the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Jul 2026
Review article: Upstream reservoir operation and downstream floodplain risk: A review of hydrodynamic modelling and optimization frameworks and case perspective on the Dong Nai–Sai Gon river basin
Abstract. Flood risk management is increasingly challenged by climate change, rapid urbanization, and the growing complexity of hydrological systems. Reservoirs play a vital role in flood mitigation, yet traditional rule-curve operations based on historical stationarity are becoming inadequate under non-stationary and compound flood conditions. This review synthesizes advances in reservoir operation, hydrodynamic flood modelling, and optimization frameworks for downstream flood risk management, drawing on literature published between 2000 and 2025. Existing studies are classified into four thematic areas: reservoir operation strategies, hydrodynamic modelling, optimization methods, and compound flooding processes. Results reveal a shift from static rule-based operations toward adaptive, forecast-informed, and data-driven approaches that improve flood control performance and reduce peak discharge. Coupled 1D–2D hydrodynamic models better capture floodplain dynamics, while evolutionary algorithms and reinforcement learning support efficient multi-objective optimization. However, a critical knowledge gap persists, as spatial flood inundation metrics and compound flood drivers remain rarely incorporated into optimization frameworks. Using the Dong Nai–Sai Gon river basin as a representative tropical, tide-influenced system, we propose an integrated framework linking reservoir operations, hydrodynamic simulation, and compound flooding processes. Future research should prioritize hybrid physics–AI approaches, real-time data integration, and climate-resilient reservoir management.
- Preprint
(1914 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3711', Anonymous Referee #1, 10 Aug 2026
-
RC2: 'Comment on egusphere-2026-3711', Anonymous Referee #2, 20 Aug 2026
Review article: Upstream reservoir operation and downstream floodplain risk: A review of hydrodynamic modelling and optimization frameworks and case perspective on the Dong Nai–Sai Gon river basin
The overall objective of this manuscript is to systematically review and synthesize recent advances (spanning from the years of 2000–2025) at the intersection of reservoir operation strategies, hydrodynamic flood modelling, and optimization frameworks. Using the Dong Nai-Sai Gon river basin as a representative tropical, tide-influenced case study, the paper aims to identify critical knowledge gaps regarding compound flooding and to propose an integrated conceptual framework that transitions reservoir management from static, rule-based operations toward adaptive, risk-informed, and AI-driven floodplain protection. The manuscript has highlighted several pivotal findings drawn from the literature. Regarding the limitations of traditional operations, the authors revealed that conventional, static rule curves are increasingly inadequate under non-stationary climatic conditions, historically leading to uncoordinated releases and resulting in a 30–40% performance deficit compared to adaptive strategies. In terms of advancements in modelling and optimization, coupled 1D-2D hydrodynamic models have significantly enhanced the representation of lateral floodplain dynamics, while advanced multi-objective optimization methods (such as evolutionary algorithms and deep reinforcement learning) can reduce peak discharges by 25–30% and greatly improve system reliability. By considering the compound flooding gap, critical research disconnected persists where spatial floodplain inundation metrics and compound flood drivers (the non-linear interaction of fluvial discharge, intense local rainfall, and tidal backwater) are rarely embedded directly into reservoir optimization objective functions. Finally, since the future AI-driven integration was critically discussed, the study outlined a progressive transition toward “Modern FIRO” (Forecast-Informed Reservoir Operations), proposing a framework that leverages real-time data, hybrid physics-machine learning models, and stochastic multi-hazard optimization to manage extreme, compound flood events in complex deltaic environments.
The manuscript offers a highly relevant and timely synthesis of reservoir operation strategies, hydrodynamic modelling, and compound flooding dynamics. The categorization of the literature into four distinct thematic areas provides a robust structure for navigating a complex interdisciplinary field, and the conceptual integration framework, particularly Figures 3B and 4, is a standout contribution that clearly visualizes the transition from static rule curves to AI-driven modern operations. This work effectively highlights the pressing need to move beyond isolated 1D stage-discharge relationships and stationary assumptions, making it a valuable addition to the literature on climate-resilient water management. Although the literature review is relatively successful in noting the above valuable findings in the field, I suggest a moderate revision before the manuscript could be considered for publication under NHESS:
- First, while the methodological framework outlined in Figure 1 is visually intuitive, the narrative description of the literature screening process requires greater transparency. The manuscript notes that the initial searches yielded 297 papers from primary databases and 260 from supplementary sources, yet the final synthesis matrix details only 20 primary studies. To strengthen the rigor and reproducibility of this review, it is necessary to explicitly define the quantitative or qualitative rationale used to distill these hundreds of initial records down to the 20 studies featured in Table 1. Clarifying whether these studies were selected for their specific focus on tropical monsoon basins, their benchmark algorithmic contributions, or citation impact will elevate the academic rigor of the methodology section.
- Second, the manuscript would benefit from weaving the case perspective of the Dong Nai-Sai Gon basin more organically throughout the earlier sections, rather than reserving it predominantly for Section 6.3. Introducing the specific compound flooding challenges of Ho Chi Minh City and the Dau Tieng reservoir earlier would ground the theoretical discussions of optimization and 2D modelling in a tangible reality. Furthermore, in the specific context of the Dong Nai–Sai Gon basin, the authors should incorporate additional discussion of upstream flooding in the Dong Nai River, as highlighted by recent runoff forecasting research (10.1080/02626667.2025.2461705), as well as the bidirectional interactions between tidal flows and riverine flows in the downstream Dong Nai–Sai Gon system. These hydrodynamic interactions are particularly relevant to compound flooding and should be explicitly considered in the proposed integrated framework. Previous studies have demonstrated the importance of hydrodynamic and morphodynamic processes in the lower basin and estuarine system (10.1016/j.jhydrol.2022.127572; 10.1080/00221686.2025.2456727), while recent research has further highlighted the strong influence of tidal dynamics on peak water-level forecasting in the tidal-dominated Dong Nai–Sai Gon basin (10.1080/02626667.2026.2671902).
- Third, to enhance the practical feasibility of the proposed framework, the authors should consider incorporating an assessment of data availability in Vietnam. In the proposed Integrated Framework for the Dong Nai–Sai Gon (DN–SG) basin, the authors mention weather radar, IoT, and digital twins. However, in practice, Vietnam’s monitoring network remains relatively sparse, while data sharing between reservoirs (e.g., Dau Tieng and Tri An) is sometimes interrupted or does not provide real-time data. The authors should include a brief discussion of data-related challenges and how AI models can help address the problem of data scarcity.
- Fourth, although Deep Reinforcement Learning (DRL) algorithms can achieve very high predictive or decision-making accuracy, they are often characterized as “black-box” models, making it difficult to explain the rationale behind a particular flood-release decision. Meanwhile, reservoir managers require transparency, interpretability, and safety assurance before implementing operational decisions. The authors should therefore further discuss Explainable Artificial Intelligence (XAI) in the Future Directions section to strengthen the framework’s credibility and acceptance among policymakers and reservoir managers.
- Fifth, when multi-objective optimization algorithms such as NSGA-III are applied to reservoir operation, reducing downstream flood risk will inevitably involve some trade-offs with other objectives, such as hydropower generation and dry-season water storage. The manuscript should place greater emphasis on how this economic–safety trade-off is quantified and evaluated within the objective functions.
- Finally, while the advocacy for AI, machine learning, and digital twins in the “Modern FIRO” framework is forward-looking, a critical discussion on the feasibility of deploying these computationally heavy and data-reliant frameworks in developing or data-scarce regions would provide a more balanced, pragmatic conclusion to this excellent review.
Citation: https://doi.org/10.5194/egusphere-2026-3711-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 154 | 39 | 14 | 207 | 13 | 15 |
- HTML: 154
- PDF: 39
- XML: 14
- Total: 207
- BibTeX: 13
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The authors presented a relatively well-structured and relevant manuscript addressing the increasingly complex interplay between reservoir operation, rainfall and downstream flooding, taking the Sai Gon – Dong Nai basin as a deep dive section. The paper has a clear and relevant overall structure, moving logically from reservoir operation, to hydrodynamic modelling, to optimization, and finally to compound flooding and the basin case study. The literature overview and connection between traditionally separate strands of literature on how to model compound floods address an important gap and would likely be appreciated by the modelling community.
I have two main comments for the authors
Minor editing suggestions:
physics–AI approaches: please specify what do we mean with AI approaches.
unpredictability of precipitation and associated runoffs: What about cascading system and inter-dependency amongst reservoirs on the same stream. See main comment 1.
lack the feedback information: What do we mean with this?
High-resolution hydrodynamic models: Spatially, or temporally?
difficult trade-offs: Please specify what do you mean with this.
Figure 1: Please specify if this is for Sai Gon – Dong Nai, or globally
Table 1 can be made more compact by reducing line spacing, please introduce the abbreviations