Search for: All records

This paper advances machine learning (ML)-based streamflow prediction by strategically selecting rainfall events, introducing a new loss function, and addressing rainfall forecast uncertainties. Focusing on the Iowa River Basin, we applied the stochastic storm transposition (SST) method to create realistic rainfall events, which were input into a hydrological model to generate corresponding streamflow data for training and testing deterministic and probabilistic ML models. Long short-term memory (LSTM) networks were employed to predict streamflow up to 12 h ahead. An active learning approach was used to identify the most informative rainfall events, reducing data generation effort. Additionally, we introduced a novel asymmetric peak loss function to improve peak streamflow prediction accuracy. Incorporating rainfall forecast uncertainties, our probabilistic LSTM model provided uncertainty quantification for streamflow predictions. Performance evaluation using different metrics improved the accuracy and reliability of our models. These contributions enhance flood forecasting and decision-making while significantly reducing computational time and costs. 

Floods are among the most destructive natural hazards, with damages expected to intensify under climate change and socio-economic pressures. Effective reservoir operation remains a critical yet challenging strategy for mitigating downstream impacts, as operators must navigate nonlinear system dynamics, uncertain inflow forecasts, and trade-offs between competing objectives. This study proposes a novel end-to-end data-driven framework that integrates process-based hydraulic simulations, a Transformer-based surrogate model for flood damage prediction, and reinforcement learning (RL) for reservoir gate operation optimization. The framework is demonstrated using the Coralville Reservoir (Iowa, USA) and two major historical flood events (2008 and 2013). Hydraulic and impact simulations with HEC-RAS and HEC-FIA were used to generate training data, enabling the development of a Transformer model that accurately predicts time-varying flood damages. This surrogate is coupled with a Transformer-enhanced Deep Q-Network (DQN) to derive adaptive gate operation strategies. Results show that the RL-derived optimal policy reduces both peak and time-integrated damages compared to expert and zero-opening benchmarks, while maintaining smooth and feasible operations. Comparative analysis with a genetic algorithm (GA) highlights the robustness of the RL framework, particularly its ability to generalize across uncertain inflows and varying initial storage conditions. Importantly, the adaptive RL policy trained on perturbed synthetic inflows transferred effectively to the hydrologically distinct 2013 event, and fine-tuning achieved near-identical performance to the event-specific optimal policy. These findings highlight the capability of the proposed framework to provide adaptive, transferable, and computationally efficient tools for flood-resilient reservoir operation. 

Abstract Computational hydrological models and simulations are fundamental pieces of the workflow of contemporary hydroscience research, education, and professional engineering activities. In support of hydrological modelling efforts, web-enabled tools for data processing, storage, computation, and visualization have proliferated. Most of these efforts rely on server resources for computation and data tasks and client-side resources for visualization. However, continued advancements of in-browser, client-side compute performance present an opportunity to further leverage client-side resources. Towards this end, we present an operational rainfall-runoff model and simulation engine running entirely on the client side using the JavaScript programming language. To demonstrate potential uses, we also present an easy-to-use in-browser interface designed for hydroscience education. Although the use case presented here is self-contained, the core technologies can extend to leverage multi-core processing on single machines and parallelization capabilities of multiple clients or JavaScript-enabled servers. These possibilities suggest that client-side hydrological simulation can play a central role in a dynamic, interconnected ecosystem of web-ready hydrological tools. 

Abstract Although prior studies have evaluated the role of sampling errors associated with local and regional methods to estimate peak flow quantiles, the investigation of epistemic errors is more difficult because the underlying properties of the random variable have been prescribed using ad‐hoc characterizations of the regional distributions of peak flows. This study addresses this challenge using representations of regional peak flow distributions derived from a combined framework of stochastic storm transposition, radar rainfall observations, and distributed hydrologic modeling. The authors evaluated four commonly used peak flow quantile estimation methods using synthetic peak flows at 5,000 sites in the Turkey River watershed in Iowa, USA. They first used at‐site flood frequency analysis using the Pearson Type III distribution with L‐moments. The authors then pooled regional information using (1) the index flood method, (2) the quantile regression technique, and (3) the parameter regression. This approach allowed quantification of error components stemming from epistemic assumptions, parameter estimation method, sample size, and, in the regional approaches, the number ofpooledsites. The results demonstrate that the inability to capture the spatial variability of the skewness of the peak flows dominates epistemic error for regional methods. We concluded that, in the study basin, this variability could be partially explained by river network structure and the predominant orientation of the watershed. The general approach used in this study is promising in that it brings new tools and sources of data to the study of the old hydrologic problem of flood frequency analysis.