Two-Dimensional Flood Modeling of a Piping-Induced Dam Failure Triggered by Seismic Deformation: A Case Study of the Doğantepe Dam

Abstract

This study presents a scenario-based, two-dimensional flood modeling approach to assess the potential downstream impacts of a piping-induced dam failure triggered by seismic activity. The case study focuses on the Doğantepe Dam in northwestern Türkiye, located near an active branch of the North Anatolian Fault. Critical deformation zones were previously identified through PLAXIS 2D seismic analyses, which served as the physical basis for a dam break scenario. This scenario was modeled using the HEC-RAS 2D platform, incorporating high-resolution topographic data, reservoir capacity, and spatially varying Manning’s roughness coefficients. The simulation results show that the flood wave reaches downstream settlements within the first 30 min, with water depths exceeding 3.0 m in low-lying areas and flow velocities surpassing 6.0 m/s, reaching up to 7.0 m/s in narrow sections. Inundation extents and hydraulic parameters such as water depth and duration were spatially mapped to assess flood hazards. The study demonstrates that integrating physically based seismic deformation data with hydrodynamic modeling provides a realistic and applicable framework for evaluating flood risks and informing emergency response planning.

1. Introduction

Dams are large-scale engineering structures serving multiple purposes such as water storage, flood control, irrigation, and energy production. Over time, these structures become vulnerable to various risks due to both natural disasters and structural degradation. In particular, earthquakes, intense rainfall, soil liquefaction, piping, and insufficient spillway capacity threaten the structural integrity of dams, potentially causing sudden failures that can lead to significant loss of life and property [1,2,3,4,5].

Dam failures often result from a combination of multiple engineering, geological, and hydrological factors. Floods generated after such failures can rapidly spread over large areas, causing severe impacts especially in settlements located near dams [6,7,8]. Therefore, scientifically based modeling of the temporal and spatial propagation characteristics of floods following dam failures has become a critical necessity for risk reduction and disaster management [9,10]. In agricultural regions with intensive land use, floods not only cause water inundation but also exacerbate environmental and economic impacts through soil erosion, damage to fertile farmland, and sediment transport processes [11,12].

Numerical hydraulic modeling is among the most commonly employed methods for such scenarios. Widely used modeling tools like HEC-RAS enable the analysis of flood behavior through both one-dimensional and two-dimensional simulations. Two-dimensional modeling, particularly under complex topographic conditions, contributes to more accurate determination of hydraulic parameters such as flow direction, velocity, and water depth [13,14,15]. Additionally, integrating HEC-RAS outputs with geographic information systems allows for detailed mapping of flood impact zones [16,17,18].

The flood generation mechanism is a key factor in modeling studies. Different failure types—such as piping, overtopping, and slide—significantly influence flood characteristics, making it essential to define appropriate input parameters for each scenario [19,20,21,22,23,24,25,26]. In critical scenarios, especially those involving sudden failure, outputs like flood wave arrival time, peak discharge, and inundation area play a decisive role in early warning system design [1,27].

Uncertainty in parameters used for dam failure modeling directly affects model accuracy. Consequently, recent studies have proposed approaches that account for scenario uncertainties and assess model sensitivities [28,29]. Moreover, deformation-focused analyses have been demonstrated to effectively identify initial conditions of earthquake-induced failures. Areas where deformation interacts with the surface are considered initiation points for piping-type failures, with flood propagation modeled as a time-dependent process accordingly [3,20].

Numerous studies have demonstrated that HEC-RAS 2D modeling can successfully simulate flood behavior related to various failure types. Model outputs—such as water depth, flow velocity, and propagation duration—are interpreted to provide decision-support data for areas at risk of disaster [14,17]. Some research has also highlighted the impact of surface detail on flood propagation by comparing digital surface models and digital elevation models to improve modeling accuracy [12]. Within this framework, modeling embankment dams using finite element methods and comparing these models with monitoring data is among the approaches applied for assessing structural behavior [30].

Located near the North Anatolian Fault Zone, one of Türkiye’s active seismic belts, the Geyve Doğantepe Dam is a critical structure exposed to high seismic hazard. Previous studies have analyzed the dam’s behavior under seismic loads using PLAXIS 2D (version 2024.2) software and have identified zones of structural weakness with engineering-based findings [31]. Building upon these seismic analysis results, the present study models a potential piping-type sudden failure scenario using HEC-RAS 2D (version 6.6) and evaluates flood impacts in the downstream region via two-dimensional hydrodynamic methods in the event of dam collapse. This study integrates seismic analysis to define physically based failure initiation points in HEC-RAS 2D modeling, improving the realism of simulated flood scenarios.

2. Materials and Methods

2.1. Study Area

The study area is the Doğantepe Dam, located within the boundaries of Geyve district in Sakarya Province, in the Marmara Region of Türkiye. The dam is situated on the Karakaya Stream, approximately 10 km from the center of Geyve district (Figure 1). According to the Turkish Earthquake Hazard Map, the region falls within the first-degree seismic zone and is located about 7 km south of the North Anatolian Fault Zone’s southern branch.

Although the Doğantepe Dam was constructed primarily for irrigation purposes, its seismotectonic setting and geological conditions render it a significant water structure with a considerable risk of earthquake-induced failure. Numerous settlements located downstream of the dam are vulnerable to flooding in the event of a potential failure. This situation necessitates an interdisciplinary approach not only to the structural safety of the dam but also to the assessment of flood propagation following a failure.

The Doğantepe Dam is designed as a concrete-faced rock-fill dam, and its structural and geometric characteristics are summarized in

Table 1. Technical specifications of Doğantepe Dam.

Feature	Value
Dam type	Concrete-faced sand–gravel embankment
Dam height (from thalweg)	50.50 m
Crest length	209.90 m
Parapet crest elevation	235.50 m
Minimum water level	214.15 m
Normal water level	230.60 m
Maximum water level	234.27 m
Reservoir volume at normal level	2.89 hm3

. The dam body is situated in a position that poses a potential flood threat to the settlements downstream.

The Doğantepe Dam is located in a region characterized by a high density of active fault lines and is situated approximately 700 m from the Geyve Fault. This geological feature is considered one of the primary factors influencing the dam structure’s behavior under seismic loading (Figure 1). Furthermore, the permeable sand–gravel fill material used in the dam body exhibits limiting characteristics in terms of internal stability, increasing the risk of failure mechanisms such as piping, slope instability, and deformation when subjected to earthquake effects [24]. Within the scope of this study, the topographic features of the dam site were evaluated using a digital elevation model and employed as fundamental data for flood propagation analysis (Figure 2). Terrain slope and surface morphology play a decisive role in determining the direction and propagation time of the flood wave following failure.

2.2. Structural Behavior of the Dam Body Under Earthquake Loading and Failure Scenario

Realistic modeling of flood impacts following dam failure requires defining the failure scenario based on physical foundations. The failure scenario developed in this study is based on previous numerical analysis findings regarding the behavior of the Doğantepe Dam under seismic loading. In this analysis, the dam body’s response to seismic loads was evaluated using PLAXIS 2D finite element software, and regions that may weaken the dam’s structural stability were identified [31].

Ground motions used in the dynamic analyses were selected according to the criteria in the “Guide for Seismic Parameter Selection in Dam Design” published by the State Hydraulic Works (DSİ) [32]. Real earthquake records matching the target scenario and contributing most to the target spectrum were preferred, with at least three different records selected. These records were taken from the PEER Ground Motion Database, considering key seismological parameters such as target magnitude (Mw 7.4–7.6), fault mechanism (strike-slip), source distance (~50 km), and site conditions (Vs30 ≥ 760 m/s). They were scaled using the SeismoMatch 2025 software.

The dam’s dynamic behavior was analyzed using PLAXIS 2D. Scaled earthquake records were applied under empty, minimum, and normal (operational maximum water level) reservoir conditions to evaluate the dam’s response under different loading states. Material models and parameters were chosen to reflect the dam’s geotechnical and structural properties. The embankment zones were modeled with the Hardening Soil model. The plasticity models in PLAXIS 2D simulate permanent deformations but do not fully capture the actual damping behavior of soils. To achieve more realistic results, additional Rayleigh damping was defined with a damping ratio of 5% in the model.

Based on these dynamic analysis results, the subsequent breach modeling focused specifically on the full reservoir level as the failure scenario. Within the dynamic analysis, ground motion records from the 1999 Düzce Earthquake were used as input for time-dependent nonlinear deformation calculations. Simulations conducted under the Mohr–Coulomb soil model and rigid boundary conditions revealed significant horizontal displacements, particularly near the upstream slope. The maximum deformation was detected at coordinates X = −3.833 m and Y = 230.6 m, which was identified as a potential weak zone.

The dam’s geometric structure and hydraulic operating levels were used as fundamental data during the flood modeling process. The cross-sectional view presented in Figure 3 shows the upstream and downstream slope gradients along with the minimum (214.15 m), normal (230.60 m), and maximum (234.27 m) water level elevations.

Piping is a failure mechanism characterized by the progressive development of internal erosion within the soil, which can lead to sudden and uncontrolled collapse of the dam body. This mechanism is particularly prone to activation under seismic effects in loosened and permeable soil conditions, posing a significant safety risk for embankment dams constructed with permeable materials [33,34,35]. Such internal deformations are often difficult to detect through direct observation and can rapidly progress, causing abrupt structural failures. Therefore, sudden failure scenarios caused by piping are considered priority cases in flood modeling studies [36].

The failure scenario developed in this study is based on the aforementioned analysis results and conceptualizes the earthquake-triggered internal erosion mechanism of piping. The simulation’s initial condition assumes the dam’s maximum operating water level (234.27 m), and the failure is represented by a breach opening at the weak zone identified in the PLAXIS analysis, with a base elevation of 210 m. It is assumed that the piping progresses over time and that the embankment collapse is completed within a short duration. This physical scenario serves as the fundamental input data for the breach modeling implemented in the HEC-RAS 2D environment.

2.3. Definition of Breach Parameters

For reliable determination of flood propagation in dam failure modeling, the breach geometry must be defined based on physical principles. In particular, parameters such as breach width, formation time, and side slope angle play a critical role in the development of the flood wave in piping-type internal erosion scenarios [15,37].

In this study, the piping-type failure scenario was constructed within the 2D module of HEC-RAS 6.6. software, based on the maximum deformation zone identified in previous PLAXIS 2D analyses. The definition of breach parameters was guided by one of the empirical methods supported by HEC-RAS, namely the Von Thun & Gillette approach. This method provides empirical relationships developed for embankment dams that estimate breach geometry based on physical variables such as dam height and reservoir volume [33,38].

Within this methodology, the average breach width (B_ave) is calculated using the following equation.

B_ave = 2.5·H_w + C_b

(1)

In this equation, B_ave represents the average breach width in meters; H_w is the effective dam height (m), defined as the elevation difference between the reservoir water surface and the dam base at the time of failure; and C_b is an empirical constant that varies according to the reservoir volume. According to guidance provided by FEMA [35], this constant is recommended as 6.1 m for small reservoirs, 18.3 m or 42.7 m for medium-sized reservoirs, and 54.9 m for large reservoirs. These parameters enable realistic modeling of breach geometry based on the dam’s physical characteristics and play a fundamental role in defining the initial conditions for flood simulation [15,33].

The breach formation time (t_f) can be estimated in different ways depending on the erosion resistance of the dam body. According to the Von Thun & Gillette method, this time is calculated using the average breach width (B_ave) and the effective dam height (H_w) through the following equations. If the dam has high erosion resistance, the formation time is expressed by the following formula:

$$t_f={\displaystyle \frac{B_{ave}}{4.H_w+61}}$$

(2)

For embankments with low erosion resistance, the formation time is expressed as follows:

$$t_f={\displaystyle \frac{B_{ave}}{4.H_w}}$$

(3)

Here, t_f, represents the breach formation time in hours. The erosion resistance criterion is generally determined based on literature recommendations considering the permeability, compaction level, and homogeneity of the embankment material [15,33,35]. Within the scope of this study, considering the sand–gravel fill characteristics of the Doğantepe Dam and the loosened soil conditions following the earthquake, a low erosion resistance scenario was selected, and the model was developed accordingly.

The breach formation time t_f is expressed in hours [33]. The side slope angle (z) is typically recommended in the literature to vary between 0.5 H:1 V and 1 H:1 V and is user-defined within the HEC-RAS interface.

In this study, the weak zone identified from PLAXIS analyses was used as a reference to define the initiation section of piping in HEC-RAS, and the breach geometry was modeled based on the Von Thun & Gillette method. In comparison, empirical methods such as those proposed by MacDonald and Froehlich rely primarily on statistical relationships based on dam height and reservoir volume, and were developed using datasets dominated by overtopping failure events [23,26,39]. These approaches offer generalized average predictions but do not explicitly represent site-specific failure mechanisms such as internal erosion or piping. In contrast, the Von Thun & Gillette approach is recommended in guidelines for modeling internal erosion scenarios [40]. It incorporates embankment type, material properties, and erosion resistance. When combined with site-specific analyses, such as the seismic-based weak zone identification used in this study, it offers a more suitable solution for piping-type failure modeling.

Table 2. Parameters for modeling the piping breach mechanism in dam failure.

Parameter	Description
Breach Location	Seismic deformation zones (PLAXIS 2D).
Breach Base Elevation	Weak zone from seismic assessment.
Breach Width	Von Thun & Gillette empirical method.
Side Slope Ratio (Z:1)	Typical for internal erosion breaches.
Formation Time	Low erosion resistance scenario.
Trigger Reservoir Level	Maximum operational level.
Breach Progression Type	Sudden failure representing piping.
Failure Mechanism	Piping-type internal erosion scenario.
Material Erosion Resistance	Low category, sand–gravel fill.

below summarizes the input parameters defined in the HEC-RAS model to represent this site-specific piping breach scenario.

2.4. HEC-RAS 2D Modeling

To realistically represent the spatial and temporal distribution of flood propagation in dam failure scenarios, this study employs the HEC-RAS 2D numerical modeling framework. The software operates based on a volume-conserving solver system derived from the shallow water equations, known as the Saint-Venant equations, and can simulate high-energy flow events such as sudden releases over complex topographies with high resolution [15,41]. HEC-RAS 2D is widely used for two-dimensional analyses of rapidly evolving hydraulic phenomena like dam breaks and spillway floods, and its accuracy has been demonstrated in various studies [16].

The model’s computational framework is grounded in the Saint-Venant equations, which describe the time-dependent behavior of two-dimensional flow based on the principles of mass and momentum conservation [42].

$${\displaystyle \frac{\partial h}{\partial t}}+{\displaystyle \frac{\partial \left(uh\right)}{\partial x}}+{\displaystyle \frac{\partial \left(vh\right)}{\partial y}}$$

(4)

$${\displaystyle \frac{\partial \left(uh\right)}{\partial t}}+{\displaystyle \frac{\partial \left(u^{2}h\right)}{\partial x}}+{\displaystyle \frac{\partial \left(uvh\right)}{\partial y}}+gh{\displaystyle \frac{\partial h}{\partial x}}=S_{fx}$$

(5)

$${\displaystyle \frac{\partial \left(vh\right)}{\partial t}}+{\displaystyle \frac{\partial \left(uvh\right)}{\partial x}}+{\displaystyle \frac{\partial \left(v^{2}h\right)}{\partial y}}+gh{\displaystyle \frac{\partial h}{\partial y}}=S_{fy}$$

(6)

The modeling domain was defined to encompass the Doğantepe Dam and the downstream valley, represented using a high-resolution digital elevation model (DEM). A computational grid composed of square cells with an approximate resolution of 20 × 20 m was generated over the digital surface. To ensure numerical stability, cell density was increased in hydraulically critical areas (Figure 4).

In the model, surface roughness was defined spatially using variable Manning’s n coefficients. For this purpose, the 2024 Land Use/Land Cover (LULC) raster data provided by Esri, based on Sentinel-2 satellite imagery, was utilized [43]. Manning’s n values were assigned to land cover classes within the raster data according to Chow (1959) and USGS (2019) guidelines [44]. The Manning’s n values used in this study were assigned to the land cover classes listed in

Table 3. Land cover classes and Manning roughness coefficients (n).

Land Cover Class	Manning’s n
Bare	0.025
Gravel Surface	0.03
Crops	0.045
Rangeland	0.04
Forest	0.10
Built Area	0.12

.

This classification process was performed using the Reclassify tool in ArcGIS. The resulting raster dataset was integrated into the HEC-RAS 2D environment, assigning spatially varying roughness coefficients to each cell (Figure 5). The Manning’s n values were selected from ranges recommended in the literature and assigned to accurately represent the known local field conditions. The study area is well understood in terms of its history of flooding and local hydraulic characteristics, and these local observations and prior modeling experience supported selecting coefficients consistent with local topography, land use, and channel features.

The failure scenario was modeled in the HEC-RAS 2D environment based on the weak zone identified through PLAXIS 2D analyses. In this process, the piping-type internal erosion mechanism was defined, and the breach geometry was configured using the “Storage Area Breach” module. Parameter determination was guided by the Von Thun & Gillette approach; geometric features and formation time were appropriately modeled considering the dam’s physical and geotechnical characteristics. The breach bottom and slope parameters were systematically defined according to the dam topography, with the triggering condition based on the water level. The modeling framework and breach configuration are presented in Figure 6.

Boundary conditions of the model were defined as a fixed water level on the reservoir side and normal depth conditions downstream. The numerical solution was performed using HEC-RAS’s solver based on the shallow water equations. The simulation duration was set to 6 h, with a computational time step of 5 s. Outputs were recorded at 1 min intervals, establishing the model’s temporal resolution accordingly.

The model results include hydraulic parameters such as water depth, flow velocity, flood propagation duration, and inundation extent. These outputs were exported in raster format and analyzed within a geographic information system (GIS) environment.

3. Results

The flood propagation resulting from the earthquake-induced piping failure of the dam body was numerically evaluated through simulations conducted with the HEC-RAS 2D model. The raster outputs obtained during the modeling process include hydraulic parameters such as water depth, flow velocity, and flood propagation duration. These parameters were processed and spatially visualized using a geographic information system (GIS)-based analysis environment.

The breach outflow hydrograph used as the upstream boundary condition in the two-dimensional flood propagation model is shown in Figure 7. This hydrograph was generated in HEC-RAS based on the defined breach parameters, including breach width, formation time, and bottom elevation. It describes the time-dependent variation of discharge (m³/s) resulting from the dam failure and determines the characteristics of the downstream flood wave. The hydrograph is characterized by a rapid rise to a peak discharge followed by a recession limb, representing the sudden release of the reservoir volume during the breach event. This boundary condition was applied in the two-dimensional flood propagation model.

The water depth data derived from the flood simulation were analyzed at three different time steps (30, 45, and 50 min) to assess the temporal development of the flood propagation. The model outputs indicate that the flood initially remains confined to the river channel but gradually expands towards low-gradient areas, spreading over a wider region.

At the 30 min mark of the simulation, floodwaters had advanced along the river channel downstream of the dam, with the flood extent largely constrained by channel morphology due to topographic conditions. Water depths predominantly ranged between 0.8 and 1.4 m, with localized accumulations exceeding 1.5 m occurring only near the riverbed in limited areas. At this stage, settlements and surrounding agricultural lands remained unaffected by flooding.

By the 45th minute, the floodwaters had overtopped the river channel and begun spreading toward nearby low-gradient plains. Significant water accumulation was observed particularly in agricultural areas, where depths were calculated between 1.5 and 2.2 m. In local depressions, water depth approached 2.5 m, and the flood impact exhibited a more widespread surface coverage.

At 50 min, flood propagation reached its maximum extent. According to the model results, the entire settlement area was encompassed within the flood boundaries at this time step. Water depths in these regions mostly ranged from 2.0 to 2.6 m, with accumulation zones reaching up to 2.8 m in low-gradient areas. At this stage, the flood reached its peak both in terms of inundation area and water accumulation. It was determined that all settlement areas were submerged within the model domain at this time.

The temporal variation of water depth indicates that the flood progressed rapidly and intensively along the channel during the first 30 min. After the 45th minute, the flood began to spread horizontally over wider areas, influenced by slope conditions. Notably, water depth increased significantly in low-lying regions, and the model results reveal that these areas are at high risk of flooding. The temporal distribution of flood depths is presented in Figure 8.

The maximum water depth map obtained from the modeling study was used to spatially assess the flood impact caused by the dam failure scenario in the downstream area. This map shows the highest water levels calculated at each point throughout the simulation, revealing the limits reached by the water and the accumulation depths in these areas, independent of the temporal progression of the flood. Figure 9 shows the overall flood extent and its progression downstream, illustrating the flood’s expansion along the river channel from Geyve center.

Analysis of the modeling results reveals that water depth values vary depending on land use types and topographic conditions. Figure 10 illustrates the maximum water levels calculated for the Geyve residential area and agricultural fields in the mid-downstream region. Water depths in residential areas typically range between 1.5 and 2.8 m, with localized accumulations exceeding 3.0 m in some low-lying zones. In agricultural areas, water depth exhibits a more uniform distribution, ranging between 2.5 and 3.5 m.

The simulation results show the distribution of water depths across different areas beyond the horizontal extent of flood propagation. The color scale represents water levels in meters, enabling simultaneous numerical and visual assessment of regional variations. Notably, elevated water levels concentrated around settlement areas indicate a higher likelihood of prolonged inundation during the flood.

The simulation results reveal that flow velocity varies significantly depending on the flood extent and regional factors such as land use and topography. To evaluate the development of the flood wave immediately downstream of the dam breach and its attenuation behavior along the river reach, a longitudinal velocity profile was generated from the HEC-RAS 2D simulation results. As shown in Figure 11, this profile illustrates the variation in surface flow velocity as a function of distance from the breach location. The results indicate that the flood wave initially exhibits higher velocity values near the breach point but gradually attenuates in the downstream direction due to changes in channel geometry and topographic slope.

This longitudinal velocity profile indicates that the flood wave reaches surface flow velocities of up to approximately 6 m/s immediately downstream of the dam breach. These velocities gradually decrease to around 2–3 m/s along the river reach due to changes in channel geometry and topographic slope. The model outputs indicate that maximum surface flow velocities provide important information regarding both the direction and intensity of flood propagation following dam failure. The highest velocity values recorded in each cell throughout the simulation were analyzed to generate a spatial distribution map. Figure 12 illustrates the overall distribution of maximum velocities during the flood, along with a detailed view of conditions around the Geyve urban area.

The flood wave advancing downstream from the dam moves at higher velocities, especially along sections following the river channel. Model results indicate that velocity values near the main flow path occasionally exceed 6.0 m/s, with localized peaks surpassing 7.0 m/s observed in narrow valley constrictions. In these areas, the flow demonstrates rapid and directional propagation aligned with the current.

As the flood wave approaches the center of Geyve district, a general decrease in flow velocities is observed due to the reduction in surface slope and increased urbanization. Detailed analysis reveals that velocity values around the settlement area typically range between 0.5 and 2.5 m/s, while increases up to 3.0 to 4.5 m/s are recorded in areas where the flow is directed into narrow streets. These velocity variations are attributed to the guiding effects of topography and structural obstacles on flood flow.

Four characteristic cross sections were identified to allow a more detailed assessment of flood wave progression and velocity profile variations under different topographic and hydraulic conditions. These sections include the narrow channel area near the dam breach, as well as the reach before the settlement area where the channel slope decreases and meandering patterns occur, thereby representing zones with distinct flood propagation behaviors. Figure 13 presents the locations of cross sections over the 2D surface velocity distribution map, together with the calculated water level and velocity profiles at each section.

The profiles at cross sections A and B indicate that, near the dam breach, flow velocities are higher while water levels remain relatively lower, reflecting the confined channel and steeper gradient in this upstream area. In contrast, cross sections C and D, located further downstream before the settlement area, show increased water levels accompanied by reduced flow velocities, consistent with a lower channel slope and broader, flatter topography. These cross-sectional results clearly illustrate the variation in flood wave characteristics along the modeled reach and describe differences in flow conditions.

The distribution of maximum flow velocity reveals that flood propagation exhibits significant regional variations not only in water depth but also in velocity profiles. The effects of factors such as surface slope, channel geometry, and urban density on flow velocities are particularly pronounced in settlement areas. These variations lead to notable changes in the movement and spread of water during the flood, intensifying the flood’s overall impact.

4. Discussion

The numerical and spatial analyses obtained in this study indicate that piping-type failures induced by seismic activity can cause high-energy floods in the downstream region of the dam within a short period. The increase in maximum water depth and flow velocity highlights the critical importance of the initial minutes following the failure. Therefore, dam safety should be assessed not only in terms of structural integrity but also within the framework of the predictability of flood propagation after the failure.

In embankment dams with permeable fill, sudden failures caused by internal erosion mechanisms—particularly piping—can lead to high-discharge floods with extensive impact zones. Such failures are typically triggered by factors such as soil loosening, permeable fill material, and stress variations induced by seismic activity [40,41]. Empirical and numerical modeling studies have demonstrated that in these scenarios, water depths may exceed 3 m, and flow velocities may surpass 5 m/s [40,42,43]. Exceeding these threshold values suggests that flood events triggered by piping-type failures may pose significant risks to residential areas and agricultural lands. The findings obtained in this study are consistent with these tendencies, allowing for the spatiotemporal simulation of one of the most adverse flood scenarios. In particular, the fact that the flood wave propagates within a matter of minutes underscores the necessity of integrating early warning systems and disaster management protocols into dam safety practices.

In addition to nonstructural measures, structural options for the downstream area should also be considered. Engineering work may include strengthening levees in high-risk sections of the river, maintaining or widening some channel sections to assist flow. When possible, bridge openings can be modified to reduce bottlenecks. Maintenance or upgrades of drainage and filter systems for the dam are important to help control seepage and reduce the risk of internal erosion or pipe failure, especially given the potential for seismic impacts. Combining these structural and non-structural measures provides a more complete approach to reducing flood risk to downstream residential and agricultural areas.

Two-dimensional hydrodynamic modeling techniques are primarily preferred for simulating flood propagation due to their ability to realistically represent interactions with terrain topography [40,41]. These methods have been found particularly effective in assessing the influence of downstream factors such as urban development density, complex channel morphology, and abrupt changes in terrain slope on flood behavior [7]. In addition to open-source software like HEC-RAS, commercial platforms such as MIKE21 and FLOW-3D are also widely used in similar applications. The choice of modeling software typically varies depending on factors such as user interface, data compatibility, and the scale of the study [42,43].

In this study, the HEC-RAS 2D version 6.6 was selected because it is developed by the United States Army Corps of Engineers (USACE) and supported by a breach module, allowing for scenario-based modeling of piping-induced dam failures based on physically defined parameters. The model outputs were transferred to a geographic information system (GIS)-based analysis environment, enabling spatial assessment of flood-related parameters such as areas subject to inundation, flow propagation duration, and maximum water depth. These outputs are valuable not only for engineering analyses but also for supporting emergency response strategies and informing disaster management planning.

In this model, the dam failure was assumed to initiate from a single breach point. While this assumption simplifies the analysis process, it does not account for real-world scenarios in which failure may occur through multiple breach locations or combined deformation mechanisms. This limitation is particularly relevant in large-scale embankment dams, where compound failures are more likely to develop. The literature emphasizes the importance of incorporating uncertainty analyses, Monte Carlo simulations, and sensitivity assessments in dam breach modeling to enhance reliability and evaluate a wider range of scenarios [5,40]. However, the scenario adopted in this study is based on the structural weakness identified in previous PLAXIS 2D seismic analyses of the Doğantepe Dam. Accordingly, the breach location and failure mechanism were modeled considering the most plausible development of piping triggered by earthquake-induced deformation.

The modeling approach adopted in this study enabled a physically grounded and practically applicable analysis of piping-induced dam failure scenarios. By integrating vulnerability zones identified through PLAXIS 2D into the HEC-RAS 2D environment, a realistic flood propagation model based on soil behavior was developed. This integrated framework extends beyond structural safety evaluations, allowing for the anticipation of potential flood impacts and supporting disaster preparedness efforts in areas exposed to inundation.

5. Conclusions

This study demonstrates that piping-induced dam failure scenarios triggered by seismic activity can be realistically analyzed through two-dimensional hydrodynamic modeling supported by physically based data. In the case of the Geyve Doğantepe Dam, the integration of vulnerability zones identified using PLAXIS 2D into the HEC-RAS 2D environment enables spatiotemporal evaluation of flood propagation following a potential breach.

The simulation results indicate that the flood wave rapidly reaches downstream areas, with water depths and flow velocities exceeding engineering design thresholds, particularly in residential zones. Additionally, prolonged inundation is observed in low-lying areas, indicating a higher potential for flood impact. These findings suggest that piping-type failures should be considered high-risk scenarios, and emergency response planning should be structured accordingly.

The presented modeling approach offers a practical method not only for engineering-based hydraulic analyses but also for developing scenario-driven insights to support flood preparedness in vulnerable regions. Evaluating flood behavior across time and space contributes to improving early intervention strategies. Future studies may enhance the reliability and applicability of this approach by simulating multiple breach locations, conducting sensitivity analyses on breach parameters, and repeating the modeling under varying topographic conditions.