulation results are “Unsatisfactory” (NSE (Nash-Sutcliffe efficiency coefficient) ranged from 0.17 to 0.32 and RSR (root mean square standard deviation ratio) ranged from 0.81 to 0.92). Therefore, the GDA correction method is used to correct the original IMERG data, and then the accuracy analysis and runoff simulation are carried out. The results show that the accuracy of the corrected IMERG data is better than that of the original data (R range from 0.886 to 0.987 and POD range from 94.08 to 100), and the simulation results of the corrected IMERG data are “Satisfactory” (NSE is over 0.55 and RSR is approximately 0.65). Therefore, the corrected data have a certain applicability in long-term continuous runoff simulations.

Runoff prediction plays an extremely important role in flood prevention, mitigation, and the efficient use of water resources. Machine learning runoff prediction models have become popular due to their high computational efficiency. To select a model with a better runoff simulation and to validate the stability of the model, the following studies were done. Firstly, the support vector machine Model (SVM), the Elman Neural Network Model (ENN), and the multi-model mean model (MMM) were used for the runoff prediction, with the monthly runoff data from 1963–2007 recorded by the Pingtang hydrological station in the Chengbi River Karst Basin, China. Secondly, the comprehensive rating index method was applied to select the best model. Thirdly, the indicators of the hydrologic alteration–range of variability approach (IHA-RVA) was introduced to measure the model stability with different data structure inputs. According to the comprehensive rating index method, the SVM model outperformed the other models and was the best runoff prediction model with a score of 0.53. The overall change of the optimal model was 10.52%, which was in high stability.

Accurate hydrological simulation plays an important role in the research of hydrological problems; the accuracy of the watershed hydrological model is seriously affected by model-parameter uncertainty and model-input uncertainty. Thus, in this study, different calibration methods and rainfall inputs were introduced into the SWAT (Soil and Water Assessment Tool) model for watershed hydrological simulation. The Chengbi River basin, a typical karst basin in Southwest China, was selected as the target basin. The indicators of the NSE (Nash efficiency coefficient), Re (relative error) and R² (coefficient of determination) were adopted to evaluate the model performance. The results showed that: on the monthly and daily scales, the simulated runoff with the single-site method calibrated model had the lowest NSE value of 0.681 and highest NSE value of 0.900, the simulated runoff with the multi-site method calibrated model had the lowest NSE value of 0.743 and highest NSE value of 0.953, increased correspondingly, indicating that adopting the multi-site method could reduce the parameter uncertainty and improve the simulation accuracy. Moreover, the NSE values with IMERG (Integrated Multisatellite Retrievals for Global Rainfall Measurement) satellite rainfall data were the lowest, 0.660 on the monthly scale and 0.534 on the daily scale, whereas the NSE values with fusion rainfall data processed by the GWR (geographical weighted regression) method greatly increased to 0.854 and 0.717, respectively, and the NSE values with the measured rainfall data were the highest, 0.933 and 0.740, respectively, demonstrating that the latter two rainfall inputs were more suitable sources for hydrological simulation.