Tài liệu hạn chế xem trước, để xem đầy đủ mời bạn chọn Tải xuống
1
/ 11 trang
THÔNG TIN TÀI LIỆU
Thông tin cơ bản
Định dạng
Số trang
11
Dung lượng
305,62 KB
Nội dung
VNU Journal of Science, Earth Sciences 23 (2007) 160‐169 A numerical model for the simulation of wave dynamics in the surf zone and near coastal structures Vu Thanh Ca* Center for Marine and Ocean-Atmosphere Interaction Research, Vietnam Institute of Meteorology, Hydrology and Environment Received 07 March 2007 Abstract. This paper describes a numerical model for the simulation of near shore wave dynamics and bottom topography change. In this part, the nearshore wave dynamics is simulated by solving the depth integrated Boussinesq approximation equations for nearshore wave transformation together with continuity equation with a Crank‐Nicholson scheme. The wave runup on beaches is simulated by a scheme, similar to the Volume Of Fluid (VOF) technique. The wave energy loss due to wave breaking and shear generated turbulence is simulated by a k − ε model, in which the turbulence kinetic energy (TKE) generation is assumed as the sum of those respectively due to wave breaking and horizontal and vertical shear. The verification of the numerical model against data obtained from various indoor experiments reveals that the model is capable of simulating the wave dynamics, turbulence and bottom topography change under wave actions. The simulation of turbulence in the surf zone and near coastal structures enable the model realistically simulates the contribution of suspended sediment transport into the bed topography change. Keywords: Wave dynamics; Wave runup; Wave energy; Surf zone; Boussinessq model. 1. Introduction1 Nadaoka [9] found by indoor experiments that during wave breaking, large vortices were formed and rapidly extended both vertically and horizontally. Ting and Kirby [15‐17] by conducting experiments with different wave conditions found that the advective and diffusive transports of TKE play a major role in the distribution of turbulence, especially under plunging breaker. They also found that under spilling breakers (the breaking of relatively steep waves on a gentle slope), the time variation of TKE was relatively small, and the time average transport of TKE was directed offshore. Under plunging breakers (the breaking of less steep waves on a gentle slope), there was a large time variation of Extensive researches on the wave dynamics, sediment transport and bottom topography change in the nearshore area, especially in the surf zone [1‐5, 7, 9, 12, 14‐17] have elucidated various aspects of coastal processes, such as the dynamics of wave breaking, characteristics of turbulence in the surf zone, structure of the undertow, the development of bottom boundary layer under breaking waves, the rate of bed load transport, uptake of bed material for suspension, settling rate of suspended sediment etc. _ * Tel.: 84‐913212455. E‐mail: vuca@vkttv.edu.vn 160 160 Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 TKE, and its time averaged transport is directed on‐shore. For situations with negligible alongshore sediment transport, the status of a beach depends on the cross‐shore transport of sediment, which is closely related with wave conditions. If the shoreward transport of sediment by incoming waves exceeds the offshore transport of sediment by retreating waves and the undertow, there will be a net onshore transport of sediment, resulting in beach accretion. Otherwise, the beach is in equilibrium state or eroded. During a storm, turbulence generated by the breaking of a relatively short wind wave has not been significantly dissipated when a new wave arrives and breaks. Thus, the time variation of TKE is relatively small, and the combination of wave‐induced flow and undertow may transport TKE and suspended sediment offshore. This results in the offshore‐directed transport of sand during storm and the associated beach erosion. On the other hand, post storms, turbulence generated by the breaking of a long period ‐ small amplitude swell has significantly dissipated when the wave retreats. Thus, there is a large time variation of TKE, and the peaks in turbulence intensity and suspended sediment concentration coincide with incoming waves. Accordingly, onshore transport of TKE and suspended sediment by incoming waves exceeds the offshore transport by retreating waves and the undertow. This results in a net onshore transport of suspended sediments and helps explaining the onshore‐directed transport of sediment during calm weather and the consequent post storm beach recovery. Schaffer [14] and Madsen [7] developed models for the simulation of the nearshore wave dynamics based on Boussinesq approximation equations. The wave energy loss due to breaking is simulated by employing a surface roller model. Due to the instability of the numerical code resulting from the treatment of the surface roller wave energy loss, Schaffer [14] had to use a smoothing technique to stabilize the solution. Rakha et al [12, 13] presented a quasi‐2D and a quasi‐3D phase resolving hydrodynamic and sediment transport models. In these models, the horizontal transport of TKE, and the associated transport of suspended sediment are neglected. However, as discussed previously, results of Nadaoka et al [9] and Ting and Kirby [16] show that the horizontal transport of TKE in the surf zone is very important and should not be neglected. Thus, without accounting for this, it is not easy to simulate the beach erosion during storm and the consequent recovery after the storm. Nadaoka and Ono [10] presented a depth‐ integrated k‐model where the TKE production rate was evaluated with a Rankine eddy model. In this model, the TKE dissipation rate and the eddy viscosity was evaluated by employing an empirical length scale. The model had not been verified against experimental data. Also, wave runup on beach, which is mainly responsible for the erosion of foreshore during storms, is not simulated in this model. Regarding all the above mentioned facts, the purpose of this study is to develop a numerical model that can simulate the nearshore wave dynamics, including wave breaking and wave runup, the generation, transport and dissipation of TKE. 2. Governing equations of the numerical model for nearshore wave dynamics In this study, the near‐shore wave Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 dynamics are simulated by solution of two‐ dimensional depth integrated Boussinesq approximation equations, including bottom friction and wave energy loss due to wave breaking and shear. The main equations of the numerical model are written as: ∂q x ∂q y ∂η + + = ∂x ∂y ∂t (1) ∂q x ∂ ⎛ q x2 ⎞ ∂ ⎛ q x q y ⎞ ∂η ⎟ + gd + ⎜⎜ ⎟⎟ + ⎜⎜ ∂t ∂x ⎝ d ⎠ ∂y ⎝ d ⎟⎠ ∂x h3 ⎡ ∂ ⎛ qx ⎞ ∂ ⎛ q y ⎞⎤ ⎜ ⎟⎥ + (2) ⎢ ⎜ ⎟+ ⎢⎣ ∂x ∂t ⎝ h ⎠ ∂x∂y∂t ⎜⎝ h ⎟⎠⎥⎦ ∂ 3q y ⎞ h ⎛ ∂ 3q ⎟ − M bx + f c Qq x = − ⎜ 2x + ⎜ ⎝ ∂x ∂t ∂x∂y∂t ⎟⎠ d2 ∂q y ∂ ⎛ q x q y ⎞ ∂ ⎛ q 2y ⎞ ∂η ⎟ + ⎜ ⎟ + gd + ⎜⎜ ∂t ∂x ⎝ d ⎟⎠ ∂y ⎜⎝ d ⎟⎠ ∂y h3 ⎡ ∂ ⎛ q y ⎞ ∂ ⎛ q x ⎞⎤ ⎜ ⎟ (3) + + ⎜ ⎟⎥ ⎢ ⎣⎢ ∂y ∂t ⎜⎝ h ⎟⎠ ∂x∂y∂t ⎝ h ⎠⎦⎥ ⎛ ∂ 3q y ∂ q x ⎞⎟ f ⎜ + − M by + c2 Qq y = ⎜ ∂y ∂t ∂x∂y∂t ⎟ d ⎝ ⎠ where q x and q y are respectively the depth − h2 integrated flow discharges in x and y directions; η is the water surface elevation; d is the instantaneous water depth; h is the still water depth; f c is the bed friction coefficient; Q is the total discharge, defined as Q = q x2 + q 2y ; and M bx and M by represent the wave energy loss due to breaking, evaluated by introducing an eddy viscosity and expressed as: ∂(q x / d ) ⎤ ∂ ⎡ ∂(q x / d ) ⎤ ∂ ⎡ + df Dν t df Dν t ∂x ⎢⎣ ∂x ⎥⎦ ∂y ⎢⎣ ∂y ⎥⎦ (4) ∂ qy / d ⎤ ∂ ⎡ ∂ qy / d ⎤ ∂ ⎡ = ⎢df Dν t ⎥+ ⎢df Dν t ⎥ ∂x ⎣ ∂x ⎦ ∂y ⎣ ∂y ⎦ M bx = M by ( ) ( ) In Eq. (4), ν t is the eddy viscosity; and f D is an empirical coefficient, determined based on the calibration of the numerical model. 161 When waves are breaking on beach, a part of the lost wave energy is transformed into turbulence energy. At the beginning of the wave breaking process, the turbulence is confined into a small portion of the breaking wave crest, the surface roller; after that, turbulence eddies rapidly expand in vertical and horizontal directions [9, 15‐17]. The turbulence under wave breaking is very complex and fully three‐dimensional. Thus, a 3D model is required for a proper simulation of turbulence processes here. However, such a model would require an excessive computational time and at the moment is not suitable for a practical application. On the other hand, based on results of Nadaoka et al [9], Ting and Kirby [15‐17], it can be estimated that in the surf zone, the time scale for turbulence energy transport in the vertical direction is much shorter than that in the horizontal directions. Thus, the simulation of the transport of TKE in the horizontal direction is more important than that in the vertical direction. Therefore, in the present study, the TKE is assumed uniformly distributed in the whole water depth, and the depth‐integrated equations for the production, transport and dissipation of the TKE and its dissipation rate read: ∂k ∂uk ∂vk + + = Pr − ε ∂t ∂x ∂y (5) ∂ ⎡ dν t ∂(k / d ) ⎤ ∂ ⎡ dν t ∂(k / d ) ⎤ + ⎢ ⎥+ ⎢ ⎥, ∂x ⎣ σ t ∂x ⎦ ∂y ⎣ σ t ∂y ⎦ ∂ε ∂uε ∂vε ∂ ⎡ dν ∂ (ε / d ) ⎤ + + = ⎢ t ⎥ ∂t ∂x ∂y ∂x ⎣ σ ε ∂x ⎦ (6) ∂ ⎡ dν t ∂ (ε / d ) ⎤ ε + ⎢ ⎥ + (C1ε Pr − C 2ε ε ) ∂y ⎣ σ ε ∂y ⎦ k where k and ε are respectively the depth integrated TKE and its dissipation rate; u and v are respectively phase‐depth averaged flow velocities in x and y directions; σ t , σ ε , C1ε , C2ε are closure coefficients. In 162 Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 Eq. (6), Pr is the TKE production rate, which is assumed as a summation of the TKE production due to bottom friction Prb , horizontal shear Prs and wave breaking Prw as: Pr = Prb + Prs + Prw (7) With known values of k and ε , the eddy viscosity is evaluated as: ν t = Cε k / (dε ) , (8) where Cε (= 0.09) is constant. The scheme for the simulation of wave runup and rundown on the beach is explained in the next section. By employing this scheme, the present model can simulate the wave setup, set down on the beach, and the erosion of foreshore during storm events. 3. Boundary and initial conditions and numerical scheme 3.1. Boundary and initial conditions It is possible to use a weekly wave reflected boundary condition such as the Summerfeld radiation condition at the offshore boundary to let reflected waves freely going out of the computational region. However, this linear wave theory based boundary condition, when applied in combination with a nonlinear wave model, does not ensure mass conservation and may lead to an accumulation or lost of water inside the computational region. Thus, in this study, water surface elevation under waves is given at the offshore boundary. Wave‐absorbing zones are introduced at the lateral boundaries to minimize wave reflection. The bed friction coefficient f c in these zones is assumed constant within first five meshes from the lateral boundaries, and then increases linearly with the distances from the boundaries towards the ends of the wave absorbing zones. Finally, at the ends of the wave absorbing zones, the Summerfeld radiation condition for long waves are introduced to let remaining waves going out of the computational region. A free slip boundary condition is applied at surfaces of the coastal structures. Zero gradients of k and ε are assumed at the offshore, lateral boundaries and at surfaces of coastal structures. A scheme similar to that of Hibberd and Peregrine [5] is used to compute the wave runup on the beach. A sketch of the scheme is shown in Fig. 1. In this scheme, when the shore is approached, all the dispersion terms in Eqs. (2) and (3) are turned off. Additionally, a cell side wetted function, defined as the wetted portion over the total length of a cell side, and a cell wetted area function, defined as the wetted portion over the total cell area are introduced to account for the fact that water flows only in wetted parts of the cells on the instantaneous shoreline. Then, the continuity equation (Eq. 1) and momentum equations (Eqs. 2 and 3) can be derived by a method similar to Vu et al [19] and become: ∂f q ∂f q ∂Sη y x + x y + =0 (9) ∂x ∂y ∂t ∂q x ∂ ⎛ Sq x2 ⎞ ∂ ⎛ Sq x q y ⎞ ⎜ ⎟+ ⎟ ⎜ + S ∂x ⎜⎝ d ⎟⎠ S ∂y ⎜⎝ d ⎟⎠ ∂t ∂ (q x / d ) ⎤ ∂η ∂ ⎡ − dν t S ∂x ⎥⎦ ∂x S ∂x ⎢⎣ ∂ (q x / d ) ⎤ f c ∂ ⎡ dν t S + Qq x = 0, − S ∂y ⎢⎣ ∂y ⎥⎦ d + gd (10) Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 ⎞ ∂ ⎛⎜ Sq y ⎞⎟ ⎟⎟ + ∂t ⎠ S ∂y ⎜⎝ d ⎟⎠ ∂ qy / d ⎤ ∂η ∂ ⎡ + gd − ⎢dν t S ⎥ ∂y S ∂x ⎣ ∂x ⎦ ∂ qy / d ⎤ fc ∂ ⎡ Qq y = − ⎢ dν t S ⎥+ ∂y ⎦ d S ∂y ⎣ ∂q y + ∂ ⎛ Sq y q x ⎜ S ∂x ⎜⎝ d ( ( ) (11) ) where f x and f y are respectively the cell side wetted functions corresponding to x and y directions, and S is the cell area wetted function. Fig. 1. The coordinate system and method for the evaluation of a wetting and drying boundary. The procedure for determining the cell side wetted function and the cell area wetted function in the numerical scheme will be discussed in the next section. A still water is assumed at the beginning of the computation. With this, all variables are set equal to zero initially. 3.2. Numerical scheme Equations (1‐3) and (5‐6) are integrated numerically on a spatially staggered grid system, where components of the flow discharge are evaluated at surfaces, and bed elevation, k and ε are evaluated at the centers of control volumes. The sketch of the coordinates and computational mesh is shown in Fig. 1. As it will be discussed later, 163 in the present scheme, the water level inside a cell is evaluated at the center of the wetted area inside the cell. A second order accurate Crank‐Nicholson scheme is employed for the time discretization for all equations, and a central differencing scheme is employed for spatial discretization of Eqs. (1) to (3). The spatial disretization for advection terms of Eqs. (5) and (6), governing the transport, diffusion, generation and dissipation of k and ε , follows the third order accurate QUICK scheme, and that for the diffusion terms follows the central differencing scheme. As the discretization scheme is implicit, an iterative scheme similar to the SIMPLE scheme of Patankar [11] is employed. At the beginning of a new time step, the computation of the flow discharges requires the still unknown water level and eddy viscosity. Thus, at first, the water level at each new time step is assumed equal to the value at the previous time step. Then, Eqs. (2) and (3) are solved to get the flow discharges in x and y directions, respectively. The new values of the flow discharges are substituted into the continuity equation to compute the new water level. Also, with the new water level, the thickness of the surface roller is evaluated. Then, Eqs. (5) and (6) are integrated to get k and ε , and consequently the new coefficient of eddy viscosity. All newly obtained water level, flow discharges and coefficient of eddy viscosity are substituted back into Eqs. (2) and (3) to compute the new components of the flow discharge. The procedure is repeated until converged solutions are reached. The wetted periphery inside a computational mesh at the intersection between the water surface and the beach, the cell side wetted function and the cell area wetted function at each time step are Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 164 evaluated explicitly based on the water level, bed elevation and the bed slope in two directions. The procedure for this is shown in Fig. 1. The bed elevations at cell corners (such as points A, B, C and D in Fig. 1) are evaluated as the average value of the bed elevation at four adjacent points. For example, the bed elevation at point C in this figure is evaluated as: bc = bi , j + bi , j +1 + bi +1, j +1 + bi +1, j , (12) where bc is the bed elevation at point C, and bi,j, bi,j+1, bi+1,j+1 and bi+1,j are respectively the bed elevations at the center of cells (i,j), (i,j+1), (i+1,j+1) and (i+1,j). The water level at a cell side is averaged from the water levels at two adjacent cells. For example, the water level on the side BC of cell i,j in Fig. 1 is evaluated as: ηbc = ηi , j + ηi , j +1 , (13) where ηbc , ηi , j and ηi , j +1 are respectively water levels at the cell side BC, and in the cells (i,j) and (i,j+1). If one of adjacent cells to a cell side is completely dry (with the value of the area wetted function equal to zero), the average water level at the cell side is assumed equal to the water level at the wetted cell. Based on the bed elevation at its two ends and the average water level on a cell side, the intersected point between the water surface and the cell side, and the wetted portion of the side are determined. When the average water level on the cell side is higher than the bed elevation at its two ends, the side is considered totally submerged into the water, and the corresponding value of the cell side wetted function is 1. For other cases, value of the cell side wetted function equals to the ratio of the length of the wetted portion over the total length of the cell side. After getting all the wetted points on four sides of the cell, the wetted periphery and the wetted area inside a cell are determined by connecting two adjacent wetted points with a straight line. This wetted periphery is shown by the dotted line in Fig. 1. The wetted area in cell i,j in this figure is the portion of the cell from the dotted line to offshore. The wetted periphery and area inside the cell are kept constant for a time step. 4. Model verification 4.1. Wave transformation and characteristics of turbulence due to wave breaking on a natural beach To verify the accuracy of the numerical model on the simulation of the wave transformation on a natural beach, existing experimental data on the wave dynamics in the nearshore area obtained by Ting and Kirby [15‐17] are used. The experiments were carried out in a two‐dimensional wave flume of 40m long, 0.6m wide and 1.0m deep. A plywood false bottom was installed in the flume to create a uniform slope of 1 on 35. Regular waves with heights and periods equal to 12.7cm, 2s and 8.7cm, 5s are used as incoming waves respectively for spilling breaker and plunging breaker experiments. Fig. 2 shows the sketch of the Ting and Kirby [15‐17] experiments. Computation was carried out with the same conditions of the experiments. The critical water surface slope for a broken wave to be recovered φ0 is set equal to 60, according to Madsen et al [7]. Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 Wave generator fD = a + b 0.38m 35 Fig. 2. Experiments by Ting and Kirby [15‐17]. As cited by various authors [2, 4], when waves are breaking, a major part of the lost wave energy is dissipated directly in the shear layer beneath the surface roller, and only a minor part of it is transformed into turbulent energy. Thus, a turbulence model may underestimate the wave energy lost due to breaking. To account for this, an empirical coefficient f D was introduced in Eq. 4. Calibrations were carried out to find the best value of this coefficient. Vu et al [18] found a constant value of 1.5 for this coefficient for their one‐dimensional model. However, their computational results show that the coefficient does not provide adequate wave energy dissipation, and the computed wave heights after breaking is significantly larger than the observed ones. As mentioned previously, wave breaking happens with a sudden loss of wave energy. This in a numerical model can be simulated by a sudden increase in the “energy dissipation coefficient” f D As the breaking wave progresses onshore, the growth of TKE may accompany an increase in the coefficient. On the other hand, turbulence length scale, and the corresponding turbulence intensity decrease with water depth, leading to a decrease in the coefficient. Thus, in this study, the coefficient is assumed suddenly increases at the breaking point, then gradually increases towards the shore, and then decreases with the decrease in the water depth in the following form: ⎛ hm ⎞ ⎟ ⎜ ⎜ h ⎟ , ⎝ mb ⎠ (14) where a and b are constants, to be determined from calibration; x and xb are respectively the coordinates in the on‐ offshore direction at the point under consideration and the breaking point; hm and hmb are the corresponding mean water depths at the respective points. Fig. 3 shows the comparison between on‐ offshore distributions of time averaged mean water surface elevation, minimum water surface elevation, maximum water surface elevation, and wave height for the spilling breaker, computed by the model (with f D evaluated following Eq. (14), a = 0.05 and b = ), and observed by Ting and Kirby [15, 16]. 0.2 Height (m) 0.4m x − xb hmb 165 Bed Comp Etaav Comp Etamax Comp Etamin Comp Waveh Obs Wavh Obs Etaav Obs Etamax Obs Etamin 0.1 -0.1 -0.2 -0.3 -0.4 10 11 12 13 Horizontal Distance (m) Fig. 3. Comparison between observed and computed time averaged wave height, highest, lowest and mean water surface elevation for spilling breaker. Experimental data from Ting and Kirby [15, 16]. It can be seen in Fig. 3 that the model can accurately predict the wave breaking point and provides adequate wave energy dissipation after breaking. The maximum, minimum and mean water levels at all points in the computational region are also predicted by the model with good accuracy. The general satisfactory agreement between computed and observed data shown in the Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 figure suggests that the model can simulate nearshore wave processes, such as wave energy loss due to breaking, wave setup, setdown etc. with acceptable accuracy. Figures (4) to (7) respectively show the time variation of ensemble averaged (phase‐ averaged) non‐dimensional water surface elevation, depth‐averaged horizontal flow velocity, TKE, and advective transport rate of TKE, computed by the model and observed by Ting and Kirby [15, 16] at (x − xb ) / hmb = 7.642 The time t in the figures is non‐dimensionalized by wave period T. For convenient, the same coordinate system in Ting and Kirby [15‐17] is employed in this study. The computed time variation of ensemble‐averaged water surface elevation fluctuation, non‐dimensionalized by local mean water depth hm (equal the sum of local still water depth and mean water surface fluctuation η ), shown in Fig. 4 agrees very well with observed between computed variation of phase horizontal flow dimensionalized by data. The agreement and observed time and depth‐averaged velocity, non‐ the local long‐wave commonly known that just after wave breaking, turbulence is concentrated only inside the surface roller, and flow in the region below remains irrotational. Thus, a depth‐integrated model for the generation, transport and dissipation of TKE cannot be considered as a good approximation for this situation. However, despite of all inadequate assumptions and approximations, order of TKE predicted by the model, shown in Fig. 6, agrees well with the observed one. Regarding difficulties in predicting the TKE under wave breaking with a numerical model, it can be said that the numerical model can predict the TKE and its advective transport with satisfactory accuracy. 0.5 0.4 0.3 (ζ -)/h 166 0.2 0.1 -0.1 -0.2 celerity c (defined as c = g (hm + H ) , with H 0.4 0.6 0.8 t/T Fig. 4. Computed and observed phase‐averaged water surface elevation at (x‐xb)/hb=7.462. Spilling breaker. 0.4 0.3 0.2 /c as the deepwater wave height) also agrees satisfactorily with observed data. The agreement between computed and observed phase and depth‐averaged non‐dimensional TKE and its advective transport is less satisfactory than that of the water level or flow velocity. It must be noted that the computation of TKE employs a depth‐ integrated k − ε model, which involves many approximation assumptions and may not accurately predict the TKE production, transport and dissipation under a complex situation such as wave breaking. Among all, the weakest point of this model might be the depth‐integrated approximation. It is 0.2 0.1 -0.1 -0.2 0.2 0.4 0.6 t /T 0.8 Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 Fig. 5. Computed and observed phase‐depth averaged horizontal flow velocity at (x-xb)/hb=7.462. Spilling breaker. - k/c (X10 3) 1.5 The agreement between computed and observed advective transports of TKE, shown in Fig. 7, is better than that for the TKE itself. Results of Ting and Kirby [15, 16] show that there is a tendency of offshore (negative) transport of TKE. The computational results by the present model also reveals the same tendency; however, as shown in Fig. 8, the residual advective offshore transport of TKE evaluated by the numerical model is significantly smaller than the observed one. From the general agreement between computed and observed values of various wave characteristics, it can be remarked that the numerical model can simulate wave transformation in the nearshore region with an acceptable accuracy. 0.006 0.005 k /c 0.004 0.003 0.002 0.001 0 0.2 0.4 0.6 0.8 t /T Fig. 6. Computed and observed phase‐depth averaged relative turbulent intensity at (x-xb)/hb=7.462. Spilling breaker. 167 1.2 0.5 -0.5 -1 0.2 0.4 0.6 0.8 t /T Fig. 7. Computed and observed phase‐depth averaged relative advective transport rate of TKE in the horizontal direction at (x-xb)/hb=7.462. Spilling breaker. 4.2. Wave runup on beach To verify the accuracy of the simulation by the present numerical model on the wave runup on beach, experimental data of Mase and Kobayashi [8] are used. The sketch of the experiment is shown in Fig. 10. As shown in the figure, the experiments were carried out in a wave flume with the length of 27 m, depth of 0.75 m and width of 0.50 m. An irregular wave generator is installed at one end of the wave flume. At the other end is a model beach with a foreshore slope of 1/20. The water depth in front of the slope is set constant and equal to 0.47 m. The wave runup on the beach is recorded by a wave meter. Wave groups used in the experiments are expressed as: η 1 = cos[2π (1 + δ ) ft ] + cos[2π (1 − δ ) ft ] (15) η max 2 = cos(2πδft )cos(2πft ), where η max is the amplitude of the incoming waves, f is the wave frequency, and ∆ is the variation in the relative wave frequency. During the experiments, η max was taken as 5 cm. Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 168 Water Surface Elevation (m) 0.05 turbulence generated by wave breaking and shear. As the model is a depth‐integrated, two‐dimensional in the horizontal directions, the computational time is relatively short. Thus, the application of the model for simulation of wave transformation in the field, especially in the vicinity coastal structures and inside harbours is very promising. 0.025 - 0.025 - 0.05 10 15 Time (sec) 20 25 Fig. 8. Computed and observed wave runup height. T = 2.5 s, ∆ = 0.1. Fig. 8 shows an example of comparison between observed and computed wave runup for different wave periods. It can be seen in the figures that the computed wave runup heights agree very satisfactorily with the observed values. The computational results (not shown) also reveal that short period waves are dissipated much more rapidly on the beach compared with long period waves. The very satisfactory agreement between computed and observed wave runup heights reveals that the numerical model can accurately simulate wave runup on beaches. The model is also verified for its applicability of computing waves near coastal structures. 5. Conclusions A numerical model has been developed for the simulation of the wave dynamics in the near shore area and in the vicinity of coastal structures. It has been found that the numerical model can satisfactorily simulate the wave transformation, including wave breaking, wave runup on the beach, and References [1] D. Cox, N. Kobayashi, Kinematic undertow model with logarithmic boundary layer, Journal of Waterway, Port, Coastal, and Ocean Engineering 123/6 (1997) 354. [2] W.R. Dally, C.A. Brown, A modeling investigation of the breaking wave roller with application to cross‐shore currents, Journal of Geophysical Research 100 (1995) 24873. [3] A.G. Davies, J.S. Ribberink, A. Temperville, J.A. Zyserman, Comparisons between sediment transport models and observations made in wave and current flows above plane beds, Coastal Engineering 31 (1997) 163. [4] R. Deigaard, Mathematical modelling of waves in the surf zone, Prog. Report ISVA 69 (1989) 47. [5] S. Hibberd, H.D. Peregrine, Surf and runup on beach: A uniform bore, Journal of Fluid Mechanics 95 (1979) 323. [6] C.W. Hirt, Nichols, Volume of fluid method for the dynamics of free boundaries, Journal of Computational Physics 39 (1981) 201. [7] P.A. Madsen, O.R. Sorensen, H.A. Schaffer, Surf zone dynamics simulated by a Boussinesq type model. Part 1: Model description and cross‐ shore motion of regular waves, Coastal Engineering 33 (1997) 255. [8] H. Mase, N. Kobayashi, Low frequency swash oscillation, Journal of Japan Society of Civil Engineers II‐22/461 (1993) 49. [9] K. Nadaoka, M. Hino, Y. Koyano, Structure of the turbulent flow field under breaking waves in the surf zone, Journal of Fluid Mechanics 204 (1989) 359. Vu Thanh Ca / VNU Journal of Science, Earth Sciences 23 (2007) 159‐168 [10] K. Nadaoka, O. Ono, Time‐Dependent Depth‐ Integrated Turbulence Closure Modeling of Breaking Waves, Coastal Engineering ACSE (1998) 86. [11] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere Publ. Co., London, 1980. [12] K.A. Rakha, R. Deigaard, I. Broker, A phase resolving cross shore sediment transport model for beach profile evolution, Coastal Engineering 31 (1997) 231. [13] K.A. Rakha, A quasi‐3D phase‐resolving hydrodynamic and sediment transport model, Coastal Engineering 34 (1998) 277. [14] H.A. Schaffer, P.A. Madsen, R. Deigaard, A Boussinesq model for waves breaking in shallow water, Coastal Engineering 20 (1993) 185. [15] F.C.K. Ting, J.T. Kirby, Observation of undertow [16] [17] [18] [19] 169 and turbulence in laboratory surf zone, Coastal Engineering 24 (1994) 51. F.C.K. Ting, J.T. Kirby, Dynamics of surf zone turbulence in a strong plunging breaker, Coastal Engineering 24 (1995) 177. F.C.K. Ting, J.T. Kirby, Dynamics of surf zone turbulence in a spilling breaker, Coastal Engineering 27 (1996) 131. Vu Thanh Ca, K. Tanimoto, Y. Yamamoto, Numerical simulation of wave breaking by a k‐ε model, Proceedings of Coastal Engineering, JSCE 47 (2000) 176. Vu Thanh Ca, Y. Ashie, T. Asaeda, A k‐ ε turbulence closure model for the atmospheric boundary layer including urban canopy, Boundary‐Layer Meteorology 102 (2002) 459. ... the? ? wave? ? dynamics? ? in? ? the? ? near? ? shore area and? ? in? ? the? ? vicinity of? ? coastal? ? structures. It has been found that the? ? numerical? ? model? ? can satisfactorily simulate the? ? wave? ? transformation, ... to wave? ? breaking on a? ? natural beach To verify the? ? accuracy of? ? the? ? numerical? ? model? ? on the? ? simulation? ? of? ? the? ? wave? ? transformation on a? ? natural beach, existing experimental data on? ?the? ?wave? ?dynamics? ?in? ?... develop a? ? numerical? ? model? ? that can simulate the? ? nearshore wave? ? dynamics, including wave? ? breaking and? ? wave? ? runup, the? ? generation, transport? ?and? ?dissipation? ?of? ?TKE. 2. Governing