Experimental Modeling of Three-Dimensional (3D) Partial Dam-Break Flows: A Review

Abstract

The growing threat of dam-break events, fueled by aging infrastructure and climate change, necessitates comprehensive risk management and mitigation strategies. Experimental studies on partial dam-break flows are pivotal for understanding the complex dynamics of these events, particularly in assessing flood risk and refining predictive models. This review synthesizes current experimental investigations on three-dimensional (3D) partial dam-break flows, with an emphasis on breach dynamics, wave impacts, and the role of urban structures. It highlights the challenges in capturing high-resolution 3D flow characteristics and the advancements in measurement techniques such as particle tracking velocimetry and ultrasonic distance meters. The paper discusses the integration of experimental data with numerical models to validate and improve predictive capabilities, stressing the need for continuous refinement of experimental setups and computational approaches. Gaps in the current literature, including the under-representation of irregular breach geometries and complex terrain, are identified, and future research directions are proposed to address these shortcomings. This work underscores the importance of hybrid measurement techniques and interdisciplinary collaboration to enhance dam-break modeling accuracy and flood risk mitigation.

1. Introduction

In recent decades, the risks associated with dam failures have escalated due to the combined effects of aging infrastructure, climate change, and insufficient maintenance [1,2]. The collapse of the Fundão tailings dam in Mariana, Brazil, on 5 November 2015, released approximately 55–62 million cubic meters of mine waste, resulting in 19 fatalities, the destruction of villages, and long-term socio-economic and environmental impacts affecting over 200 municipalities [3,4]. Just four years later, the Brumadinho disaster in January 2019 caused approximately 270 deaths when a tailings dam failure generated a catastrophic mudflow that destroyed settlements and contaminated river systems, prompting regulatory reforms aimed at phasing out upstream dam designs [5]. Conventional water supply dams face similar challenges. On 7 February 2017, California’s Oroville Dam experienced severe damage to its spillways during prolonged rainfall events, triggering the evacuation of approximately 190,000 residents and repair costs that required substantial public investment [6]. Likewise, the near-failure of Toddbrook Reservoir in England in August 2019—following intense rainfall and the partial collapse of its spillway—led to emergency evacuations in what became one of the United Kingdom’s significant dam-related emergency responses in recent decades [7]. These cases collectively underscore the grave consequences of dam breaches, including loss of life, environmental degradation, and massive economic disruption [8]. Moreover, climate-induced increases in extreme and persistent rainfall further exacerbate the vulnerability of inadequately maintained or poorly designed dams, with studies showing overtopping risks could increase by factors of 2.4 to 17 under future warming scenarios, making dam safety management and risk mitigation an urgent global priority [9,10,11].

Statistical analyses suggest that smaller dams, which often lack sufficient discharge capacity and structural integrity, are more prone to failure. However, the impact of large dam failures is considerably more disastrous. The increasing frequency of dam failures globally has prompted urgent calls for enhanced risk management strategies, including better safety monitoring, emergency preparedness, and infrastructure adaptation to address the evolving realities of climate change [12]. Despite advances in engineering and safety protocols, the inherent risks associated with dam infrastructure require continued vigilance and adaptive strategies to mitigate the potential for catastrophic events [13,14]. Collectively, these incidents underscore the urgent need for comprehensive risk assessments and the implementation of both structural and non-structural mitigation measures to safeguard against future dam-break events [15].

The study of partial dam-break flows relies heavily on experimental investigations, which provide indispensable data for understanding complex three-dimensional (3D) dynamics and validating numerical models, given the scarcity of reliable field observations [16]. One-dimensional (1D) and two-dimensional (2D) models, widely applied for their computational efficiency, are effective in flood routing and inundation mapping but remain limited in capturing vertical structures, turbulence, and non-hydrostatic effects [17,18]. In contrast, three-dimensional (3D) models, though computationally demanding, offer a comprehensive representation of flow physics, including velocity distributions and localized impact forces, thereby ensuring higher fidelity in both risk assessment and engineering applications [19]. The three-dimensional nature of these flows cannot be captured fully through two-dimensional studies alone, making 3D experimental research indispensable for flood risk management and model validation [17]. For example, Elkholy et al. [20] employed advanced measurement techniques such as PTV and ultrasonic velocity profilers to capture the three-dimensional velocity fields and water-surface elevations—vital data for understanding the flow dynamics following a partial dam breach. These 3D measurements are critical for accurately modeling the flow patterns, which are often unpredictable and highly sensitive to the spatial variations in the flow field.

Similarly, Aleixo et al. [21] emphasized the importance of capturing 3D velocity profiles over the entire flow depth using PTV to assess the downstream impacts of dam-break flows. The incorporation of 3D dynamics into these experimental studies allows for a more comprehensive understanding of the flow’s behavior in both near-field and far-field regions, which is crucial for accurate flood risk assessments. These experimental studies also serve as benchmarks for validating numerical models, particularly since real-world data from dam-break events are often scarce and fraught with uncertainty. Kocaman et al. [22] used experimental data to validate numerical models simulating dam-break flows over irregular topographies, emphasizing the critical role of laboratory tests in enhancing model accuracy and ensuring that these models can account for the complexities of 3D flow behavior.

Moreover, experimental studies contribute to the refinement of predictive models for flood management. Liu et al. [23] developed a hydraulic model for partial dam breaks, demonstrating a high prediction accuracy when compared to experimental results. Their work underscores the pivotal role of 3D experiments in advancing predictive modeling and improving the reliability of flood risk assessments. Taha and Mihardja [24] consolidated findings from various experimental setups, illustrating their practical applications in hydraulic engineering and flood risk management. Furthermore, Aureli et al. [25] noted that laboratory experiments, particularly those focused on 3D flow dynamics, are indispensable for understanding the physical mechanisms behind dam-break processes and for developing new technologies aimed at mitigating the impacts of flood waves.

Despite the advancements, experimental studies of partial dam-break flows present significant challenges due to the inherent complexity and transient nature of these events. A major challenge is the accurate capture of the three-dimensional dynamics of the flow, which involves rapidly changing velocities, water-surface elevations, and the interaction of flow patterns at different depths. Techniques such as PTV and stereoscopic video cameras have been employed to measure these parameters, but they require sophisticated equipment and precise calibration to ensure accuracy [26,27]. Another challenge lies in the development of predictive models that can accurately simulate the 3D flow characteristics during a partial dam break. Hydraulic models based on sluice gate flow have been proposed, incorporating time and submergence correction coefficients to improve prediction accuracy; however, these models still face limitations in capturing the full complexity of 3D real-world scenarios [23]. Physical modeling, such as the distorted physical model of the Urkmez Dam, offers valuable insights into flow behavior in urban environments, yet scaling issues and the need to replicate environmental features like buildings and roads add layers of complexity [28]. Numerical methods, including the volume of fluid (VOF) method and large eddy simulation (LES), provide high accuracy but require substantial computational resources and careful validation against experimental data [29,30]. Furthermore, deep learning techniques like Time Series Convolutional Neural Input Networks (TSCNIN) show promise in reconstructing transient flow fields, but their effectiveness depends heavily on the availability of comprehensive datasets for training and validation [31,32].

The inherent variability and unpredictability of dam-break flows, coupled with the need for high-resolution 3D data and advanced modeling techniques, make the study of partial dam-break flows a challenging yet essential area of research for effective flood risk management and mitigation strategies [33,34,35].

Therefore, this review systematically examines and analyzes existing experimental research on partial dam-break flows. By compiling current experimental data and recent research advances, this review not only serves as a resource for researchers engaged in numerical simulations to test and validate new computational methods but also provides comparative datasets and highlights unresolved research questions for scholars involved in experimental studies. Additionally, this paper identifies key flow characteristics and phenomena within the context of partial dam-break research that have either received adequate attention or require further investigation. Field data are excluded from this review due to their typical scarcity and high degree of uncertainty. The remainder of this paper is structured as follows: Section 2 reviews and summarizes existing experimental datasets and research results related to partial dam-break flows; Section 3 evaluates specific processes where experimental data are currently lacking or inadequate and proposes recommendations for future experimental research directions.

2. State of the Art of Available Partial Dam-Break Datasets

A typical setup for partial dam-break studies is shown in Figure 1, illustrating an experimental facility designed to simulate a partial breach scenario. This setup typically includes a rectangular flume with a vertically movable gate, which can be quickly removed to release a large volume of still water stored upstream. Various parameters can be adjusted in these experiments to address different research objectives, such as model scale, dam thickness, breach geometry, bed slope, surface roughness, and the presence of downstream obstacles. For example, Figure 2 presents an alternative experimental facility for dam-break flow studies, where the focus was on breach size, featuring a dam body thicker than those in conventional models.

An important aspect of experimental design is the rapid removal of the gate, since the opening process itself may significantly influence the resulting flow field. To minimize such effects, early studies commonly employed a pulley–weight system, in which the gate was connected to a counterweight that fell freely to accelerate the gate upward, thereby reducing its interference with the flow [36,37,38,39]. More recently, alternative solutions have been introduced. As shown in Figure 2, a more advanced mechanism integrates a high-power motor, cable system, and gear reducer, allowing the acceleration and motion of the gate to be intelligently controlled. This design provides greater flexibility and ensures a smoother, faster gate opening, thereby further reducing the impact of the gate operation on the flow dynamics.

To ensure a systematic and rigorous review process, the selection of studies was restricted to experimental investigations of three-dimensional partial dam-break flows. The retrieved papers were subsequently filtered according to multiple criteria, including model size, initial water depth, bed roughness, slope, breach characteristics, and downstream arrangements. Finally, the eligible studies were analyzed with respect to model configuration, measured data, measuring techniques, and research focus. The overall workflow of selection, filtering, and analysis is illustrated in Figure 3.

The references collected in the literature review are categorized based on the objectives of the experimental investigations and are organized into distinct tables.

Table 1. Experimental setup and initial parameters of 3D partial dam-break flow tests.

Reference	L	W	${\mathit{h}}_{\mathit{u}}$	${\mathit{h}}_{\mathit{d}}$	Bed Roughness	Slope	$\mathit{L}/\mathit{W}$	${\mathit{h}}_{\mathit{d}}/{\mathit{h}}_{\mathit{u}}$	Breach Width Ratio	Breach Characteristics	Downstream Characteristics	Year
Tingsanchali and Rattanapitikon [40]	4	1.9	0.1, 0.2, 0.25	0	smooth	0, 0.005	2.11	0.00	0.05	symmetric	N	1993
Fraccarollo and Toro [41]	3	2	0.6	0	smooth	0, 0.07	1.50	0.00	0.20	symmetric	N	1995
Jovanović and Djordjević [42]	1	0.8	0.15	0	smooth	0.00	1.25	0.00	0.13	half symmetric	N	1995
Stelling and Duinmeijer [43]	31	8.3	0.6	0, 0.05	smooth	0.00	3.73	0.05–0.08	0.05	symmetric	N	2003
Shige-eda and Akiyama [44]	2.79	2.6	0.4, 0.2	0	smooth	0.00	1.07	0.00	0.19	asymmetric	square pillars	2003
Eaket et al. [45]	4.75	2.31	0.1, 0.2, 0.3	0, 0.05, 0.1	smooth	0.00	2.06	0, 1/6, 1/3	0.39	symmetric	N	2005
Soares-Frazão et al. [46]; Soares-Frazão and Zech [47]	35.8	3.6	0.4	0.011	smooth	0.00	9.94	0.03	0.28	symmetric	three urban district layouts	2006
Bukreev [37]	7.2	0.202	0.145, 0.16	0	smooth	0.00	35.64	0.00	0.50	symmetric	bottom, a step, a shelf, and a threshold	2006
Szydlowski and Twarog [48]	6.75	3	0.21	0	smooth	0.00	2.25	0.00	0.17	symmetric	urban district layout with aligned buildings	2006
Liang et al. [49]	25	1.6	0.235	0.059	smooth	0.00	15.63	0.25	0.09	symmetric	column	2007
Aureli et al. [50]	2.6	1.2	0.15	0.01	smooth	0.00	2.17	0.07	0.25	symmetric	insubmersible obstacle	2008
Yang et al. [51]	28	1.6	0.4	0.12	smooth	0.00	17.50	0.30	0.13	symmetric	N	2010
Aureli et al. [33]	2.6	1.2	0.030–0.064	0.0068–0.0157	smooth	0.00	2.17	0.20	0.25	symmetric	insubmersible obstacle	2011
Aureli et al. [52]	2.6	1.2	0.07–0.13	0	smooth	0.00	2.17	0.00	0.25	symmetric	insubmersible obstacle	2015
Elkholy et al. [20]	11	4.3	0.25, 0.5, 0.75	0	smooth	0.00	2.56	0.00	0.10	symmetric	N	2016
Honglu Qian et al. [53]	80	1.2	0.25, 0.3	0.1	mobile beds	0.00	66.67	0.4, 1/3	0.33	symmetric	three sediment samples	2018
Cordero et al. [54]	4	1	0.1, 0.15, 0.2	0	smooth	0, 0.21	4.00	0.00	0.2–0.4	symmetric	N	2018
Liu et al. [55]	6	0.24	0.055, 0.13	0	smooth	0.04	25.00	0.00	0.10	symmetric	building	2018
Chumchan and Rattanadecho [56]	0.984	0.484	0.15	0	smooth	0.00	2.03	0.00	0.21	symmetric	two configurations of blocks	2020
Kocaman et al. [22]	1	0.5	0.15	0	smooth	0.00	2.00	0.00	0.20	symmetric	insubmersible obstacle	2020
Dong et al. [57]	20.5	3	0.09, 0.19, 0.29	0	smooth	0.00	6.83	0.00	0.33	symmetric	six configurations of idealized urban street	2021
Kocaman et al. [58]	1	0.5	0.15	0.015, 0.030	smooth	0.00	2.00	0.1, 0.2	0.20	symmetric	N	2021
Erduran et al. [59]	6	1.2	0.25, 0.3	0.025, 0.030	smooth	0.00	5.00	0.10		symmetric	N	2024
Elisa Beteille et al. [27]	16	2	0.250–0.251	0	smooth	0.00	8.00	0.00	0.25	symmetric	Three different configurations of obstacles	2025

Note: L is the facility length, W is the facility width, h_u is the upstream water depth, h_d is the downstream water depth, and N indicates that no obstruction structures are present downstream.

presents studies that focus on the fundamental physical characteristics of partial dam-break floods. These studies primarily aim to investigate the generation and propagation mechanisms of dam-break waves. While most experiments are conducted in smooth horizontal flumes, some explore mildly sloping downstream beds, and a small number consider mobile beds downstream.

Table 1 comprises 14 columns. Columns 1 and 14 list the reference sources and their year of publication, respectively.

Columns 2 and 3 present the experimental model’s length (L) and width (W), with the length referring to the entire model, including the upstream reservoir and downstream flume. Since the near-field characteristics of partial dam-break floods are the primary focus, most studies employ relatively short model lengths, with only a few using extended flume setups.

Columns 4 and 5 list the initial upstream and downstream water depths, while Column 8 shows the ratio of downstream to upstream water depth. Both dry-bed and wet-bed scenarios have been investigated in the literature. In wet-bed cases, the depth ratio is typically below 1/3, with only a few exceptions showing higher ratios.

Columns 6 and 7 indicate the bed roughness and slope of the flume. In most collected cases, the flume bed is smooth and horizontal. A few studies introduce slight bed slopes, and only one case involves a mobile bed condition downstream.

Columns 9 and 10 report the ratio of breach width to flume width and the location of the breach. In the majority of cases, this ratio is below 0.2, and in some cases below 0.5. Most breaches are located at the center, while a few are arranged asymmetrically.

Column 11 refers to the downstream boundary conditions. Special downstream settings are introduced to replicate realistic scenarios, such as flood impacts on downstream walls, buildings, or urban structures.

Table 2. Measured data, measuring technique, and focus of 3D partial dam-break flow tests.

Reference	Measured Data	Measuring Technique	Focus
Tingsanchali and Rattanapitikon [40]	Wave front propagation; water depth hydrographs; outflow; downstream outflow velocity	Video camera; water depth gauges; mini-current meter	Develops a 2D depth-averaged hydrodynamic model to simulate dam-break wave propagation over a dry flood plain. Accurately predicts wave-front propagation, spreading pattern, and depth hydrographs under various flow conditions. Uses a two-step predictor–corrector finite-difference scheme. Validates results via laboratory experiments and conducts sensitivity analysis on parameters such as Manning’s roughness coefficient.
Fraccarollo and Toro [41]	Pressure; water depth; flow velocity components	Pressure transducers; capacitance wave meters; electromagnetic velocity meters	Focuses on experimental and numerical assessment of a shallow-water model for 2D dam-break problems. Presents experimental/numerical results on dry-bed flows. Evaluates model validity using a Godunov-type scheme. Utilizes a physical laboratory model to validate simulations.
Jovanović and Djordjević [42]	Water depth profiles	Water depth probes; video camera	Verifies the MacCormack explicit scheme for simulating unsteady open-channel flows. Examines performance in 1D and 2D dam-break scenarios. Compares simulations with lab data, emphasizing accuracy and conservation. Demonstrates ease of use and suitability for dam-break modeling.
Stelling and Duinmeijer [43]	Water depth hydrographs; waterfront propagation	Water depth resistance probes; video camera	Presents a numerical method for simulating shallow-water flows, including hydraulic jumps and bores, using staggered grids and implicit integration. Ensures mass conservation and eliminates flooding-drying issues. Validates accuracy via lab experiments on dam break and flooding.
Shige-eda and Akiyama [44]	Wave front; flow depths; surface velocity; forces	Video camera; PTV; load cell	Numerically and experimentally investigates the behaviour of 2D flood flows, with and without structures. Specifically examines hydrodynamic forces on structures. Verifies the FUF-2DF model against data on front positions, flow depths, surface velocities, and hydrodynamic forces. Demonstrates that the model can predict structural forces with reasonable accuracy.
Eaket et al. [45]	Water-surface profiles; flow velocity	PTV; video camera	Investigates the use of video stereoscopy for measuring unsteady dam-break flows. Determines 3D surface profiles and velocities using stereo images. Develops a prototype system to measure hydraulic parameters and guide optimal setup. Explores accuracy and limitations for depth and velocity in dynamic conditions.
Soares-Frazão et al. [46]; Soares-Frazão and Zech [47]	Water levels; water-surface profiles; surface velocity	Water level gauges; video camera; Voronoï PTV	Provides data on how an isolated obstacle (building) affects a dam-break wave. Describes setup with a rectangular obstacle downstream of a simulated dam. Uses water level gauges, Doppler velocimeters, and imaging to analyze surface velocity. Supports validation of models for transient flows over urban topographies (IMPACT project). Analyzes wave interaction with the building, including flow deflection and hydraulic-jump formation.
Bukreev [37]	Water level profiles; depth hydrographs; longitudinal and vertical velocities	Video camera; PIV	Presents experimental data on discharge after partial/total dam breaks. Shows that partial breaks with rectangular breaches yield higher discharge per unit width. Investigates flow regimes, critical depth, and velocity classification. Analyzes velocity patterns and quasistationary transitions within breach sections. Highlights effects of reflected waves and confirms supercritical flow behavior.
Szydlowski and Twarog [48]	Water depth	Pressure transducers; water depth gauge	Focuses on numerical simulations of flood wave propagation in urban areas due to river embankment failure. Predicts flash-flood parameters to identify inundation zones using shallow-water equations and a finite-volume method. Investigates both a lab-scale model and a real urban topography. Assesses building influence by comparing cases with and without structures. Provides a dataset for hazard mapping and flood mitigation.
Liang et al. [49]	Water depth hydrographs; wave front position; flow velocity	Video camera	Develops an efficient shock-capturing numerical model for flood flows from levee breaches. Uses a predictor–corrector MacCormack scheme with TVD enhancement. Validates the model on lab experiments, showing strong agreement. Demonstrates practical application on the Thames River, highlighting capability in simulating flood flows over complex terrain.
Aureli et al. [50]	Water-surface profiles; water depth	Video camera; ultrasonic distance meters	Presents experimental and 2D numerical results of dam-break tests with sudden sluice-gate removal. Acquires water depth data via backlit imaging and ultrasonic sensors. Discusses method accuracy (within 20% deviation in most cases). Simulates using a 2D MUSCL-Hancock finite-volume model. Concludes the model and effectively reproduces main flow features despite local differences.
Yang et al. [51]	Water depth; surface velocity	PIV; pressure probe; video cameras	Uses a 3D unsteady-Reynolds model with volume of fluid (VOF) technique to predict near-field dam-break flows and impact forces. Validates against lab data for pressure, velocity, and water depth. Demonstrates the need for 3D modeling in turbulent, non-hydrostatic near-field flow scenarios.
Aureli et al. [33]	Water depth	Video camera; ultrasonic distance meters	Introduces an imaging technique that measures the free surface of rapidly varying dam-break flows using light absorption and digital image processing. Involves coloring the water as a variable-density filter and capturing images with a downstream-facing digital camera and backlit setup. Assesses validity by comparing water depth time series from this technique with those from ultrasonic distance meters. Indicates accuracy comparable to ultrasonic transducers. Demonstrates its effectiveness for collecting high-resolution, spatially distributed data. Recommends JPEG format for optimal time resolution.
Aureli et al. [52]	Impact force	Load cell	Evaluates the force exerted by a dam-break wave impacting a structure. Compares three mathematical models: a 2D depth-averaged model, a 3D Eulerian two-phase model, and a 3D SPH model. Conducts experiments simulating flip-through impacts against a rigid structure. Analyzes load time history and repeatability. Shows 3D models reproduce key features better than 2D models. Performs sensitivity analysis and discusses uncertainty. Makes dataset available online.
Elkholy et al. [20]	Pressure head; water-surface elevations; surface velocity; velocity profile	Pressure sensors; PTV; ultrasonic velocity profiler	Presents experimental results on partial-breach dam-break flows. Uses PTV and stereoscopic cameras to measure free-surface velocity and 3D velocity fields. Employs pressure transducers and ultrasonic velocity profilers for pressure heads and velocity profiles. Provides a non-dimensional relationship for upstream reservoir water elevation. Reports on wave front speed, lateral propagation, and pressure-head variations.
Honglu Qian et al. [53]	Water level; bed topography; sediment composition	Water level probe; video camera	Presents a new experimental dataset on partial dam-break floods over mobile beds. Measures water level evolution, bed topography, and sediment composition. Studies effects of uniform and non-uniform sediment on bed deformation. Analyzes influence of upstream/downstream water level differences. Finds that non-uniform sediment is more erodible and results in more severe scour and deposition.
Cordero et al. [54]	Water surface; waterdepth time series; water depth profiles	Video camera	Focuses on dam-break wave propagation along a hillslope using a physical model. Reconstructs water surface, assesses flooded area shape, and determines wave arrival time. Captures images with a high-resolution CMOS camera, mapping pixel intensity to water depth. Produces spatially distributed data to validate dam-break flood estimation methods.
Liu et al. [55]	Water level	Pressure gauge; ultrasonic distance meters	Uses physical model experiments to replicate flood flow around/through a house. Measures water levels with pressure and ultrasonic sensors. Analyzes the influence of house orientation and proximity on flood impact. Examines door forces from flow velocity and pressure.
Chumchan and Rattanadecho [56]	Flow images	Video camera	Studies dam-break flow through complex obstacles using experiments and simulations. Applies FVM and LBM with LES and VOF for free-surface tracking. Compares high-speed camera data with numerical results. Finds LBM more computationally efficient than FVM.
Kocaman et al. [58]	Wave front; water depth	Video camera; ultrasonic distance meters	Examines 3D dam-break wave propagation in enclosed domains under varying bottom conditions. Includes laboratory experiments and FLOW-3D simulations solving RANS and SWEs. Compares measurements with models. Analyzes positive/negative wave interaction and reflection. Highlights tail-water depth effects, an underexplored area.
Dong et al. [57]	Water hydrographs; flow velocity time series; drainage discharge time series	Ultrasonic distance meters; electromagnetic velocity meter; electromagnetic flowmeters	Combines laboratory experiments and numerical modeling to study flash-flood processes in urban streets. Analyzes how different layouts and infrastructures influence inundation via water depth and flow velocity measurements. Uses a 2D shallow-water model, examining mesh resolution and inlet discharge formulas. Shows sewer systems reduce surface water depth and flood-wave speed. Proposes modeling strategies for urban flood prediction.
Kocaman et al. [58]	Water surface; water depth time series	Video camera; ultrasonic distance meters	Investigates 3D dam-break wave propagation under dry and wet bottom conditions. Includes lab experiments and FLOW-3D simulations (RANS + SWEs). Compares measurements with simulations. Analyzes positive/negative waves and boundary reflections. Emphasizes tail-water depth effects, a less-studied aspect.
Erduran et al. [59]	Water surface; flood shock wave propagation	Video camera	Focuses on experimental and numerical studies of partial dam-break waves. Examines wave speed and pattern with dry and wet downstream beds and varying reservoir levels. Uses horizontal flume experiments and image-processing techniques. Simulates with FLOW-3D and a 2D model. Compares 2D and 3D model results to experimental data for understanding wave behavior.
Elisa Beteille et al. [27]	Water hydrographs; instantaneous free surface flow velocity around obstacles	Resistive wave gauge; acoustic wave gauge; LSPIV	Presents experimental datasets on dam-break flows over various obstacle configurations. Establishes benchmarks for validating numerical models in urban flash floods. Compares obstacle size and layout effects. Demonstrates performance of Code-Saturne CFD solver and VOF method. Emphasizes need for high-quality validation data and reveals challenges in matching simulation with experimental observations.

Note: Finite Volume Method (FVM), Lattice Boltzmann Method (LBM), volume of fluid (VOF), large eddy simulation (LES), particle tracking velocimetry (PTV), ultrasonic velocity profilers (UVP), Reynolds-Averaged Navier--Stokes (RANS), Shallow-Water Equations (SWEs), and Smoothed Particle Hydrodynamics (SPH)

comprises four columns. Columns 2 and 3 summarize the measured data types and the corresponding measurement techniques. The commonly recorded data types in current research include the following:

• Wavefront propagation and water depth variation curves such as wavefront position, water level fluctuations, and corresponding flow velocities;
• Water-surface and depth profiles including depth variation curves, free surface profiles, and velocity profiles;
• Flow field or surface contour images describing the spatial evolution of flood waves;
• Multi-point time series of water level or depth capturing dynamic changes at various spatial locations;
• Pressure and impact force measurements related to flood forces, such as pressure fluctuations on the flume bed or impact on obstacles;
• Flow velocity and surface velocity measurements involving temporal variations of surface or local velocity fields;
• Sediment transport and bed morphology used to analyze bed evolution and sediment dynamics under flood action.

The commonly used measurement techniques include the following:

• Video-based techniques such as standard high-speed cameras (ranging from 25 fps to 300 fps), digital video recorders, and particle tracking velocimetry (PTV, Voronoï PTV);
• Water level sensors including resistive gauges, capacitive wave gauges, ultrasonic rangefinders, and pressure transducers;
• Flow velocity and profile measurement devices such as miniature current meters, electromagnetic flowmeters, ultrasonic velocity profilers, and Acoustic Doppler Velocimeters (ADV);
• Load and pressure sensors used to capture forces and pressures induced by flood waves;
• Advanced image processing and sensing technologies such as RGB-D (Red-Green-Blue-Depth) sensors, digital imaging systems, and Large-Scale Particle Image Velocimetry (LSPIV).

Column 4 summarizes the main research focus. Most investigations prioritize the development and validation of hydrodynamic models, particularly two-dimensional shallow-water or depth-averaged frameworks, often employing advanced numerical techniques such as MacCormack, Godunov-type, or finite-volume schemes. Rigorous verification and validation are typically conducted through comparison with comprehensive laboratory datasets, ensuring the reliability and applicability of computational approaches. Parallel to numerical advancements, many studies emphasize the design and implementation of laboratory experiments to capture critical parameters such as water-surface elevation, velocity fields, and impact forces. Innovations in measurement techniques—including PIV, and synchronized multi-instrument setups—are frequently highlighted for their role in improving the accuracy and richness of experimental datasets. Another significant theme centers on the analysis of dam-break dynamics under various breach geometries and boundary conditions, including partial versus total breaches and the influence of roughness or structural features. The impact of obstacles, buildings, and urban topography is increasingly recognized as a key factor, with several studies focusing on urban flood modeling and the quantification of inundation hazards in built environments. Finally, a subset of research delves into the measurement and modeling of local pressure and force distributions, providing valuable insights into the interaction between flood waves and structures.

3. Results and Discussion

The study of three-dimensional (3D) partial dam-break flows encompasses a wide range of research fields, including breach dynamics, wave impacts on hydraulic structures, and downstream morphological evolution [30]. These investigations highlight the complex, multidisciplinary approach necessary for understanding and managing the intricacies of dam-break events. Experimental efforts are typically designed to replicate real-world scenarios by varying key parameters, such as upstream water depth, bed slope, and breach width, which are essential for accurately modeling flood dynamics [54,60]. Recent advancements in laboratory research now include the integration of urban features and obstacle interactions, improving the relevance of these studies for practical engineering applications [61]. However, physical model dam-break tests are highly effective but are subject to scale-induced distortions, such as the dominance of viscous and capillary forces [62,63], poor air-entrainment scaling [64], limited measurement resolution [65], geometric simplifications [51,66], and mismatches in sediment mobility [67]. Additionally, laboratory and boundary constraints can alter depths, velocities, and entrainment rates [51,68]. In particular, Froude scaling is often employed to preserve the gravity-inertia balance; however, this choice prevents the simultaneous matching of Reynolds and Weber numbers, leading to systematic scale effects that can modify flow depth, velocity, and two-phase behavior [62,64]. Small model depths and velocities amplify viscous and capillary forces compared to prototype values, resulting in measurable deviations when extrapolating the results to full scale [66].

A statistical analysis of recent experimental studies, as summarized in Table 1, reveals that most studies employ narrow flumes, ranging in length from 1 to 80 m and widths between 0.2 and 8.3 m. Upstream water depths typically vary from 0.15 to 0.4 m, while the downstream beds are generally dry or only slightly wet. The majority of studies utilize flat bed slopes, with a length-to-width ratio typically under 10, although some studies explore larger configurations. Breach widths generally account for approximately 20% of the flume width, with breach ratios ranging from 5% to 50%. Symmetric breaches along the channel centerline are predominant, while asymmetric and semi-symmetric configurations remain underexplored. A significant portion of studies employs smooth, fixed beds, while the use of mobile or rough beds is less common. While some experiments maintain standard downstream conditions, an increasing number are incorporating obstacles and complex boundary features, reflecting a growing interest in flood–structure interactions and urban flood dynamics. These diverse experimental setups have been fundamental in model validation while increasingly pushing toward the reproduction of real-world complexities.

Overall, standardized configurations—such as smooth beds, symmetric breaches, and dry downstream conditions—facilitate experimental repeatability and robust model validation. However, critical gaps remain, particularly for extreme breach ratios (i.e., <0.05 or >0.5), deeper downstream water depths, and steeper bed slopes. Realistic configurations, including mobile beds, asymmetric breaches, rough surfaces, and shaped dam bodies, remain under-represented in the experimental literature. Recent studies have increasingly incorporated urban layouts, stepped boundaries, and intricate obstacle arrangements, marking a shift toward more practical, engineering-oriented experimentation. Future research should focus on asymmetric breach dynamics, sediment–flow interactions, and propagation over irregular terrain to facilitate more comprehensive flood risk assessments.

A summary of the key measured parameters, measuring technique, and thematic research focus is presented in Table 2. Accurate hydraulic measurement remains a central component of dam-break research. Key parameters such as water-surface elevation (hydrographs, time series, and profiles), wavefront propagation, and flow velocity fields are crucial for analyzing flood wave dynamics and flow-structure interactions. Recent studies have expanded to include pressure measurements, impact forces, and bed topography, offering a more comprehensive perspective on the complex, multiphysical processes involved. However, in 3D partial dam-break tests, capturing near-field hydraulic parameters remains a significant challenge. Internal velocity components, vorticity, longitudinal water depth variations, and air-entrained hydraulic jumps near sidewalls remain inadequately measured, with no well-established datasets or methodologies available.

Progress in non-intrusive measurement techniques over the past decade has been particularly notable. Image-based methods, such as LS-PIV, PTV, and ultrasonic velocity profiling, have dramatically improved the accuracy of free-surface and surface velocity measurements [47,69,70]. These optical techniques enable detailed, flow-preserving measurements of temporal evolution [33], achieving unprecedented spatial and temporal resolution for surface flows, thereby enhancing our ability to investigate complex hydrodynamics [20,71]. Despite these advances, these techniques face limitations due to lighting requirements, image resolution, and the challenges of post-processing. Moreover, subsurface flow measurements remain constrained.

Intrusive instruments—such as water depth gauges, pressure probes, and electromagnetic velocity meters—remain indispensable for capturing depth, pressure, and velocity data. While these tools may disrupt flow to some extent, they are essential for 3D breach tests, particularly in capturing internal velocity distributions. For unsteady flows, both intrusive and non-intrusive techniques face challenges in capturing key features during transient, high-speed events. Leng et al. [72] proposed using repeated measurements to enhance accuracy, though such methods tend to be resource-intensive.

Moreover, the measurement instruments commonly used in dam-break experiments have their own limitations and error sources:

High-speed cameras: Although they can capture the rapid evolution of the free surface, their accuracy is highly sensitive to lighting conditions, viewing angles, and refraction. The tradeoffs among frame rate, exposure (motion blur), particle image signal-to-noise ratio, and particle displacement between frames collectively affect localization and trackability [73]. Multi-pulse timing directly controls pathline accuracy, where poor timing (a too large $\mathsf{\Delta }t$ or badly chosen multi-pulse scheme) increases reconstruction error in fast transient flows. Rapid shuttering to avoid motion blur reduces the per-frame signal and worsens particle image localization uncertainty [74]. Data storage and post-processing are demanding, while particle tracking may suffer from occlusion and imaging noise. Single-camera depth methods (defocusing, astigmatic) trade depth sensitivity against robustness, and depth calibration curves can produce large depth errors if optical conditions change during fast free-surface events [75,76].

Acoustic Doppler Velocimeter (ADV): This provides point velocity measurements, but its performance deteriorates significantly under strong turbulence, air entrainment, or sediment-laden flows, leading to spurious signals or data gaps. Furthermore, its single-point measurement cannot fully represent the flow field, limiting its ability to capture spatial flow variations in complex dam-break scenarios.

Wave gauges: These are useful for monitoring local water level variations, but are highly affected by bubbles and floating debris in high-velocity dam-break flows. Point wave-gauges and intrusive probes in high-velocity, aerated dam-break flows are vulnerable to measurement corruption when gauges interact with air–water interfaces, producing corrupted signals in highly aerated impact events. Protruding instrumentation can modify the local flow and become damaged or fouled by debris, limiting reliability in violent dam-break scenarios [77]. In addition, they only provide point information and cannot resolve the overall water-surface morphology.

In summary, while each instrument offers unique advantages, their inherent limitations and error sources highlight the necessity of complementary measurements, careful calibration, and error-propagation analysis to enhance the reliability of experimental results. The propagation of errors from particle localization and tracking to velocity fields and derived quantities is non-linear, with particular spatial/temporal error patterns producing disproportionately large downstream errors [78]. Methods that explicitly model localization uncertainty or avoid worst-case error patterns have been shown to improve robustness substantially [74].

Looking ahead, hybrid measurement systems that combine optical, acoustic, computational, and communication-based methods show significant promise in overcoming existing limitations. For example, particle tracking could evolve to incorporate GNSS RTK, IMU, and SLAM technologies to track particles in real time, reconstructing their spatiotemporal trajectories and offering new insights into complex flows and sediment transport dynamics [79].

When calibrated rigorously using experimental datasets, numerical models have shown strong predictive capabilities [22,80]. In current research, the development and application of numerical simulation methods constitute one of the core directions in dam-break flood studies. Early works mainly focused on one- and two-dimensional shallow water equation (SWE) models [40,41,42], with governing equations expressed as:

$${\displaystyle \frac{\partial \mathbf{U}}{\partial t}}+{\displaystyle \frac{\partial \mathbf{F}\left(\mathbf{U}\right)}{\partial x}}+{\displaystyle \frac{\partial \mathbf{G}\left(\mathbf{U}\right)}{\partial y}}=\mathbf{S}\left(\mathbf{U}\right),$$

(1)

where

$$\mathbf{U}=\left[\begin{array}{c}h\\ hu\\ hv\end{array}\right],\mathbf{F}\left(\mathbf{U}\right)=\left[\begin{array}{c}hu\\ h{u}^2+\frac{1}{2}g{h}^2\\ huv\end{array}\right],\mathbf{G}\left(\mathbf{U}\right)=\left[\begin{array}{c}hv\\ huv\\ h{v}^2+\frac{1}{2}g{h}^2\end{array}\right],$$

with $\mathbf{U}$ denoting water depth and flux terms. Based on finite difference, finite volume, Godunov-type methods, and the explicit MacCormack scheme, these studies simulated the wave-front propagation, water depth evolution, and velocity distribution of dam-break floods [43,44].

With increasing computational power, more efficient and stable algorithms have been developed, such as the MacCormack scheme combined with TVD (Total Variation Diminishing) techniques, the MUSCL-Hancock scheme, and staggered-grid implicit methods [50,51]. These algorithms ensure mass conservation and non-negative water depth, while improving the ability to capture rapidly varied flows (e.g., hydraulic jumps and shock waves). The numerical flux is often expressed in a high-resolution form:

$$F_{i+\frac{1}{2}}=F(U_i,U_{i+1})-\frac{1}{2}\varphi \left(\theta \right)(U_{i+1}-U_i),$$

(2)

where $\varphi \left(\theta \right)$ is the limiter function.

In recent years, research has increasingly turned to three-dimensional numerical simulations [20,52,53]. A typical approach is based on the Reynolds-averaged Navier–Stokes (RANS) equations:

$${\displaystyle \frac{\partial u_i}{\partial t}}+u_j{\displaystyle \frac{\partial u_i}{\partial x_j}}=-{\displaystyle \frac{1}{\rho }}{\displaystyle \frac{\partial p}{\partial x_i}}+\nu {\displaystyle \frac{{\partial }^2{u}_i}{\partial x_j^2}}-{\displaystyle \frac{\partial \overline{u_i^{\prime }{u}_j^{\prime }}}{\partial x_j}},$$

(3)

combined with the volume of fluid (VOF) method for free surface tracking:

$${\displaystyle \frac{\partial C}{\partial t}}+\mathbf{u}·\nabla C=0,$$

(4)

where C represents the fluid volume fraction. These methods, supplemented by the LES (large eddy simulation), LBM (Lattice Boltzmann Method), and SPH (Smoothed Particle Hydrodynamics), demonstrate significant advantages in capturing non-hydrostatic pressure distributions, turbulent structures, and local impact forces [55,56,58], albeit at the cost of higher computational demand. Overall, the evolution of numerical methods has enhanced model accuracy and applicability, providing robust technical support for simulating flood propagation and disaster assessment under diverse scenarios [54].

Although three-dimensional approaches such as RANS, LES, and SPH offer superior capability in capturing local-scale details, they remain computationally expensive [81,82]. Marsooli and Wu [81] demonstrated that 3D RANS+VOF models coupled with sediment transport provide improved accuracy for near-front dynamics and morphological changes compared to depth-averaged models, but at significantly greater computational cost. Similarly, Hien and Chien [83] showed that 3D CFD (RANS/LES) produces more realistic impact forces and local flow patterns than 2D shallow-water models, particularly for obstacle interactions, though this enhanced accuracy comes with substantial computational overhead. By contrast, traditional one- and two-dimensional shallow water models are advantageous in terms of efficiency, as they can reproduce general flood propagation and inundation extent with far lower CPU requirements while maintaining acceptable accuracy for many planform inundation metrics [84].

Future research is expected to move toward the development of multi-scale hybrid frameworks: employing two-dimensional models for large-scale flood propagation to ensure computational efficiency, while incorporating three-dimensional or particle-based methods in localized regions to resolve impact forces, turbulence structures, and complex obstacle interactions [85,86,87]. Balstrøm and Jensen [85] successfully demonstrated such an approach through their target-specified sub-modeling framework, which combines selective high-resolution 2D sub-models with coarser models to achieve large speedups with minimal accuracy loss. The SERGHEI modeling framework [86] exemplifies this multi-scale hybrid design philosophy by integrating modular hydrodynamics, transport, and morphodynamics with scalable multi-module execution capabilities on both CPUs and GPUs.

Furthermore, coupling hydrodynamic models with processes such as sediment transport, bank erosion, and debris motion will substantially enhance the realism of simulations [88,89,90]. Biswal and Moharana demonstrated the critical importance of sediment transport coupling in dam-break modeling, showing significant effects on flow dynamics during initial stages [88]. Tayfur further illustrated how coupled sediment–flow models over movable beds provide more realistic representations of dam-break scenarios involving earthfill dams and associated erosion processes [89].

In addition, using efficient algorithm programming to simulate dam-break flows can efficiently predict hydrological parameters [91]. And with the rapid advancement of computational power, parallel computing and GPU-accelerated algorithms are anticipated to become central tools for large-scale urban and regional flood risk assessments. Khrapov and Khoperskov [92] demonstrated the effectiveness of GPU-parallel implementations for self-consistent modeling of shallow water dynamics and sediment transport, while Bates et al. [93] showed significant acceleration in urban flood modeling using GPU-parallel non-uniform structured grid approaches. Leveraging cloud computing and distributed platforms, near-real-time dam-break scenario prediction can be achieved, significantly improving the practicality of numerical simulations in disaster prevention and emergency response.

This integration of experimental and numerical approaches enhances our understanding of both macro- and micro-scale processes in dam-break dynamics and sediment transport. However, numerical models still often rely on simplifying assumptions, such as homogeneous dam materials, simplified sediment–water interactions, and neglecting multiphase flow effects [89]. These assumptions lead to discrepancies between model predictions and observed behaviors, underscoring the need for continuous refinement of models and the development of more advanced simulation tools. Maranzoni [82] provides a comprehensive review highlighting these limitations in current 3D dam-break modeling approaches, emphasizing how simplified assumptions about material properties and flow physics can affect model accuracy.

A significant challenge in advancing model reliability is the scarcity of high-quality, comprehensive experimental datasets. Vosoughi et al. [94] explicitly noted that experimental observations for dam breaks with high sediment loads are sparse and presented new laboratory datasets to improve model validation, directly highlighting this critical gap in available validation data. Addressing this issue will require increased investment in high-resolution measurement technologies and stronger collaboration between experimental and modeling communities. Only through such interdisciplinary integration can we develop robust, validated models that support predictive flood analysis and improve hydraulic infrastructure management.

4. Conclusions

This review of 24 experimental studies (1993–2025) confirms that, despite the broad range of scenarios investigated, several critical gaps remain unaddressed. In particular, no experiment to date has systematically examined the influence of breach width (often characterized by a breach width coefficient) on the ensuing flood dynamics, leaving the role of breach size in partial dam-break flows poorly quantified. Likewise, physical models have yet to incorporate realistic three-dimensional dam-body geometries: nearly all experiments rely on simplified planar breaches in straight flumes, without representing a dam’s full longitudinal extent or structural complexity. As a result, current laboratory setups—while invaluable for fundamental insights—remain idealized and may not fully capture certain flow behaviors and scale effects present in real-world dam failures.

Another major limitation of the existing experiments is the scarcity of detailed internal flow-field measurements. Many studies have provided benchmark data for validating numerical models, but these datasets typically focus on bulk metrics such as water-surface elevations, wavefront travel times, or breach outflow hydrographs. Far fewer experiments report comprehensive velocity fields or turbulence measurements, especially in near-field zones immediately downstream of the breach where flow is highly three-dimensional and transient. The absence of high-resolution velocity data in these critical regions diminishes the utility of the experiments for advanced model calibration. It constrains our ability to develop and validate high-fidelity numerical simulations of dam-break floods, since key hydrodynamic processes in the immediate breach vicinity are not empirically captured.

Improving data accessibility is equally essential to accelerate progress in this field. At present, many experimental results remain available only in limited forms (e.g., figures within publications or proprietary reports), which hampers knowledge dissemination and cross-comparison. The establishment of open-access experimental databases for partial dam-break flows is urgently needed. Centralized, freely available repositories of well-documented test data would enable researchers and engineers worldwide to readily utilize past results for model benchmarking, validation, and further analysis. Such shared databases will not only prevent the redundancy of effort in future studies but also foster a more collaborative and cumulative scientific approach, benefiting both the academic community and practicing hydraulic engineers.

Despite these shortcomings in the current literature, the comprehensive nature of the present review provides a strong foundation for future advancements. By systematically cataloging and critically evaluating three decades of experiments, this review is, to the best of our knowledge, one of the most extensive compilations to date on partial dam-break flows. It offers a valuable reference for identifying specific knowledge gaps and highlights a suite of well-characterized test cases that can serve as rigorous benchmarks for numerical model development. These synthesized insights bridge the gap between physical and computational modeling, guiding modelers in selecting validation cases and informing experimentalists where new efforts are most needed. Importantly, the review also underscores clear priorities for improving experimental practice and data management, which are pivotal for translating laboratory findings into real-world flood mitigation strategies.

Future experimental studies should prioritize the key scientific and methodological gaps identified in this review, with a focus on the following directions:

• Incorporating realistic three-dimensional dam-body geometries and longitudinal scales in physical models to better replicate structural complexity and to systematically evaluate their influence on breach dynamics and downstream flow evolution.
• Expanding the range of tested scenarios to include extreme and complex conditions, such as heterogeneous dam materials, multiple or progressive breaches, and irregular topographies, thereby enriching the experimental database with more representative and practically relevant cases.
• Enhancing flow-field measurements in highly turbulent near-breach zones, where the flow is strongly three-dimensional and unsteady, through the use of fine-scale velocity and turbulence data to support high-fidelity numerical model calibration.
• Developing and deploying advanced, non-intrusive measurement techniques—such as high-speed 3D PIV, Laser-Induced Fluorescence (LIF), and PTV—to capture detailed spatiotemporal hydrodynamic information with improved resolution and accuracy.
• Establishing open-access, standardized experimental data repositories for partial dam-break flows, with detailed metadata and documentation, to enable global researchers to reuse, benchmark, and cross-validate datasets, fostering a more collaborative and cumulative scientific approach.

With these developments, the next generation of experiments will substantially improve the physical realism, calibration accuracy, and predictive value of dam-break simulations. This progress will, in turn, strengthen our capacity for risk-informed design, flood hazard mitigation, and emergency response planning, thereby providing robust scientific support for managing real-world dam-failure scenarios.