Modelling The Morphodynamics of Tidal Channels: Gianluca Vignoli
Modelling The Morphodynamics of Tidal Channels: Gianluca Vignoli
Modelling The Morphodynamics of Tidal Channels: Gianluca Vignoli
Gianluca Vignoli
2004
Doctoral thesis in Environmental Engineering ( XV cycle ) Faculty of Engineering, University of Trento Year: 2004 Supervisor: Prof. Marco Tubino Cotutor: Guido Zolezzi
What is now proved was once only imagind William Blake - The Marriage of Heaven and Hell (1790-3)
Acknowledgements The author wishes to thank Marco Tubino and Guido Zolezzi for their friendship and for the introduction in the complex world of scientic research. The author is thankful to Marco Toffolon for his collaboration in the analysis of some topics of this thesis. This work has been developed within the tidal research group of the Department of Civil and Environmental Engineering of Trento, which is composed by Marco Tubino, Marco Toffolon, Ilaria Todeschini and myself. Some results have been presented in preliminary form at international and national conferences: 3rd RCEM 2003-Barcelona, Spain; XXX IAHR 2003Thessaloniki, Greece; RiverFlow 2002-Louvain-La-Neuve, Belgium; Idra2000, Genova and 28 Convegno Nazionale di Idraulica e Costruzioni Idrauliche-2003, Potenza.
vi
Contents
1 Introduction 2 On tide propagation in convergent and non-convergent channels 2.1 2.2 2.3 2.4 2.5 2.6 2.7 Formulation of the 1D problem . . . . . . . . . . . . . . . . . . . . . . . . . . . Numerical solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . External parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tide propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Non-linear effects on the average water level . . . . . . . . . . . . . . . . . . . . Marginal conditions for tide amplication . . . . . . . . . . . . . . . . . . . . .
1 11 15 17 18 19 22 27 30 35 35 37 38 40 48 51 51 56 57 58 61 61 63 65
3 Large scale equilibrium proles in convergent estuaries 3.1 3.2 3.3 3.4 3.5 Long term equilibrium proles in convergent estuaries . . . . . . . . . . . . . . Formulation of the problem and numerical scheme . . . . . . . . . . . . . . . . Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bottom equilibrium proles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4 Local-scale model for tidal channels 4.1 4.2 4.3 Formulation of the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Boundary conditions in the longitudinal direction . . . . . . . . . . . . . Closure and empirical inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 A three dimensional numerical model for suspended sediment transport 5.1 5.2 5.3 Vertical coordinate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Numerical solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
vii
Contents
5.3.1 5.3.2 5.3.3 5.3.4 5.3.5 5.4 5.5 The numerical scheme of Casulli and Cattani (1994) . . . . . . . . . . . Evaluation of the shear velocity and of the eddy-viscosity coefcients . . Numerical scheme for the advection-diffusion equation . . . . . . . . . . Exner equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Numerical procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 73 73 77 78 79 79 79 81 82 87 87 91 93 95
Boundary conditions in the longitudinal direction . . . . . . . . . . . . . . . . . Numerical tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.1 5.5.2 Vertical velocity prole in uniform ow . . . . . . . . . . . . . . . . . . Vertical velocity and concentrations prole with perturbed ow . . . . . 5.5.2.1 5.5.2.2 5.5.2.3 The analytical solution of Tubino et al. (1999) . . . . . . . . . Results under non-linear conditions . . . . . . . . . . . . . . . Comparison between the numerical scheme for the concentrations eld with and without splitting . . . . . . . . . . . . . .
6 Meso-scale bed forms: an application to uvial and tidal bars 6.1 6.2 6.3 Sand bars: linear theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The steady case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The analytical solution of Bolla Pittaluga and Seminara (2003a) . . . . . . . . . 113 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 122
Bibliography
viii
List of Figures
1.1
The Venice lagoon. (Consorzio Venezia nuova <http://www.salve.it>) . . . . . . The Western Scheldt. (Image downloaded from the internet) . . . . . . . . . . . Sediment grain size distribution in the Western Scheldt. (Image downloaded from the internet) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4 8 9 12 15
Vertical velocity prole during the ow reversal. . . . . . . . . . . . . . . . . . . Channels in the Venice lagoon. (Courtesy of Walter Bertoldi, 2003) . . . . . . . Funnel shaped estuaries, from Seminara et al. (2001a) . . . . . . . . . . . . . . . Sketch of the estuary and basic notation. . . . . . . . . . . . . . . . . . . . . . . Contour plot of the ratio U0 /Ug evaluated using the (2.23) (continuous lines) and the algebraic mean of peak values of ood and ebb velocity at the mouth evaluated using the numerical model (dotted lines). The different dotted lines are obtained through different values of in order to obtain a wider range for . Lb = 1/6 10km, D0 = 10m, ks = 30 90m1/3 s1 (ks = gCh Rh ), transparent boundary
condition landward, Le = 300km, a0 = 0.01 0.1 0.5 1 2 3m. . . . . . . Free surface proles along a sample estuary with length Le = 300km for non
22
2.4
convergent ( = 0, left) and convergent channel ( = 7.4, right). = 0.2, ks = 1/6 45m1/3 s1 (ks = gCh Rh ), D0 = 10m , transparent boundary condition. . . . . 2.5 Amplitude of the leading order Fourier components of the time series of free surface elevation in each cross section evaluated using (2.32) (left) and (2.33) (right) in a non-convergent channel = 0. The amplitudes are scaled using the value h1
1/6 of mode 1. = 0.1, D0 = 10m, ks = 45m1/3 s1 (ks = gCh Rh ), and transparent
24
boundary condition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
26
ix
List of Figures
2.6 Amplitude of the leading order Fourier components of the time series of free surface elevation in each cross section for a convergent = 7.4 (left) and nonconvergent = 0 channels (right). The amplitudes are scaled using the value h1
1/6 of mode 1. = 0.1, D0 = 10m, ks = 45m1/3 s1 (ks = gCh Rh ), fully non-linear
closure and transparent boundary condition. . . . . . . . . . . . . . . . . . . . . 2.7 Tidally averaged value of the free surface slope evaluated through equations 2.35 and 2.36 or from the longitudinal slope h0 /x of the mode 0 of the spectrum
1/6 of the free surface elevation. = 0.1, ks = 45m1/3 s1 (ks = gCh Rh ), D0 = 10m,
26
transparent boundary condition. . . . . . . . . . . . . . . . . . . . . . . . . . . 2.8 Tidally averaged value of the free surface level evaluated through equation 2.37 or from the amplitude h0 of the mode 0 of the spectrum of the free surface elevation.
= 0.1, ks = 45m1/3 s1 (ks = gCh Rh
1/6
28
29
2.9
different values of , as obtained through the numerical model. The interpolating power laws are reported in the plot with the corresponding correlation coefcient R2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.10 The coefcients m (left) and k (right) of equation 2.39 as function of the amplitude ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.11 Marginal conditions for the amplication of tidal amplitude in the plane, for different values of , as obtained through the numerical model. The interpolating 33 32 31
Marginal conditions for the amplication of tidal amplitude in the plane, for
power laws are reported in the plot with the corresponding correlation coefcient R2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.12 Marginal conditions in term of the modied dimensionless parameter dened in equation 2.41 The interpolating power law is reported with the corresponding correlation coefcient R2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 3.2 3.3 Boundary condition in the case of drying area. . . . . . . . . . . . . . . . . . . . Boundary condition in the case of wetting area. . . . . . . . . . . . . . . . . . . The long term evolution of the bottom prole of a convergent estuary for different values of channel length. Lb = 120km, D0 = 10m, a0 = 4m, Ch = 20, Ds = 101 mm; Le = 160km (a), Le = 480km (b). The longitudinal coordinate x+ is scaled with the length Le , the bottom elevation + is scaled with the reference depth D0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 34 39 39
List of Figures
3.4 The equilibrium bottom proles in short (a) and long (b) channels with different boundary conditions at the mouth of the tidal channel: vanishing sediment ux (dashed lines) and equilibrium sediment ux (solid lines). D0 = 10m, a0 = 4m, Ch = 20, Ds = 101 mm; Le = 160km, Lb = 120km (a); Le = 280km, Le = 40km (b). The longitudinal coordinate x+ is scaled with the length Le , the bottom elevation + is scaled with the reference depth D0 . . . . . . . . . . . . . . . . . . . . . . 3.5 Degree of asymmetry, maximum, minimum and residual values of ow velocity (a), normalised with the average of its peaks at the mouth, and of sediment ux (b), normalised with gD3 , along the estuary, the longitudinal coordinate x+ s is scaled with Le . Le = 160km, D0 = 10m, a0 = 4m, Ch = 20, ds = 101 mm; different values of convergence length: Lb , (a1 , b1 ), Lb = 160km, (a2 , b2 ), 3.6 Maximum, minimum and residual values of ow velocity U +, Lb = 10km, (a3 , b3 ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . along the estuary after one cycle (dashed line) and at equilibrium (continuous line), for different boundary conditions at the seaward and: (a) vanishing sediment ux; (b) equilibrium sediment ux. Le = 40km, Lb = 20km, D0 = 5m, a0 = 2m, Ch = 20, ds = 101 mm; the velocity U + is normalized with the maximum value at the mouth at equilibrium and the longitudinal coordinate x+ is scaled with Le . . . . . . . . . 3.7 Maximum, minimum and residual sediment ux q+ s scaled with gD3 s along the estuary after one tidal cycle (dashed line) and at equilibrium (continuous line), for different boundary conditions at the seaward end: (a) vanishing sediment ux; (b) equilibrium sediment ux. Data as in gure 3.6. The longitudinal coordinate x+ is scaled with Le . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.8 3.9 Equilibrium length of the estuary La as a function of the initial length Le , for different values of convergence length. D0 = 10m, a0 = 4m, Ch = 20, ds = 101 mm. 48 Dimensionless equilibrium length Lc /D0 as a function of the dimensionless degree of convergence D0 /Lb , for different values of D0 . = 0.4, Ch = 20, ds = 101 mm. 4.1 4.2 5.1 5.2 5.3 5.4 5.5 Sketch of the channel and notation. . . . . . . . . . . . . . . . . . . . . . . . . 49 52 56 47 46 44 43
Suspension of sediment: relevant uxes. . . . . . . . . . . . . . . . . . . . . . . Computational domain in natural scale (left) and in logarithmic deformed scale (right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The boundary tted coordinate system. . . . . . . . . . . . . . . . . . . . . . . . Computational cell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Computational grid: horizontal spacing. . . . . . . . . . . . . . . . . . . . . . . Lagrangian approach, an example of trajectory. . . . . . . . . . . . . . . . . . .
62 64 66 66 67
xi
List of Figures
5.6 Computational grid: vertical spacing. Near the bed the half cell allows to impose the no-slip condition, at the free surface the whole cell allows to impose the vanishing stress condition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.7 Vertical velocity prole under uniform ow condition: analytical solution (red solid line) and numerical solution (black dots). 0 = 1, R p = 4, Ds = 105 , = 0.1, nz = 50. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.8 Numerical error in the estimate of velocity prole under uniform ow condition: norm E1 (left), mean value (right). 0 = 1, R p = 4, Ds = 105 . . . . . . . . . . . 5.9 Numerical error in the estimate of velocity prole under uniform ow condition: norm E1 (left), mean value (right). 0 = 1, R p = 10, Ds = 2 105 . . . . . . . . . 5.10 Comparison of present numerical results with the analytical solution of Tubino et al. (1999): vertical proles of the amplitude U1 (left) and phase lag (right) of the perturbation of longitudinal velocity. nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green) and the analytical solution (black). The phase lag is measured with respect to the peak of bed elevation, the amplitude U1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 102 . . . . . . . . . . . . . . . . . . . 5.11 Comparison of numerical results with the analytical solution of Tubino et al. (1999): vertical proles of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration. nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green) and the analytical solution (black). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 102 . . . . . . . . . . . . . . . . . . . . . . . . 5.12 Comparison of numerical results with the analytical solution of Tubino et al. (1999): vertical proles of the amplitude U1 (left) and phase lag (right) of the perturbation of the longitudinal velocity. nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green) and the analytical solution (black). The phase lag is measured with respect to the peak of bottom prole, the amplitude U1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 10, Ds = 2 105 , = 0.1, = 15, ny = 32, nz = 100, A = 102 . . . . . . . . . . . . . . 84 83 82 81 80 80 69
xii
List of Figures
5.13 Comparison of numerical results with the analytical solution of Tubino et al. (1999): vertical structure of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration. ny = 16 (cyan), ny = 32 (red), ny = 64 (blue), ny = 128 (green) and the analytical solution (black). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the = 0.1, = 15, nx = 32, nz = 100, A = 102 . . . . . . . . . . . . . . . . . . . 5.14 Comparison of numerical results with the analytical solution of Tubino et al. (1999): the phase and the amplitude of the perturbations of longitudinal (u , Au ) and transverse (v , Av ) components of velocity and of suspended sediment concentration (c , Ac ) with respect to the wave-number . The phase lag is measured with respect to the peak of bottom prole, the amplitude A is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, = 20, Ds = 105 , R p = 4, nz = 100, A = 102 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.15 Dependence on the longitudinal grid spacing of the numerical results for the vertical proles of the amplitude U1 (left) and phase lag (right) of the perturbation of the longitudinal velocity component: nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green). The phase lag is measured with respect to the peak of bottom prole, the amplitude U1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 0.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.16 Dependence on the longitudinal grid spacing of the numerical results for the vertical proles of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration: nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 0.5. . . . . . . . . 5.17 Dependence on the longitudinal grid spacing of numerical results for the vertical proles of the amplitude U1 (left) and phase lag (right) of the longitudinal component of the velocity: nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green). The phase lag is measured with respect to the peak of bottom prole, the amplitude U1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 10, Ds = 2 105 , = 0.1, = 15, ny = 32, nz = 100, A = 0.5. . . 87 86 86 85 dimensionless amplitude of bottom prole A . 0 = 1, R p = 10, Ds = 2 105 , 84
xiii
List of Figures
5.18 Dependence on longitudinal grid spacing of numerical results for the vertical proles of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration: nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . R p = 10, Ds = 2 105 , = 0.1, = 15, ny = 32, nz = 100, A = 0.5. . . . . . . . . . . . 5.19 Vertical concentration prole under uniform ow and suspended load conditions evaluated without the splitting procedure: analytical solution (red solid line) and numerical solution (black dots). 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, nz = 100. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.20 Comparison of the numerical results obtained with the splitting procedure or the direct solution of the equation for the concentration eld: vertical proles of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration. Analytical solution (black), splitting method (red) and direct solution method (cyan). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, nx = 64, ny = 32, nz = 100, A = 102 . 89 5.21 Comparison of the numerical results obtained with the splitting procedure or the direct solution of the equation for the concentration eld: difference between the values of the phase lag predicted with the numerical solution with splitting (red dots) and without splitting (black dots) and those computed with the analytical solution. 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 102 . 90 6.1 Free bars in the Rio Branco, South America. (Image Science and Analysis Laboratory, NASA-Johnson Space Center. 18 Mar. 2005. "Earth from Space - Image Information." <http://earth.jsc.nasa.gov/sseop/EFS/photoinfo.pl?PHOTO=STS61C33-72> 28 Apr. 2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 A typical Fourier spectrum of the equilibrium bar topography with dominant bedload. k longitudinal modes, m transverse modes. . . . . . . . . . . . . . . . . . . 6.3 Marginal stability curves for bar formation: is width ratio, is the longitudinal wave-number. 6.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 93 92 89 88
The maximum growth-rate max is plotted versus the width ratio for different transverse modes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
xiv
List of Figures
6.5 Comparison between the time development of the amplitude of the leading components of bar topography under bed-load dominated conditions 0 = 0.1, = 20, Ds = 102 , R p = 11000 and with dominant suspended load 0 = 1, = 15, Ds = 105 , R p = 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.6 The Fourier spectrum of the equilibrium bar topography with dominant suspended load: k denotes longitudinal modes, m transverse modes 0 = 1.25, = 12, Ds = 6.7 2 105 , R p = 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Time development of the amplitude of the leading components of the Fourier representation of bed topography for different values of Shields stress = 11, 6.8 The computed values of width ratio em at which local emergence of bar structure is observed are plotted versus Shields stress for two different values of particle Reynolds number. 6.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 The equilibrium bed topography for 0 = 1.25, R p = 10, = 12, Ds = 2 105 . . 100 Ds = 2 105 , R p = 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 98 97
6.10 The equilibrium bed topography for 0 = 1.5, R p = 10, = 13, Ds = 2 105 . . . 101 6.12 The equilibrium bed topography for 0 = 2, R p = 10, = 12, Ds = 2 105 . . . . 102
6.11 The equilibrium bed topography for 0 = 1, R p = 10, = 10, Ds = 105 . . . . . 101
6.13 The equilibrium bed topography for 0 = 1.5, R p = 4, = 14, Ds = 105 . . . . . 102 6.14 The equilibrium bed topography for 0 = 1, R p = 4, = 15, Ds = 105 . . . . . . 103 6.15 The equilibrium bed topography for 0 = 2, R p = 4, = 12, Ds = 105 . . . . . . 103 6.16 Ratio between the suspended load and the bed load, as a function of the sediment particle Reynolds number R p , for different values of the Shields stress = 0.5 (a), = 1 (b), = 1.5 (c), = 2 (d). The computation is performed using the standard closure relationships of van Rijn (1984), which are not valid for the higher values of R p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 6.17 Time sequence of bed topography during the tidal cycle, under bed load dominated condition. 0 = 0.1, = 13, R p = 11000, Ds = 102 . . . . . . . . . . . . . . . 106 6.18 The bed topography under suspended load dominated condition. = 2, Ds = 105 , R p = 4, = 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 6.19 The time development of the Fourier component A11 of the bed prole for = 2, Ds = 105 , R p = 4, = 15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 6.20 The time development of the Fourier components A11 , A21 , A31 , A41 of the bed prole for = 2, Ds = 105 , R p = 4, = 13. . . . . . . . . . . . . . . . . . . . . . 108 7.1 7.2 Suspension of sediments beyond an abrupt change of the bed boundary condition. 112 Vertical concentration and velocity proles. . . . . . . . . . . . . . . . . . . . . 113
xv
List of Figures
7.3
Vertical proles of the perturbation C1 at different cross sections: Lb = 10km,
= 0.5m, D = 5m and R p = 10, = 0.023 (left), R p = 4, = 0.042 (right). 0 0 Dotted line: numerical solution; continuous line: analytical solution; dashed line: analytical solution assuming 0,x = 0. . . . . . . . . . . . . . . . . . . . . . . . 117 7.4
Vertical proles of the perturbation C1 at different cross sections: Lb = 2.5km, = 1.5m, D = 10m, R p = 4, = 0.37 (left); Lb = 5km, = 0.5m, D = 5m, 0 0 0 0
R p = 4, = 0.085 (right). Dotted line: numerical solution; solid line: analytical solution; dashed line: analytical solution assuming 0,x = 0. 7.5 Longitudinal proles of the perturbation qs1 :
Lb
. . . . . . . . . . . 117 0 = 0.5m, D = 5m 0
= 10km,
R p = 10, = 0.023 (left), R p = 4, = 0.042 (right). Dotted line: numerical solution; continuous line: analytical solution. . . . . . . . . . . . . . . . . . . . 118 7.6
Longitudinal proles of the perturbation qs1 : Lb = 5km, = 1.5m, D = 10m; 0 0
R p = 10, = 0.10 (left), R p = 4, = 0.19 (right). Dotted line: numerical solution; continuous line: analytical solution. . . . . . . . . . . . . . . . . . . . . . . . . 118 7.7 Difference between the numerically and the analytically evaluated amplitude of the rst (left) and second (right) mode of the Fourier spectrum, as a function of . Dots R p = 4 and crosses R p = 10. . . . . . . . . . . . . . . . . . . . . . . . . . 119 7.8 Difference between the numerically and the analytically evaluated phase lag of the rst (left) and second (right) mode of the Fourier spectrum, as a function of . Dots R p = 4 and crosses R p = 10. . . . . . . . . . . . . . . . . . . . . . . . . . 120
xvi
List of Tables
2.1
Values of the amplitude ratio for various estuaries, evaluated using the data reported by Lanzoni and Seminara (1998). . . . . . . . . . . . . . . . . . . . . . . 14
xvii
List of Tables
xviii
List of symbols
Symbol a a, b, d a0 A An B B B B0 c0 cx , cy , cz C Ce Cf Cf0 Ch C0 C0 C1 LT 1 [L] [L] [L] [L] [L] dimensionless reference level for the concentration eld coordinate of the trajectory foot tide amplitude matrix normalized Fourier component tide asymmetry angle between the shear stress and the bed load vector ratio between local and turbulent effects channel width channel half width asymptotic channels width channel width at the mouth aspect ratio frictionless celerity Courant numbers volumetric sediment concentration equilibrium sediment concentration near the bed friction factor reference friction factor dimensionless Chezy coefcient analytical contribution to the concentration depth averaged concentration numerical contribution to the suspension
Eq. 4.26c 5.12a, 5.12b and 5.12c 2.1 5.19 and 5.21 2.34 3.6 4.23a 7.2 2.2 4.1a 2.2 2.2 4.2a 2.15 5.41 4.3d 4.26a 4.28 4.2c 2.4 5.34 7.9 5.32
xix
Symbol C1 CFL D D0 D 0 Ds D s es E f F0 0 H n U n g G G1 , G2 h h hn k ks Kv Lb
Lb
Eq. perturbation of the concentration Courant-Friedrichs-Levi number convergence parameter = Cf0 dimensionless ow depth [L] [L] [L] mean depth at the mouth reference ow depth dimensionless grain size grain size ratio between advection and settling effects weighting function equivalent roughness vector dimensionless tide amplitude vertical concentration prole Froude number angle of repose of bed material vertical structure of the concentration phase lag for the free surface elevation phase lag for the ow velocity LT 2 gravity accelerations vector Rouse numbers [L] [L] [L] bed elevation dimensionless bed elevation free surface elevation dimensionless free surface elevation Fourier component of the free surface elevation unit vector in the vertical direction L T 1 [L] [L] [L]
1 3
7.12 2.8 2.20 4.3a 4.1b 2.1 4.1b 4.2b 4.2b 7.7 5.42 4.26c 5.25 2.1 5.35 4.3a 4.25 7.9 2.9 2.10 2.3a 5.23 and 5.24 5.36a and 5.36b 3.1 4.1b 2.3a 4.1b 2.9 4.10
Gaukler-Strickler coefcient von Karman constant convergence length dimensional wavelength length of the estuary 2.2 7.17 2.12
Le
Symbol Lg i, j, k if j n n ns N T p P p1 , p2 T Q Q0 Qbx , Qby qs qs1 0 cr cr r r rt , rx , ry Rh Rp s t t T [T ] [L] L2 T 1 L3 T 1 LT 1 [L] frictionless wavelength dimensionless wave length eigenvalues numerical indexes mean bed slope ow dissipation direction normal to the bed load vector unit vector normal to the bed unit vector normal to the free surface vertical structure for the eddy viscosity dimensionless eddy viscosity sediment porosity vertical structure for the eddy diffusivity forcing terms dimensionless eddy diffusivity ow discharge dimensionless parameter dimensionless bed load vector solid discharge perturbation of the suspended load dimensionless Shields stress effective Shields stress reference Shields stress critical value of Shields stress for vanishing bed slope critical value of Shields stress empirical coefcient empirical coefcient coefcients for the coordinate transformation hydraulic radius particle Reynolds number direction the bed load vector time dimensionless time transport parameter
Eq. 2.16 7.2 2.7 4.3a 2.3a 4.24 4.11 4.18 4.8 4.3a 3.1 4.9 7.16 4.3d 2.3a 4.12 4.29a 3.1 7.18 4.22, 3.3 4.27 4.2c 4.25 4.25 4.24 4.25 5.4a, 5.4b and 5.4c 2.4 4.2d 4.25 2.3a 4.1d 4.26b
Symbol T0 T0 Tbed Tb Ts u, v, w un U U U U0
U0
Eq. [T ] [T ] [T ] [T ] [T ] tide period reference time scale time scale for bed evolution time scale for bed evolution due to bed load time scale for bed evolution due to suspended load dimensionless shear stress ML1 T 2 shear stress dimensionless velocity dimensionless shear velocity LT 1 LT 1 Fourier component of the ow velocity ow velocity vector of the dimensionless x-component of the ow velocity LT 1 LT 1 LT 1 LT 1 local maximum velocity velocity scale reference velocity frictionless velocity scale vector of the dimensionless y-component of the ow velocity T 1 L2 tidal frequency ratio between advective and turbulent effects cross sectional area contravariant component of the vertical velocity dimensionless particle settling velocity [L] longitudinal coordinate pointing landwards plain dimensionless coordinate dimensionless vertical boundary tted coordinate ratio between friction and inertia vertical dimensionless coordinate dimensional reference level for the velocity prole Rouse number 2.32 2.21 4.1c 2.17 5.18 2.9 7.2 2.3a 5.5 4.30 2.2 4.1a 4.4 2.19 4.1b 5.14 7.2 5.1 4.7 4.1c 4.7 2.10 2.4 5.18 2.11 4.1d 3.5 6.2 6.1
Ug V w Ws x x, y z z0 Z0
Eq. 5.2
1 Introduction
An estuary is a water body located close to the sea, whose main distinctive feature is the occurrence of the tide. Following Perillo (1995) an estuary is a semi-enclosed coastal body of water that extends to the effective limit of tidal inuence, within which sea water entering from one or more free connections with the open sea, or any other saline coastal body of water, is signicantly diluted with fresh water derived from land drainage, and can sustain euryhaline (i.e. able to tolerate a wide range of salinity) biological species from either part or the whole of their life cycle. Tides take place on a time scale of the order of magnitude of a day, which is relatively short with respect to the typical time scale of the morphodynamic behaviour of a tidal system. As an example typical patterns of tidal oscillation in the Venice lagoon are reported in gure 1.1. Two main phases occur: the ood phase, when the ow is directed from sea to land, and the ebb phase, when the ow goes in the opposite direction. Many types of natural environments can be classied as tidal systems. Their shape and dynamics depend on many factors, like the amplitude of tidal oscillations, the presence of subsidence, the mean sea level rise, the typology of sediment. Figures 1.2 and 1.3 show two typical examples of tidal systems, a lagoon (the Venice lagoon) and a tide dominated estuary (the Western Scheldt). A glance to these pictures suggests that tidal systems can be extremely different from one another. The classications proposed recently by Seminara et al. (2001a) and by Perillo (1995) are mainly based on morphological criteria and on the amplitude of the tidal range, and consider most of the morphological features observed around the world in tidal systems, like lagoons, tide dominated estuaries, delta estuaries and tidal rivers. In simple descriptive terms tide dominated estuaries are those in which tidal currents play the dominant role, while density driven circulations are nearly absent because strong tidal effects are able to destroy vertical stratication. Tide dominated estuaries are usually funnel-shaped and characterised by the presence of sand waves, intertidal ats and salt marshes (see gure 1.3). The periodic tidal currents can store large volumes of water in the estuary at high tides, which are followed by drainage at low tides. The total volume of water exchanged during the tidal cycle is known as the tidal prism. In tide-dominated estuaries the tidal prism is at least an order of magnitude greater than the volume of water discharged by the river daily (Harris et al., 1993).
1. Introduction
Figure 1.1: Tide oscillations in Venice lagoon. (Comune di Venezia <http://www.comune.venezia.it/maree/astro.asp>) Coastal lagoons are inland water bodies connected to the sea by one or more restricted inlets from which a complex network of tidal channels may originate. Delta estuaries are shoreline protuberances, formed where river supplied sediments have accumulated in standing bodies of water faster than it can be redistributed by basinal processes such waves, currents, tides or submarine failures (Write, 1985). Tidal rivers are those systems that are affected by tidal actions but salt intrusion may be limited to the mouth or it is totally absent. Normally these estuaries are associated to large river discharge. Despite this high variability, for the purpose of the present work it is important to focus on the recurring elements of natural tidal systems. Besides the forcing effect due to tide oscillations, another relevant and common characteristic of such systems is that sediments are typically very ne; an example is given in gure 1.4 where the grain size distribution of the Western Scheldt is reported. Sediments can be transported as wash load, suspended and bed load. The rst mechanism involves only the nest fraction (clay particles): the vertical concentrations proles of wash load are fairly homogeneous. Suspension occurs as the result of two counteracting mechanisms, namely the ability of turbulence to raise sediment grains and their tendency to settle due to gravity. Grains smaller than 0.15mm can be entrained into suspension as soon as they begin to move, while larger
1. Introduction
Figure 1.3: The Western Scheldt. (Image downloaded from the internet)
1. Introduction
() Figure 1.4: Sediment grain size distribution in the Western Scheldt. (Image downloaded from the internet) grains move dominantly as bed-load at low values of ow velocity and then go into suspension when velocity attains higher values during the tidal cycle. The dynamics of a tidal channel is a complex phenomenon that reveals the presence of several spatial and temporal scales. The spatial scales in tidal systems may range between the embayment length and the ow depth and can be dened according to the classication introduced by de Vriend et al. (2000): micro-scale (or small-scale): the level of the smallest morphological phenomena associated with water and sediment motion (ripple and dune formation); meso-scale: the level of the main morphological features (depositional bars, channels and shoals); macro-scale: the level at which the meso-scale features interact; mega-scale: the level at which the principal elements of the entire system interact (i.e. the estuary considered as a whole water body). It is worth noticing that observed bed-forms in tidal systems may range from centimeter-size, ple and Rhodes, 1995). current ripples, through decimeter-meter-size, dunes, and meter-size, bar f orms (see Dalrym-
1. Introduction The time scale of bed evolution Tb is in general independent of the tidal period T0 and is mainly related to ow and sediment characteristics. Depending upon the spatial scale under investigation, the suitable time scale of bed development may range from 102 to 106 tidal periods ( month century). The lower limit corresponds to the behaviour of meso-scale bedforms (bars) in small channels, while the upper limit corresponds to mega-scale phenomena. Moreover the time scale associated to the formation of ripples and dunes is comparable to the tidal period. Another common element, which is relevant for the present work, is that many tidal systems are frictionally dominated, which means that the role of inertia is much smaller compared to that played by friction. The classication of Perillo (1995) points out that the morphological characteristics are inuenced by the interaction of channels with shoals and intertidal zones, like tidal ats and salt marshes. An intertidal at is a deposit emerging during low tide and submerged during high tide, while salt marshes are environments in the intertidal zone where a muddy substrate can support a varied and normally dense stand of halophytic plants (Reineck, 1972; Allen and Pye, 1992). The morphodynamics of tidal systems is still not completely understood, though many contributions have been recently proposed to investigate their behaviour (Friedrichs and Aubrey, 1994; Lanzoni and Seminara, 1998; Schuttelaars and de Swart, 1999; Schuttelaars and de Swart, 2000; Lanzoni and Seminara, 2002). Due to the fact that uvial systems have been more deeply investigated in the past decades, many authors have tried an extension of river theories to tidal contexts (Seminara and Tubino, 2001; Solari et al., 2002). The relevance of understanding the morphological behaviour of tidal channels lies in the fact that such systems may have a strong impact on human activities, like in the case of the city of Bruges, formerly an important commercial harbour, which lost its commercial role during the 15th century when the tidal channel which joined the city to the sea was lled with sediments. Also note that several important harbours in Europe are located on tide-dominated estuaries (Antwerp, Hamburg, Bordeaux, London). The present study investigates the morphological behaviour of tidal channels, like tide-dominated estuaries and channels in coastal lagoons, through different models that reect the morphodynamics at the relevant spatial and temporal scales. We rst focus the attention on the mega-scale morphodynamic behaviour of tidal channels. The main aim is to determine an equilibrium conguration of the bed prole as a function of geometrical parameters (channel depth, width, degree of convergence, etc.) and hydrodynamic parameters (tidal excursion, friction factor). Note that the existence of equilibrium congurations in morphodynamical systems is a delicate concept, that has been widely discussed in the literature (e.g. Seminara, 1998). Then we move our attention to the meso-scale morphodynamic behaviour of tidal channels,
1. Introduction with the specic aim of investigating the development process of bar-forms, which are prominent morphological features within most estuaries. Due to their large size relative to channel width, these bedforms have an important role on the dynamics of an estuary and strongly affect its use for human activity. The study of the formation of depositional bars is a key ingredient for the comprehension of the morphodynamics of both river and tidal systems, as pointed out by many authors (e.g. de Vriend et al., 2000, Tubino and Seminara, 1990), since the behaviour of such systems is mainly determined by the non-linear interaction of free- and forced-bars, the forcing effects being essentially related to the presence of channel curvature and width variations. Understanding the mechanism of formation and migration of bars and their response to forcing effects is then the preliminar requirement for the development of models able to predict the evolution of tidal channels. The problem of the development of bedforms in ne sediment systems (like tidal channels) has been studied in the past by many authors, namely in the case of small scale bedforms (i.e. dunes) both in uvial and tidal contexts; for example, Southard and Boguchwal (1990) provide an extensive set of observations in laboratory umes, while Dalrymple et al. (1978) and Rubin and McCulloch (1980) propose a classication based on eld observations. Dune size, shape, orientation and migration speed have been studied by several authors (Allen, 1968, 1982; Nordin, 1971; Yalin, 1964, 1977, 1987; Rubin and Hunter, 1987; Rubin and Ikeda, 1990; van Den Berg, 1987). Dunes are very often superimposed over large scale bedforms and induce signicant dissipative effects on the ow. On the contrary, the mechanism of bars formation in ne sediment systems is not widely understood as for dunes. Referring to the uvial case, indeed, their development process has been investigated only recently in sandy rivers (Tubino et al., 1999; Watanabe and Tubino, 1992). While the main results are discussed in detail in Chapter 6, it is sufcient to recall here that when sediment transport mainly occurs as suspended load, several distinctive features arise with respect to gravel bed rivers. In the latter case, which has been widely investigated in the last thirty years (see for example Colombini et al., 1987), free bars exhibit a wavelength nearly equal to 6 7 times the width of the channel. In the case of the Adige river near Trento, Italy (the closest example to the place where the present study has been carried out), whose averaged width is 80 m, a bar length of nearly 500 m is predicted. When suspended load is dominant, theoretical predictions of reaches, where ne sediments are observed, and the width may be of the order of few hundreds of meters (e.g. the lower reach of Po river, the largest river in Italy), the predicted wavelength of free bars would range up to 25 30 km! This poses a severe limitation to eld observations. A recent analysis of Federici and Seminara (2003a) indeed shows that, for the development of a regular train of migrating free bars, the channel should be straight for a length of the order free bar wavelength may increase up to 50 60 channel widths. This implies that in lower river
1. Introduction of several bar wavelengths. Hence, in order to observe at least some free bar wavelengths, in the case of gravel bed rivers the channel must keep approximately straight for few kilometres, as may occur in articially straightened channels and also in natural rivers; on the contrary ne sediment channels should keep straight for tens of kilometres, a condition which can be hardly met in natural rivers. We also note that while many contributions are now available to understand the formation of point forced bars in curved uvial channels under bed load dominated conditions (for example Ikeda et al., 1981, Tubino and Seminara, 1990, Seminara et al., 2001b), only few of them consider the effect of suspended sediment transport and its crucial implications in the planimetric development of meandering channels. Tidal bar-forms display a shape similar to that of uvial bars, with oblique and longitudinally oriented depositional fronts. Several eld observations, like those on the Salmon River estuary, the Bay of Fundy (Dalrymple and Rhodes, 1995) and along the tidal creeks of South Carolina (Barwis, 1978), suggest that the bar length ranges around 6 times the channel width, a similar scaling of the river case. Dalrymple and Rhodes (1995) assert that ow reversal does not alter the process of bar formation (like in the case of dunes) and propose a distinction between "repetitive bar-forms", which are very similar to uvial bars, and "elongate tidal bars", which display more complex structures and are observed, for example, along the Thames (Robinson, 1960) and in the Bay of Fundy (Knight, 1980; Dalrymple et al., 1990). As stated in the review of Dalrymple and Rhodes (1995) the dynamics of estuarine bar-forms is still more poorly understood than their uvial counterparts. It is worth noticing that in an estuary only a part of the tidal cycle contributes to the morphodynamic behaviour of the channel, because close to the ow reversal the shear stress falls below the threshold value for sediments transport. The hydrodynamics of uvial systems is characterised by negligible values of local ow accelerations, as occur in tidal channels when the ow velocity attains its maximum value. In these conditions, corresponding to the peak of sediment transport, local values of ow acceleration are negligible. When the ow velocity is close to zero, settling dominates over turbulence. Differences with respect to the river case dramatically increase at the ow reversal, since ow accelerations are strong: in these cases the vertical distribution of longitudinal velocity often exhibits a peculiar structure, displaying opposite directions at the free surface, where inertia effects are larger, and near the bed, where inertia vanishes (see gure 1.5). In the last two decades mathematical approaches, mostly based on perturbative methods, have been successfully applied to the study of the dynamics of free and forced-bar in uvial systems and the planimetric evolution of river channels. Such approaches have been recently extended to tidal systems by Seminara and Tubino (2001), Solari et al. (2002) and Toffolon (2002). The above theories mostly refer to the case of frictionally dominated tidal channels with constant width and follow a so-called "local" approach, whereby a reach length of the order of few
1. Introduction
10 9
1. Introduction
Figure 1.6: Channels in the Venice lagoon. (Courtesy of Walter Bertoldi, 2003)
whose results seem quite different from the theoretical predictions of Seminara and Tubino (2001). The complexity of the above picture is further enhanced by the tendency of ne sediment bars to emerge from the free surface forming islands. A similar behaviour is observed also in gravel bed rivers at relatively high values of the width to depth ratio. A possible explanation for such phenomenon can be related to the reduced stabilising effect of gravity when suspended load is dominant. In fact, suspended load is not directly inuenced by gravitational effects, as pointed out by Talmon et al. (1995), since suspended sediments dont move "over" the bed but "within" the water column, while bed load is affected by gravity which moves sediments downslope from the top of bars to the pools. In gravel bed rivers this provides a stabilising contribution which inhibits the growth of bars and prevents the formation of islands. The high complexity inherent in the dynamics of tidal systems can be considered one of the main reasons why results of the existing mathematical models still do not agree on many subtopics. Sound predictions of the morphodynamic evolution of tidal channels require a deep understanding of the physics involved and a careful model setup. Namely, at least a suitable predictive model of the morphodynamic behaviour of tidal channels shall incorporate the following crucial aspects: the strongly non-linear character of the physical processes and the presence of suspended load. The equations for the ow eld and the bottom evolution are non-linear, while most of the theoretical results presently available have been obtained through linearized models. Hence, in the present work suitable numerical models are introduced
1. Introduction to solve the complete set of equations without linearizations. Furthermore, when suspended load is included, results for bottom topography crucially depend on the ability of the model to reproduce adequately the adaptation of the concentration proles to variable ow conditions, which implies, at least within the context of a meso-scale analysis, the adoption of a three dimensional formulation whereby vertical concentration proles can be estimated. Hence, two different models have been developed in the present work, a 1D model and a 3D model. The former is introduced to investigate the propagation of the tidal wave in convergent and non-convergent estuaries and to study large scale morphodynamic processes. The three dimensional model is then used to characterise the role of suspended sediment load on the formation of estuarine and river free bars. More specically the model is applied within the context of a "local" approach to study the stability of bar-forms. Hence, the procedure adopted herein allows one to investigate not only the initial process of bar formation (when the amplitude of bed-form is small and a linearized approach is also applicable) but also its nite amplitude behaviour, which cannot be studied within a linear framework. For the sake of simplicity and of wider generality the numerical models are not applied to real estuaries, though the ranges of values of the relevant parameters have been selected in order to simulate real cases. The models are applied to rectangular channels, both with convergent and non convergent geometry, and subjected to a simple semi-diurnal M2 tide, without over-tides. It is useful to recall that the gravitational effect due to the moon is purely sinusoidal, therefore in the open sea the free surface oscillation is very close to a sinusoidal function (commonly referred to as the M2 tide). Over-tides are measured as the difference with respect to such pure sinusoidal behaviour and appear as the higher frequency components of the tidal oscillation. They originate from non-linear effects due to water ow, that characterise every real case.
The present work is organised as follows. In Chapter 2 the main distinctive features of tide propagation in convergent estuaries are summarised; in Chapter 3 the large scale morphodynamical processes in convergent estuaries are investigated through 1D numerical model. Chapter 4 is devoted to the formulation of the three-dimensional model for the study of meso-scale development of tidal channels; the detailed description of the numerical model follows in Chapter 5. In Chapter 6 the process of free bars formation in sandy rivers and tidal channels is investigated. Finally, in Chapter 7 a comparison is presented between the results obtained using different models for the suspended sediment load along with some concluding remarks on the predictive ability of approximate models.
10
In this chapter we focus on the main hydrodynamic properties of tidal channels, namely those associated with tide propagation in tide dominated estuaries. We will show that a fully non-linear hydrodynamic model is required to capture the most relevant elements of their behaviour. Tide dominated estuaries usually have a funnel shape (gure 2.1) characterised by the presence of sand waves, intertidal ats and salt marshes. The morphodynamic behaviour of such systems depends strongly by upstream propagation of the tide. The celerity at which tide moves along the estuary is governed by the shallow water equations and is therefore an increasing function of water depth. As a result of this depth dependence, tides are deformed during their propagation: ood velocities are greater than ebb velocities and the ood-phase is generally shorter than the ebb-phase. The upstream decrease of the cross sectional area (depth and width of the channel) forces the tide to increase its amplitude during upstream propagation; however, frictional dissipation tends to counteract this effect, decreasing the amplitude of tidal wave. Hypersyncronous estuaries are those where the effect of channel convergence is dominant and the wave is amplied; around the world many examples of such systems can be found, like the Scheldt (Netherlands - see gure 1.3). Besides the geometrical characteristics of tidal channels, the hydrodynamics of estuaries is strongly affected by many other elements, like the presence of short waves incoming from the sea or due to the wind. Moreover in some cases the river discharge at the landward end of the channel plays an important role and may induce stratication phenomena and density driven circulations. Here we restrict the analysis to a widespread class of tidal inlets, namely the well-mixed estuarine channels. This kind of morphological large-scale elements includes those estuaries and lagoon channels where the tidal forcing is so strong that stratication does not occur. The absence of the salt wedge allows one to consider a constant water density and to describe the ow eld using the usual equations of single-phase uid. Understanding the hydrodynamics of tidal channels is relevant for many environmental issues. In particular, for the evaluation of the consequences of both natural and anthropogenic modi-
11
12
2. On tide propagation in convergent and non-convergent channels cations, it is essential to describe the role of several basic factors (length of the estuary, friction, channel convergence, bed altimetry, river discharge) on the properties of the tidal wave (see also Toffolon, 2002). The problem of the propagation of the tidal wave in convergent channels has been tackled by several authors in recent years (e.g. Friedrichs and Aubrey, 1994; Friedrichs et al., 1998; Lanzoni and Seminara, 1998), following the contribution of Jay (1991), who has rst revisited the problem of tide amplication due to channel convergence, originally investigated by Green (1837). Though these theories provide valuable results for the comprehension of the basic mechanisms, they mostly rely on the assumption that some parameters can be considered small so that the mathematical description of the system can be simplied through linear or weakly non-linear analysis. In the present work we try to remove the above restriction and to investigate large amplitude effects on tidal wave propagation, tackling the fully non-linear problem through a suitable numerical model. We recall that among the few fully non-linear models that have been recently applied to these issues (Hibma et al., 2001, 2003) most of them have been devoted mainly to characterise the morphodynamic behaviour of tidal channels rather than the process of tide propagation. Tidal waves are driven by water oscillations imposed at the channel mouth, which is connected with the outer sea. Almost all the non-linear terms appearing in the governing equations for the ow eld are proportional to the ratio = a0 D0 (2.1)
between the tidal amplitude a0 and the average water depth D0 , where the subscript 0 denotes the values at the channels mouth. The above mentioned analytical solutions assume the ratio to be small enough for its effect to be negligible, at least at the leading order of approximation. We may notice that ranges between 0 and 1 and in many real estuaries it can reach relatively large values, as shown in table 2.1. Notice that tidal wave generates over-tides along its propagation. For the sake of simplicity, in the present work we neglect the effect of external over-tides and force the system with a purely sinusoidal semi-diurnal M2 tide at the mouth of the estuary. Over-tides, like the quarter-diurnal M4 , occur at the mouth of estuaries when the offshore shelf is relatively wide and at; moreover the presence of a wider wave spectrum at the channel inlet may affect the overall hydrodynamics of the system, while increasing the number of degrees of freedom in the analysis. Accounting for the effects of over-tides is beyond the aim of the present analysis.
13
Estuary Outer Bay of Fundy Bristol Channel Hudson Irrawaddy Rotterdam Waterway Columbia Potomac Delaware Soirap Fraser Khor Elbe Severn Tees Gironde Thames Scheldt Hoogly St. Lawrence Fleet Ord Conwy Tamar
a0 [m] 2.1 2.6 0.69 1.0 1.0 1.0 0.65 0.64 1.3 1.5 1.3 2.0 3.0 1.5 2.3 2.0 1.9 2.1 2.5 0.6 2.5 2.4 2.6
D0 [m] 60.0 45.0 9.2 12.4 11.5 10.0 6.0 5.8 7.9 9.0 6.7 10.0 15.0 7.5 10.0 8.5 8.0 5.9 7.0 1.5 4.0 3.0 2.9
0.04 0.06 0.08 0.08 0.09 0.10 0.11 0.11 0.16 0.17 0.19 0.20 0.20 0.20 0.23 0.24 0.24 0.36 0.36 0.40 0.63 0.80 0.90
Table 2.1: Values of the amplitude ratio for various estuaries, evaluated using the data reported by Lanzoni and Seminara (1998).
14
15
2. On tide propagation in convergent and non-convergent channels 1994; Lanzoni and Seminara, 1998), according to the relationship: B = B + (B0 B ) exp x Lb (2.2)
where Lb is the convergence length and the asymptotic width B is included to set a minimum width landward, also in the case of strong convergence and long estuary. The standard onedimensional shallow water equations are used, which read: Q + t x Q2 + g h + g j = 0, x (2.3a)
Q + = 0, t x
(2.3b)
where t is time, x the longitudinal coordinate, Q is the water discharge, is the area of the cross dened in terms of bottom elevation as shown in gure 2.2; the frictional term is evaluated in the following form: j= Q |Q| 2 g2Ch Rh (2.4) section, h is the free surface elevation, g is gravity; furthermore, the water depth D = H is
having denoted with Ch the dimensionless Chzy coefcient and with Rh the hydraulic radius: Rh = BD . B + 2D
As for the longitudinal bottom prole, it is commonly observed that in tidal channels the ow depth decreases landward. Prandle (1991) founds that the behaviour of the width and depth of real estuaries can be described in terms of power laws. Bottom proles may be chosen analytically (e.g. linear, exponential) or evaluated through morphological models as pointed out in Chapter 3. In the present chapter, as rst step of the analysis, we assume the bed to be horizontal, in order to reduce the number of independent variables. The most suitable form for the numerical solution of the ow equations is (2.3), because it is a semi-conservative form. In order to point out the role played by different terms it is convenient to rewrite the system in non-conservative form. Moreover channel convergence can be written as follows: 1 dB 1 B B 1 = = , B dx Lb B Lb (2.5)
which is valid when the asymptotic width B is much smaller than the actual width B(x). In this case the dependence of the solution on the actual width of the channel can be ruled out, as it can be readily seen from equations (2.3a-2.3b), rewritten in non-conservative form in terms of the
16
2. On tide propagation in convergent and non-convergent channels cross-sectionally averaged velocity U = Q/: U h U +U + g + g j = 0, t x x U D UD D +D +U = 0, t x x Lb g j, while the last term in (2.6b) is related to channel convergence. (2.6a)
(2.6b)
Please note that the non linear terms in (2.6a) are the advective term U U and the frictional term x
The eigenvalues (2.7) of the system (2.3) are always distinct and generally have opposite sign in tidal system. Two boundary conditions are required by the system (2.3), one corresponding to each eigenvalue. This implies that boundary conditions must be imposed at each sides: the mouth of the channel and its landward end. Equations (2.3a-2.3b) are discretized through nite differences, with spatial step x and time step t, and are solved numerically using the explicit MacCormack (1969) method. The method consists of two steps: prediction and correction. The former is made using a forward difference while the latter with backward nite difference. This numerical method has a second order accuracy both in space and in time. The stability condition requires the Courant-Friedrichs-Levi number (CFL) not to exceed the unity: CFL = max t 1 x (2.8)
where max = max (|1 | |2 |) is the leading eigenvalue. Since the tidal wave tends to break during
its propagation due to non-linear terms, namely those associated with the effect of friction and convergence, a suitable articial viscosity is introduced through a TVD lter (Total Variation Diminishing) in order to remove the spurious oscillations around discontinuities, that are typical of second-order central schemes (see, for instance, Garcia-Navarro et al., 1992). Results can be given in term of Fourier modes, because in each section of the estuary the
solution for the ow eld is a periodic function, with a period equal to that of the tidal oscillation
17
2. On tide propagation in convergent and non-convergent channels imposed at the seaward boundary. Hence, results for the free surface elevation and for the ow velocity can be given the form of a standard Fourier representation: h (x,t) = h0 + hn (x) sin nt + H , n U (x,t) = u0 + un (x) sin nt + U , n
n=1 n=1
(2.9) (2.10)
of free surface elevation and velocity, respectively, and n is the phase lag of the n th harmonic. In equations (2.9 and 2.10) n = 1 corresponds to the semi-diurnal M2 -tide, n = 2 to M4 -tide and so on.
where = 2/T0 is the tidal frequency, T0 is the tidal period, hn and un are the n th components
At the landward boundary we impose a suitable relationship between the free surface level and water discharge. It is difcult to impose such condition, because many estuaries strongly interact and superpose to the nal reaches of rivers. Frequently the presence of a peculiar geometrical conguration induces a reection of the tidal wave. Two limit cases can be recognised: 1. the reecting barrier: this situation corresponds to the assumption of vanishing discharge at the landward end of the computational domain Q (x = Le ,t) = 0, and determines the complete reection of the tidal wave. 2. the transparent condition: this condition refers to a situation such that no obstacles are present at the landward end and the tidal wave exits from the computational domain without being deformed or reected. In this case the model reproduces the typical behaviour of a (2.12)
18
2. On tide propagation in convergent and non-convergent channels long estuary, for which the effect of the landward boundary condition is not relevant. The transparent condition is given considering the outgoing characteristic curve, corresponding to the positive eigenvalue 1 , that reads: h dh (x = Le ,t) h = + 1 dt t x Q dQ (x = Le ,t) Q = + 1 dt t x (2.13a)
(2.13b)
Recalling that to ensure stability requirement the Courant-Friedrichs-Levi number C = 1 t/x must not exceed the unity and approximating the characteristic curve by a linear function the conditions (2.13) in discrete form become: hk+1 = hk C + (1 C) hk N1 N N Qk+1 = Qk C + (1 C) Qk N1 N N represent respectively the present and the following time step. Notice that, as pointed out by Friedrichs and Aubrey (1994), the landward boundary condition is less important in the case of strongly convergent channels. In addition to the physically based boundary conditions the numerical scheme requires other two conditions, called ctitious, which have been imposed using the forcing term, if it is possible as in the case of the transparent conditions, or otherwise by solving the system of the ow eld. (2.14a) (2.14b)
where the subscript N represents the index of the last section and the superscripts k and k + 1
19
2. On tide propagation in convergent and non-convergent channels quently, all the parameters dened in the following only imply an evaluation of a velocity reference value in terms of external quantities. A rst relevant external parameter is the dimensionless tidal amplitude , dened in (2.1). When frictional effects are negligible, the linearized theories (see section 2.5 for further details) are suitable tools for the understanding of the ow eld; the linear solution in an innitely long channel with constant width is a wave function whose celerity is: c0 = gD0 . (2.15)
which can be taken as a reference length scale. Furthermore, when the role played by the frictional term is negligible, a frictionless velocity Ug can be dened in terms of external parameters in the following way: Ug = gD0 = a0 Lg . D0 T0 (2.17)
The relative role of friction can be quantied, like in Lanzoni and Seminara (1998), in terms of the ratio between frictional terms R and inertial terms S in momentum equation: U0 T0 R = 2 . S Ch D0 (2.18)
Notice, however, that the above ratio is dened in term of a priori unknown velocity scale U0 . In order to dene the ratio R/S in terms of external variables we use Ug (dened in 2.17) instead of U0 to write = a0 Lg Lg = 2 . 2 D2 Ch 0 Ch D0 (2.19)
The parameter represents the ratio between friction and inertia. Since is linearly proportional to , it follows that high values of can be related in almost all cases to frictionally dominated correspond to very deep estuaries, like the Bay of Fundy (ow depth 60m) or the Bristol Channel estuaries. Notice that in real estuaries may range between 4 5 and 100. The smaller values
(ow depth 45m), which are relatively rare to be found around the world. The values of most To account for the effect of width variation along the estuary, a dimensionless convergence
Lg , Lb
(2.20)
20
2. On tide propagation in convergent and non-convergent channels where Lb is the convergence length dened in (2.2).
The reference velocity U0 can be dened in a straightforward way only for some specic cases. In the case of validity of linear theories, when both and channel convergence are small, the scale of velocity U0 coincides with Ug , furthermore, in strongly convergent estuaries a suitable velocity scale can by readily obtained in the following form (Toffolon, 2002) U0 = 2 Ug , 4 (2.21)
where the limit = 4 has been tacked into account as the lowest, according to Jay (1991) and Lanzoni and Seminara (1998). In the case of strongly dissipative and weakly convergent channels the role played by inertia is negligible, friction is balanced by the gravitational term and the following velocity scale is found: U0 = 2
1 3
Ug
(2.22)
Recently Toffolon (2002) has proposed a more general, albeit simplied, formulation to evaluate a reference value of velocity at the mouth of the estuary, in terms of external parameters. Such formulations can be applied to both strongly or weakly convergent estuaries as well as to friction dominated or frictionless channels. It is based on the analysis of the relative weight of the various terms in the momentum and continuity equations (2.6a - 2.6b): through dimensional considerations, it leads to the following relationship for the velocity scale U0 , dened as the algebraic mean of peak values of ood and ebb velocity at the mouth of the estuary: U0 = where 1 1 Ug , 1 3 (2.23)
Equation (2.23) includes and improves the relationships proposed to cover the asymptotic case of strongly-convergent and weakly-dissipative estuaries and its dual case. Equation (2.23) indeed reduces to (2.21) in the case of strongly convergent channels, corresponding to high values of , for strongly dissipative estuaries (high values of ) the (2.22) is obtained. Also notice that (2.23) refers to constant depth channels and is valid provided the frictionless limit U0 < Ug is satised. The approximate velocity scale (2.23) is plotted, as a function of the parameters and , in gure 2.3, where also a comparison is made with the velocity evaluated numerically at the mouth of the
= 1 +
1+
1/3
4 . 3
(2.24)
21
20
0.4
0.3
15
0.5
10 5
0.4
0.6
0.5 0.6
0 0
10
20
30
40
50
60
70
Figure 2.3: Contour plot of the ratio U0 /Ug evaluated using the (2.23) (continuous lines) and the algebraic mean of peak values of ood and ebb velocity at the mouth evaluated using the numerical model (dotted lines). The different dotted lines are obtained through different values of in order to obtain a wider range for . Lb = 10km, D0 = 10m, 1/6 ks = 30 90m1/3 s1 (ks = gCh Rh ), transparent boundary condition landward, Le = 300km, a0 = 0.01 0.1 0.5 1 2 3m. estuary, as the algebraic mean of the peak values of ood and ebb phase at the mouth: 1 2 U (x = 0) f lood + |U (x = 0)ebb |max .
max
22
2. On tide propagation in convergent and non-convergent channels the following scales, where the superscript + denote the dimensionless quantities: t = T0t + , x = Lg x + , U = UgU + , D = D0 D+ , H = a0 H + . System (2.6a-2.6b) can be rewritten in the following form in the case of vanishing channel convergence: U + U + H + U +2 + U + + + + + 2 + = 0 t + x x gCh D D+ U + D+ + D+ + + U + + = 0 t + x x equation (2.25) are negligible and the system can be rewritten in the following form: U + H + + + =0 t + x H + U + + + =0 t + x which leads to the classic wave equation H + H + =0 t +2 x+2 (2.29) (2.27) (2.25) (2.26)
(2.28)
for the water level; a similar wave equations may be written also for the ow velocity. The solution of the (2.29) strongly depends on the boundary conditions. Assuming a purely M2 tidal forcing at the channel mouth, it is possible to recover the simple wave behaviour H + = sin [2 (x+ t + )] U + = sin [2 (x+ t + )] (2.30)
assuming an open landward boundary (innitely long channel), while the solution becomes a standing wave H+ =
+ cos (2 (Le x+ )) sin 2t + + cos 2Le
U+ =
(2.31)
23
1.5 0
0.5
1.5 x[m]
2.5 x 10
3
5
Figure 2.4: Free surface proles along a sample estuary with length Le = 300km for non convergent ( = 0, left) and convergent channel ( = 7.4, right). = 0.2, ks = 45m1/3 s1 1/6 (ks = gCh Rh ), D0 = 10m , transparent boundary condition.
+ when a reective landward barrier (U (x+ = Le ) = 0) is considered. Please note that water level 2m1 4 Lg ,
and velocity are, respectively, in phase in the former case and out of phase of in the latter, 2 with m natural number. For these channel Le =
3Lg 4 , ...)
the system. The (2.31) is then no more valid, because the denominator is vanishing. When the tide amplitude is greater the non-linear terms play a non negligible role. Two most relevant effects can be highlighted, that dont appear in (2.30) and (2.31). First of all the tidal amplitude is damped due to frictional effects. Moreover during its propagation the tidal wave doesnt keep a regularly sinusoidal as higher order harmonic components are generated. The latter effect is related to the shallow water character of the system (2.6a-2.6b): the wave celerity is an increasing function of the ow depth. In convergent estuaries tidal waves strongly modify their shape, amplify and tend to form shock waves, which propagate quite rapidly in the landward direction. As a result the wave prole is deformed as indicated in gure 2.4. Results of numerical simulations suggest that the frictional non-linear term plays an important role, greater than that of the advective term. In order to point out the role of the frictional term a comparison between the results obtained using different closures for the dissipation term is shown in gure 2.4, where longitudinal proles of the free surface are plotted, both for non-convergent (left) and convergent ( = 7.4, right) estuaries. The linear solution is obtained using a linearized formulation for the frictional term j, which can be written, using the rst order term of the Lorenz (1922) transformation, in the following form: j= U 8 , U 3 k2 R4/3
s h
(2.32)
24
2. On tide propagation in convergent and non-convergent channels 1 gCh Rh 6 and U is a characteristic maximum value of local velocity, evaluated at each
where ks =
cross section, as the algebraic mean of the peak values of ood and ebb phase: U (x) f lood + |U (x)ebb |max
U (x) =
max
It is worth noticing that the solution obtained using (2.32) does not generate shocks. On the contrary, higher order formulations for the frictional term, like that proposed by Dronkers (1964), lead to some results comparable to those obtained using the complete non-linear formulation. When U is a periodic function with zero mean, Dronkers relationship can be expressed using Chebyshev polynomials as follows: j= U U 16 U 2 +2 4/3 15 k2 R U U s h
3
(2.33)
A quantitative comparison of tidal wave characteristics, for different choices of model parameters, can be given in terms of the amplitude of the leading Fourier components of the time series of free surface elevation in each cross section. Moreover the Fourier representation allows to investigate the generation of over-tides (e.g. higher order components), which arise from non-linear effects in the governing equations during tide propagation, though the forcing tide is purely M2 sinusoidal. Over-tides are well represented in gures 2.5 and 2.6 where the amplitudes of the Fourier components are scaled by the amplitude of the semi-diurnal harmonic h1 : An = hn . h1 (2.34)
Results of gure 2.5 are obtained through different closures for the frictional term and for different values of the convergence length. Also the tendency of the tidal wave to break during its propagation is related to non-linear effects driven by the shallow water character of the governing equations, by the frictional term and by channel convergence. At the mouth of the channel (x = 0) over-tides are not present, because the boundary condition is a purely sinusoidal oscillation, while moving inside the estuaries several higher order harmonic component are present. The role of the higher order harmonics may be signicant especially if the friction term is represented in non-linear form, as shown comparing gures 2.5 and 2.6 (right). When a linear closure is adopted for the frictional term gure 2.5 (left) the spectrum is characterised by few harmonic components and the occurrence of mode 2 is mainly related to inertial effects. When a non-linear closure is adopted, the amplitude of higher order harmonics is larger and may induce the wave to break. This is shown in gure 2.5 (right) and 2.6 (right) where the Dronkers approximation
25
0.2
0.15
0.2
0.15 A
0.1
0.1
0.05
0.05
0 0
50
100
150 x[km]
200
250
300
0 0
50
100
150 x[km]
200
250
300
Figure 2.5: Amplitude of the leading order Fourier components of the time series of free surface elevation in each cross section evaluated using (2.32) (left) and (2.33) (right) in a nonconvergent channel = 0. The amplitudes are scaled using the value h1 of mode 1/6 1. = 0.1, D0 = 10m, ks = 45m1/3 s1 (ks = gCh Rh ), and transparent boundary condition.
0.25
0.2
mode 2 mode 3 mode 4 mode 5
0.2
0.15
0.15 A
A 0.1
0.1
0.05
0.05
0 0
50
100
150 x[km]
200
250
300
0 0
50
100
150 x[km]
200
250
300
Figure 2.6: Amplitude of the leading order Fourier components of the time series of free surface elevation in each cross section for a convergent = 7.4 (left) and non-convergent = 0 channels (right). The amplitudes are scaled using the value h1 of mode 1. = 0.1, 1/6 D0 = 10m, ks = 45m1/3 s1 (ks = gCh Rh ), fully non-linear closure and transparent boundary condition.
26
2. On tide propagation in convergent and non-convergent channels (2.33) and the fully nonlinear formulation (2.4) for j have been adopted, respectively. Moreover a non-linear term in the continuity equation can arises from channel convergence; if Lb is small enough the term
UD Lb
spectrum, as shown in gure 2.6 (left). Figure 2.6 allow a comparison between the behavior of the free surface in convergent (left) and non-convergent (right) estuaries.
U d + g U x x h j = x
Z t+T0
t
hd + g
Z t+T0
t
jd = 0
(2.35)
since the advective term U U is almost negligible everywhere (square brackets denote the average x over the tidal cycle). The relationship (2.35) describes the variation of the mean slope of the free surface elevation along the estuary. The tidally averaged frictional term can be written as: j = 1 T0
Z t+T0 Q () |Q ()|
t
2 gB2Ch D3 ()
(2.36)
where the hydraulic radius has been approximated by the ow depth. Using the simplest description for the tidal wave, valid in the frictionless case, Q cos (t) , D 1 + cos (t )
where the discharge is assumed to be sinusoidal due to mass conservation requirement, we obtain side is positive in the range (/2, /2), for any value of . It is important to note that, within that the right hand side of (2.35) vanishes when the phase lag = /2; moreover, the right hand
the framework of this simplied approach, the mean slope of the free surface is related to the variation of the ow depth, which introduces a non-linearity through the frictional term. A comparison between the value < j > obtained from (2.36) and the mean slope of the leading order component of the Fourier representation of the free surface, obtained using the fully non-
27
x 10
2 0
50
100
150 x [km]
200
250
300
Figure 2.7: Tidally averaged value of the free surface slope evaluated through equations 2.35 and 2.36 or from the longitudinal slope h0 /x of the mode 0 of the spectrum of the 1/6 free surface elevation. = 0.1, ks = 45m1/3 s1 (ks = gCh Rh ), D0 = 10m, transparent boundary condition.
linear closure (2.4), is shown in gure 2.7. It is worth noticing that the two curves do not coincide due to the non completely negligible contribution of the advective term U U . A similar comparx ison can be given in terms of the value of mode h0 of the spectrum of the free surface prole along the estuary and the value of the mean free surface elevation obtained through the following equation: h (x) =
Z x
0
j d
(2.37)
Such a comparison is given in gure 2.8 for a convergent estuary ( = 7.4). In this case the contribution of the advective term is even more negligible since the longitudinal scale is larger with respect to the case of non-convergent estuaries, as also pointed out by Lanzoni and Seminara (1998) and Friedrichs and Aubrey (1994); hence, in this case the estimates of the mean water level obtained through the numerical model or using the simplied formulation (2.37) are fairly close.
28
0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 mode 0 Integral of <j> 0.02 0 0.02 0
<H> [m]
0.5
1.5 x[km]
2.5 x 10
3
5
Figure 2.8: Tidally averaged value of the free surface level evaluated through equation 2.37 or from the amplitude h0 of the mode 0 of the spectrum of the free surface elevation. 1/6 = 0.1, ks = 45m1/3 s1 (ks = gCh Rh ), D0 = 10m, transparent boundary condition.
29
D0 D(x)
1/4
(2.38)
The Greens law is based on energy conservation considerations and relies on two unrealistic assumptions, namely that energy dissipation is negligible and convergence is much weaker than the tidal wavelength. In particular, the frictionless character of the relationship does not allow one to describe the wave damping. According to (2.38) any degree of channel convergence or decrease of ow depth should result into an amplication of the tidal wave, which is obviously not true in real estuaries. Jay (1991), Friedrichs and Aubrey (1994) (see also Friedrichs et al. 1998) and Lanzoni and Seminara (1998) have proposed suitable extensions of Greens law that take into account also the role of friction. Their analytical approaches are based on the assumption that the parameter is relatively small. As discussed in the preceding sections the order of magnitude of is strictly related to ; hence such condition corresponds to consider weakly dissipative estuaries. On the other hand, in strongly dissipative cases, tide propagation has to be treated as a strongly non-linear process. The above theories, which more or less implicitly consider marginal conditions for tide amplication, lead to relationships for the marginal state that can be cast in the following form: = km . (2.39)
For weakly convergent and weakly dissipative estuaries Lanzoni and Seminara (1998) and Friedrichs et al. (1998) have found a linear dependence, hence m=1. On the other hand, for the case of strongly convergent channels, Friedrichs and Aubrey (1994) have found that marginal conditions are attained when the actual celerity c of the tidal wave is equal to the frictionless celerity c0 .
30
25
y = 0.8228x R2 = 0.9982
0.6132
20
y = 1.1029x0.5893 2 R = 0.9979
15
10
0 0 50 100
150
200
250
Figure 2.9: Marginal conditions for the amplication of tidal amplitude in the plane, for different values of , as obtained through the numerical model. The interpolating power laws are reported in the plot with the corresponding correlation coefcient R2 .
31
m
0.8 0.6
0.6
0.8
0.4 0.2 0
0.4
0.5
0.6
0.1
0.2
0.3
0.4
0.5
0.6
Figure 2.10: The coefcients m (left) and k (right) of equation 2.39 as function of the amplitude ratio . Adapting their relationship to our notation, we nd 4 = 1/2 . 3 (2.40)
Hence for strongly convergent estuaries the exponent of power law (2.39) reduces to m = 0.5. In the present work the marginal conditions for tide amplication have been determined through numerical experiments and comparing the resulting tidal amplitude at a given section with that imposed at the mouth. An iterative algorithm has been developed in order to obtain, within a given tolerance, the value of the length Lb for which the difference was minimised. Simulations have been performed placing both a reective barrier landward in a very long channel and the transparent condition, obtaining comparable results. More than 200 simulations have been conCh [10 30], paying particular attention to the choice of typical conditions of real estuaries. ducted in a wide range of values of the parameters, namely D0 [2.5m 50m], [0.005 0.6], The marginal conditions obtained numerically are plotted in gure 2.9 in terms of the dimensionless ratio between friction and inertia and the degree of convergence , for different values of the amplitude ratio . Below the numerical points the wave is damped during its propagation landward, while above them it is amplied. It is worth noticing that the larger are the dissipative effects along the estuary (large ), the stronger is the required degree of convergence to achieve the marginal conditions. On the contrary, a relatively weak variation of channel geometry can produce wave amplication in weakly dissipative estuaries. From gure 2.9 it appears that numerical points corresponding to a given value of can be tted fairly well through power law curves of the form (2.39), whose coefcients m and k only depend on the parameter . The above dependence embodies the effect of the nite amplitude of
32
3
y = 0.9504x R2 = 0.999
0.9131
y = 0.9629x0.858 R2 = 0.9979
=0.015
1
=0.010 =0.005
Figure 2.11: Marginal conditions for the amplication of tidal amplitude in the plane, for different values of , as obtained through the numerical model. The interpolating power laws are reported in the plot with the corresponding correlation coefcient R2 .
the tidal wave on its amplication. The numerical ndings are summarised in gure 2.10 where m and k are plotted as a function of : it is shown that the exponent m depends strongly on nite amplitude effects such that its value decreases sharply as increases and reaches an almost constant value, nearly equal to 0.6, for relatively large values of . Numerical results for the weakly convergent and weakly dissipative case are represented in more detail in gure 2.11. As pointed out before, analytical results suggest that in this case a linear relationship should hold, with m=1. Numerical results conform to this behaviour only for very small values of . For commonly observed values of the tidal range within this class of estuaries the power law is non-linear. This is even clearer when considering that, as shown in gure 2.10, the tendency of m towards unity is almost vertical. Notice that the linear solution would be satisfactory in the case of an horizontal asymptotic trend of the curve towards m=1. Also notice that the landward boundary condition becomes more and more important in the numerical model as we approach the frictionless limit, hence when we consider very small values of and . In the case of strong convergence a signicant inuence of the tidal amplitude on wave characteristics can be expected along with a substantial deviation from the linear behaviour. In this case
33
10
y = 0.9884x R2 = 0.9959
0.1 0.1 1
10
100
Figure 2.12: Marginal conditions in term of the modied dimensionless parameter dened in equation 2.41 The interpolating power law is reported with the corresponding correlation coefcient R2 . the approximate analytical solution (see equation 2.40) leads to a value of the exponent m =0.5 those which are typically encountered in real estuaries according to the data reported in Table 1. range (0.6-1.2) obtained through numerical simulations and reported in gure 2.10. It is worth noticing that if we introduce a modied parameter (, ) = k()m() , (2.41) Notice, however, that the coefcient k 2.31 of the relationship (2.40) does not fall within the which is fairly close to the numerical result (m 0.6) within a wide range of values of , namely
using the values of the coefcients k and m given in gure 2.10, the numerical points which dene the marginal conditions for tide amplication collapse on a single logarithmic plot as shown in gure 2.12.
34
investigated. As pointed out in the preceding chapter, channel convergence may strongly affect the hydrodynamics of estuaries (see also Friedrichs and Aubrey, 1994 and Friedrichs et al., 1998). Its role on the morphological evolution of tidal channels, which was neglected in previous morphological analysis, (e.g. Schuttelaars and de Swart, 2000), has been recently highlighted by Lanzoni and Seminara (2002). In their work a classication of tidal channels is proposed, based on the relative role of the different mechanisms than control tidal hydrodynamics, as discussed in Chapter 2. However, the analysis of Lanzoni and Seminara (2002) is mainly oriented to characterise the behaviour of the solution in the asymptotic limits dened by the dominance of one of these effects; therefore, it is restricted to a limited number of situations and does not allow one to fully
35
3. Large scale equilibrium proles in convergent estuaries recognize the role of the different parameters involved. The intermediate conditions in between the asymptotical behaviours, are the most common in natural systems. In the present work we employ a one-dimensional model, starting from initially horizontal bed and neglecting the effect of intertidal areas. We will show how under these conditions the tidal channels are typically ood dominated in a large part of their length and the net ux of sediment is mainly directed landward. Due to the velocity asymmetry, a net upstream sediment ux is present for example in the Ord River estuary (Australia) (Wright et al., 1973), in the Salmon River estuary (Canada) (Dalrymple et al., 1990), and in the Severn estuary (UK) (Murray and Hawkins, 1977). Results indicate that, when river discharge is not dominant and a reective boundary is assigned at the landward end, a sediment front forms and slowly migrates landward until it leads to the emergence of a beach, which generally inhibits a further channel development. Hence, an inclined equilibrium bed prole is established, which is bounded by a beach at the landward end. This condition determines an asymptotic intrinsic length of the estuary. It is worth noticing that both in the present work and in previous analysis (Schuttelaars and de Swart, 2000; Lanzoni and Seminara, 2002) the resulting equilibrium bed prole shows an increasing bottom elevation in the landward direction. The above scenario is also conrmed by the experimental observations of Bolla Pittaluga et al. (2001). Also notice that the ow eld over the equilibrium prole is characterised by symmetrical ood and ebb phases; therefore, the net sediment ux, averaged over a tidal cycle vanishes. In present analysis we investigate in detail the dependence of such equilibrium length on the relevant physical parameters which characterise the tide and the channel geometry. Furthermore, we analyse the role of the seaward boundary condition for sediment transport on channel development. Two different scenarios can be identied as the role of channel convergence increases. 1. in weakly convergent estuaries the equilibrium length coincides with the initial length of the channel, which is dened as the distance from the mouth to the landward boundary condition, where a reective barrier is imposed; 2. when channel convergence is strong the landward beach may form within an internal section of the channel, provided that the landward boundary is located sufciently far from the mouth. In this case the equilibrium length is shorter and mainly depends on the convergence length and on the mean channel depth. In the former case the nal length of the channel coincides with the initial imposed length, while in the latter case the equilibrium length is controlled by channel convergence. Another relevant result is that the equilibrium prole doesnt depend appreciably on the type of boundary condition
36
3. Large scale equilibrium proles in convergent estuaries for the sediment ux imposed at the mouth. In the perspective of the present results the analysis of Lanzoni and Seminara (2002) seems specially suitable to describe the case of relatively short estuaries.
where p is the sediment porosity and qs is the total solid discharge per unit width. In the present analysis Engelund and Hansen formula is adopted which accounts for both bed and suspended load; in dimensional form it reads: qs = where denotes the Shields stress =
2 U 2 /Ch = , (s ) gDs gDs 2 gD3 0.05Ch 5/2 s
(3.2)
(3.3)
which is computed in terms of the local values of the bed shear stress or of the ow velocity U = Q/. In (3.3 and 3.2) is water density, s and Ds are sediment density and diameter, respectively, and Ch is the dimensionless Chezy coefcient, = (s ) / and g is the gravity acceleration. Substituting from (3.1) into (2.3b) the set of governing equations can be cast in the following form: Q + t x Q2 + g h Q |Q| + = 0, 2 x Ch Rh (3.4a)
qs 1 h 1 Q + + =0 t B x (1 p) b x (1 p) 1 bqs + =0 t B x
(3.4b) (3.4c)
The introduction of the continuity equation for the sediments doesnt change the mathematical nature of the system, in fact it is easy to show that system (3.4) is hyperbolic, like system (2.3); the number of eigenvalue in this case is three. The numerical scheme employed for the resolution
37
3. Large scale equilibrium proles in convergent estuaries is similar to that discussed in Chapter 2 and consists in the MacCormak TVD algorithm in the form proposed by Garcia-Navarro et al. (1992).
38
39
3. Large scale equilibrium proles in convergent estuaries In order to consider wetting and drying areas the numerical algorithm has been modied in the following form. In the drying phase, when the free surface elevation is decreasing, the maximum ow discharge in every cell is evaluated through geometrical considerations. Let consider a situation like that represented in gure 3.1: in this case the value of the predicted free surface elevation, and then of ow discharge, is not compatible with the position of the bed prole. of water (volumes 1 and 2 for the cell i and i 1 , respectively) stored between two consecHence, the maximum ow discharge in the cell i and i 1 is evaluated in term of the volume
utive sections, divided by the time step of the computation. A similar procedure is adopted
for the adjacent cells. If the numerical algorithm evaluates a ow discharge smaller than the maximum value previously dened, the numerical result is kept as the correct value: under gure 3.1. On the other hand, when the numerical scheme evaluates a discharge greater than the maximum discharge, the latter value is taken as the right one and the free surface elevation is imposed equal to the bottom elevation (sections i and i 1 in gure 3.1). In the wetting phase the cells is reintroduced in the computational domain when the free surface elevation, which is assumed horizontal beyond the last active cell (see gure 3.2), is larger than the bed elevation. The ow discharge in the new submerged cell is evaluated through geometrical considerations. We may notice that this approximated procedure requires relatively small values of the Courant number (nearly equal to 0.5) in order to preserve the stability of the scheme. Also notice that the above procedure allows one to satisfy the continuity equation within the domain subject to wetting and drying, while the momentum equation is not solved therein. this condition the depth doesnt vanishes, as in the case of sections from i 5 to i 2 in
40
(1 p)U0 L0 gD3 s
(3.5)
where U0 and L0 are suitable scales for the velocity and the length of the estuary, respectively 100km, D = 0.1 0.2mm) we obtain that Tbed ranges between 10 500 years! s (see discussion in Chapter 2). For reasonable choice of parameters (U0 1 2ms1 , L0 10 According to the results of numerical simulations the morphological evolution of a tidal channel can be described as follows. Starting from an initial horizontal bed prole a sediment front is formed where the divergence of the sediment ux is larger and negative. The front migrates landward and amplies until it reaches the last section, where it may be reected. For values of the relevant parameters which are typical of real estuaries, after a fairly long period of time, say of the order of hundreds of years, the system tends to an equilibrium conguration, which is characterised by a bottom prole displaying an upward concavity, as also found by Lanzoni and Seminara (2002). As shown in gure 3.3, the morphological evolution of the channel may follow two different behaviours, depending upon the role of the physical constraint which is posed by the nite length of the estuary (notice that in both cases the channel is convergent). Results reported in gure 3.3a correspond to a relatively short estuary. In this case the asymptotic conguration is characterised by the formation of a beach at the landward end, such that the nal length coincides with the imposed length of the estuary Le . This is the case treated by Lanzoni and Seminara (2002). In longer channels (gure 3.3b) the sediment front, which migrates landward, may stop at a certain distance from the mouth because a beach is formed inside the channel. This condition generally prevents the further development of the channel. In this case the equilibrium bottom prole establishes within a fraction of the total length of the estuary. Hence, for given values of the relevant parameters an equilibrium length of the estuary La can be dened, which is achieved provided that the landward boundary condition is located sufciently far from the mouth. Notice that results presented herein refer to relatively large values of and hence of solid discharge qs , which implies a much faster evolution of the system. In gure 3.4 the effect of the seaward boundary condition for the sediment transport is investigated. Bed proles, evaluated through the condition of vanishing inow of sediment at the mouth of the estuary, are plotted with dashed lines, while continuous lines denote results obtained with sediment inux from the sea equal to the transport capacity computed according to the local and instantaneous ow conditions at the mouth. We may notice that the former condition implies a larger scour at the mouth, whose effect on the equilibrium solution depends on the initial length of the estuary: in short channels, where the equilibrium bottom prole extends over the whole estuary, a larger scour at the mouth leads to a slight increase of the equilibrium slope; in long channels, where the asymptotic length La can be achieved, the slope is roughly constant and the
41
0.5
(a)
0
-0.5
-1
-1.5
after 2 years after 10 years after 50 years after 100 years equilibrium
-2
x+
0.6
0.8
0.5
(b)
0
-0.5
-1 after 2 years after 10 years after 20 years after 50 years after 100 years after 130 years equilibrium 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1.5
-2
-2.5
x+
Figure 3.3: The long term evolution of the bottom prole of a convergent estuary for different values of channel length. Lb = 120km, D0 = 10m, a0 = 4m, Ch = 20, Ds = 101 mm; Le = 160km (a), Le = 480km (b). The longitudinal coordinate x+ is scaled with the length Le , the bottom elevation + is scaled with the reference depth D0 .
42
1 0.5 0 -0.5
1
(a)
0.5 0 -0.5
(b)
+ -1
-1.5 -2 -2.5 -3 -3.5 0 0.2 0.4 x + 0.6 0.8 1
+ -1
-1.5 -2 -2.5 -3 -3.5
0 0.2 0.4 x+ 0.6 0.8 1
Figure 3.4: The equilibrium bottom proles in short (a) and long (b) channels with different boundary conditions at the mouth of the tidal channel: vanishing sediment ux (dashed lines) and equilibrium sediment ux (solid lines). D0 = 10m, a0 = 4m, Ch = 20, Ds = 101 mm; Le = 160km, Lb = 120km (a); Le = 280km, Le = 40km (b). The longitudinal coordinate x+ is scaled with the length Le , the bottom elevation + is scaled with the reference depth D0 .
equilibrium beach slightly migrates landward. It is worth noticing that the longitudinal location of the sediment front in the channel depends strongly on the degree of convergence, inasmuch as the hydrodynamic behaviour of the system is inuenced by this factor. Also notice that the non-symmetric (ood-dominated) character of many estuaries is driven toward a more symmetrical conguration when the bottom proles increases its slope (Friedrichs and Aubrey, 1994; Toffolon, 2002). Recalling that sediment transport mainly depends on ow velocity through an exponent larger than 1, it is possible to relate the residual sediment transport to the peak values of velocity during the ebb and ood phases; for this purpose it is convenient to dene the degree of asymmetry as follows : = log U f lood max . |Uebb max | (3.6)
The inuence of the degree of convergence on sediment transport, and hence on the morphological evolution, is shown in gure 3.5, where characteristic values of both velocity U and sediment ux qs are plotted at the initial stage of evolution (horizontal bed). We note that the asymmetry exhibited by the peak values of velocity, which is related to the residual sediment ux, may induce a different behaviour depending upon the degree of convergence of the estuary. In weakly convergent estuaries, the asymmetry decreases almost monotonically in the landward direction and the sediment front starts to form close to the landward boundary. On the other hand, when the degree of convergence increases, a maximum of velocity and residual transport occurs inside the channel, which determines there a larger sediment ux gradient; as a consequence the sediment
43
1.5 1 0.5 U+ 0
(a1)
150
(b1)
100 50 qs
+
0
maximum minimum residual
-50
-0.5 -1 0 0.2 0.4 x+ 0.6 0.8 1
0.8
1
(a2)
150 100 50
(b2)
0.5 0
qs 0
maximum minimum residual asymmetry
0.8
(a3)
(b3)
0.2
0.4 x +
0.6
0.8
0.2
0.4 x +
0.6
0.8
Figure 3.5: Degree of asymmetry, maximum, minimum and residual values of ow velocity (a), normalised with the average of its peaks at the mouth, and of sediment ux (b), normalised with gD3 , along the estuary, the longitudinal coordinate x+ is scaled with s Le . Le = 160km, D0 = 10m, a0 = 4m, Ch = 20, ds = 101 mm; different values of convergence length: Lb , (a1 , b1 ), Lb = 160km, (a2 , b2 ), Lb = 10km, (a3 , b3 ).
44
3. Large scale equilibrium proles in convergent estuaries front develops in an internal section of the channel. It is worth noticing that in the numerical simulations we impose an impermeable barrier at the landward end of the estuary; hence, the equilibrium conguration requires that the net sediment ux vanishes everywhere. In fact, when the tidally averaged asymptotic condition is achieved and the mean bed level keeps constant, the sediment continuity equation (3.1), written in terms of the residual sediment transport recalling also the (2.5), reads: qs qs + =0 x Lb (3.7)
where the square brackets denote tidal average. With the imposed boundary condition at the landward end, equation (3.7) only admits of the solution qs = 0. In gure 3.6 we plot the maximum, minimum and residual values of ow velocity U as a function of the landward coordinate, as obtained at the beginning of the simulation (when the bed is horizontal) and at equilibrium, for different boundary conditions at the mouth of the estuary. We note that at equilibrium the peak value of velocity keeps almost constant along the whole estuary, as it is commonly observed in real estuaries. A similar plot is given in gure 3.7 in terms of the sediment ux. It is interesting to note that the equilibrium conditions asymptotically reached by the system are dynamical, since they are achieved only in terms of net transport. Results reported in gure 3.7 also clarify the inuence of the seaward boundary condition for sediment transport. When sediment inow is set equal to zero, sediment transport must vanish at the mouth in order to reach a stable conguration. This implies a larger scour at the entrance and hence a smaller velocity. On the contrary, when sediment inow balances the equilibrium transport capacity associated with the local hydrodynamic conditions, the sediment transport can maintain a non-vanishing peak value even at equilibrium. Results of numerical simulations concerning the equilibrium length of the estuary are summarised in gure 3.8 in dimensional form, in terms of the initial length Le , i.e. the distance from the mouth of the landward boundary, and the convergence length Lb . The nal length La is dened as the distance between the mouth and the beach, at equilibrium. As pointed out before, when the channel is not sufciently long, that is for small values of Le , the equilibrium length La coincides with the physical dimension imposed to the system: the corresponding points fall on the bisector line of the graph. On the contrary, when the channel is long enough the system can reach an equilibrium length which depends on the degree of convergence of the channel: the stronger is channel convergence (i.e. small values of Lb ), the shorter is the equilibrium length La . Numerical results also suggest that a transition zone occurs close to the bisector line, in which La still depends on the initial physical dimension, beyond which an asymptotic value of the equilibrium length is reached, for a given value of Lb . Hence, we can dene this asymptotic length as the convergence-induced equilibrium length Lc . Notice that the width of the transition zone increases as the convergence
45
(a)
U+
x + 0.5
0.6
0.7
0.8
0.9
(b)
1.5 1 0.5
U+ 0
-0.5 -1 -1.5 -2 0 0.1 0.2 0.3 0.4
x+
0.5
0.6
0.7
0.8
0.9
Figure 3.6: Maximum, minimum and residual values of ow velocity U + , along the estuary after one cycle (dashed line) and at equilibrium (continuous line), for different boundary conditions at the seaward and: (a) vanishing sediment ux; (b) equilibrium sediment ux. Le = 40km, Lb = 20km, D0 = 5m, a0 = 2m, Ch = 20, ds = 101 mm; the velocity U + is normalized with the maximum value at the mouth at equilibrium and the longitudinal coordinate x+ is scaled with Le .
46
80
(a)
60
40
20
+ 0
qs
-20
-40
x + 0.5
0.6
0.7
0.8
0.9
80
(b)
60
40
20
+ qs 0
-20
-40
x+
0.5
0.6
0.7
0.8
0.9
Figure 3.7: Maximum, minimum and residual sediment ux q+ scaled with gD3 along the s s estuary after one tidal cycle (dashed line) and at equilibrium (continuous line), for different boundary conditions at the seaward end: (a) vanishing sediment ux; (b) equilibrium sediment ux. Data as in gure 3.6. The longitudinal coordinate x+ is scaled with Le .
47
Lb
L a [km]
300 250 200 150 100 50 0 0 100 200 300 400 500 600 700
800
900
1000
L e [km]
Figure 3.8: Equilibrium length of the estuary La as a function of the initial length Le , for different values of convergence length. D0 = 10m, a0 = 4m, Ch = 20, ds = 101 mm. length increases. The convergence-induced equilibrium length also depends on the initial depth D0 . In fact, when the initial depth is large, a longer reach is required to let the beach to emerge. However, the above dependence can be ruled out once the equilibrium length Lc is scaled with the initial depth D0 , as shown in gure 3.9. Also notice that in weakly convergent channels or in constant width channels the beach does not always emerge. However, a suitable equilibrium length can be dened also in this case in terms of the distance from the mouth of the leading edge of bottom prole (the corresponding points are denoted by x in gure 3.8).
3.5 Discussion
In this chapter we have investigated the long-term morphological evolution of estuarine channels through a relatively simple one-dimensional numerical model. The nal equilibrium prole is the natural response of the system to the external forcing effects due to tide propagation and channel convergence. After a long period of time, say of the order of the centuries, an equilibrium conguration is achieved, which displays the presence of a beach at the end of the channel or inside the estuary, depending upon the length and the degree of convergence. The channel convergence can be considered as a forcing term, as in the uvial case, where
48
45 40
35 30
D 0 = 20 m D 0 = 10 m D0 = 5 m
Lc/ D0
20 15 10 5 0 0 0.1 0.2
D 0 /L b
0.3
0.4
Figure 3.9: Dimensionless equilibrium length Lc /D0 as a function of the dimensionless degree of convergence D0 /Lb , for different values of D0 . = 0.4, Ch = 20, ds = 101 mm. width variations can induce the development of bed forms. In the case of tidal channels, the convergence, which exerts an important effect on the hydrodynamics (see Chapter 2), also plays a signicant role on bed evolution, increasing the ability of the tidal ow to form deposits inside the channel. In the case of a short channel the behaviour of the bottom prole is the following: a ux of sediment takes place inside the channel, because in tide dominated estuaries the hydrodynamics is generally ood dominated, which induces a landward transport of sediment. At the landward end of the channel a reective boundary condition is imposed; hence, the ow discharge and the sediment ux vanish. Due to the decrease of sediment discharge close to the landward end, its gradient may attain high values and a sediment front is formed. Then, the front tends to migrate landward, until the nal section is reached and a beach at the end of the estuary establishes. In long convergent channels the beach forms within an internal section, because as the length of the channel increases the maximum gradient of solid discharge moves away from the landward end. From gure 3.3b we may note that close to the channel end a smaller secondary front is still present due to the effect of the vanishing ux condition. The equilibrium conguration has been dened as the condition in which the tidally averaged sediment ux vanishes or, alternatively, the bottom elevation attains a constant value. The dynamic equilibrium achieved by the channel is characterised by negligible residual values of the sediment transport and almost constant values of the velocity along the estuary. As described above, the
25
0.5
0.6
49
3. Large scale equilibrium proles in convergent estuaries nal equilibrium length of the channel is found to depend mainly on two parameters, namely the physical length of the channel and the degree of convergence. In particular, for relatively short and weakly convergent estuaries, a beach is formed at the landward end of the channel and the equilibrium length coincides with the initial length of the channel, that is the distance of the landward boundary from the mouth. When the initial length of the channel exceeds a threshold value, which decreases for increasing values of the degree of convergence of the channel, the beach forms within an internal section of the estuary and the nal equilibrium length is much shorter and mainly governed by the convergence length. Present results also suggest that imposing the condition of vanishing sediment input ux from the outer sea during the ood phase determines a larger scour in the seaward part of the estuary, though the overall longitudinal bottom prole does not differ much from which one obtained with the alternative boundary condition of sediment inux equal to the equilibrium transport capacity of the channelized ow at the mouth. Numerical results are in fairly good agreement with the experimental observations of Bolla Pittaluga et al. (2001) and Bolla Pittaluga (2002), which refer to a laboratory ume, 30 cm wide and 24 m long, with constant width. In fact, the experimental runs lead to equilibrium proles which are similar to those predicted by the present model, displaying a landward slope, an upward concavity and a beach at the landward end. Furthermore, the maximum scour at the mouth is nearly equal to the mean ow depth. Finally, it is worth noticing that several factors have been neglected, which may play an important role in the morphological evolution of tidal channels. In particular, the input of fresh water and sediment discharge at the landward boundary is not considered. One may argue that the above effect is likely to counteract the formation of emerged areas and, consequently, it may affect both the length and the structure of the nal bottom conguration. However, the slope of the equilibrium prole seems to be mainly determined by the hydrodynamic behaviour of the estuary and by the tendency of the system to minimise the asymmetry between the ood and the ebb phases; hence, we expect that within the seaward part of the estuary results are not likely to change, at least qualitatively, also in the presence of river discharge.
50
As pointed out in Chapter 1, the morphodynamic equilibrium of tidal channels, as well as that of rivers, can be investigated at different scales. In this chapter we focus the attention to the meso-scale; more specically we investigate the instability process which leads to the formation of free bars in systems characterised by ne sediments, like tidal channels or the lower reaches of alluvial rivers. The mechanisms of erosion and deposition associated with meso-scale feature like estuarine bars often determine complex pathways of sediment transport, such that some channels may be dominated by ebb transport and other by ood transport. A "local" analysis is performed, which implies that we restrict our attention to a conveniently short reach of the channel, whose length is of the order of few tens of channel widths. To characterise adequately the suspended sediment transport a three dimensional model for the ow and the concentration eld is adopted, coupled with an evolution equation for the bed topography. Further discussion on the suitability of the three dimensional approach with respect to the much simpler two-dimensional model is also contained in Chapter 7.
51
Figure 4.1: Sketch of the channel and notation. the channel half width B to scale the planimetric coordinates x and y ; a reference ow depth D to scale the vertical coordinate z , the local free surface elevation 0 h , bed elevation and ow depth D ;
a reference ow velocity U0 to scale the velocity components. Furthermore, a characteristic time scale T0 = B /U0 is introduced. Hence, we write:
(x, y) = (z, , h, D) =
(x , y ) B
(4.1a)
(z , , h , D ) D 0 (u, v, w) U0
(u, v, w) = t=
t T0
The reference values U0 and D can be readily dened in the river case as the depth averaged 0
velocity and depth of the uniform ow, for given discharge, bed slope and sediment size. In dimensionless form such reference state is completely determined once the following dimensionless parameters are given: = B , D 0 (4.2a)
52
Ds =
(4.2b)
0 =
2 U0 C f 0 , gD s
(4.2c)
Rp =
gD3 s ,
(4.2d)
where D is sediment diameter, is the relative density of sediment, is the cinematic viscosity s of fresh water, g is gravitational acceleration and C f 0 is the friction coefcient of the reference uniform ow. For tidal channels the reference state can not be identied simply as in the river case. In particular, as discussed in Chapter 2 the velocity scale in a tidal channel is not an external variable and depends on the relative importance of the main ingredients, which control the propagation of a tidal wave, namely inertia, friction and degree of convergence of the channel. Furthermore, it also depends on the landward boundary condition. The reference state can be dened using for example the velocity scale proposed by Toffolon (2002) and a tidally averaged value for the ow depth. In the simplied approach introduced by Seminara and Tubino (2001), which mainly apply to tidal channels like the main channels of Venice lagoon, a reference state is dened assuming a sinusoidal oscillation for the ow velocity and keeping constant depth during the tidal cycle. The ow eld and the concentration eld in a tidal channel are investigated here within the framework of a three dimensional model where the standard shallow water approximation is adopted. The hydrostatic approximation is justied provided the channels is wide enough and the tidal wavelength largely exceeds the ow depth, as generally occurs in tidal systems. Notice that in strongly convergent estuaries, like for example the Araguari river or the Gironde estuary, the tidal wave can break during its propagation and form shock wave, which cannot be studied within the context of the shallow water approximation. However, this phenomenon affects only locally the morphodynamic behaviour, since it is conned within a short length of the channel and lasts for a relatively short time. In dimensionless form, and keeping only the signicant turbulent uxes, the Reynolds equations, the ow continuity equation and the transport equation for the suspended sediment read: u,t +uu,x +vu,y +wu,z (T u,z ) ,z + h,x iF = 0, Fo2 F02 h,y = 0, Fo2 (4.3a)
(4.3b) (4.3c)
53
4. Local-scale model for tidal channels C,t + (uC) ,x + (vC) ,y + [(w Ws )C] ,z + (T C,z ) ,z = 0, (4.3d)
where (u, v, w) denote the three-dimensional velocity eld, C is the volumetric concentration of
suspended particles, = C f 0 , Ws is particle fall velocity, scaled with U0 , and F0 is the Froude
number of the reference ow. Furthermore, the subscripts (,t), (, x) ,(, y) ,(, z) denote partial derivative with respect to t, x, y and z, respectively, and iF is the mean bed slope in the longitudinal direction. Closure assumptions for the eddy viscosity T and diffusivity T are introduced assuming that the slowing varying character of the ow eld both in space and in time leads to a sequence of equilibrium states. Therefore we employ a self-similar structure for T and T , written in term of a boundary tted coordinate: = z . D (4.4)
Furthermore, the scales of the eddy viscosity T and diffusivity T are given in term of the local and instantaneous values of ow depth and shear velocity. Hence, we write: T = T = T = u DN () , C f 0U0 D 0 T = u DP () , C f 0U0 D 0 (4.5)
(4.6)
where and are the dimensional values of eddy viscosity and diffusivity and u is the shear T T
C f 0U0
u =
1 C f 0U0
| | ,
(4.7)
is the dimensional bed shear stress and N() and P() are the vertical distributions of T and T at equilibrium, for which the relationships proposed by Dean (1974) and McTigue (1981) are employed. They read: N () = P () = Kv (1 ) , 1 + 2A2 + 3B3 if if < 0.314 0.314 (4.8)
0.35 0.11
(4.9)
where A = 1.84, B = 1.56 and Kv = 0.4 is the von Karman constant. Sediments in uvial and tidal channels are transported as bed load and suspended load. Usually the former mechanism is assumed to be conned in a thin layer close to the bed (the bed layer) and it is evaluated as a direct function of the bed shear stress. Suspended load occurs in the
54
4. Local-scale model for tidal channels upper part of the water layer (i.e. over the reference level sketched in gure 4.2), where value of sediment concentration depends on advective and diffusive effects. The main distinctive feature of suspended load, with respect to bed load, is the non-instantaneous response to ow variations, since it requires a relatively large adaptation length to achieve local equilibrium with changing hydraulic conditions. The advection-diffusion equation (4.3d), in which sediments are considered as passive tracers except for their tendency to settle, is the differential equation that reproduces this delay. The net sediment exchange between the two layers is due to turbulent and settling effects, as shown in gure 4.2. The suitability of this approach requires small values of sediment concentration and Richardson number, and a particle grain size much smaller than turbulent length scales. Close to the bottom, where sediment concentration attains larger values due to settling effect, these condition are only approximately satised. The formulation of the problem is completed through the introduction of the continuity equation for the sediments which describes the development of bottom topography; considering the bed load uxes and also the net sediment ux through the reference level, the mass balance for sediments reads: (1 p) ,t +Q0 (Qbx,x + Qby,y ) + + kWs (C Ce ) n
=a
= 0,
where p is sediment porosity, (Qbx , Qby ) is bed-load vector, scaled with vector in the z-direction, and n= aD,x ,x ; aD,y ,y ; 1
2
(4.11)
2
is the unit vector in the direction normal to the bed. Furthermore, a is the conventional dimensionless level, scaled with the ow depth, where the bed boundary condition for the evaluation of the concentration eld is imposed. Finally the dimensionless parameter Q0 is given in this form: Q0 =
gDs / (U0 D ) . 0
3
(4.12)
The last term in (4.10) accounts for the net ux of sediment exchanged between the bed layer and the water column (suspended load), which is computed as the difference between the actual value of local concentration at the reference level and the value Ce that concentration eld may attain at equilibrium with the local and instantaneous ow conditions (see below the discussion of the boundary conditions).
55
u = v = 0,
(4.13)
with z0 dimensionless conventional reference level where vanishing of ow velocity is imposed (see 5.14); cinematic condition and vanishing stress at the free surface ( = 1): w uh,x vh,y h,t = 0, v,z = 0, cinematic condition at the bed ( = z0 ): w u,x v,y ,t = 0; (4.16) u,z = 0; (4.14) (4.15)
Further boundary conditions are required for the concentration eld at the free surface and at the bed. The rst condition is that of vanishing sediment ux at the free surface ( = 1): WsCk + C f 0 T C ns = 0, (4.17)
56
4. Local-scale model for tidal channels with ns unit vector normal to the free surface: ns = h,x ; h,y ; 1 . (4.18)
1 + (h,x )2 + (h,y )2
At the bed two different choices for the boundary condition have been discussed in the literature (see, e.g., Van Rijn (1985)). The simplest way is to impose the so-called "concentration boundary condition", which implies that at the reference level a the concentration attains the reference value Ce , which is evaluated using the local and instantaneous values of the ow characteristics. The second procedure is known as "gradient boundary condition" and prescribes a given sediment ux at the reference level a. This condition is more suitable under non uniform conditions, as pointed out by Parker (1978), since it doesnt force the concentrations eld to attain its equilibrium value near the bed and leads to a smooth adaptation of the concentration led to changing ow conditions. Note that in the case of uniform ow and concentration eld (Rouse solution) the two conditions are equivalent. In the present work the "gradient boundary condition" is adopted, where the entrainment condition for the net ux of sediment at the reference level ( = a) is related to the difference between the local value of the concentration at the reference level and the equilibrium value Ce : C f 0 T C n u Ws k nCe = 0, sediment : v = Qby = 0 (y = 1). (4.20) (4.19)
Finally we impose the sidewalls of the channel to be impermeable both to the ow and to the
4.2.1
Channel geometry is reported in gure 4.1. Besides the boundary conditions at the sidewalls, at the free surface and at the bed, suitable boundary conditions must be imposed in the longitudinal direction. The three dimensional formulation is able to reproduce the typical morphological structures at the meso-scale. The most important bed forms observed at this scale are bars, which display alternate sequences of scours and deposits. However, the development of regular trains of free bars requires a relatively long straight channel reach, say of the order of several bar wavelengths. In a numerical solution, like that discussed in the following chapters, this would imply the use of a fairly long computational domain, resulting in large computational time. However, when periodic boundary conditions are adopted the longitudinal extension of the domain can be set equal to bar wavelength. The application of periodic boundary conditions is straightforward; for the generic variable,
57
4. Local-scale model for tidal channels denoted by , periodicity can be written as: (x = Ld ,t, y, z) = (x = 0,t, y, z) (4.21)
having denoted by Ld the dimensionless longitudinal length of the domain, scaled by the channel width. In the uvial case a given bed slope is set as the forcing term of the ow, while in tidal context the forcing term due to gravity, which is represented by term i f in (4.3a), is assumed to be a sinusoidal function of time, which reproduces an oscillating bed shear stress, as employed by Seminara and Tubino (2001).
where is the local bed stress vector evaluated in the form described in paragraph 5.1. Gravitational effects on bed load transport are taken into account, like in many other contributions (see for example Tubino et al., 1999), through a simplied semi-empirical approach which is formally justied only for small values of local bed slopes (Ikeda, 1982b; Kovacs and Parker, 1994). With reference to gure 4.1 we write: Qbx = |Qb | [x cos () y sin ()] || |Qb | [x sin () + y cos ()] || r tan () = , n Finally bed load intensity |Qb | is assumed to be a function of cr , with cr = cr + r1 s 1 (/n)2 , tan2 () (4.25) (4.23a)
Qby =
(4.23b)
(4.24)
where s is directed as the local bed stress vector, while n is orthogonal to the above direction.
58
4. Local-scale model for tidal channels where is the angle of repose of the bed material and cr is the standard critical value of Shields stress for vanishing slope. According to experimental observations the empirical constants r and r1 fall in the range (0.3 0.6) and (0.1 0.2), respectively. Calculation are performed employing the Van Rijn (1985) relationships for the reference level a and the reference concentration Ce . They read:
3
Ds T 2 Ce = 0.015 a R0.2 p T= cr cr
(4.26a)
(4.26b) (4.26c)
a = max (0.01, es )
where es is the equivalent dimensionless roughness which accounts for the effect of small scale bed forms, like dunes or ripples, which are often superimposed on meso-scale topography when the sediment is ne, and is the effective Shields stress acting on the bed. Its denition arises from the stress-partition procedure which is typically adopted to model resistance effects induced by small scale bed-forms (see Einstein, 1950), which assumes that the total bed stress can be viewed as the sum of friction (the effective shear stress) and form drag. The local values of the effective shear stress acting on the bed is computed, in dimensionless form, that is in term of the Shields stress , using the empirical formulation proposed by Engelund and Fredsoe (1982), namely = 0.06 + 0.31.5 , where is computed using the total shear stress. The above approach is also used to compute the friction coefcient C f in the following form: C2 = f 6 + 2.5 ln 2.5Ds (4.28) (4.27)
Implicit in this procedure is the assumption that small scale bed-forms respond instantaneously to changing ow conditions, as suggested, for instance, by eld observations in the Severn estuary where mega-ripples crests can be rebuilt within a tidal cycle, resulting in amplitude changes by as much as 2 metres (Harris and Collins, 1985). Furthermore, the intensity of bed load transport |Qb | is evaluated using Parkers formula: |Qb | = 0.00218 2 G ()
3
(4.29a)
59
G=
if if
1 < 1.65 1
1.5878
(4.29b)
(4.29c)
Finally a t of the experimental curve reported by Parker (1978) is used to estimate the dimensional particle fall velocity, in the following form: Ws = where A = log10 (Rp).
2 3 4 gD 10(1.181+0.966A0.1804A +0.003746A +0.0008782A ) s
(4.30)
60
In the present chapter we describe the nite difference numerical model which is adopted for the solution of the ow and concentration eld. In order to keep the required detail to reproduce correctly suspended sediment transport and its effect on bed morphology the problem is tackled within the context of a three-dimensional framework, as discussed in the preceding chapter. Many numerical schemes have been developed so far for the solution of the ow eld using a 3D approach, while only few contributions are available in the literature in which a 3D approach is used for the solution of the concentration eld. In the last years the increasing interest for the solution of environmental problems, related to estuarine and coastal sediment dynamics and to the transport of heavy metals and toxic waste through their adsorption on sediment particles, has motivated the development of more rened models, like those proposed by Lin and Falconer (1996), in which a semi analytical scheme is proposed for the simulation of the concentration eld in the employed to investigate the over-deepening process which occurs when two straight channels are joined thought a 180 degree bend. In the present model we follow the numerical procedure proposed by Casulli and Cattani (1994), according to which the non-linear terms appearing in the momentum equations are discretized using a lagrangian approach. Furthermore, the concentration eld is solved using an original semi-analytical scheme. The resulting numerical scheme seems sufciently accurate and efcient to represent adequately the ow and concentration elds of uvial and tidal systems and their role in the development of bed topography. Humber estuary, and by Wu et al. (2000), where a k approach for the turbulence closure is
61
Figure 5.1: Computational domain in natural scale (left) and in logarithmic deformed scale (right).
suspended particles. The 3D approach adopted herein allows for the direct evaluation of the bed shear stress, which takes the following simplied form when the shallow water approximation is introduced: = (x , y ) = T v u , T z z (5.1)
z
2 where is the dimensionless shear stress, scaled with C f 0U0 , and z is a suitable dimension-
less level for the evaluation of the bed shear stress, which can be assumed to be of the same order of magnitude of the grain size Ds . Analogously, the evaluation of turbulent uxes in the transport equation for suspended particles requires the numerical estimate of vertical gradients of velocity "near the bed". Hence, using the natural vertical boundary tted coordinate , as dened in (4.4), the relatively small vertical spacing needed to reproduce correctly the solution near the bed would involve an exceedingly large number of computational points, in particular near the free surface: there, the velocity gradient attains smaller values and a sparser numerical grid is sufcient to obtain the required accuracy. Hence, in the present model a suitable logarithmic vertical coordinate is introduced in the following form: = ln () = ln z D (5.2)
through which a grid is obtained whose density decreases from the bed to the free surface (see gure 5.1). The above transformation leads to a vertical domain falling within the range [ln (z0 ) , 0] . The logarithmic vertical coordinate also minimizes the numerical truncation errors. Let us consider the simple case of a uniform ow. In this case the vertical velocity prole, which is logarithmic in the variable , becomes a linear function when the variable is adopted. The trun-
62
5. A three dimensional numerical model for suspended sediment transport cation error of the numerical scheme, which is rst order accurate in the vertical direction, as will be pointed out in the following sections, is proportional to the second derivative of velocity with respect to in some internal point between the grid points; hence, the error becomes vanishingly small when the logarithmic coordinate is used. It is worth noticing that in tidal ows, where conditions are typically non uniform, the vertical velocity proles often display for most of the tidal cycle a self similar structure, that is a vertical logarithmic structure with the mean velocity varying in time due to the tidal oscillation. Hence, also in this case the use of the variable may provide a better numerical approximation. As for the computational efciency, the number of grid points which are found to be necessary for a good accuracy of the numerical simulation when using the natural vertical coordinate , for logarithmic coordinate the number of grid point to achieve the same accuracy reduces by one order of magnitude. reasonable choices of ow parameters, is about 200 300 in the vertical direction, while using the
5.2 Equations
In order to solve equations (4.3a), (4.3b) and (4.3c) we rst introduce the transformation into the new coordinate system, in which we use the vertical coordinate dened in (5.2). The coordinate system is non-orthogonal and boundary tted; the computational domain is transformed into a nice rectangular box. The following transformation rules are introduced: e rt t t D e rx x x D e ry y y D e z D where the coefcients rt , rx , and ry are, respectively: rt = rx = D z D + = + e t D t t t z D D + = + e x D x x x (5.4a) (5.3a)
(5.4b)
63
ry =
z D D + = + e y D y y y
(5.4c)
In order to write the governing equations into the new system of coordinate it is convenient to introduce the contravariant vertical component of the velocity. Following Pielke (1994) we dene the contravariant vertical component of the velocity as: w = w rt urx vry (5.5)
It is worth noticing that w represents the vertical velocity in terms of base vectors that are tangent to the surface along which the coordinate is constant, as represented in gure 5.2. Using the
t contravariant vertical component of velocity, the vertical Courant number cz = w , where
is the vertical spacing and t the time step, keeps generally very small in the computation. In particular, the Courant number cz vanishes at the bed and at the free surface, because bed and free
surface are steram lines and coincide with the surfaces = ln (z0 ) and = 0, respectively. In the new coordinate system, and adopting (5.5), equations (4.3) become: u,t +uu,x +vu,y + e e wu, D D e e wv, D D e T u, D e T v, D +
,
(5.6a)
+
,
(5.6b)
u,x +v,y +
e (w + rt ) , = 0, D
(5.7)
64
5. A three dimensional numerical model for suspended sediment transport and the boundary conditions read:
u = v = 0,
(5.8)
The kinematic conditions at the bed and at the free surface are automatically satised provided w = 0 there. Finally, the advection diffusion equation for the concentration of suspended sediment (4.3d) becomes: (CD) ,t + (DuC) ,x + (DvC) ,y +e [(w Ws )C] , e e T C, D , = 0, (5.10)
The boundary conditions for the above equation are discussed in section 5.3.3.
(squares in gures 5.4 and 5.6), the y-component at point i, j + 1 , k (crosses in gure 5.4) and 2
1 the z-component at point i, j, k + 2 (diamonds in gure 5.6). The free surface elevation and the
bed elevation are dened at integer i and j (circles in gure 5.4); the sediment concentration is dened at the centre of each computational cell (stars in gure 5.6). The equations (5.6) and (5.7), with boundary conditions (5.8) and (5.9), are then solved through a semi-implicit nite difference scheme. For the solution of the ow eld the Eulerian Lagrangian approach is adopted, in the no slip condition at the bed. form proposed by Casulli and Cattani (1994), which has been suitably modied to account for the
65
66
5.3.1
The numerical method used here is semi-implicit, in particular the non-linear terms of equations (5.6) are solved explicitly using a Lagrangian approach, while the pressure term and the viscous term are solved implicitly. The rst step of the procedure consists in the evaluation of the total derivative
n+1 ui+ 1 , j, k un 1 a, jb, kd e du i+ 2 2 = u,t +uu,x +vu,y + , wu, dt D t
(5.11)
where a, b, c dene the starting point of the trajectory of the considered water particle, as shown in gure 5.5, and are computed explicitly: a= b= d=
Z Z
un dt vn dt
e n w dt D
The trajectories are evaluated using a three-linear interpolation algorithm; therefore, the numerical scheme is rst order accurate in the spatial coordinates x, y, . It is worth noticing that if the time step t is relatively small, that is the Courant number is less that one, the Lagrangian scheme reduces to an up-wind difference scheme, while for relatively large values of the time step, that is for Courant numbers larger that one, the Lagrangian approach keeps stable, while the numerical estimate worsen. The trajectories are integrated using a time substep t smaller than the time step t adopted in the numerical solution, in order to cross at each substep at most one cell border (see
67
5. A three dimensional numerical model for suspended sediment transport gure 5.5). The remaining terms of the momentum equations (5.6) are discretized using central differences; in order to obtain a good accuracy in time the pressure term is evaluated at time step n + , the weight falling within the range that the more tends to
1 2, 1 2,1
accurate in time but unfortunately it is unstable. The momentum equations in discrete form read:
n+1 ui+ 1 , j,k un 1 a, jb, kc i+
2
n+ n+ hi+1, j hi, j
F02 x
iF + F02
ek Dn 1 , j i+
2
1 e k+ 2 n 1 , j,k+ 1 T,i+ 2 2
1 e k 2 n 1 , j,k 1 T,i+ 2 2
(5.13a)
F02 y
ek Dn j+ 1 i,
2
1 e k+ 2 n j+ 1 ,k+ 1 T,i, 2 2
1 e k 2 n j+ 1 ,k 1 T,i, 2 2
(5.13b)
The eddy viscosity coefcients and the ow depth are evaluated explicitly at time step n such that the resulting algebraic system is linear. In equations (5.13), for k = 1 the values of ow velocity components at the bed appear, namely ui+ 1 , j,0 and vi, j+ 1 ,0 , which are set equal to zero, according
1 to the no-slip boundary condition at the bed. At the free surface, k = nz + 2 , the vanishing stress
2 2
condition is imposed u, = v, = 0. In the vertical direction the ow domain is divided into nz + 1 2 parts, as shown in gure 5.6. Below the level k =
1 2
the bed is introduced to impose the no-slip boundary condition. The level k = 0 correspond to the conventional level 0 for the velocity, which is evaluated using the reference state: z0 = e
0.777+KvC f 02
1
,
1
(5.14) . (5.15)
This approach allows one to take into account also the equivalent bed roughness which is used to simulate the presence of ripples or dunes, using a suitable value of C f 0 . Notice that the variations of the reference level 0 due to changing ow conditions are assumed to be negligible in the computation.
68
Figure 5.6: Computational grid: vertical spacing. Near the bed the half cell allows to impose the no-slip condition, at the free surface the whole cell allows to impose the vanishing stress condition.
The generic term appearing in the equations, denoted by in the following, is evaluated at the time step n + using the following linear combination: n+ = (1 ) n + n+1 . Equations (5.13a) and (5.13b) can be rewritten in vectorial form:
n+1 An 1 , j Ui+ 1 , j = Gn 1 , j i+ i+
2 2 2
(5.16)
F02 x F02 y
(5.17a)
n+1 An j+ 1 Vi, j+ 1 = Gn j+ 1 i, i,
2 2 2
(5.17b)
n+1 n+1 where the vectors Ui+ 1 , j and Vi, j+ 1 are respectively:
2 2
z 2 un+1 1 i+ 2 , j,nz 1 n+1 Ui+ 1 , j = .... 2 n+1 ui+ 1 , j,2 2 n+1 ui+ 1 , j,1 2
n+1 Vi, j+ 1
2
n+1 vi, j+ 1 ,n
(5.18)
69
z z 2 2 p s di di+ 1 , j,n 1 di+ 1 , j,n 1 1 i+ 2 , j,nz 1 z z 2 2 An 1 , j = ... ... ... i+ 2 p i s di+ 1 , j,2 di+ 1 , j,2 di+ 1 , j,2 2 2 2 p i di+ 1 , j,1 di+ 1 , j,1 2 2
p di+ 1 , j,n
s di+ 1 , j,n
p di+ 1 , j,n z 2
Dn 1 , j enz i+ 2
tn 1 , j,n 1 T,i+
2 z
(5.19)
Dn 1 , j e i+
2
1 z 2
(5.20a)
p di+ 1 , j,k =
2
k = nz 1...1
(5.20b)
s di+ 1 , j,k =
2
k = nz ...2
(5.20c)
i di+ 1 , j,k =
2
k = nz 1...1
(5.20d)
An j+ 1 i,
2
2 z 2 z p s di di, j+ 1 ,n 1 di, j+ 1 ,n 1 1 i, j+ 2 ,nz 1 2 z 2 z = ... ... ... p s i di, j+ 1 ,2 di, j+ 1 ,2 di, j+ 1 ,2 2 2 2 p i di, j+ 1 ,1 di, j+ 1 ,1 2 2
p di, j+ 1 ,n
s di, j+ 1 ,n
p di, j+ 1 ,n = Dn j+ 1 enz i,
2 z 2
tn j+ 1 ,n 1 T,i, 2 z 2 + n 1 n Di, j+ 1 e
2 z
(5.21)
(5.22a)
p di, j+ 1 ,k =
2
tn j+ 1 ,k 1 T,i, 2 2 + k 1 n Di, j+ 1 e
2 2
k = nz 1...1
(5.22b)
s di, j+ 1 ,k =
2
tn j+ 1 ,k 1 T,i, 2 2 k 1 n Di, j+ 1 e
2 2
k = nz ...2
(5.22c)
70
5. A three dimensional numerical model for suspended sediment transport tn j+ 1 ,k+ 1 T,i,
1 Dn j+ 1 e k+ 2 i, 2
2 2
i di, j+ 1 ,k 2
, k = nz 1...1
(5.22d)
1 F02 x
hn j hn j + iF Dn 1 , j t enz i, i+1, i+
2 2
1 i+ 2 a, jb, nz 1c
1 F 2 x hn j hn j + iF Dn 1 , j t enz 1 i, i+1, i+
0
...
un un
1 i+ 2 a, jb, 2c
t t vn
1 F 2 x hn j hn j + iF Dn 1 , j t e2 i, i+1, i+
0 2
1 i+ 2 a, jb, 1c
1 F 2 x hn j hn j + iF Dn 1 , j t e1 i, i+1, i+
0 2
Gn j+ 1 i,
2
, ,
(5.23)
1 ia, j+ 2 b, nz c
vn ia, j+ 1 b, nz 1c 2 t
1 F02 y
hn j+1 hn j i, i,
Dn j+ 1 t enz i, 2 Dn j+ 1 t enz 1 i,
2
1 F 2 y hn j+1 hn j i, i,
0
...
vn
1 ia, j+ 2 b, 2c
(5.24)
t t
vn
1 F 2 y hn j+1 hn j i, i,
0
Dn j+ 1 t e2 i,
2
ia, j+ 1 b,1c 2
1 F02 y
hn j+1 hn j i, i, 1
Dn j+ 1 t e1 i, 2
The ow continuity equation (4.3c) is integrated over the ow depth in the following form:
Z h
nz 1 e E = ... 2 e e1
Z h
enz
(5.25)
+z0
u,x dz +
+z0
(5.26)
Using the Leibnitz rule and recalling the kinematic conditions at the bed (4.16) and at the free surface (4.15) we then obtain: (h ) + t x
Z h
+z0
udz +
Z h
+z0
vdz = 0.
(5.27)
71
5. A three dimensional numerical model for suspended sediment transport Finally, introducing the transformation (5.2) we obtain: (h ) + D t x
Z 0
ue d +
ln z0
D y
Z 0
ve d = 0.
(5.28)
ln z0
The numerical algorithm adopted in the present work is decoupled, namely the ow eld and the bottom topography are evaluated independently. Therefore, as a rst step the ow eld is evaluated, keeping a constant bed level; then the new bed topography is updated with reference to the computed ow eld. According to this procedure equation (5.28), in discrete form, becomes:
n+1 hi, j hn j i,
1 x
nz
nz
nz
k=1
k=1
k
1 y
Dn j+ 1 i,
2
k=1
n+ vi, j+ 1 ,k ek Dn j 1 i,
2 2
k=1
n+ vi, j ,k e
1 2
nz
= 0.
(5.29)
n+1 n+1 Using the vectorial notation, recalling the relationship (5.16) and substituting Ui+ 1 ,j and Vi,j+ 1
from equation (5.17) the continuity equation can be given the following discrete form:
n+1 hi, j hn j i,
2 t 2 F02 x2
2
Dn 1 , j i+
2
ET A1 E +
1 i+ 2 , j
1 i+ 2 , j
Dn 1 , j i
2 2
ET A1 E +
i 1 , j 2
t Dn 1 , j ET A1 Gn i+ 2 x
Dn 1 , j ET A1 Gn i 2 t 2 F02 y2
2
1 i 2 , j
(1 ) t Dn 1 , j Un 1 , j E Dn 1 , j Un 1 , j E i+ 2 i 2 i 2 i+ 2 x ET A1 E
1 i, j 2 1 i, j+ 2
Dn j+ 1 i,
2
1 i, j+ 2
Dn j 1 i,
2
ET A1 E +
t Dn j+ 1 ET A1 Gn i, 2 y
Dn j 1 ET A1 Gn i,
1 i, j 2
(1 ) t Dn j+ 1 Vn j+ 1 E Dn j 1 Vn j 1 E = 0, i, 2 i, 2 i, 2 i, 2 y (5.30)
n+1 where the unknowns are the values of the local free surface elevation hi, j and the coefcients are
where represents the scalar product. (5.30) represents a linear system of nx x ny equations,
known and depend only on the ow eld and ow depth at time step n. The system is symmetric and is solved using a conjugate gradient algorithm. We may note that the continuity equation is discretized in conservative form. Once the free surface elevation is known, equations (5.17a-5.17b) are used in order to evaln+1 n+1 uate the ow velocities components ui+ 1 , j,k and vi, j+ 1 ,k . The vertical components wi, j,k+ 1 are
2 2 2
72
The above equation is applied to each water column, setting at the bed the no slip condition,
n+1 n+1 wi, j, 1 = 0, and computing the values of wi, j,k+ 1 from the bed (k = 1) to the free surface (k = nz ).
2 2
5.3.2
The numerical computation requires estimations of local and instantaneous values of bed shear stress, which are used to compute the eddy viscosity, the bed load vector and the reference concentration which is needed in the boundary condition for the concentrations eld at the bed. Recalling the denition of the shear velocity u = 1 C f 0U0 | | ,
the shear velocity is evaluated, using (5.1), through the following expression: 1 Cf0 e D u
2
u =
C f 0 T
u =
e N () Cf0
where N () is the function proposed by Dean (1974) and reported in (4.8) and = ln (z ), where z is set equal to the equivalent roughness. Once the shear velocity is known, it is possible to evaluate the eddy viscosity in the following form: T = u DN () .
5.3.3
Due to the effect of settling velocity Ws , the vertical advection term in equation (5.10), In particular the Courant number is fairly large close to the bed, where the vertical grid spacing is small and the term e attains high values. Under these conditions the numerical scheme doesnt perform satisfactorily, as typically occurs when advection is dominant. The overcome the above e (w Ws )C1 , , leads to relatively high values of the vertical Courant number cz = e
t.
73
5. A three dimensional numerical model for suspended sediment transport difculty a semi-analytical procedure for the solution of the advection-diffusion equation for the suspended sediment is introduced, in which the concentration eld is splitted into two parts: C = C0 +C1 , where C0 is the solution of the following ordinary differential problem: WsC0 + e C f 0 |U| P ()C0, = 0, D = ln (a) , (5.33a) (5.33b) (5.32)
C0 = Ce ,
and represents a Rouse-type vertical prole of concentration. More specically, C0 is the distribution that the concentration eld would attain at equilibrium within a uniform ow characterised by the local and instantaneous ow conditions. Hence C0 represents the contribution of the concentration eld in phase with the local shear stress; assuming for the eddy diffusivity the relationship (4.9) and recalling that = ln , C0 becomes: C0 , ds , R p , a = Ce , ds , R p , a f (, G1 , G2 , a) (5.34)
where the function Ce ( , ds , R p , a) represents the reference concentration evaluated in terms of the local values hydrodynamic parameters. Furthermore the function f can be given the form:
a e G1
f (, G1 , G2 , a) =
if exp e 0.314
G2
G1 a 0.314
(5.35)
if
G2 =
(5.36b)
74
5. A three dimensional numerical model for suspended sediment transport Substituting equations (5.32) and (5.33a) into (5.10) we obtain the following equation for the component C1 : (C1 D) ,t + (DuC1 ) ,x + (DvC1 ) ,y +e [(w Ws )C1 ] , + e e T C1 , D , = (5.37)
It is worth noticing that, through the right hand side of (5.37), the component C0 produces a forcing term for C1 which can be computed analytically evaluated through (5.34). The boundary conditions associated with (5.37) are: the entrainment condition for the net ux of sediment at the reference level ( = ln (a)): C f 0 T C1 n + C f 0 T hC0 n u Ws k nCe = 0, (5.38)
where is the differential gradient operator, which is dened using (5.3b), (5.3c) and (5.3d), and h is the horizontal gradient operator: h = e e rx , ry x D y D ; (5.39)
The component C1 , which quanties the spatial delay of sediment concentration with respect to the bed shear stress which is due to the advection and diffusion processes, is determined numerically through a mixed algorithm. In the horizontal directions we adopt the Lax scheme, in the form proposed by Leveque (1996); this numerical scheme is explicit, conservative and second order accurate in space; moreover it takes into account the diagonal uxes with respect to the axis and cell orientation. hence, the method is stable within a wide range of Courant number: ut 1, x vt 1. cy = y cx =
(5.41)
In the vertical direction an explicit scheme cannot be used because the vertical advection term
75
5. A three dimensional numerical model for suspended sediment transport e (w Ws )C1 , leads to very high values of the vertical Courant number, as pointed out be-
fore. Hence, an implicit scheme is required. In the present work, due to the relatively high contrialgorithms is adopted in order to guarantee the stability, according to the following scheme (central) + (1 ) (up wind) , where the weighting coefcient is locally prescribed in order to warrant the stability of the
bution of the vertical advective term, a linear combination of up wind and central di f f erence
scheme, though reducing the accuracy of the computation, according to the following relationship (Casulli and Greenspan, 1988): = max 0, min 1, 2e P D |w Ws | . (5.42)
The above procedure allows one to include the up wind correction only where it is strictly necessary.
Using an explicit scheme for the horizontal uxes, equation (5.37) reads: (C1 D) ,t e
1 n+ 2
e T C1 , D
n+
n+ +e [(w Ws )C1 ] , +
= (DuC1 ) ,n (DvC1 ) ,n x y
(5.43)
Vertical terms in equation (5.43) are solved using the implicit scheme; hence we obtain: ek n+1 n+1 n+ n+1 n+ n+1 Di, j C1,i, j,k + wi, j,k+ 1 Ws C1UP,i, j,k wi, j,k 1 Ws C1DOW N,i, j,k t 2 2 Cf0 1 1 n+1 n+1 n+1 n+1 T,i, j,k+ 1 e k+ 2 C1,i, j,k+1 C1,i, j,k T,i, j,k 1 e k 2 C1,i, j,k C1,i, j,k1 - n 2 2 Di, j =
ek n n n+ n n+ n Di, jC1,i, j,k (1 ) wi, j,k+ 1 Ws C1UP,i, j,k wi, j,k 1 Ws C1DOW N,i, j,k + t 2 2 C f 0 (1 ) 1 1 n n n n T,i, j,k+ 1 e k+ 2 C1,i, j,k+1 C1,i, j,k T,i, j,k 1 e k 2 C1,i, j,k C1,i, j,k1 + n 2 2 Di, j (5.44)
+explicit terms
where C1UP,i, j,k and C1DOW N,i, j,k are evaluated using the mixed upwind and central di f f erencing
76
5. A three dimensional numerical model for suspended sediment transport algorithm: C1UP,i, j,k = C1,i, j,k+1 +C1,i, j,k + (1 ) 2 C1,i, j,k+1 C1,i, j,k
n+ if wi, j,k+ 1 Ws < 0
2
(5.45a)
(5.45b)
When = 1 the proposed scheme reduces to a central difference, while the condition = 0 implies a rst-order upwind differencing. Results of computational tests suggest that in our case is always very close to 1. Smaller values of can be expected for large values of sediment diameter, which leads to high value of the settling velocity Ws , or for low values of ow velocity, which may be the case of tidal ows during ow reversal. Notice that the decomposition (5.32) guarantees a good computational accuracy since the dominant component C0 is determined through an analytical procedure. Using the explicit scheme for the horizontal uxes and the implicit procedure for the vertical ux leads to the solution of simple three-diagonal algebraic systems, one for each water column. The solution for each system is obtained using the standard LU decomposition algorithm. The above numerical procedure, which is summarised by (5.44) and (5.45), has been tested under different conditions, as discussed in the next sections. Notice that the standard test presented by many authors, which checks the ability of the model to reproduce the equilibrium concentration proles in uniform ow, is here unnecessary because the proposed decomposition (5.32) automatically satises the above condition. In fact, in the case of uniform ow the contribution of C1 vanishes and the concentration prole C=C0 coincides with that evaluated analytically without any numerical approximation.
5.3.4
Exner equation
The continuity equation for the sediment (4.10) requires the denition of bed load components Qbx and Qby . They are evaluated in terms of the local values of Shields stress using (4.29a); furthermore, the effect of gravity is accounted for through the relationships (4.23a) and (4.23b). The continuity equation is solved using a nite difference algorithm
n+1 i, j n j i,
(1 p)
+ Q0
n+ n+ Qbx,i+ 1 , j Qbx,i 1 , j
2
n+ n+ Qby,i, j+ 1 Qby,i, j 1
2
n+ + kWsC1,i, j,na n = 0,
(5.46)
77
5. A three dimensional numerical model for suspended sediment transport where na represent the index at the reference level a where the bed boundary condition for
the concentration is imposed. The quantities at time step n + are evaluated using the relationship (5.16):
n+1 i, j n j i,
(1 p)
(1 )
+ Q0
Qn 1 , j Qn 1 , j bx,i+ bx,i
2
Qn j+ 1 Qn j 1 by,i, by,i,
2
+ +, (5.47)
n+1 n+1 n+1 Since the values of Qbx and Qby depend on the local bed elevation i, j , which is still unknown
5.3.5
Numerical procedure
The numerical procedure adopted in the solution for the ow eld, the concentration eld and the bed topography is the following: starting from the nth time step the trajectories and the velocity component un 1 a, jb, kd and i+ vn j+ 1 b, kd are evaluated; ia,
2 2 2 2
and
are computed ;
once the ow eld is completely determined, the shear velocity and the eddy-viscosity coefcients are evaluated; the analytical contribution to the concentration eld (5.34) is determined; the explicit uxes for the concentration equation are evaluated; equation (5.44) is solved for each i, j; bed load components Qbx and Qby are determined;
78
5. A three dimensional numerical model for suspended sediment transport nally, exner equation (5.47) is solved; All the quantities Di+ 1 , j , Di, j+ 1 are evaluated using linear interpolation.
2 2
boundary conditions are imposed. Hence, the solution for each dependent variable can be given the standard Fourier representation: f (x, y, ,t) = Fkm (,t) exp i kx + m y 2 nx ny where =
2 Ld /B
nx ny
(5.48)
The st numerical test is designed to check the ability of the numerical model to reproduce the vertical velocity prole under uniform ow condition, namely the standard logarithmic prole suitably corrected to account for the wake effect. Adopting the relationship (4.8) for the eddy viscosity the analytical solution reads: uanalytical () = Cf0 ln Kv z0 + A2 + B3 (5.49)
79
0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Figure 5.7: Vertical velocity prole under uniform ow condition: analytical solution (red solid line) and numerical solution (black dots). 0 = 1, R p = 4, Ds = 105 , = 0.1, nz = 50.
10
10
10
E1
10
E m%
10
1
10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Figure 5.8: Numerical error in the estimate of velocity prole under uniform ow condition: norm E1 (left), mean value (right). 0 = 1, R p = 4, Ds = 105 .
80
10
10
10
10
E m%
10
1
E1
10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Figure 5.9: Numerical error in the estimate of velocity prole under uniform ow condition: norm E1 (left), mean value (right). 0 = 1, R p = 10, Ds = 2 105 . The numerical model reproduces the vertical analytical velocity prole as shown in gure 5.7, where the solution is plotted versus the natural vertical coordinate . It is worth noticing that the use of a logarithmic vertical coordinate implies a large number of numerical points close to the bed, which allows for a satisfactory reproduction of the velocity prole where the gradient is large. A quantitative estimate of the numerical error is obtained in terms of the norm E1 =
k=1
|unumerical uanalytical | ,
nz
and the mean value of the truncation error Em % = 100 nz (unumerical uanalytical ) . nz k=1
Comparisons are made for different choice of ow parameters: the results are reported in gure 5.8, for = 1, R p = 4, Ds = 105 , and in gure 5.9, for = 1, R p = 10, Ds = 2 105 , where E1 (left) and Em % (right) are reported as functions of the vertical grid spacing. We should note that the scheme is rst order accurate; hence, the error is sufciently small for values of less than 0.2, that is nz ranging between 30-40 depending on ow parameters.
5.5.2
The numerical model is then tested under non-uniform ow conditions. In this case only few solutions are available, because the complexity and the non-linear character of the momentum
81
U1
136
134
132
130
128
126
Figure 5.10: Comparison of present numerical results with the analytical solution of Tubino et al. (1999): vertical proles of the amplitude U1 (left) and phase lag (right) of the perturbation of longitudinal velocity. nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green) and the analytical solution (black). The phase lag is measured with respect to the peak of bed elevation, the amplitude U1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 102 . equations do not allow to derive analytical solutions except for particular simplied conditions. The validation of the numerical model, under non-uniform ow conditions, is made through the comparison of numerical results with the analytical solution of Tubino et al. (1999), which is valid for straight channels with a perturbed bottom whose amplitude is small with respect to the water depth.
5.5.2.1 The analytical solution of Tubino et al. (1999) In this contribution the Authors consider an innitely long straight channel with xed banks and a varying bottom, subject to steady boundary conditions. The problem is cast in three dimensional form, in terms of longitudinal, transverse and vertical coordinates x, y and z , as described in Chapter 4; the solution for the ow and the concentration eld is obtained analytically, using a perturbative approach which is based on the assumption that the ratio between the amplitude of bed perturbation and the mean ow depth is a small parameter. Under these conditions, when the bed is perturbed using a regular function like, for example a Fourier mode 11, all the dependent variables (ow velocity, free surface elevation, bed shear stress, concentration, etc.) display a perturbation with respect to the uniform basic state which exhibits the same regular structure, with different amplitudes and phase lags with respect to the bottom prole. A phase lag = 0 corresponds to
82
C1
88
87
86
85
Figure 5.11: Comparison of numerical results with the analytical solution of Tubino et al. (1999): vertical proles of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration. nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green) and the analytical solution (black). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 102 .
a perturbation in phase with the bottom prole, that is the peak value of the variable is located at bar crest, while = 90 corresponds to a perturbation whose maximum value is located where the bottom prole crosses the average bed level, i.e. where bed perturbation vanishes. In gures 5.10, 5.11, 5.12 and 5.13 a comparison is pursued between the results of numerical computations and theoretical ndings of Tubino et al. (1999). The comparison is given in terms of the amplitude (left) and phase lag (right) with respect to the peak of bottom prole of the perturbations of velocity and of suspended sediment concentration with respect to the uniform basic state. The gure 5.14 shows that a satisfactory agreement with the analytical solution is achieved both at the free surface ( = 1) and close to the bed ( = 0.01). It is worth noticing that the accuracy of the numerical solution doesnt change even for relatively large values of bar wavenumber (see gure 5.14), whence numerical diffusion is strongly inhibited by the numerical procedure adopted herein. Furthermore the numerical model correctly reproduces the shift of the peak of sediment concentration from negative to positive values of as increases. The shift is larger at the free surface where advective terms are stronger.
83
U1
136
134
132
130
Figure 5.12: Comparison of numerical results with the analytical solution of Tubino et al. (1999): vertical proles of the amplitude U1 (left) and phase lag (right) of the perturbation of the longitudinal velocity. nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green) and the analytical solution (black). The phase lag is measured with respect to the peak of bottom prole, the amplitude U1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 10, Ds = 2 105 , = 0.1, = 15, ny = 32, nz = 100, A = 102 .
1 0.9 0.8 0.7 0.6
1 0.9 0.8 0.7 0.6
C1
105
100
Figure 5.13: Comparison of numerical results with the analytical solution of Tubino et al. (1999): vertical structure of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration. ny = 16 (cyan), ny = 32 (red), ny = 64 (blue), ny = 128 (green) and the analytical solution (black). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 10, Ds = 2 105 , = 0.1, = 15, nx = 32, nz = 100, A = 102 .
84
Figure 5.14: Comparison of numerical results with the analytical solution of Tubino et al. (1999): the phase and the amplitude of the perturbations of longitudinal (u , Au ) and transverse (v , Av ) components of velocity and of suspended sediment concentration (c , Ac ) with respect to the wave-number . The phase lag is measured with respect to the peak of bottom prole, the amplitude A is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, = 20, Ds = 105 , R p = 4, nz = 100, A = 102 .
85
0.5 0.4 0.3 0.2 0.1 0 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
U1
142
140
138
136
Figure 5.15: Dependence on the longitudinal grid spacing of the numerical results for the vertical proles of the amplitude U1 (left) and phase lag (right) of the perturbation of the longitudinal velocity component: nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green). The phase lag is measured with respect to the peak of bottom prole, the amplitude U1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 0.5.
C1
98
96
94
92
Figure 5.16: Dependence on the longitudinal grid spacing of the numerical results for the vertical proles of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration: nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 0.5.
86
0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5
U1
142
140
138
Figure 5.17: Dependence on the longitudinal grid spacing of numerical results for the vertical proles of the amplitude U1 (left) and phase lag (right) of the longitudinal component of the velocity: nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green). The phase lag is measured with respect to the peak of bottom prole, the amplitude U1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 10, Ds = 2 105 , = 0.1, = 15, ny = 32, nz = 100, A = 0.5. 5.5.2.2 Results under non-linear conditions The comparisons presented in section 5.5.2.1 are made between the numerical solution and the analytical linear solution, which is valid provided the amplitude of bed perturbation is small with respect to the mean ow depth. In this case the numerical solution smoothly converges to the analytical solution when the grid spacing is sufciently small. A check of the model behaviour under non-linear conditions is made using a regular bed perturbation with an amplitude equal to half ow depth. In this case the linear analytical solution is no longer valid; hence in gures 5.15, 5.16, 5.17 and 5.18, the test is made in terms of the numerical results obtained with different longitudinal grid spacing. 5.5.2.3 Comparison between the numerical scheme for the concentrations eld with and without splitting The semi-analytical scheme adopted in the present numerical model for the concentration eld is based on the splitting of the concentration into two parts C0 and C1 ; the former is evaluated analytically, while the latter is computed numerically using the procedure described in section 5.3.3. One may wonder whether the splitting procedure does indeed improve the results with respect to the direct numerical solution of equation 5.10. The comparison between the two approaches is made solving equation 5.10 with the same nu-
87
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
C1
0 125
120
115
110
105
100
Figure 5.18: Dependence on longitudinal grid spacing of numerical results for the vertical proles of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration: nx = 16 (cyan), nx = 32 (red), nx = 64 (blue), nx = 128 (green). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . R p = 10, Ds = 2 105 , = 0.1, = 15, ny = 32, nz = 100, A = 0.5.
merical scheme adopted for the splitting-method, suitable modied to take into account a different bed boundary condition, since when splitting is not introduced the whole sediment ux exchanged between the bed and the water column has to be considered. The rst numerical test is made in order to verify that the numerical scheme is able to reproduce the equilibrium Rouse prole (5.35) under uniform ow and sediment transport conditions; the result is reported in gure 5.19 for = 1, R p = 4, Ds = 105 , nz = 100. A comparison between the two approaches under non-uniform conditions is given in gure 5.20 where the numerical results are compared with the analytical solution of Tubino et al. (1999). We may notice that the amplitute and the perturbation of the concentration eld is quite well reproduced by both models, while in terms of the phase lag the solution obtained using the splitting procedure adopted herein seems more accurate, in particular in the upper part of the water column. A possible explanation of the above behaviour is the following. Using equation (5.37) the bed boundary condition is implicitly embodied in the right hand side, which is a forcing term for the unknown C1 , and is acting over the whole depth. On the contrary, using equation (5.10) the bed boundary condition is imposed at the reference level a, which corresponds to the numerical node na , and its effect over the depth is affected by the propagation of numerical approximations. The better performance of the splitting-method can be appreciated from the results reported in gure
88
Figure 5.19: Vertical concentration prole under uniform ow and suspended load conditions evaluated without the splitting procedure: analytical solution (red solid line) and numerical solution (black dots). 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, nz = 100.
1 0.9 0.8 0.7 0.6
1 0.9 0.8 0.7 0.6
C1
88
87
86
85
Figure 5.20: Comparison of the numerical results obtained with the splitting procedure or the direct solution of the equation for the concentration eld: vertical proles of the amplitude C1 (left) and phase lag (right) of the perturbation of the concentration. Analytical solution (black), splitting method (red) and direct solution method (cyan). The phase lag is measured with respect to the peak of bottom prole, the amplitude C1 is scaled with the dimensionless amplitude of bottom prole A . 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, nx = 64, ny = 32, nz = 100, A = 102 .
89
128
64
32
nx
16
Figure 5.21: Comparison of the numerical results obtained with the splitting procedure or the direct solution of the equation for the concentration eld: difference between the values of the phase lag predicted with the numerical solution with splitting (red dots) and without splitting (black dots) and those computed with the analytical solution. 0 = 1, R p = 4, Ds = 105 , = 0.1, = 15, ny = 32, nz = 100, A = 102 . 5.21, where the difference between the analytical solution and numerical solutions =
nz 1 |numerical analytical | , nz na k=na
is reported as a function of the longitudinal grid spacing. Finally, it is worth noticing that the splitting-method better reproduces the typical delay of suspended load with respect to the local bottom shear stress, which affects crucially the stability of meso-scale bed forms as will be pointed out in the following chapter.
90
In this chapter the formation of free bars in rivers and tidal channels is investigated using the 3D model presented in Chapter 5. Bars induce a fairly regular sequence of scours and deposits along the channel. The spontaneous development of these bed-forms in almost straight reaches seems to require only the ability of the ow, when perturbed by a non uniform bottom prole, to enhance the altimetric form through selective processes of erosion and deposition. With reference to a regular bed wave this implies that ow and sediment transport must reach their maximum just upstream the crests so that deposition occurs at the crests and immediately downstream, leading to bed-forms whose amplitude increases in time while migrating in the downstream direction. The increase of the amplitude of bar fronts is mainly counteracted by the downward pull of gravity on sediment particles along transverse slopes. The latter effect becomes weaker as the width to depth ratio of the channel increases, as transverse slopes are gentler when the transverse scale largely exceeds the vertical scale, like in natural channels. As a result a threshold value c exists below which free bars are not expected (and observed) to form. The above scenario mainly applies to almost straight channels whose width is nearly constant and not exceedingly large, with well sorted sediments mainly transported as bed load. Under these circumstances various theoretical works (e.g. Blondeaux and Seminara, 1985; Fredsoe, 1978; Colombini et al., 1987; Struiksma and Crosato, 1989; Shimizu and Itakura, 1989) and a large number of experimental observations in laboratory umes (e.g. Kinoshita, 1961; Chang et al., 1971; Ikeda, 1982a; Jaeggi, 1984; Fujita and Muramoto, 1985; Garcia and Nio, 1993) suggest the following picture. The instability process which leads to bar formation is not strongly size-selective in the longitudinal direction, since different longitudinal modes within the unstable range are characterised by almost similar growth rate; on the contrary the transverse mode selected by the instability process depends strongly on the width ratio of the channel: as a result, in gravel bed rivers bars generally display an alternating structure, while central bars or higher order transverse modes are not likely
91
Figure 6.1: Free bars in the Rio Branco, South America. (Image Science and Analysis Laboratory, NASA-Johnson Space Center. 18 Mar. 2005. "Earth from Space - Image Information." <http://earth.jsc.nasa.gov/sseop/EFS/photoinfo.pl?PHOTO=STS61C-33-72> 28 Apr. 2005)
to form spontaneously, in the absence of some forcing mechanism, unless the channel is fairly wide. Notice that predicted and observed values of the longitudinal wavelength of bars fall in the range of 5-12 channel widths. When the width ratio is not exceedingly large with respect to the critical value c and the channel is long enough to allow their development, regular trains of bars which migrate almost steadily in the downstream direction are invariably observed in ume experiments. The occurrence of such equilibrium conguration, which displays typical asymmetries like diagonal depositional fronts and deeper pools, is mainly the consequence of the "low degree" of non linearity displayed by the system, which is clearly witnessed by the scarcity of relevant components in the two dimensional Fourier representation of bottom prole at equilibrium (see gure 6.2). Theoretical results of Colombini et al. (1987) suggest that for values of sufciently close to c non-linear interactions lead to a periodic solution with steady equilibrium amplitude; however, non-linearity is weak in that the growth of higher harmonics is passively driven by the development of the fundamental alternate-bar mode. As suggested by Schielen et al. (1993) the above solution may be unstable and lead to quasi periodic solutions, though a straight reach with a length of few hundred widths is required to appreciate the associated modulation of bottom conguration. Finally, as the width ratio increases the nonlinear competition between different modes becomes stronger and may lead to the occurrence of complex transverse structures (Fujita, 1989; Colombini and Tubino, 1991). Moreover, local emergence of bar structures, which invariably occurs for
92
Figure 6.2: A typical Fourier spectrum of the equilibrium bar topography with dominant bed-load. k longitudinal modes, m transverse modes. larger values of , generally prevents the establishment of an equilibrium conguration.
93
Figure 6.3: Marginal stability curves for bar formation: is width ratio, is the longitudinal wave-number. the interactions with small scale features, like ripples and dunes, such that the longitudinal bottom proles were signicantly distorted by high frequency components. As for tidal free bars, laboratory observations are not known to the present authors. On the other hand, eld observations suggest that, in spite of the oscillating character of the ow and of the dominant suspended load, the dependence of bar wavelength on channel width in tidal environments seems to conform to the trend exhibited by alternate bars in gravel bed rivers (Dalrymple and Rhodes, 1995). However, ascertaining the role played by several forcing mechanism is particularly relevant for tidal networks. In fact, such systems often display over a length of few kilometres the same degree of geometrical non uniformity which is typically distributed over an entire river basin; hence, bar development in tidal channels may be strongly inuenced by forcing factors, like the nite length and the changing geometry of the channel as well as the interaction with adjacent channels and the exchange of ow and sediments with tidal ats. Theoretical results for bar formation in both sandy rivers (Tubino et al., 1999) and tidal channels (Seminara and Tubino, 2001) have been recently derived within the context of a linear framework. Both analysis refer to an innitely long straight channel, with bed composed by a ne homogeneous sediment; they essentially differ for the different character of the basic ow whose stability is investigated: the basic ow is steady in Tubino et al. (1999), while the channel is subject to the propagation of a tidal wave in the analysis of Seminara and Tubino (2001). In particular, in the latter work, at the leading order of approximation, local inertia and spatial variations of tidal wave are found to be negligible at the scale of bars: hence, free bars feel the tidal wave as an oscillatory longitudinally uniform ow. The reader is referred to the above papers for further details on the structure of the solution. It is important to recall here that both solutions display several distinc-
94
6. Meso-scale bed forms: an application to uvial and tidal bars tive features with respect to the case of gravel beds. In fact, as Shields parameter increases the threshold value c for bar formation tends to vanish as a result of the vanishing role of the stabilising contribution of gravity. When suspended load is dominant bar stability crucially depends on the longitudinal wavelength of bars. In particular, shorter bars are damped while longer bars are enhanced, which implies a distortion of marginal curves for bar formation as qualitatively reproduced in gure 6.3. In tidal channels the above distortion is less pronounced due to the variation of Shields stress during the tidal cycle. As a result, when suspended load becomes dominant bar perturbations falling within the most unstable range of bed-load dominated gravel rivers are damped while alternate bars as long as 25-30 channel widths are expected to form, which correspond to dimensionless bar wave-numbers ranging about 0.1-0.12. Also notice that, while in the case of dominant bed load the alternate bar mode is the fundamental transverse mode in a wide range of values of , the linear theory of Tubino et al. (1999) suggests that in sandy streams various transverse modes are characterised by almost similar growth rates, as shown in gure 6.4. These results can be interpreted as an indirect suppressing effect of suspended load on bar stability in that a long straight reach is required to allow for the development of regular trains of bars: for channel widths of the order of few hundred meters a river reach should keep straight and relatively uniform over a length of several tens of kilometres. Also notice that uniformity of channel geometry seems to be an essential requirement for the spontaneous development of migrating bars: in fact, results of theoretical and experimental investigations (Kinoshita and Miwa, 1974; Tubino and Seminara, 1990; Whiting and Dietrich, 1993; Repetto and Tubino, 1999) suggest that spatial variations of curvature and channel width may strongly inhibit the migration of free bars along the channel. The results of gure 6.4 suggest that to determine the nite amplitude structure of bed topography, a weakly non linear analytical approach, like that introduced by Colombini et al. (1987), may turn out to be inadequate when suspension is dominant. In this case bar development is more likely to be controlled by the simultaneous amplication and non-linear competition between several unstable transverse mode. The resulting strong non linearity may imply that local emergence of bar structure may occur even for values of relatively close to the critical value c ; furthermore, the system may not reach an asymptotic equilibrium conguration displaying the simple bed structure depicted in gure 6.2. Under these circumstances one should resort to a fully non linear numerical approach like that proposed in Chapter 5.
95
Figure 6.4: The maximum growth-rate max is plotted versus the width ratio for different transverse modes. convection, whose effect is larger for shorter wavelengths. Hence, as increases the contribution of suspended load to the growth rate shifts from positive to negative as the peak of longitudinal transport exceeds bar crests: under this condition sedimentation occurs at bar pools and bar topography is damped. Linear analyses also suggest that the prediction of marginal stability conditions under suspended load dominated conditions is very sensitive to the model adopted and strongly depends on the ability of the model to reproduce adequately convective effects. Results of numerical simulations under fully non-linear conditions suggest that bar development with suspended load is similar to that observed under bed-load dominated conditions, provided the aspect ratio is fairly close to the threshold value for bar formation. Under these conditions an equilibrium conguration is achieved, which mainly arises from the damping effect of higher harmonics on the growth of the fundamental alternate-bar mode, as found by Colombini et al. (1987) for dominant bed-load. The resulting bed topography is still characterised by the dominance of the rst alternate mode, though the amplitude of higher harmonics is typically larger when suspended load is present: in particular, second order transverse modes (central bars modes) may attain an amplitude which is comparable to that of alternate bars. This is shown in gure 6.5 where the time development of the amplitude of the leading transverse modes of the Fourier representation of bed topography is reported: 11 denotes the alternate bar mode, while 02 + 22 are second order transverse modes which represent, respectively, a transverse deformation of the bed, in the form of a central deposit which doesnt display any longitudinal variation, and a central bar mode with a longitudinal wavelength equal to half the length of the fundamental 11 mode. Figure 6.5 allows for a comparison between the results obtained under bed-load dominated
96
Figure 6.5: Comparison between the time development of the amplitude of the leading components of bar topography under bed-load dominated conditions 0 = 0.1, = 20, Ds = 102 , R p = 11000 and with dominant suspended load 0 = 1, = 15, Ds = 105 , R p = 4. conditions and those with dominant suspended load: numerical simulations refer to different values of reference dimensionless parameters, though they are similar in terms of the distance of from the critical conditions for bar formation. Also notice that both the amplitudes of the Fourier components and time are scaled in the gure using the amplitude and the growth rate of the fundamental 11 component, respectively. A large number of signicant components characterises the spectrum of bed topography at equilibrium when suspended load is dominant, as shown in gure 6.6. As increases non-linear interactions, which are stronger with suspended load, lead to complex bed congurations; numerical results suggest that an equilibrium conguration is no longer achieved; furthermore, the superposition of transverse modes soon leads to local emergence of bar structures. Figure 6.7 suggests that for given value of particle Reynolds number the role of second order transverse modes increases for higher values of Shields stress, that is of the intensity of suspended load (the amplitudes of the Fourier components and the time are scaled as in gure 6.5). The values of the width ratio em at which local emergence of bar structure occurs in numerical computations are plotted in gure 6.8 versus Shields stress , for different values of particle Reynolds number R p . Results are given in terms of the relative distance of the condition of bar emergence from the critical condition for bar formation. Notice that the role of suspended load increases as increases and R p decreases. It appears that when suspended load is large the local emergence of bar structure, which may imply the formation of central islands and the transition to non-migrating complex bed-forms, may occur at relatively low values of width ratio.
97
Figure 6.6: The Fourier spectrum of the equilibrium bar topography with dominant suspended load: k denotes longitudinal modes, m transverse modes 0 = 1.25, = 12, Ds = 2 105 , R p = 10.
98
Figure 6.7: Time development of the amplitude of the leading components of the Fourier representation of bed topography for different values of Shields stress = 11, Ds = 2 105 , R p = 10. Equilibrium bed topographies obtained for values of the aspect ratio falling below the critical values for bar emergence are given in gures 6.9-6.15 for different values of the dimensionless parameters. Figures 6.9-6.12 are obtained with the particle Reynolds number R p = 10, which corresponds to a dimensional grain size of 0.2 mm. In this case the topography displays a diagonal arrangement of bar fronts like that typically observed in the case of gravel bed river bars. As the role of suspended load increases and becomes dominant bar pattern changes as shown in gures 6.13-6.15 where a value of R p = 4 has been used, which corresponds to a grain size of 0.1 mm (the reader is referred to gure 6.16 where the effect of R p on the ratio of suspended to bed load is reported). In this case diagonal fronts are nearly absent and downstream slopes become gentler.
99
Figure 6.8: The computed values of width ratio em at which local emergence of bar structure is observed are plotted versus Shields stress for two different values of particle Reynolds number.
40
14 12 10 8 6 4
0.5
4106
20
0.13
13
0.1
3527
0.
0.39605
0 0
2 10
0.3960
20
30
40
50
4 0. 0.2 0 0 7053 0. . 13 13 20 52 2 7
0.264 0
202
0.13
202
60
Figure 6.9: The equilibrium bed topography for 0 = 1.25, R p = 10, = 12, Ds = 2 105 .
100
0.2
64
0.
04
39
0.2 6
058
5 60
0 . 27 0. 0 13 30.4058 5 52 7
16
0
0.39605
0 .13 0
20
02 0.132
02 0.132
0.26404
0.26404
0.2
6 0.2
04 64
4 40
16
0 0 0.1027 9 0.2055 8
14 12 10 8 6 4 2
558
0.
20
8 55
0.20
00
0.
30
5 0.205 7 3
79
79
79 56 02 103 1 . 0. 00.20712
69 10 .3 0
02
0 0 .1
142
02
0.4
0.205
58
10
20
30
40
50
60
Figure 6.10: The equilibrium bed topography for 0 = 1.5, R p = 10, = 13, Ds = 2 105 .
16
24
12 10 8
94
0.1
0 0.124 42
0.26329 0 0 0.13165
0.24884
0.
25
0.24884
394
0.12442
14
0.
0.373
0.24884
94 394 0. 0 .26329 0 .1 31 65
6 4 2
44
3 0.1
16
5
0.12442
732
0. 12
10
0.3
20
30
40
0. 12
0.3
949
73
25
44
0.3
3 26 0 0 0.0.39 494
0.1 29
3165
50
Figure 6.11: The equilibrium bed topography for 0 = 1, R p = 10, = 10, Ds = 105 .
37 0.1 325 24 42
42
60
101
00
0 0
0. 12
00
02 44
0.1
0.
30
83
0.30837
0.1
79 02 0.1 0
8 55 20 0.
0.20558
0.30837
24
0.12442
42
.26 32 9
4 0.2
4 88
0 0
16
0
0.2765 8
0 0.
5 14 69 9 .0382 . 0 0 001
12 10
0.
0
8 6 4 2
0.
8 13
29
0 0 .0
10
20
30
40
50
0.
0.2
45
69 0 1
06
91
45
0.0
13
0 0.1 .0 96 98 4 0 0 21 4 8 0 .0 69 14 5
.290.39287 46 5
82
Figure 6.12: The equilibrium bed topography for 0 = 2, R p = 10, = 12, Ds = 2 105 .
16
0
12
0 0.045 0
14 0 .180
22
2
8 6 4 2
0.04 505 0 6
0.18
022
17 35 1 0. 0.1 8
02
10
20
30
40
50
Figure 6.13: The equilibrium bed topography for 0 = 1.5, R p = 4, = 14, Ds = 105 .
102
0 0 .04505 6 0
60
0 0 0 .0 0. 988 148 08 21
0.0 05
0.090112
45
0.13517 11 0 09 10 0.
56
02
05
0.
18
0.1
14
3 74 20 0.
45 91 0 06 0. 0
8 0.2765 3 0.2074
0.069 145 0
60
0 0 5056 0.04
0
0. 04
0.0 450
50
56
56
0.0 45
4 .0 0 94
0 0. 494
04
04
16
0.13466
05
14 12 10 8 6 4 2
0 0.0962
0.4
9 03
0
0
.1
8 0.4039 2 93 .26 0
0 0.2 .1 88 92 61 41
5 20 96 .0 0 0 6 0.1346
0.13466
0.2 693 2
0.13466
0
0. 205 096 0. 19 2
82
50
66
0.13466 0
0.1
34
09 0.
19
1 24
0.
5 620 0 .2 88 6 38482 1
20
0.
69 32 40 39 8
0.13466
0.13466
0
40
.2
61 88 0.384
41
0.2 693 2
0
0
10
30
60
16 14 12 10 8
0.13122
0
0.
0
0.13122
0.13122
0 0
39
36
7 45 62 .2
00.2449 .16 32 6
2
0.39367
24 . 0 0810.160. 326 63 2 0
122
49
0.26
.0
81
245
63 0
0.13
0 .0 0 81 63
4 2
24 5 0.39367
163 26 0.2449
0 0
0.2 624 5
6 32 .16 49 0 24 . 0
10
0.
.0 0 6 81
0 0 32
26
0.
0.3
936
20
30
40
50
60
103
2.5
Qs/Qb
Qs/Qb
(a)
1.5
5 4 3
0.5
2 1 0 (b)
30 ds=1E5 ds=1E4 ds=1E3 ds=1E5 ds=1E4 ds=1E3
0 0
18 16 14 12
10
15
20
25
30
35
40
10
15
20
25
30
35
40
Rp
Rp
25
20
Qs/Qb
10 8 6 4
Qs/Qb
(c) 5 10 15 20 25 30 35 40
15
10
5 2 0 0 0 0 (d) 5 10 15 20 25 30 35 40
Rp
Rp
Figure 6.16: Ratio between the suspended load and the bed load, as a function of the sediment particle Reynolds number R p , for different values of the Shields stress = 0.5 (a), = 1 (b), = 1.5 (c), = 2 (d). The computation is performed using the standard closure relationships of van Rijn (1984), which are not valid for the higher values of Rp.
104
(1 p) D 0 , Ws S0
(6.1)
(6.2)
We may notice that even under suspended load dominated conditions, the effect of bed load cannot be neglected because it always affects substantially the behaviour of bed-forms. Results of numerical simulations suggest that two main behaviours can be recognised, depending upon the relative values of the time scale of bed development and the tidal period. When the time scale of bed development is shorter than the tidal period, as it is typical of relatively coarse particles, the main transport mechanism is the bed load. In this case bed forms migrate slowly during the tidal period, but on the average their celerity vanishes, as shown in gure 6.17 where the bed topography corresponding to quarters of the tidal cycle is plotted. In the second case, namely when the time scale of bed development is much longer than the tidal period, as it is typical of ne sediment tidal channels, suspended load is dominant, the bed elevation is a slow time variable and bed forms dont migrate during the tidal cycle, display symmetrical shapes and mainly grow close to the peak values of the ood and ebb phases, as shown in gure 6.19. Notice that in this case residual effects are absent due to the imposed symmetry of the basic ow. Numerical results are in fairly good agreement with eld observations in the Severn estuary (Harris and Collins, 1985), which suggest that larger bed-forms such as
105
16
0.4
0. 2 0.1
14 12 10 8
.4 0. 3 0.1
0.2
0.
0.1 0. 0 1
0.1
0.4
.4
0.5
0.2
0.
5 16
10
15
20
25
30
.1 0 0 0.2 0.3
14 12 10 8 6
0
0
0. 1
0.2
0.3
0.1 0
0.4
0.4
.3
0.4
.3
0.
0.1
0 0 .1
0.3
.2
0.1
0.5
5 16
10
15
20
25
30
0.4
.4
.2 0.
14 12 10 8 6 4 2
0 .1 0
0.2
0.3
0.1
0.3
0 .4 0 .3
0.1
0.2
0.1 0
0.2 0. 3
0.2 0.3
0.4
.5
.4
0.1
.4 0 0
10
0.4
0.3 0.4
0 .5
5 16
15
20
0 .4 0 .5
25
30
0 0. 1 0.1
.1
14 12 10 8 6
0.1
0.1
0.2
0. 2
0.1
0.1
0.3
0.
0.3 0.2
0.3
0
0.1
0 0.1
0.2
0
0.2 0.1 .1
0.
0 0.1
0.4
0.4
20 25
2 5 10
15
30
Figure 6.17: Time sequence of bed topography during the tidal cycle, under bed load dominated condition. 0 = 0.1, = 13, R p = 11000, Ds = 102
106
0.2 0.3
.3
0.2 0.3
0.1
0.1
0
0.2 0.1 0 0 .1
.4
.3 0.4 0 0.5
0.2
0.4
0.1
0
0.1
0.3
0.1 0 .1 0 0 0 . 3 .4
0.2
0.1
0.3
4 0.
0.5
0.5 0.4
1 0. 0
1 0. 0 0.1 0.2
0.1
.3 0
0.3
0.1
0
0.1 2 0. .3 0
.2 0
0.3
0.2 0.1
.2 0 .1 0 0.1
0.2
0.4
0.1
0.2
0.4
.3
0.1
0.1 0.2
0.3 0.4
0.1
0.1 0 0.1
0.1
0.4
0.3
0.2
0.1
0.1
0.1 0 .1 0 .2 3 0 0.
0.1
0
0
0. .2
0 1
0.2 .1 0
0.2
0.1 1 0. 0
0.3 0.4
0.4
14 12 10
8 6 0.2
.05 0 0
0.1
0 00
0. 15
0.
0.05 0.1
0.05
0.2
0.3.25.215 0 00.
2 0. 0..2 10 5 5 0 .1
0.05 0
0.05 0
0.05
0 0.05
4 0.3 0.35 2
0.3
0.3 5
20
40
60
80
100
120
Figure 6.18: The bed topography under suspended load dominated condition. = 2, Ds = 105 , R p = 4, = 13. sand bars are stable even at spring-neap tidal frequency. Furthermore, long-term asymmetry of bedforms can only be produced and maintained when tidal currents are asymmetric, with a tidal phase dominating over the other (Harris and Collins, 1985). Numerical results clearly suggest that the development of bed topography is related to the growth of the bed forms close to the peak values of ood and ebb phase. Hence, present results do not seem to support the hypothesis introduced in the analytical solution of Schuttelaars and de Swart (1999) where bed evolution is assumed to depend only on the net sediment transport averaged over a tidal cycle. In gure 6.20 the time behaviour of the amplitude of different longitudinal bar modes is reported at the onset of bar development (notice that the longitudinal length decreases with the integer k). As shown in gure 6.19 mode amplication exhibits an oscillating behaviour, whose amplitude depends strongly on the longitudinal scale of the mode. In particular shorter wavelengths (green and blue solid lines in the gure), whose tidally averaged growth rate is smaller, may undergo an instantaneously faster growth which is then followed by a severe damping, while longer wavelengths (black and red solid lines in gure), which fall within the most unstable range, exhibit a more regular evolution. The overall process of bar formation doesnt seem to differ much with respect to the uvial case; also in this case the emergence of bar structure may occur for values of the aspect ratio which are relatively close to the critical value c . The main distinctive feature of the tidal case is the vanishing value of tidal averaged celerity, at least under the forcing effect of a symmetrical tide considered herein. Non vanishing values of bar celerity could be associated with tidal asymmetries; however, these effects are not likely to produce large values of migration speed like those observed in gravel bed experiments in straight umes. The above result suggests that the interaction between free and forced responses may be dif-
0 0 0.05
0.05
.3 0
1 0. 5 0.0
15 0. 1 0. 0.05
0
0.1 0 .15 0.2
05 0.
.1 0.15
0.05
0 .2
0.2 0.150.25
0.05
0.05 0.1
107
0.01004
flood
ebb
flood
ebb
flood
A
0.01 0
2000
4000
6000
8000
10000
12000
14000
16000
18000
Figure 6.19: The time development of the Fourier component A11 of the bed prole for = 2, Ds = 105 , R p = 4, = 15.
0.01003
k, m
Figure 6.20: The time development of the Fourier components A11 , A21 , A31 , A41 of the bed prole for = 2, Ds = 105 , R p = 4, = 13.
108
6. Meso-scale bed forms: an application to uvial and tidal bars ferent with respect to the uvial case, where free bars migration in the downstream direction may inhibit local bank erosion. In the tidal case the presence of quasi-steady alternate bars may provide a mechanism to produce the local erosion of the bank, with the consequent deection of the channel axis eventually leading to a meandering conguration.
109
110
Estuarine and coastal suspended sediment dynamics is a complex phenomenon, whereby sediments undergo a sequence of processes such as erosion, deposition, advective and diffusive transport. Suspended sediment motion is inherently a three- dimensional process, though most of the estuarine and coastal models are two-dimensional: hence, they consider only the horizontal directions and imply the solution of the depth-integrated sediment transport equation (see for example Galappatti and Vreugdenhil, 1985). Two dimensional models are simpler and require much less computational effort with respect to a 3D formulation. However, within the context of a 2D approach, only the averaged (or depth integrated) sediment concentration is known; hence, the value of the concentration close to the bed, or the upward ux, which is required for the evaluation of the sediment erosion or deposition rate, must be related to the averaged concentration. When a 2D approach is used, the ratio between the mean concentration and the value at the bed is typically assumed to be equal to the corresponding ratio for an equilibrium concentration prole. Implicit in this procedure is the assumption that vertical concentration proles can adjust instantaneously to changing ow conditions; therefore, the model can be safely applied only to situations in which the differences between the local concentration prole and the equilibrium prole, computed in terms of the local hydrodynamic conditions, are conveniently small. The vertical concentration proles in ne sediment systems, and hence the morphological predictions with suspended load, are very sensitive to the choice of the procedure and of the reference value of concentration used to set the bed boundary condition, as pointed out by van Rijn (1984). Under uniform conditions the vertical concentration prole can be represented through the well known Rousean distribution (5.34); on the other hand, changes in the boundary conditions can modify signicantly such distribution, as in the case of the transition from one equilibrium state to another due to an abrupt change of the bed boundary condition, a problem which has been investigated by Hjelmfelt and Lenau (1970) among others (see gure 7.1). Deviation of the vertical prole from the local equilibrium prole, which is dened as the Rouse distribution corresponding to the local and instantaneous hydraulic conditions, can be fairly large under non-uniform condi-
111
Figure 7.1: Suspension of sediments beyond an abrupt change of the bed boundary condition. tions, since suspended load requires a relatively large adaptation length to respond to changing hydraulic conditions. The role of this adaptation process on bed stability and bedforms dynamics under suspended load dominated conditions has been highlighted in many contributions (see the original work of Engelund and Fredsoe, 1982) for small scale bedforms and the recent contributions of Tubino et al. (1999) and Seminara and Tubino (2001) for meso-scale bedforms. However, few analytical formulations are presently available to account for the effect of ow and concentration non-uniformities on the concentration proles (see Armanini and Silvio, 1988, Galappatti and Vreugdenhil, 1985). Bolla Pittaluga and Seminara (2003a) have recently revisited the approximate solution proposed by Galappatti and Vreugdenhil (1985), which was aimed at deriving a suitable two-dimensional closure for sediment transport. Suspended sediment transport under non uniform conditions has been also analysed recently through three-dimensional numerical models, like that discussed in Chapter 5 or those proposed by Lin and Falconer (1996) and Wu et al. (2000). These models have been applied to relatively simple cases; in fact, their use is limited by the large computational time which is required for long term simulations, since the three dimensional formulation needed when a two dimensional model is applied to the same context. Hence, a three dimensional model can be hardly used to describe the long term morphodynamic behaviour of a river reach or an estuary, since it would require prohibitively long numerical simulations. For instance, we may note that a simulation, like those presented in Chapter 6 to investigate the development of bars in tidal channel under suspended load dominated condition, may require nearly one month of computational time on a Xeon 2 GHz processor. In the steady case numerical simulations are faster (nearly one day is required to achieve the equilibrium conguration) since a longer time step can be used. The above limitation is the main reason why nding an analytical formulation for the vertical implies a quite large number of grid nodes, say 10 100 times the number of grid nodes which are
112
Figure 7.2: Vertical concentration and velocity proles. concentration prole, to be included within two-dimensional morphological models, would be highly desirable. However, in order to achieve a reasonable accuracy in the prediction of the local morphological response of uvial and tidal systems, the analytical solution must represent adequately the typical delay of suspended load with respect to the local bottom shear stress. In this chapter we pursue a comparison between the analytical asymptotic solution of Bolla Pittaluga and Seminara (2003a) and the results of the present numerical model. For the sake of simplicity, the comparison is performed with reference to a plane ow, that is the model is applied non-uniform ows; we then consider a steady ow over a sinusoidal bottom prole, for different values of the amplitude and wavelength. to two-dimensional x z context. As a rst step, we restrict our analysis to the case of spatially
113
(z, , h, D) = (u, w) = t=
where u and w are the velocity components in the longitudinal and vertical direction, respectively, is the bed level, h and D are the free surface elevation and ow depth, respectively. Furthermore
L0 , D and U0 are suitable scales for the longitudinal coordinate, the ow depth and the ow 0
D 0 , L0 C f 0
Z0 =
Ws , C f 0U0
L0 , D 0
(7.2)
where is the Von Krmn constant, C f 0 is the reference friction coefcient and Ws the dimensional settling velocity of sediments, the momentum equation in longitudinal direction, ow continuity and the advection diffusion equation for the suspended sediment can be written in the following form for plane ow: u u u + u + w t x z = h u + z , 2 x z z F0 (7.3)
u w + = 0, x z wC uC C + + t x z Z0 C C = T . z z z
(7.4) (7.5)
The boundary conditions associated with equations (7.3), (7.4) and (7.5) are those discussed in section 4.1: the no slip condition for the velocity at the bed, the dynamic condition of vanishing shear stress and the kinematic condition at the free surface, the condition of vanishing sediment ux at the free surface and the gradient boundary condition for sediment concentration at the bed. Following the approach introduced by Bolla Pittaluga and Seminara (2003a), it is possible to provide analytical solutions of the vertical concentration proles under the assumption of slowly varying conditions. In fact, as discussed by the Authors, the dimensionless parameters and appearing in the differential equation (7.3) are fairly small when the ow and the concentration eld display a slow variability both in time and in space, which implies that both T0 and L0 are fairly large. Such conditions are typically encountered in tidal ows and ood waves. A formal perturbation solution of the differential problem, made of equation (7.5) along with the boundary conditions, can be obtained by expanding the concentration C in powers of the small parameter
114
7. Vertical concentration proles in non-uniform ows in the following form: C = C0 + C1 + O(2 ) where = and Z denotes the local Rouse number.
U0 D 0 = L ws 0 Z
(7.6)
(7.7)
Substituting the latter expansion into the differential equation (7.5) and equating likewise powers of , a sequence of differential problems at the various orders of approximation in the small parameter is found.
At the leading order of approximation the solution (5.34) is found, which corresponds to the classical Rouse type concentration prole: C0 = Ce , Ds , R p , a f (, Z, a) written in term of the boundary tted vertical coordinate =
z D .
only on the closure relationship employed for the eddy diffusivity T ; in particular, when the closure proposed by McTigue (1981) (4.9) is used, the function f takes the form (5.35). Equation (5.34) can also be expressed in terms of the depth averaged concentration C0 in the following form: C0 = C0 (x,t)0 (, Z, a), where 0 (, Z, a) = I(Z, a) = 1 (1 a) f (, Z, a) , I(Z, a) f (, Z, a)d. (7.9)
(7.10a) (7.10b)
Z 1
a
At the next order of approximation, the leading contribution of the spatial non-uniformity of the ow eld on the vertical concentration prole is obtained. Recalling that: C0 C0 0 = 0 () +C0 . x x x the solution for C1 can then be written in the form: C1 = D C0 C11 +C0C12 , x (7.12) (7.11)
115
7. Vertical concentration proles in non-uniform ows where the functions C1 j ( j = 1, 2) are the solutions of the following boundary value problems: dC1 j d 1 T kZ0 D d d T dC1 j +C1 j = 0 kZ0 D d dC1 j =0 d where p1 = u 0 , p2 = u + dC1 j = p j () , d ( = 1) , ( = a) , 0 . x (7.13) (7.14) (7.15)
(7.16)
The slowly varying character of the ow eld allows one to introduce a self similar logarithmic structure for the velocity u to compute the forcing terms p1 and p2 appearing in (7.16). It is worth noticing that in the original formulation of Bolla Pittaluga and Seminara (2003a) the term proportional to 0,x in equation (7.12) is neglected, though this approximation is not formally justied within the framework of the perturbation scheme adopted by the Authors. However, as pointed out before, the neglected effect doesnt seem to contribute signicantly to the approximate solution.
7.2 Results
As a rst step of present analysis a comparison is pursued between numerical and analytical solutions under steady ow conditions, for the case of an imposed bed prole which changes in the longitudinal directional according to the following sinusoidal form: = sin 2 0 x . Lb (7.17)
Simulations are performed with values of the wavelength Lb ranging between 2500m and 10000m
and values of the mean ow depth ranging between 5m and 15m. Notice that in this case the length
scale L0 , dened in (7.1a), is xed and coincides with the wavelength Lb . In the numerical solution
periodic boundary conditions are imposed for the ow eld and for the concentration eld, at the upstream and downstream ends of the longitudinal domain. The comparison between the results of the two models is performed according to the following procedure: the depth-integrated solution for the concentration is computed through the numerical model; then, the solution is supplemented to the asymptotic model for the evaluation of C0 ,x , C0 , D and 0,x ; nally, the analytical solution is determined. A comparison between the results of the numerical and analytical solution at different cross
116
0.4 0.2 x/L=0 0.0 -0.015 -0.010 -0.005 0.000 C1 0.005 0.010 0.015
x/L=3/4
Figure 7.3: Vertical proles of the perturbation C1 at different cross sections: Lb = 10km, = 0 0.5m, D = 5m and R p = 10, = 0.023 (left), R p = 4, = 0.042 (right). Dotted 0 line: numerical solution; continuous line: analytical solution; dashed line: analytical solution assuming 0,x = 0.
C1
0.00
0.05
0.10
-0.02
-0.01
0.00 C1
0.01
0.02
0.03
Figure 7.4: Vertical proles of the perturbation C1 at different cross sections: Lb = 2.5km, = 0 = 10m, R = 4, = 0.37 (left); L = 5km, = 0.5m, D = 5m, R = 4, 1.5m, D0 p p 0 0 b = 0.085 (right). Dotted line: numerical solution; solid line: analytical solution; dashed line: analytical solution assuming 0,x = 0.
117
0.05
qs1
-0.01 -0.02
0.00
0.25
0.50
0.75
1.00
qs1
0.00
0.00
-0.05
x/L
x/L
Figure 7.5: Longitudinal proles of the perturbation qs1 : Lb = 10km, = 0.5m, D = 5m 0 0 R p = 10, = 0.023 (left), R p = 4, = 0.042 (right). Dotted line: numerical solution; continuous line: analytical solution.
0.10 0.05
0.4 0.2
qs1
qs1
0.25 0.50 0.75 1.00
0.00
0.0
-0.05 -0.10
-0.2 -0.4
0.00
0.00
0.25
0.50
0.75
1.00
x/L
x/L
Figure 7.6: Longitudinal proles of the perturbation qs1 : Lb = 5km, = 1.5m, D = 10m; 0 0 R p = 10, = 0.10 (left), R p = 4, = 0.19 (right). Dotted line: numerical solution; continuous line: analytical solution.
sections in the longitudinal direction, is reported in gures 7.3 and 7.4, for different values of the parameter and of the amplitude of the sinusoidal bed prole. It is worth noticing that 0 the perturbation approach is able to reproduce at least qualitatively the vertical structure of the correction of the concentration prole, with respect to the equilibrium prole, and tends smoothly to the numerical solution as vanishes. However, the response of the analytical model, to the effect of variable ow conditions, seems somewhat exaggerated in that the analytical solution displays a much faster adaptation to the local conditions with respect to the numerical model. A further comparison is made in terms of the deviation qs1 of the suspended sediment transport from the value qs0 , which would be attained at equilibrium with the local hydrodynamic conditions. For the numerical model the value of qs1 is directly computed by subtracting the
118
8 7
20 6 5 15
A1
4 3 2 1 0 0
A2
10 5 0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Figure 7.7: Difference between the numerically and the analytically evaluated amplitude of the rst (left) and second (right) mode of the Fourier spectrum, as a function of . Dots R p = 4 and crosses R p = 10. equilibrium solution from the numerical solution; for the approximate solution we can write: qs1 = D
Z 1
a
C1 u d
(7.18)
Analytical and numerical results are fairly close for small values of (see gure 7.2), while for larger values differences may become signicant (gure 7.2). In gures 7.2 and 7.2 is given a closer comparison between numerical and analytical solutions, in terms of the difference between numerically and analytically evaluated amplitude and phase lag of the leading components of Fourier analysis of qs1 . Results are plotted as functions of (note that the fundamental rst mode has the same wavelength of the imposed bed prole Lb ). It appears that the analytical model overestimates the amplitude of the perturbation of suspended load transport by an amount which increases with ; however, for relatively small values of , say smaller than 0.1, the analytical model reproduces quite well the numerical solution. Furthermore, the phase lag between the two solutions grows rapidly with . One may argue that the differences between the two solutions could also be due to the assumption of a logarithmic structure for the velocity prole introduced in the analytical solution. However, according to the results of the numerical model, the deviations from the logarithmic velocity prole are negligible in the slowly varying context analysed herein. The results of comparisons discussed above suggest the suitability of the analytical model of Bolla Pittaluga and Seminara (2003a) at least for relatively small values of . The denition of the range of applicability of the model is important in order to incorporate the procedure in a depth averaged model to evaluate the suspended sediment ux. For instance, rather than using a
119
25
20
20 15
15
2
10 5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0
10
0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Figure 7.8: Difference between the numerically and the analytically evaluated phase lag of the rst (left) and second (right) mode of the Fourier spectrum, as a function of . Dots R p = 4 and crosses R p = 10. more complex three dimensional approach, the analytical model could be employed to investigate whether the presence of suspended load has any effect on the nature of bar instability. This challenging investigation has recently been renewed by Federici and Seminara (2003a) in the case of bed load only. In Federici and Seminara (2003b) the analysis have been extended to the case of suspended load; however, in order to have small values of , the analysis is made assuming a fairly large value of particle diameter, which implies a large Ws . According to the results reported in gure 6.16, suspended load is not dominant with such large values of R p ; furthermore, the standard closure relationships which are adopted to compute the suspended load (e.g., van Rijn (1984)) are not valid within this range. It is worth noticing that when the wavelength is relatively short, the sediment is ne (as it is typical of suspension dominated environments) or the ow depth becomes large, the analytical model may introduce a fairly large approximation. We note also that the bed conguration can be inuenced by several factors, like a meandering pattern or the presence of regulation works, which can introduce even smaller length-scales than those considered herein. The case of unsteady ows, like those occurring in estuaries and tidal channels, could also be tackled with a similar analytical approach. However, the denition of the basic ow and the identication of the factors to take into account in the perturbation solution pose additional difculties. Moreover, while the denition of the time scale T0 is straightforward, since it coincides
with the tidal period, the denition of the length scale L0 is not obvious, since it depends on the
hydrodynamic behaviour of the tidal channel, as discussed in Chapter 2. Notice, however, that according to Bolla Pittaluga and Seminara (2003b) the morphodynamics of tidal channels seems to be only slightly affected by non-equilibrium effects. The analysis of the range of applicability
120
7. Vertical concentration proles in non-uniform ows of the analytical model in tidal ows is still a matter of investigation.
121
122
Bibliography
J. Allen. Current Ripples: Their Relation to Pattern af Water and Sediment Motion. North-Holland Publ. Co., Amsterdam, 1968. J. Allen. Sedimentary Structures - Their Characteristics and Phyical Basis, volume 1. 593 pp. Developments in Sedimentology, Vol.30A. Elsevier, New York, 1982. J. R. L. Allen and K. Pye. Preface. In J. R. L. Allen and K. Pye, Saltmarshes: Morphodynamics, Conservation and Engineering ignicance, pages viiviii. Cambridge University Press, 1992. A. Armanini and G. D. Silvio. A one-dimensional model for the transport of a sediment mixture in non-equilibrium contion. Journal of Hydraulic Research, 26:275292, 1988. J. Barwis. Sedimentology of some South Carolina tidal-creek point bars, and a comparison with their uvial counterparts. In A.D. Miller, Fluvial Sedimentology, pages 129160. Can. Soc. Petrol. Geol. Mem., 1978. P. Blondeaux and G. Seminara. A unied bar-bend theory of river meanders. J. Fluid Mech., 127: 449470, 1985. M. Bolla Pittaluga. Long term morphodynamic equilibrium of tidal channels. PhD thesis, University of Genoa, December 2002. M. Bolla Pittaluga and G. Seminara. Depth integrated modelling of suspended sediment transport. Water Resources Research, 39(5):ESG 1, 2003a. M. Bolla Pittaluga and G. Seminara. Long term morphodynamic equilibrium of tidal channels. the role of settling lag, overtides and sediment supply. Water Resources Research, submitted 2003b. M. Bolla Pittaluga, N. Tambroni, C. Zucca, L. Solari, and G. Seminara. Long term morphodynamic equilibrium of tidal channels: preliminary laboratory observations. In Proceedings of 2nd RCEM Symposium, pages 423432, Obihiro (Japan), 10-14 September 2001.
123
Bibliography V. Casulli and E. Cattani. Stability, accuracy and efciency of a semi-implicit method for threedimensional shallow ow. Computers and Mathematics with Applications, 27(1):99112, 1994. V. Casulli and D. Greenspan. Numerical Analysis for Appl. Mathematics, Science and Engineering. Addison-Wesley, 1988. H. Chang, D. B. Simons, and D. A. Woolhiser. Flume experiments on alternate bars formation. J. Waterways, Harbours, Coastal Eng. Div. ASCE, 97(1):155165, 1971. M. Colombini and M. Tubino. Finite-amplitude free bars: a fully nonlinear spectral solution. In Sand Transport in Rivers, Estuaries and the Sea, pages 163169. R. Soulsby and R. Bettes eds. Balkema, 1991. M. Colombini, G. Seminara, and M. Tubino. Finite-amplitude alternate bars. J. Fluid Mech., 127: 213232, 1987. R. Dalrymple, R. Knight, and J. Lambiase. Bedforms and their hydraulics stability relationships in a tidal environment, Bay of Fundy, Canada. Nature, 275:100104, 1978. R. Dalrymple, R. Knight, B. Zaitlin, and G. Middleton. Dynamics and facies model of a macrotidal sand-bar complex, Cobequid Bay-Salmon River estuary (Bay of Fundy). Sedimentology, 37: 577612, 1990. R. W. Dalrymple and R. M. Rhodes. Estuarine dunes and bars. In G.M.E. Perillo (ed.): Geomorphology and Sedimentology of Estuaries, Development in Sedimentology, 53, pages 359422. Elsevier Science, Amsterdam, 1995. H. J. de Vriend, J. Dronkers, M. J. F. Stive, A. V. Dongeren, and Z. B. Wang. Coastal inlets and tidal basins (partly in dutch). Technical report, TU Delft, January 2000. R. B. Dean. Aero Rep., volume 74-11. Imperial College, London, 1974. J. Dronkers. Tidal Computations in Rivers and Coastal Waters. North-Holland, New York, 1964. F. Engelund and J. Fredsoe. Sediment ripples and dunes. Ann. Rev. Fluid Mech, 14:1337, 1982. B. Federici and G. Seminara. On the convective nature of bar instability. J. Fluid Mech., 99(1): 99, 2003a. B. Federici and G. Seminara. On the convective nature of bar instability: the effect of suspended load. In Proceedings of 3rd RCEM Symposium, pages 136147. Barcelona (Spain), 1-3 September 2003b.
124
Bibliography J. Fredsoe. Meandering and braiding of rivers. J. Fluid Mech., 74:609624, 1978. C. T. Friedrichs and D. G. Aubrey. Tidal propagation in strongly convergent channels. J. Geophys. Res., 99:33213336, 1994. C. T. Friedrichs, B. D. Armbrust, and H. E. de Swart. Hydrodynamics and equilibrium sediment dynamics of shallow, funnel-shaped tidal estuaries. In Physics of Estuaries and Coastal Seas, Balkema, Rotterdam, pages 315327. Dronkers and Scheffers (eds.), 1998. Y. Fujita. Bar and channel formation in braided streams. In River Meandering, pages 41746. S. Ikeda and G. Parker eds. AGU Water Res. Monograph, 1989. Y. Fujita and Y. Muramoto. Studies on the process of development of alternate bars. Technical report, D. P. R. I. Bull., Kyoto Univ., Part3, No. 314, 55-86 1985. R. Galappatti and C. B. Vreugdenhil. A depth-integrated model for suspended sediment transport. J. Hydr. Eng., ASCE, 23(4):359377, 1985. M. Garcia and Y. Nio. Dynamics of sediment bars in straight and meandering channels: experiments on the resonance phenomenon. J. of Hydraulic Res., 31(6):739761, 1993. P. Garcia-Navarro, F. Alcrudo, and J. Saviron. 1D open-channels ow simulation using TVDMacCormak scheme. J. Hydr. Eng. ASCE, 118(10):13591372, 1992. G. Green. On the motion of waves in a variable canal of small depth width. Trans. Cambridge Philos. Soc., 6(457-), 1837. P. Harris and M. Collins. Bedform distribution and sediment tranport paths in the Bristol Channel and Severn Estuary. U.K. Mar. Geol., 62:153166, 1985. P. T. Harris, E. K. Baker, A. R. Cole, and S. A. Short. A preliminary study of sedimentation in the tidally dominated Fly River Delta, Gulf of Papua. Cont. Shelf Res., 13:44472, 1993. A. Hibma, H. J. de Vriend, and M. J. F. Stive. Channel and shoal formation in estuaries. In Proceedings of 2nd RCEM Symposium, pages 463471. Obihiro (Japan), 10-14 September 2001. A. Hibma, H. Schuttelaars, and H. de Vriend. Initial formation and evolution of channels-shoals patterns in estuaries. In Proceedings of 3rd RCEM Symposium, pages 749760. Barcelona (Spain), 1-3 September 2003. A. T. Hjelmfelt and C. W. Lenau. Non-equilibrium transport of suspended sediment. J. Hydr. Div., ASCE, 96(HY7):15671586, 1970.
125
Bibliography S. Ikeda. Prediction of alternate bar wavelength and height. Technical report, Rep. Dept. Found. Eng. and Const. Eng., Saitama Univ., Vol. 12, 23-45 1982a. S. Ikeda. Lateral bedload transport on side slopes. J. Hydraulic Div., ASCE, 108(99):13691373, 1982b. S. Ikeda and T. Nishimura. Bed topography in bends of sand-silt rivers. J. Hydraulic Div. ASCE, 108(11):13971411, 1985. S. Ikeda, G. Parker, and K. Sawai. Bend theory of river meanders. Part 1. Linear development. J. Fluid Mech., 112(363-3), 1981. M. Jaeggi. Formation and effects of alternate bars. J. Hydraulic Div. ASCE, 111(11):11031122, 1984. D. A. Jay. Greens law revisited: Tidal long-wave propagation in channels with strong topography. J. Geophys. Res., 96(C11):2058520598, November 15 1991. R. Kinoshita. Investigation of channel deformation in Ishikari River. Technical report, Rep. Bureau of Resources, Dept. Science and Technology, Japan, 1961. R. Kinoshita and H. Miwa. River channel formation which prevents downstream translation of transverse bars. Shinshabo (in japanese), 94(1):1217, 1974. R. Knight. Linear sand bar development and tidal current ow in Cobequid Bay, Bay of Fundy, Nova Scotia. In The Coastline of Canada. Geol. Surv. Can. Pap., pages 8010: 123152. S.B. McCan, 1980. A. Kovacs and G. Parker. A new vectorial bedload formulation and its application to the time evolution of straight river channels. Journal of Fluid Mechanics, 127:153183, 1994. S. Lanzoni. Experiments on bar formation in a straight ume. 1. Uniform sediment. Water Resources Research, 36(11):33373349, November 2000. S. Lanzoni and G. Seminara. Long term evolution and morphodynamic equilibrium of tidal channels. J. Geophys. Res., 107:113, 2002. S. Lanzoni and G. Seminara. On tide propagation in convergent estuaries. J. Geophys. Res., 103 (C13):3079330812, 15 December 1998. R. L. Leveque. High resolution conservative algorithms for advection in incompressible ow. J. of Num. Analysis, 33(1):627665, 1996.
126
Bibliography B. Lin and R. A. Falconer. Numerical modelling of three-dimensional suspended sediment for estuarine and coastal waters. J. of Hydraulic Res., 34(4):435456, 1996. H. Lorenz. Het in rekening brengen vanden weerstand bij schommelende vloeistofbewegingen. De ingenieur, page 695, 1922. R. MacCormack. The effect of viscosity in hypervelocity impact cratering. AIAA Paper, pages 69384, 1969. D. F. McTigue. Mixture theory for suspended sediment transport. J. Hydr. Div. ASCE, 107(HY6): 659673, 1981. J. Murray and A. Hawkins. Sediment transport in the Severn Estuary during the past 8000-9000 years. J. Geol. Soc. London, 132:385398, 1977. C. Nordin. Statistical properties of dunes proles. U.S.G.S. Prof.Pap. 562-F, 1971. G. Parker. Self-formed straight rivers with equilibrium banks and mobile bed. Part 1. The sand-silt river. J. Fluid Mech., 89(1):109125, 1978. G. M. E. Perillo. Geomorphology and Sedimentology of Estuaries, volume 53. Developments in Sedimentology, Elsevier, Amsterdam, 1995. R. A. Pielke. Mesoscale meteorological modelling. Accademic press: San Diego, Calif, 1994. D. Prandle. Tides in estuaries and embayments. In Tidal hydrodynamics. John Wiley & Sons, pages 125151. B.B. Parker, 1991. H. E. Reineck. Tidal ats. In Recognition of Ancient Sedimentary Environments. SEPM Spec. Pub., pages 16: 146159. J. K. Rigby and W. K. Hamblin, 1972. R. Repetto and M. Tubino. Transition from migrating alternate bars to steady central bars in channels with variable width. In Proceedings of 1st RCEM Symposium, pages 605614. Genoa (Italy), 6-10 September 1999. A. Robinson. Ebb-ood channel systems in sandy bays and estuaries. Geography, 45:183199, 1960. D. Rubin and R. Hunter. Bedform alignment in directionally varying ows. Science, 237:276278, 1987. D. Rubin and H. Ikeda. Flume experiments on the alignment of transverse, oblique, and longitudinal dunes in directionally varying ows. Sedimentology, 37:673684, 1990.
127
Bibliography D. Rubin and D. McCulloch. Single and superimposed bedforms: a syntesis of San Francisco Bay and ume observations. Sediment. Geol., 26:207231, 1980. R. Schielen, A. Doelman, and H. E. de Swart. On the nonlinear dynamics of free bars in straight channels. J. Fluid Mech., 252(1):325356, 1993. H. M. Schuttelaars and H. E. de Swart. Multiple morphodynamic equilibria in tidal embayments. J. Geophys. Res., 105:2410524118, 2000. H. M. Schuttelaars and H. E. de Swart. Initial formation of channels and shoals in a short tidal embayment. J. Fluid Mech., 127:1542, 1999. G. Seminara. Stability and morphodynamics, volume 33. Meccanica, Kluviert Academic Publishers, 1998. G. Seminara and M. Tubino. Sand bars in tidal channels. Part 1. Free bars. J. Fluid Mech., 440: 4974, 2001. G. Seminara, S. Lanzoni, M. Bolla Pittaluga, and L. Solari. Estuarine patterns: An introduction to their morphology and mechanics. In Geomorphological Fluid Mechanics, pages 9999. N.J. Balmforth and A.Provenzale, Lecture notes in Phisics, Springer Verlag, Berlin, 2001a. G. Seminara, G. Zolezzi, M. Tubino, and D. Zardi. Downstream and upstream inuence in river menadering. Part 2-Planimetric development. J. Fluid Mech., 438:213230, 2001b. Y. Shimizu and T. Itakura. Calculation of bed variation in alluvial channel. J. Hydraul. Eng., 115 (3):367384, 1989. L. Solari, G. Seminara, S. Lanzoni, M. Marani, and A. Rinaldo. Sand bars in tidal channels. part 2. tidal meanders. J. Fluid. Mech., 451:203238, 2002. J. Southard and L. Boguchwal. Bed congurations in steady unidirectional water ows. Part 2. Synthesis of ume data. J. Sediment. Petrol., 60:658679, 1990. N. Struiksma and A. Crosato. Analysis of a 2-D bed topography model for rivers. In River Meandering, pages 153180. S. Ikeda and G. Parker eds. AGU Water Res. Monograph, 1989. A. M. Talmon, N. Struiksma, and M. C. L. M. van Mierlo. Laboratory measurements of the direction of sediment transport on transverse alluvial-bed slopes. J. of Hydraulic Res., 33(4): 495517, 1995. M. Toffolon. Hydrodynamic and morphodynamics of tidal ystems. PhD thesis, University of Trento, December 2002.
128
Bibliography M. Tubino and G. Seminara. Free-forced interactions in developing meanders and suppression of free bars. J. of Fluid Mech., 127:131159, 1990. M. Tubino, R. Repetto, and G. Zolezzi. Free bars in rivers. J. Hydraulic Res., 37(6):759776, 1999. J. van Den Berg. Bedform migration and bed-load transport in some rivers and tidal environments. Sedimentology, 34:681698, 1987. L. C. Van Rijn. Mathematical models for sediment concentration proles in steady ow. Euromech, 192(1):4968, 1985. L. C. van Rijn. Sediment Transport. Part II: Suspended Load Transport. Journal of Hydraulic Engineering, ASCE, 110(11):16131641, November 1984. Y. Watanabe and M. Tubino. Inuence of Bed Load and Suspended Load on Alternate Bar, volume 36. Proceedings of Hydraulics Engineering, JSCE, 1992. P. J. Whiting and W. E. Dietrich. Experimental constraints on bar migration through bends: Implications fo meander wavelength selection. Water Resources Research, 29(4):10911102, 1993. L. Wright, J. Coleman, and B. Thom. Processes of channel development in a high-tide-range environment: Cambridge Gulf-Ord River delta, Western Autralia. J. Geol., 81:1541, 1973. L. Write. River Deltas. In R.A. Davis, Jr., Coastal Sedimentary Environments(2nd ed.), pages 176. Springer-Verlag, New York, 1985. W. Wu, W. Rodi, and T. Wenka. 3D numerical modeling of ow and sediment transport in open channels. J. Hydraulc Eng., ASCE, 126(1):415, 2000. M. Yalin. Geometrical properties od sand waves. In Proceedings of Am. Soc. Civil Eng., pages 90: 105119, 1964. M. Yalin. Mechanics of Sediment Transport, 2nd edtion. Pergamon Press, Toronto, 1977. M. Yalin. On the formation mechanism of dunes and ripples. In Euromech Colloq. Proc., page 261, 1987.
129