Study of Dam-break Due to Overtopping of Four Small Dams in the Czech Republic

717 ACTA UNIVERSITATIS AGRICULTURAE ET SILVICULTURAE MENDELIANAE BRUNENSIS Volume 63 81 Number 3, 2015 http://dx.doi.org/10.11118/actaun201563030717 STUDY OF DAM-BREAK DUE TO OVERTOPPING OF FOUR SMALL DAMS IN THE CZECH REPUBLIC Zakaraya Alhasan1, Jan Jandora1, Jaromír Říha1 1 Institute of Water Structures, Faculty of Civil Engineering, Brno University of Technology, Veveří 95, 602 00 Brno, Czech Republic Abstract ALHASAN ZAKARAYA, JANDORA JAN, ŘÍHA JAROMÍR. 2015. Study of Dam-break Due to Overtopping of Four Small Dams in the Czech Republic. Acta Universitatis Agriculturae et Silviculturae Mendelianae Brunensis, 63(3): 717–729. Dam-break due to overtopping is one of the most common types of embankment dam failures. During the fl oods in August 2002 in the Czech Republic, several small dams collapsed due to overtopping. In this paper, an analysis of the dam break process at the Luh, Velký Bělčický, Melín, and Metelský dams breached during the 2002 fl ood is presented. Comprehensive identifi cation and analysis of the dam shape, properties of dam material and failure scenarios were carried out a er the fl ood event to assemble data for the calibration of a numerical dam break model. A simple onedimensional mathematical model was proposed for use in dam breach simulation, and a computer code was compiled. The model was calibrated using the fi eld data mentioned above. Comparison of the erodibility parameters gained from the model showed reasonable agreement with the results of other authors. Keywords: dam-break, embankment dam, numerical model, dam overtopping, peak discharge INTRODUCTION Overtopping is the most common cause of the failure of embankment dams. The statistics for embankment dam failures shows that this failure mode represents approximately 40% of all embankment dam failures (Jandora and Říha, 2008). Even if both large and small embankment dams are not usually designed as overfl ow dams, it is necessary to take into account the fact that no dam design can ensure absolute protection against overtopping. During the catastrophic fl oods in August 2002, which aff ected central and Western Europe, about 70 small dams were damaged in the Czech Republic. 10 of them were completely breached. During the fl ood events, the most frequent type of failure aff ecting small dams was breaching due to overtopping (Říha, 2004, 2006). When conducting embankment dam breach analysis, the calculation of fl ow parameters has to be performed. In the case of practical studies, numerical methods are used. The common problem with the modelling of breaching is the estimation of the erodibility parameters of dam materials. The most reliable erodibility parameter values are obtained from the analysis of breaching processes occurring during past incidents. However, the frequency of dam incidents is not so great and thus it is rather diffi cult to compile statistics for breach parameters. In this paper, the authors contribute to this topic by discussing the failures of four small dams, these being the Luh, Velký Bělčický, Melín, and Metelský dams located in the Blatná region of the Czech Republic (Fig. 3). In this paper, data from real cases are assembled and analysed. To simulate dam breach due to overtopping a one-dimensional mathematical model was proposed and computer code was compiled and tested. The erodibility parameters were determined via model calibration during which calculated and observed peak outfl ow discharges and breaching times were compared. The obtained parameter values were compared with previously published data. State of the Art Review A lot of publications about dam breaks have been made available since 1990. Singh (1996) presented a list of dam failures and their causes in his monograph. He provided an illustration of the 718 Zakaraya Alhasan, Jan Jandora, Jaromír Říha dam breaching mechanism, introduced empirical models with dimensional and dimensionless solutions, mathematical models of dam breaching, and a comparative evaluation of dam breach models. Wahl (1997) examined empirical procedures and numerical models used for the prediction of dam breach parameters and outlined a program for the development of an improved numerical model for the simulation of embankment dam breach events. Jun and Oh (1998) presented a simulation of dam-break processes during the overtopping of embankments and conducted fl ood routing analysis along downstream reaches using the NWS DAMBRK model. Good agreement with eyewitness evidence in terms of dam crest overtopping and the progression of breach formation was achieved. Other dam break simulations were presented by Tingsanchali and Chinnarasri (2001), who developed a one-dimensional numerical model for dam failure due to overtopping. In the model, one-dimensional equations of continuity and momentum for unsteady varied fl ow over steep bed slopes were solved together with sediment transport equations. The model was successfully calibrated and verifi ed using laboratory experimental data. The use of experimental data to calibrate numerical models has been addressed in numerous studies. A set of laboratory experiments were carried out by Aureli et al. (2000) to verify a numerical model under test conditions, including shock formation, reverse fl ow and wetting and drying conditions. Říha and Daněček (2000) compared the results obtained from analytical solution for diff erent breach channel shapes with the results of experiments carried out at an outdoor laboratory. Coleman et al. (2002) presented experimental results for small homogeneous embankments made of non-cohesive materials breached by overtopping under constant reservoir level conditions. Chinnarasri et al. (2003) investigated the fl ow patterns and progressive damage at a dike during overtopping a er analysing the data obtained from nine experimental runs. Results obtained from experimental tests were considered by Franca and Almeida (2004) for dam break modelling using two erosion parameters. The fi nal breach geometry and dimensions were obtained from those experiments. Toledo et al. (2006) presented the results obtained from tests concerning the failure of rock-fi ll dams due to overtopping. A similar experimental study of embankment dam breaching was carried out by Dupont et al. (2007), and laboratory tests enabled validation of the study and the completion of a numerical approach. An overview of fi eld tests and laboratory experiments was given by Morris et al. (2007), where the authors addressed the assessment and reduction of risks from extreme fl ooding caused by natural events or the failure of dams and fl ood defence structures. A series of tests investigating dike breaches due to overtopping were carried out by Schmocker and Hager (2009) to examine model limitations. Roger et al. (2009) compared data obtained from an experimental model (discharges, water level, and depth profi les of horizontal velocities) with the results of numerical computations of dike-break induced fl ows with a focus on the fi nal steady state. Soares-Frazao et al. (2012) conducted experiments concerning two-dimensional dam-break fl ows over a sand bed at Université catholique de Louvain, Belgium. One, two and three dimensional numerical dam break models are currently being developed by authors worldwide. Wang and Bowles (2006) developed an erosion and force equilibrium based three-dimensional model for the simulation of breaching of non-cohesive earthen dams due to overtopping and checked three-dimensional slope stability in the breach channel. Wang et al. (2006) developed a physically based numerical model to simulate the growth of a breach in an overtopped earth or rock-fi ll embankment. A one and two dimensional numerical dam-break fl ow model is provided by Galoie and Zenz (2011), in which the shallow water equations are solved by means of the fi nite diff erences method (FDM). To obtain an idea about the breaching mechanism, breach opening dimensions and erodibility of soils, the development of a database of real cases and experimental results is necessary. Jun and Oh (1998), Aigner et al. (2002), Alcrudo and Mulet (2007) and other authors analysed real cases of embankment dam failures. Lemperiere et al. (2006) presented a database of real world case studies of embankment dam incidents and summarized the lessons learnt from dam failures by overtopping to propose a new empirical breach peak outfl ow formula. Løvoll (2006) presented the results of 3 fi eld tests carried out on 6 m high embankment dams in Norway during the period of 2001–2003. Alcrudo and Mulet (2007) described the event that led to the breach of the Tous Dam in Spain which broke due to overtopping on October 20th 1982 and the eff ects of the fl ood. Goran and Goran (2009) presented the results of hydraulic analysis of the failure of two dams in Croatia with a special focus on the fi rst stage of breach formation when water fl ows through the initial breach and accelerates eroding soil. From this state of the art review it can be seen that a lot of studies deal with particular problems related to experimental and numerical dam break modelling. However, there is a lack of studies analysing real cases and comparing real incidents with numerical results. This is probably due to poor records available regarding the real circumstances of dam failure. This paper is contributing to the latter topic, discussing carefully compiled records from 4 failures of small embankment dams in the Czech Republic. A simple numerical model was calibrated using the data from the dams’ sites. The paper contains a proposal for a numerical procedure, including conceptual, mathematical and numerical models, descriptions of 4 dam failure cases, and fi nally a comparison of calculated dam break parameters with site data. Study of Dam-break Due to Overtopping of Four Small Dams in the Czech Republic 719 Conceptual Model In general, the problem of dam failure due to overtopping involves an extremely turbulent threedimensional fl ow comprising a mixture of water, air and soil, all with diff erent densities. This fact creates theoretical and numerical diffi culties when solving practical problems. Therefore, in this study the problem of dam failure due to overtopping is reduced from a three-dimensional (3D) problem to a one-dimensional (1D) one. To calculate the breach discharge Qb, it is necessary to suggest the probable shape of the breach, which is characterized by its shape and area Ab, the water level H in the reservoir and the breach bottom level Z above the reference level (above sea water level SWL). The trapezoidal breach opening is taken into account (Fig. 1). The slope of the breach opening sides is considered as constant during the breaching period. The erosion capacity of fl owing water can be described by empirical relations which express the changes in the breach parameters over time, namely breach bottom level Z(t) and breach bottom width b(t), as a function of water velocity and sediment discharge across the breach. The relations can be formally written as follows (Singh 1996, Jandora and Říha, 2008): d Z = f(v, qs, 1, 1), dt (1) db = f(v, qs, 2, 2), dt (2) where v .............................the cross-sectional velocity in the breach, qs ............................the specifi c discharge of sedi ment transport (dam material) and 1, 2, 1 and 2 ...are empirical parameters expressing the erodibility of dam material. Mathematical Model Solving the problem of dam failure due to overtopping involves the determination of time dependent breach outfl ow discharge Qb(t). This breach discharge is aff ected by the dimensions of breach shape, water level and the volume of reservoir storage during the failure. The unknown variables of the problem are as follows (Fig. 1): H(t) .....is the water level in the reservoir measured from a reference level, Qb(t) ....is breach discharge, v(t) .......is cross-sectional velocity in the breach, Z(t) ......is the breach bottom level measured from a reference level, b(t) .......is the breach bottom width. To determine these variables, the following fi ve equations are used. The instantaneous change in the reservoir storage volume V(t) as a function of the reservoir infl ow Qin and the outfl ow from the reservoir through the breach Qb, and through appurtenant works Qf (spillway and bottom outlet), is expressed as follows: dV = Qin − Qb − Qf. dt (3) The instantaneous change in the reservoir storage volume can be also expressed as follows: dV(t) = As(H(t))dH(t), (4) where As(H(t)) is the area of the water surface of the reservoir determined from the area-elevation curve (bathygraphic curve). When substituting equation (4) for equation (3) the following form is obtained: dH As  = Qin − Qb − Qf. dt (5) We can rearrange equation (5) to yield dH(t): dH 1 = (Qin − Qb − Qf). dt As (6) For the breach discharge Qb, the following equation holds: Qb = vAb, (7) where Ab ......the breach cross-sectional area. 1: Proposed section of dam breach opening 720 Zakaraya Alhasan, Jan Jandora, Jaromír Říha Breach discharge Qb (in case of a trapezoidal breach shape) can be determined by the overfl ow equation for a broad crested weir:  t b Q mb gHZ msHZ 3/2 5/2 2( ) ( ) , (8) where m ..............the discharge coeffi cient for a rectangular weir, mt .............the discharge coeffi cient for a triangular weir, b ...............the breach width at its bottom, g ...............the gravitational acceleration, s ................the average slope of the breach sides and, (H − Z) ..the overfl ow head. In the case of a trapezoidal breach shape the breach cross-sectional area is defi ned as: Ab = b(H − Z) + s(H − Z)2. (9) From equations (7), (8) and (9) we obtain the following equation: mb b      t b b gH Z msH Z Q vAHZsHZ 3/2 5/2 2 2( ) ( ) ()() . (10) Equations (1) and (2) can be expressed by simple relations (Singh, 1996; Jandora and Říha, 2008):   Zv t 1 1 d d, for v > vcr, (11)   bv t 2 2 d d, for v > vcr, (12) where 1, 2, 1 and 2 ...empirical erodibility parameters of dam material, vcr ..........................the critical value of fl ow velocity at which erosion of dam material starts (non-scouring velocity). Initial Conditions During the solution two signifi cant instants are distinguished (Fig. 2): t0 = 0 ....instant of time corresponding to the beginning of dam overtopping, tb ............instant of time corresponding to the beginning of dam breaching. The initial conditions of the evolution problem are as follows: b(t = t0) = b0, H(t = t0) = H0 = Zc, (13) Z(t = t0) = Z0 = Zc, where b0 .........the initial breach width, H0 ........the initial water level in the reservoir, Z0 ........the initial level of the breach bottom, Zc ..........the elevation of the dam crest at its lowest place. The values of initial conditions are determined at the beginning of dam overtopping at time t = t0 and when the non-scouring velocity is exceeded (t = tb). Numerical Solution For the approximation of the numerical solution of equations (6), (10), (11) and (12), the Newton method has been used. t refers to time step and ti+1 = ti + t is discrete time. From equation (6), the water level H(ti+1) in the reservoir can be expressed using diff erences:    +1 ( )= ()+ () () () (()) ii iniii bf si t Ht Ht Q t Q t Q t AHt . (14) For the cross-sectional velocity v(ti) in the breach (Eq. 10), the following equation holds: () ()= () i b i i b Qt vt At, (15) where:   ii iitii b Q t mbt g Ht Zt ms Ht Zt 3/2 5/2 () ()2 () () () () , (16) Ab(ti) = b(ti)[H(ti) − Z(ti)] + s[H(ti) − Z(ti)]2 (17) By the fi nite diff erence approximation of equations (11) and (12) we obtain:      ii Z Z Zt Zt v tt t 1 +1 1 d()() d,  (18) Z(ti+1) = Z(ti) − t1v(ti)1, for v(ti) > vcr, (19)      ii b b bt bt v tt t 2 +1 2 d()() d,  (20) b(ti+1) = b(ti) + t2v(ti)2, for v(ti) > vcr. (21) 2: Time axis of the failure process Study of Dam-break Due to Overtopping of Four Small Dams in the Czech Republic 721 CASE STUDY When calibrating the erodibility parameters of the proposed numerical model, the observed characteristics of real breached dams were used. In the case study, the fi eld data from 4 small dams breached during the fl ood in 2002 in the Czech Republic were used. The August 2002 Flood On August 5th 2002, a cyclone developed above the Western Mediterranean, which during August 6th proceeded to the northeast and reached the eastern Alps. At that time it supplied the south of Bohemia with heavy rainfall featuring local showers of high intensity, which temporarily ceased on August 8th in the morning. On Saturday August 10th, the cyclone regenerated above Italy and continued to the north again. During 11th and 12th August it reached the Czech Republic, where long lasting precipitation struck almost the whole country with daily totals exceeding 200 mm and with a peak daily total of 321 mm. In the area of interest (Fig. 3), the threeday total rainfall (from 11th to 13th August) was about 160 mm. The total rainfall during both rainfall episodes was 305 mm. Some Details of the Studied Dams In the area of interest (Blatná region), 4 small dams were breached: the Luh, Velký Bělčický, Melín and Metelský dams. The Luh and Velký Bělčický dams are located about 8 km north of the town of Blatná and about 2 km west of the village of Belcice, both are situated on the Závišínský stream (Fig. 4). The Melín and Metelský dams are located about 12 km to the northwest of Blatná, and situated on the Metelský stream (Fig. 5). Both the Luh and the Velký Bělčický dams have a heterogeneous body (Fig. 6a) comprising a sandy clay upstream blanket and a downstream shoulder, which in case of the Luh dam is composed of clayey sand, while sand mixed with fi ne-grained soil has been used for the shoulder at the Velký Bělčický dam (Fig. 7). The Luh dam is equipped with two spillways at the right and le side of the dam and the Velký Bělčický dam is equipped with one spillway at the right abutment. The Melín dam has a practically homogeneous body (Fig. 6b) made of sandy clay (Fig. 8). This dam is equipped with two emergency spillways at the le and right abutments and one wooden bottom outlet with a service sha located at its upstream end. 3: Map of the Czech Republic with the area of interest I: Geometric data for the studied dams Name of dam Dam height [m] Length at the dam crest [m] Width of the dam crest [m] Upstream slope Downstream slope Luh 3.8 202 7–7.8 1:1.7 1:1.5 Velký Bělčický 6.7 296 7.4 1:1.4 1:2.1 Melín 6.2 195 2.5–3.5 1:1.5 1:1.5 Metelský 8.5 433 4–6.85 1:1.2 1:2 722 Zakaraya Alhasan, Jan Jandora, Jaromír Říha The Metelský dam has a heterogeneous body (Fig. 6a) comprising a sandy clay upstream blanket and a sandy downstream shoulder (Fig. 8). The dam was equipped with two wooden bottom outlets and an emergency spillway. The dam crest of the Luh dam serves as a public asphalt road but the dam crests of the other dams are overgrown with grass and trees. The upstream slopes of all dams are covered with stone pavement and the downstream slopes are covered with grass, bushes and trees (Figs. 4 and 5). 4: The Luh (left) and Velký Bělčický (right) dam breaches 5: The Metelský (left) and Melín (right) dam breaches Study of Dam-break Due to Overtopping of Four Small Dams in the Czech Republic 723 The grain size distribution curves of the above mentioned dams are shown in Figs. 7 and 8. Some further geometric data are summarized in Tab. I. Method of Solution (Methodology) Erodibility parameters (1, 2, 1 and 2) used in the numerical model were determined based on data obtained from real dams breached due 6: Schematic cross-sections of the dam body 7: Grain size distribution curves for the Luh and Velký Bělčický dams 8: Grain size distribution curves for the Melín and Metelský dams 724 Zakaraya Alhasan, Jan Jandora, Jaromír Říha to overtopping, and those parameters were evaluated by comparison the results obtained from the numerical model with the results of model calibration, namely in terms of the maximum breach discharge and instant of peak discharge. For the model calibration the trial-and-error method was used during the performance of backward analysis. The solution procedure can be illustrated in the following steps: 1. Eyewitness testimony provided by local inhabitants was used to reconstruct the fl ood and dam break events. Based on assembled information, the initial breach width (b0) was determined together with the instants corresponding to the beginning of dam overtopping (t0) and the maximum breach discharge (tbmax). These data were used for the model calibration. 2. Site investigations were carried out to determine the fi nal breach bottom width (b) and the average slope of the breach sides (s) by measuring the breach opening geometry from which the maximum breach discharge (Qbmax) was estimated using Eq. (8). The traces le by the water level in the downstream valley, on the trees and houses downstream of the dams were indicated (Fig. 9). These data enabled reliable estimation of the maximum breach discharges during the events (see next step). 3. The maximum breach discharge was determined by comparing the value determined from Eq. (8) based on the geometry of the breach opening (Figs. 4, 5) with the value obtained from hydraulic calculations in the valley downstream of the breached dam (Fig. 9). HEC-RAS so ware was used. 4. The infl ow into the reservoirs Qin(t) was determined using the rainfall-runoff model with the use of information about the temporal progression of precipitation during the fl ood event (provided by the Czech Hydrometeorological Institute). At the same time, the necessary hydraulic calculations were carried out for appurtenant works (bottom outlets, spillways) of the mentioned dams to obtain Qf(H(t)). 5. The event was reconstructed by the numerical model described above. During the model calibration, the hydraulic and erodibility parameters were derived via a trial-and-error procedure. RESULTS AND DISCUSSION As mentioned above, fl ow and erodibility parameters were obtained from the model calibration. The outputs from the calculations were dam break hydrographs for all 4 dams. Tab. II summarizes the values of initial breach width (b0), the average slope (s) of the breach opening sides, the discharge coeffi cients (m and mt) and erodibility parameters (1, 2, 1 and 2) used in the calibrated numerical model. The fi nal results are presented in Figs. 10 and 11, where the modelled peak discharges of the breached dams are compared with values obtained from site investigations, and the time diff erence between the 9: Area below the Metelský dam after the flood Study of Dam-break Due to Overtopping of Four Small Dams in the Czech Republic 725 II: Input data for the mathematical model (numerical solution) Name of dam b0 [m] sm [-] mt [-] 1212 Luh 3 1.00 0.33 0.35 0.0001 0.001 1 1 Velký Bělčický 12 0.80 0.35 0.35 0.003 0.003 1 1 Melín 1 0.80 0.35 0.35 0.003 0.003 1 1 Metelský 1 0.95 0.35 0.35 0.003 0.003 1 1 10: Flood hydrograph and model calibration for the Luh and Velký Bělčický dam failures 11: Flood hydrograph and model calibration for the Melín and Metelský dam failures III: Results of model calibration and the mathematical model Name of dam Average breach width [m] Qbmax [m3/s] observed Qbmax [m3/s] calculated t * [hour:min] observed calculated Luh 17 15 58 64.1 0:05 Velký Bělčický 42 38 600 602.1 0:10 Melín 17 9.23 150 149 0:10 Metelský 42 + 30 ** 35 + 15 ** 550 554 0:15 * Time diff erence between the peak discharges (Figs. 10 and 11). ** Two breach openings (Fig. 5, le ).