The precision of numerical overland flow models is limited by their computational cost. A GPU-accelerated 2D shallow flow model is developed to overcome this challenge in this study. The model employs a Godunov-type finite volume method (FVM) to solve shallow water equations (SWEs) with unstructured grids, while also considering rainfall, infiltration, bottom slope, and friction source terms. The numerical simulation demonstrates that this model has well-balanced and robust properties. In an experiment of urban rain-runoff and flood, the accuracy and stability of the model are further demonstrated. The model is programmed with CUDA, and each numerical computation term is processed in parallel to adopt multi-thread GPU acceleration technology. With the GPU computation framework, this model can achieve a speeding up ration around 75 to single-thread CPU in the dam-break flow for a large-scale application.

Climate change and human activities have a great impact on the environment and have challenged the assumption of the stability of the hydrological time series and the consistency of the observed data. In order to investigate the applicability of machine learning (ML)-based rainfall–runoff (RR) simulation methods under a changing environment scenario, several ML-based RR simulation models implemented in novel continuous and non-real-time correction manners were constructed. The proposed models incorporated categorical boosting (CatBoost), a multi-hidden-layer BP neural network (MBP), and a long short-term memory neural network (LSTM) as the input–output simulators. This study focused on the Dongwan catchment of the Yiluo River Basin to carry out daily RR simulations for the purpose of verifying the model’s applicability. Model performances were evaluated based on statistical indicators such as the deterministic coefficient, peak flow error, and runoff depth error. The research findings indicated that (1) ML-based RR simulation by using a consistency-disrupted dataset exhibited significant bias. During the validation phase for the three models, the R² index decreased to around 0.6, and the peak flow error increased to over 20%. (2) Identifying data consistency transition points through data analysis and conducting staged RR simulations before and after the transition point can improve simulation accuracy. The R² values for all three models during both the baseline and change periods were above 0.85, with peak flow and runoff depth errors of less than 20%. Among them, the CatBoost model demonstrated superior phased simulation accuracy and smoother simulation processes and closely matched the measured runoff processes across high, medium, and low water levels, with daily runoff simulation results surpassing those of the BP neural network and LSTM models. (3) When simulating the entire dataset without staged treatment, it is impossible to achieve good simulation results by adopting uniform extraction of the training samples. Under this scenario, the MBP exhibited the strongest generalization capability, highest prediction accuracy, better algorithm stability, and superior simulation accuracy compared to the CatBoost and LSTM simulators. This study offers new ideas and methods for enhancing the runoff simulation capabilities of machine learning models in changing environments.

Numerical simulation models are commonly used to analyze and simulate urban waterlogging risk. However, the computational efficiency of numerical models is too low to meet the requirements of urban emergency management. In this study, a new method was established by combining a long short-term memory neural network model with a numerical model, which can quickly predict the waterlogging depth of a city. First, a numerical model was used to simulate and calculate the ponding depth of each ponding point under different rainfall schemes. Using the simulation results as training samples, the long short-term memory neural network was trained to predict and simulate the waterlogging process. The results showed that the proposed “double model” prediction model appropriately reflected the relationship between the changes in waterlogging depth and the temporal and spatial changes in rainfall, and the accuracy and speed of computation were higher than those of the numerical model alone. The simulation speed of the “double model” was 324,000 times that of the numerical model alone. The proposed “double model” method provides a new idea for the application of artificial intelligence technology in the field of disaster prevention and reduction.

The joint flood control operation of reservoir groups is a complex engineering problem with a large number of constraints and interdependent decision variables. Its solution has the characteristics of strong constraint, multi-stage, nonlinearity, and high dimension. In order to solve this problem, this paper proposes a hybrid slime mold and arithmetic optimization algorithm (HSMAAOA) combining stochastic reverse learning. Since ancient times, harnessing the Yellow River has been a major event for the Chinese nation to rejuvenate the country and secure the country. Today, flood risk is still the greatest threat to the Yellow River basin. This paper chooses five reservoirs in the middle and lower reaches of the Yellow River as the research object, takes the water level of each reservoir in each period as the decision variable, and takes the peak clipping of Huayuankou control point as the objective to build an optimization model. Then, HSMAAOA is used to solve the problem, and the results are compared with those of the slime mold algorithm (SMA) and particle swarm optimization (PSO). The peak clipping rates of the three algorithms are 52.9% (HSMAAOA), 48.69% (SMA), and 47.55% (PSO), respectively. The results show that the HSMAAOA algorithm is better than other algorithms. This paper provides a new idea to solve the problem of the optimal operation of reservoir flood controls.

In recent years, the combined effects of rapid urbanization and climate change have led to frequent floods in urban areas. Rainstorm flood risk warning systems play a crucial role in urban flood prevention and mitigation. However, there has been limited research in China on nationwide urban flood risk warning systems based on rainfall predictions. This study constructs a two-level early warning system (EWS) at the national and urban levels using a two-dimensional hydrological–hydrodynamic model considering infiltration and urban drainage standards. A methodology for urban rainstorm flood risk warnings is proposed, leveraging short-term and high-resolution rainfall forecast data to provide flood risk warnings for 231 cities in central and eastern China. Taking Beijing as an example, a refined rainstorm flood warning technique targeting city, district, and street scales is developed. We validated the methodology with monitoring data from the “7.31” rainstorm event in 2023 in Beijing, demonstrating its applicability. It is expected that the findings of this study will serve as a valuable reference for the urban rainstorm flood risk warning system in China.

Insufficient precipitation observations hinder the bias-correction of Global Climate Model (GCM) precipitation outputs in ungauged and remote areas. As a result, the reliability of future precipitation and water resource projections is restricted for these areas. Open-access quantitative precipitation estimation (QPE) products offer a potential solution to this challenge. This study assesses the effectiveness of three widely used, long-term QPEs, including ERA5, PERSIANN-CDR, and CHIRPS, in bias-correcting precipitation outputs from the CMIP6 GCMs. The evaluation involves the reproduction of precipitation distribution, streamflow simulation utility based on a hydrological model, and the accuracy of extreme indices associated with rainstorm/flood/drought events. This study selects the Beijiang basin located in the subtropical monsoon area of South China as the case study area. The results demonstrate that bias-correction using QPEs improves the performance of GCM precipitation outputs in reproducing precipitation/streamflow distribution and extreme indices, with a few exceptions. PCDR generally exhibits the most effective bias-correction utility, consistently delivering reasonable performance across various cases, making it a suitable alternative to gauge data for bias-correction in ungauged areas. However, GCM outputs corrected by ERA5 tend to overestimate overall precipitation and streamflow (by up to about 25% to 30%), while the correction results of CHIRPS significantly overestimate certain extreme indices (by up to about 50% to 100%). Based on the revealed performance of QPEs in correcting GCM outputs, this study provides references for selecting QPEs in GCM-based water resource projections in remote and ungauged areas.