Two‐dimensional modelling of free surface shallow water flow Hydraulic Reference Manual Iber v1.0 1
Views 102 Downloads 0 File size 665KB
Two‐dimensional modelling of free surface shallow water flow
Hydraulic Reference Manual Iber v1.0 19.07.2014
Two-dimensional modelling of free surface shallow water flow
REFERENCE MANUAL
1. PRESENTATION ...................................................................................... 5 2. HYDRODYNAMIC MODULE ...................................................................... 7 2.1. Introduction ........................................................................................7 2.2. Hydrodynamic Equations .......................................................................7 2.3. Bed Friction .........................................................................................8 2.4. Superficial Friction due to Wind ..............................................................9 2.5. Effective Stresses.................................................................................9 2.6. Hydrodynamic boundary Conditions ...................................................... 11 2.6.1. Closed Boundaries ........................................................................ 11 2.6.2. Open Boundaries .......................................................................... 13 2.7. Internal boundary conditions ............................................................... 14 2.7.1. Gates .......................................................................................... 15 2.7.2. Weirs .......................................................................................... 15 2.7.3. Gate-Weir Combination.................................................................. 16 2.7.4. Local Losses ................................................................................. 16 2.8. Infiltration ........................................................................................ 17 2.8.1. Green-Ampt ................................................................................. 17 2.8.2. Horton ........................................................................................ 18 2.8.3. Linear ......................................................................................... 19 2.9. Initial Abstraction............................................................................... 19 3. TURBULENCE MODULE .......................................................................... 21 3.1. Introduction ...................................................................................... 21 3.2. Scales of Turbulence in Shallow Waters ................................................. 22 3.3. Constant Turbulent Viscosity................................................................ 23
Iber. Turbulent Flow in Shallow Waters
3.4. Parabolic Profile of turbulent Viscosity ................................................... 23 3.5. Mixing Length Model ........................................................................... 24 3.6. The Rastogi and Rodi K-E model (Rastogi & Rodi, 1978) .......................... 24 3.7. Dimensional Analysis of the turbulent terms in the shallow water equations 25 4. NON-STATIONARY SEDIMENT TRANSPORT MODULE ............................ 28 4.1. Sediment Conservation Equations......................................................... 28 4.2. Bed Transport.................................................................................... 29 4.2.1. Stress Partition ............................................................................. 29 4.2.2. Bedload Discharge ........................................................................ 29 4.2.3. Bed slope correction ...................................................................... 31 4.2.4. Avalanche model .......................................................................... 32 4.2.5. Considerations on a non-erodible layer ............................................ 32 4.3. 2D Turbulent Suspended Load Module................................................... 32 4.3.1. Turbulent Suspended Load Equation ................................................ 32 4.3.2. Entrainment/Deposition (E-D) term Calculation ................................. 33 4.3.3. Sedimentation Velocity .................................................................. 35 5. NUMERICAL SCHEMES .......................................................................... 37 5.1. Calculation Mesh ................................................................................ 37 5.2. Discretization in finite volumes for the 2D-SWE Equations ........................ 38 5.2.1. Discretization of the convective flux terms ........................................ 39 5.2.2. Discretization of the bed slope source term ...................................... 42 5.3. Discretization of the Transport Equations in the K-E Turbulence Model, and in the Suspended load Transport Model ........................................................... 43 5.3.1. Depth-averaged Transport Equation ................................................ 43 5.4. Discretization of the Exner Conservation of Sediment Equation ................. 45 5.4.1. Non-erodible bed layer .................................................................. 46 5.5. Managing Wet-Dry Fronts .................................................................... 47 Page 2 of 56
Iber. Turbulent Flow in Shallow Waters
6. REFERENCES ......................................................................................... 50 7. NOMENCLATURE ................................................................................... 52
Page 3 of 56
Iber. Turbulent Flow in Shallow Waters
IBER Turbulent Flow in Shallow Waters
PRESENTATION
Page 4 of 56
Iber. Turbulent Flow in Shallow Waters
1. PRESENTATION Iber is a numerical model for simulating turbulent free surface unsteady flow and environmental processes in river hydraulics. The ranges of application of Iber cover river hydrodynamics, dam‐break simulation, flood zones evaluation, sediment transport calculation and wave flow in estuaries. At this moment Iber has 3 main computational modules: a hydrodynamic module, a turbulence module and a sediment transport module. All of them work in a finite volume non‐structured mesh made up of triangle or quadrilateral elements. In the hydrodynamic module, which constitutes the base of Iber, the depth averaged two‐dimensional shallow water equations are solved (2D Saint‐Venant Equations). The turbulent module allows including turbulent stressess in the hydrodynamic calculation, in this way allowing its use for different turbulent models for shallow waters at different degrees of complexity. The most recent version of Iber also includes a parabolic model, a mixing length model and a k‐ε model. The sediment transport module solves the bedload and the turbulent suspended load transport equations, based on the sediment mass balance evolution at the bed. This manual describes in detail the equations and models used in the different Iber computational modules, including the used numerical schemes. In order to cite the Iber model in technical documents and publications the following reference should be used: Bladé, E., Cea, L., Corestein, G., Escolano, E., Puertas, J., Vázquez‐Cendón, M.E., Dolz, J., Coll, A. (2014). "Iber: herramienta de simulación numérica del flujo en ríos". Revista Internacional de Métodos Numéricos para Cálculo y Diseño en Ingeniería, Vol.30(1) pp.1‐10
Page 5 of 56
Iber. Turbulent Flow in Shallow Waters
IBER Turbulent Flow in Shallow Waters
HYDRODYNAMIC MODULE
Page 6 of 56
Iber. Turbulent Flow in Shallow Waters
2. HYDRODYNAMIC MODULE
2.1. Introduction The hydrodynamic module solves the depth averaged Shallow Water Equations (2D‐SWE), also known as the two‐dimensional St. Venant Equations. These equations consider a hydrostatic pressure distribution and a relatively uniform distribution of the depth velocity. The hydrostatic pressure hypothesis reasonably complies in river flows and estuary wave currents. Likewise, the hypothesis of uniform depth velocity distribution usually complies in rivers and estuaries, even though there may be zones in which this requirement is not fulfilled due to local three‐dimensional flows or saline wedges. In these cases it is necessary to study the extension of these zones and its possible repercussion in the model results. At this moment, the numerical models based on the 2D‐SWE are widely used in coastal and river dynamics studies, flood zones evaluation and sediment and contamination transport calculations.
2.2. Hydrodynamic Equations The hydrodynamic module solves the conservation of mass and momentum equations in the two horizontal directions.
h hU x hU y MS t x y e Zs τ s,x τ b, x g h 2 ρ hU x hU 2x hU x U y hτ exx hτ xy gh 2 Ω sinλ U y MX t x y x x y ρ 2 x
hU y t
hU x U y x
hU 2y y
gh
hτ exy hτ eyy Zs τ s,y τ b, y g h 2 ρ 2 Ω sinλ U x MY y x y ρ 2 y
Where h is depth, Ux, Uy are the depth averaged horizontal velocities, g is the gravity acceleration, Z is the free layer elevation, τs is the free surface friction due to wind induced friction, τb is the bed friction, ρ is the water density, Ω is the earth’s rotation angular velocity, λ is the latitude of the studied point, τexx, τexy, τeyy are the effective horizontal tangential stressess, and Ms, Mx, My are respectively the terms of: mass source/drain and momentum, which are used to model precipitation, infiltration and drainage. The following terms are included in the hydrodynamic equations: Hydrostatic pressure Bed Slope Page 7 of 56
Iber. Turbulent Flow in Shallow Waters
Viscous and Turbulent tangential stresses Bed Friction Superficial friction due to wind Rainfall (Precipitation) Infiltration Likewise, wet/dry fronts, both stationary and non‐stationary, that may appear in the area of study are modelled. These fronts are fundamental in the fluvial and coastal flood zones modelling (rivers and estuaries). In this way the possibility to evaluate the extension of flood zones in rivers is introduced, as is the movement of wave fronts in estuaries and coastal zones.
2.3. Bed Friction The bed wields a friction force over the fluid that is equivalent to the friction with a wall, even though, in general the bed’s roughness has a greater value in the case of rivers and estuaries. Bed friction has a double effect on the flow equations. On one side it produces a friction force that opposes the mean velocity, and on the other side, it produces turbulence. Both effects can be characterized by the friction velocity uf, which is a form of expressing the bed’s tangential stress in velocity units:
uf
τb ρ
Where τb is the bed friction force module and ρ is the water density. In depth averaged models it’s not possible to calculate friction velocity by means of standard wall functions, as it is done in wall type boundaries, because these equations are solved in a vertical direction. This means that it’s necessary to relate friction velocity uf with depth averaged velocity by means of a friction coefficient. The bed stress can be expressed as:
τ b ρu f2 ρCf U
2
Where Cf is the bed friction coefficient. There are different expressions to approximate this coefficient. Most of them assume uniform channel flow in a logarithmic profile of depth velocities.
Page 8 of 56
Iber. Turbulent Flow in Shallow Waters
Unlike 1D models, in 2D models the hydraulic radius is not defined as the wet section area over the wet perimeter, since in 2D models it makes no sense to define a transversal section. Taking a column of fluid of length Δx and depth h, the hydraulic radius would be calculated as:
Rh
A h Δx h Pm Δx
Meaning that in 2D models hydraulic radius and depth are equivalent. The bed friction is evaluated using the Manning formula, which uses the Manning n coefficient as parameter. The Manning formula uses the following roughness coefficient:
n2 C f g 1/3 h
2.4. Superficial Friction due to Wind The friction force due to wind over the free surface can be calculated considering the wind velocity at 10 meters of height and a dragging coefficient, using the Van Dorn equation (Van Dorn, 1953).
τs ρ Cvd V10 2 Where ρ is the water density, V10 is the wind velocity at 10 meters of height and Cvd is the surface drag coefficient. By default a drag coefficient of Cvd =2.5 10‐6 is considered.
2.5. Effective Stresses The horizontal effective stresses that appear in the hydrodynamic equations include the effects of the viscous stresses, of the turbulent stresses and the dispersion terms due to the non‐homogeneity of the depth velocity profile.
τ ije τ ijv u'i u' j D ij v
Where τij are the viscous stresses, u' i u' j are the turbulent stresses (also called the Reynolds stresses), and Dij are the lateral dispersion terms:
D ij
1 Zs U i u i U j u j dz h Z b
Page 9 of 56
Iber. Turbulent Flow in Shallow Waters
The dispersion terms are not considered in the 2D‐SWE equations (remember the hypothesis of uniform depth velocity profile), because it is impossible to calculate them in a general way with a depth averaged model. Their importance will be greater as less uniform is the depth velocity profile. A typical situation in which these terms can gain importance is in channels with elbows or small curvature radii, like in confluence channels (Figure 1)
Q1
Q
Q
3
2
Figure 1. Secondary flows (left) and depth velocity profile (right). Main causes of the dispersion terms. The viscous stresses are calculated from the fluid’s kinematic viscosity, ν as:
U U j τ ijv ν i x x i j In general, the order of magnitude of the viscous stresses is much lower than the rest of the terms that appear in the hydrodynamic equations, except near the walls and in laminar flow. The turbulent stresses are much larger in magnitude than the viscous stresses, especially in recirculation zones, where turbulence production is elevated. In the case of two‐dimensional shallow equations the turbulent stresses constitute three new unknowns to calculate, added to the depth and velocities Ux, Uy they mean a total of 6 unknowns. This is what is known as turbulence closure problem, because it is necessary to solve 3 equations with 6 unknowns. Due to this, it is necessary to use a turbulence model that allows calculating these turbulent stresses. Most of the models calculate the turbulent diffusive terms from the following expression:
u'i u' j x j
U i νt x j x j
Where ν t is the turbulent viscosity calculated in the turbulence module. The problem lies in the non‐ existence of a universal turbulence model, that may allow calculating in a precise form the turbulent stresses. Through time some models have been developed with different degrees of complexity. The Boussinesq formulation is used by all turbulence modules in Iber. Page 10 of 56
Iber. Turbulent Flow in Shallow Waters
2.6. Hydrodynamic boundary Conditions In a two‐dimensional problem it is necessary to distinguish two types of boundaries: open and closed. The closed boundaries, also called wall‐type boundaries, are water‐resistant, meaning that they don’t allow water through them.
2.6.1. Closed Boundaries The presence of a wall‐type boundary generates a lateral friction force in the fluid in a similar fashion than the friction exerted by the bed’s roughness. At a closed boundary the following conditions can be used: Slip condition (null tangential stress) Wall friction Condition (wall functions) The slip condition is equivalent to undermining the friction stress generated by the wall‐type boundaries over the fluid. In hydraulic engineering, and especially in river engineering, the surface contact with lateral boundaries is much lower than the surface contact with the bed, due to the separation between horizontal and vertical scales, so the friction forces in the wall boundaries can be undermined. In this case the slip condition would hold. In problems where the horizontal and vertical dimension are similar (narrow section channels) this roughness force can have some importance in the flow development, even though in general the influence is small. If you want to consider the lateral friction, a wall‐type friction can be introduced; this consists in imposing a tangential force in the opposite direction to the flow. In this case, Iber distinguishes between smooth turbulent regime and rough turbulent regime in function of the wall’s roughness and the flow velocity close to the wall. The wall friction velocity (u*) is defined in function of the wall friction ( τ w ) as:
u*
τw ρ
The tangential velocity at the wall can be expressed as a function of the friction velocity, the roughness’s depth and the wall distance as:
u
u* Ln E y κ
y
y u* ν
Page 11 of 56
Iber. Turbulent Flow in Shallow Waters
Where y is the perpendicular distance to the wall, and E is a parameter which value depends on the flow characteristics. To calculate E, Iber considers smooth turbulent flow conditions, rough turbulent flow conditions and a transition between smooth and rough (Table 1).
KS
Regimen type
Smooth Turbulent
K S 5
Rough Turbulent
KSu* ν
u
u* Ln E y κ
E 9.0
5 < KS 70
E=
30 KS
KS 70
E=
1 0.11 + 0.033 K S
Smooth‐Rough Transition
Table 1. Wall Friction. The turbulence is defined as smooth when this relationship complies:
KS
KSu* 5 ν
Where Ks is the wall’s roughness height, a measure of the wall’s roughness, with length units. In this condition the tangential velocity at the wall can be expressed as a function of friction velocity and the kinematic viscosity, as:
u
u* yu Ln 9.0 * κ
Rough turbulent regime is applied when this relationship complies:
K S
K Su* 70 ν
In this condition, the tangential velocity at the wall can be expressed as a function of the friction velocity and the bed’s roughness’s height, as
u
u* y Ln 30 κ KS
In the transition between regimes, the tangential velocity at the wall can be expressed as a function of friction’s velocity, kinematic viscosity and roughness’s height, as:
Page 12 of 56
Iber. Turbulent Flow in Shallow Waters
u y u * Ln κ 0.11 ν + 0.033 K S u* 2.6.2. Open Boundaries In open bondaries, different types of boundary conditions can be introduced. So that the 2D‐SWE equations are correctly presented from a mathematical point of view, the number of conditions to use in the open boundaries will depend if the boundary is an inlet or an outlet, it will also depend on the type of regime at the boundary (fast/slow). In an inlet boundary it is necessary to introduce 3 boundary conditions if the regime is supercritical (one for each of the 3 St. Venant equations), while if the regime is subcritical, it is enough to present 2 conditions. In an outlet boundary, it is enough to introduce a single condition if the regime is subcritical; meanwhile, it is not necessary to establish any condition if the regime is supercritical. If the user introduces fewer conditions than those necessary, from a mathematical point of view the equations will be undetermined and no correct solution will be found. The conditions which can be entered by the user are height, velocity components, or a combination of both. In Iber different options are considered to introduce boundary conditions, as explained in Table 2. It is common in river hydraulics for the flow to be in subcritical regime in the contours of the modeled section. In this case, what is often done is to introduce free surface depth or water level in the downstream boundary. In the upstream boundary the most common is to enter a total inflow (in cubic meters per second cms) in the direction of the flow, which in general, lacking more precise data, is assumed to be perpendicular to the inlet boundary. It is also possible to introduce upstream velocities (in m/s) or specific flow (m2/s), though this occurs less often. In the case of total flow in the inlet, a unit flow distribution (m2/s) is entered, according to the following expression:
qn
h 5/3 5/3 h dy
Q
Where qn is the normal specific discharge (m2/s) at each point of the inlet, and Q is the total inlet flow at mentioned boundary. The integral in the denominator is extended along the length of the boundary. Besides height, other possibilities can be considered at the outlet, such as weirs or rating curves. The weir condition establishes a relationship between outflow and height at each point of the contour:
q = Cd (ZS ZW )1.5 Page 13 of 56
Iber. Turbulent Flow in Shallow Waters
Where Cd is the weir’s discharge coefficient, Zs is the water level upstream of the weir and Zw is the water level on top of the weir. The user must introduce Cd and Zw. The rating curve contour condition establishes a general relationship between the outflow and the water level en each contour point. This relationship is introduced by the user using a table defining values of specific discharge and water level. The set of conditions implemented in Iber in open contours are shown in Table 2. Boundary
Regime
Boundary conditions
Subcritical / Critical
Total discharge in normal direction to the boundary
Supercritical
Total discharge in normal direction to the boundary and average velocity
Subcritical / Critical
Specific discharge in normal direction to the boundary
Discharge
Inlet
a) Specific discharge in normal direction to the boundary and water depth
Unit discharge Supercritical
b) Specific discharge in normal direction to the boundary and water surface elevation a) Water depth b) Water surface elevation
Subcritical c) Weir (elevation and discharge coefficient)
Outlet
d) Rating curve Supercritical / Critical
Not necessary to introduce any condition
Table 2. Contour Conditions implemented in open contours.
2.7. Internal boundary conditions The internal conditions are used to model hydraulic structures like gates, weirs or bridges which change the conditions of the system.
Page 14 of 56
Iber. Turbulent Flow in Shallow Waters
The internal boundary conditions implemented in Iber can be used to model the following flow conditions: Flow below a gate Flow over a free falling weir Weir‐Gate combination Local loss
2.7.1. Gates The equation used is the gate drainage, which works on free or submerged flow. The data to insert are the discharge coefficient, the gate’s bed level, the gate opening and its width. By default, the discharge coefficient will have a value of Cd=0.6.
ZU ZD
h ZB Figure 2. Scheme and equations of the gate internal contour condition.
(ZD ZB ) / (ZU ZB )
Discharge Equations
Free Flow
0.00 – 0.67
Q = Cd B h 2g (Z U ZB )
Transition
0.67 – 0.80
Q = Cd B h 6g (ZU ZD )
Submerged Flow
0.80 – 1.00
Q = Cd B h 2g (ZU ZD )
2.7.2. Weirs The rectangular weir discharge equation is used, which can work for free flow and submerged flow. The data to input are the weir’s upper level, the discharge coefficient and the weir’s length. By default the discharge coefficient has a value of Cd=1.7.
Page 15 of 56
Iber. Turbulent Flow in Shallow Waters
ZU Zw
ZD
ZB Figure 3. Scheme and equations of the weir internal contour condition.
(ZD ZW ) / (ZU ZW )
Discharge Equation
Free Flow
0.67
Q = 2.598 Cd B (ZD ZW ) (ZU ZW )0.5
2.7.3. Gate‐Weir Combination In this case the condition includes the two previous ones, so both parameters must be used, the gates and weirs. The total discharged flow is taken as the sum below the gate and the flow above the weir.
ZU Zw
ZD
Q Qcompuerta Qvertedero h ZB Figure 4. Scheme and equations of the gate‐weir internal contour condition.
2.7.4. Local Losses In this case, there is a local energy loss in the flow transference between two finite volumes of value H=
v2/2g. The St. Venant Equations are the mathematical expressions of the conservation of mass
and momentum, a way to consider this loss of energy is over the friction slope. For this, an additional term is added to the friction slope, Sf, through the edge of the finite volume, this added term is equal to H/V, where V is the element’s volume. In this way, the energy loss through the two elements will be
Page 16 of 56
Iber. Turbulent Flow in Shallow Waters
equal to H+ Sf∙L, where L is the distance between element’s centre on both sides of the boundary where the local loss is implemented.
v2 H 2g H S 'f S f V
Figure 5. Scheme and equations of the local loss internal boundary condition.
2.8. Infiltration To simulate rainfall processes, it can be necessary to consider water infiltration in non‐saturated terrain for calculating runoff. Modeling infiltration of surface waters is especially important in rainfall‐runoff models. Infiltration is considered in the model using a negative term in the conservation of mass equation (loss of water):
h hU x hU y =-i t x y Where i is the real infiltration rate, measured as the minimum between the potential infiltration rate f (soils infiltration capacity at every instant, which depends on soil characteristics and conditions), and the amount of readily available surface water for infiltration.
i = min ( f ,
h ) Δt
To calculate the potential infiltration 3 infiltration common models are used: The Green‐Ampt model, the Horton model and linear model.
2.8.1. Green‐Ampt The infiltration rate, expressed in m/s, is calculated in each cell using the Green‐Ampt formula (Chow, Maidment, & Mays, 1988), in it; it is assumed that there exists a saturated front that separates the saturated soil region, immediately below the terrain, and the non‐saturated region, where there is suction.
Page 17 of 56
Iber. Turbulent Flow in Shallow Waters
As the water infiltrates, the saturated front descends and the depth of the saturated region L becomes bigger. The infiltration rate f is calculated as:
h Ψ Δθ f k s 1 L0 Δθ F
t
F f dt 0
L L0 +
F Δθ
Δθ θi
Where ks the saturated permeability of the soil, h is the depth, ψ is the suction in the non‐saturated soil region, Δθ the humidity content change of the soil as the saturated front increases, θi the soil’s initial humidity content, the soil’s total porosity, and L is the depth of the saturated region. The infiltration rate is equal to the potential infiltration rate as long as there is enough surface water to infiltrate. The parameters to introduce by the user for this module are:
The soil’s saturated permeability (ks)
The non‐saturated region’s suction (ψ)
Soil’s effective porosity (drainable) (θe)
Soil’s initial effective saturation (Se), defined as:
Se =
θi - θ r θe
Where θ r is the soil’s retention capacity (non‐drainable humidity) and θi the soil’s initial humidity. The soil’s porosity is equal to the drainable porosity plus the soil’s retention capacity (
= θe θr
).
Starting with the effective porosity and the soil’s initial effective saturation, the change in humidity content can be calculated as the saturation front advances with the formula:
Δθ = - θi = - θr - θe Se θe 1 Se All the parameters of the Green‐Ampt equation can be introduced as spatial variables (different for each element of the mesh).
2.8.2. Horton In the Horton model, the potential infiltration rate is calculated as:
f f c f 0 - f c exp k t
Page 18 of 56
Iber. Turbulent Flow in Shallow Waters
Where t is time, starting with the beginning of the precipitation. The user must introduce as model parameters an initial infiltration rate (݂ ), the infiltration rate at infinite time (݂ ) and the constant k, which defines the temporal variation of the potential infiltration rate. All the parameters of the Horton infiltration equation are spatially variable (calculated for each element of the mesh).
2.8.3. Linear The linear model considers an initial abstraction P0 (volume per unit of area), and afterwards some continuous constant losses (volume per unit of area and unit of time). The value of the initial abstraction and the continuous losses can vary per element.
Figure 6. Temporal evolution of the infiltration rate according to the linear model
2.9. Initial Abstraction If the Green‐Ampt or Linear infiltration models are used to calculate the infiltration losses, there is also the possibility to consider an initial abstraction. This initial abstraction can represent other processes like the canopy retention (because of vegetation) or land depressions, or the infiltration capacity of dry soils with an elevated porosity. The initial abstraction is defined as a volume per unit of area, and as that has length unit. This value is subtracted from the water that reaches the soil, being from rainfall or from surface runoff. That way it can be used either in areas with or without rainfall.
Page 19 of 56
Iber. Turbulent Flow in Shallow Waters
IBER Turbulent Flow in Shallow Waters
TURBULENCE MODULE
Page 20 of 56
Iber. Turbulent Flow in Shallow Waters
3. TURBULENCE MODULE
3.1. Introduction A great number of studies in hydraulic engineering implies the analysis of open channel flows, many of which can be considered shallow flows, understanding this as low depth with small vertical‐horizontal relationship. Practically all open channel flows are turbulent. While looking at a river, small swirls appear and disappear with an almost chaotic movement, showing the complexity of the turbulent motion. These swirls are the main cause of mixing processes, so they play an important role in the diffusion of soluble substances and suspended solids, among others. Despite the fact that all flows in river engineering are turbulent, in many cases the turbulence is not large enough to have a notorious influence in the mean velocity. Such is the case of rivers, estuaries and in general in coastal zones with a smooth enough geometry so that no recirculation zones appear. Nevertheless, even in this type of situations it’s important to consider turbulence while modelling these water systems, since it plays an important role in the processes of transport and mixing of contaminants and sediments. The diffusion of heat, of a solute, or suspended sediment is produced basically because of turbulence, except in laminar flow, which generally does not occur in hydraulic engineering, even less in rivers and estuaries. The turbulent diffusion coefficient is several times higher in magnitude than the molecular diffusion coefficient. Because of this, it is necessary to evaluate previously the turbulent kinetic energy to calculate the diffusive flow. One of the main characteristics of Iber is the inclusion of diverse RANS type turbulence models, which are solved in the turbulence module. The following turbulence models are included for shallow waters, enumerated by order of complexity: Constant Turbulent Viscosity Parabolic Model Mixing Length Model k‐ε (ka‐epsilon) model from Rastogi ‐ Rodi (Rastogi y Rodi, 1978) Using models with different levels of complexity allows selecting the most adequate for each case study, considering the complexity of the flow and the model. In general the mixing length model provides satisfactory results in rivers and estuaries, this way discouraging the use of other turbulence models in these cases. In hydraulic structures such as open flow channels with marked elbows and recirculation
Page 21 of 56
Iber. Turbulent Flow in Shallow Waters
zones, it may be necessary to use at least one mixing length model and one k‐ε model to test the results. The choice of the turbulence model that best fits each case is based on user experience, considering that the more complex the model is, the longer it takes to calculate and more complex the equation solution is. The objective of the turbulence models is to calculate the Reynolds stresses. Models based on the Boussinesq hypothesis (the ones enumerated before, used in Iber), the Reynolds stresses are evaluated from the following expression:
U i U j 2 k δ ij uiu j ν t x j x i 3 The turbulence model calculates the turbulent viscosity to be used in the previous expression.
3.2. Scales of Turbulence in Shallow Waters One of the main characteristics of shallow flows is the split‐up between horizontal and vertical scales, this happens because the vertical extension of the fluid (limited by depth) is much smaller than the horizontal extension. This scale split‐up is applicable in the spatial dimension, velocities, and of course the turbulence. Speaking of turbulence, its main effect is the fragmentation between tri‐dimensional (swirls) and bi‐dimensional turbulent structures. The 3D turbulence spatial scale is limited by its depth, and thus its structures are much smaller than those associated to the 2D turbulence, which are only limited by the horizontal scale. The 3D turbulence is originated mainly by the bed friction, while the origins of the 2D turbulence are the horizontal velocity gradients It’s important for the turbulence model to include not only the effects of the 3D turbulence, formed by the bed friction, but also the 2D turbulence, formed by the horizontal velocity gradients. In the shallow water models, the bi‐dimensional characteristics of the flow are implicitly considered in the transport equations when considering a homogenous depth velocity profile, while the tri‐dimensional causes of turbulence are usually included using another term which depends on the bed tangential stress. In the same way, even when considering 3D‐SWE, the turbulence model should consider turbulence anisotropy in the horizontal and vertical directions. Iber uses the following turbulence models. All of them are shallow water depth averaged turbulence models.
Page 22 of 56
Iber. Turbulent Flow in Shallow Waters
3.3. Constant Turbulent Viscosity The order of magnitude of the turbulent viscosity can approximated depending on the considered flux. Several published articles propose these turbulent viscosity approximated values in function of the considered flux. This focus is very simple, and cannot be considered as an adequate, nor realistic in any case, turbulence model, since it does not consider spatial changes of turbulent viscosity. It’s important to highlight the importance of spatially variable turbulence viscosity to the mean velocities. Furthermore, the existing tables only provide approximated values. Because of these reasons, the use of this method is discouraged, since it can lead to excessively large errors, generally coming from the use of large values of turbulent viscosity and from not considering its spatial variability. Another important downfall of the method is produced while modeling non‐stationary flow regimes, since in these cases; turbulence varies through space and time.
3.4. Parabolic Profile of turbulent Viscosity This model assumes a parabolic distribution of turbulent viscosity in depth, this way considering a mean viscosity in depth which can be calculated from the following expression:
ν t 0.068 u f h Where h is the total depth and uf is the bed friction velocity, calculated from the bed tangential stress, from the expression:
uf
τf ρ
If the Manning formula is used to calculate the bed friction, the following formula for turbulent viscosity can be used:
ν t 0.068 g n U h 5/6 This means that the turbulent viscosity depends on the local depth, the depth averaged velocity and the manning coefficient. For better results, a coefficient is introduced to adjust the turbulent viscosity, which can be handled by the user.
ν t C m 0.068 g n U h 5/6
Page 23 of 56
Iber. Turbulent Flow in Shallow Waters
3.5. Mixing Length Model In the shallow waters mixing length model, the turbulent viscosity is calculated based on the local characteristics of the flow using the following expression: 2
ν t min 0.267 κ h, κ d wall
2
u 2S ij S ij 2.34 f κh
Where κ=0.41 is the von Karman constant. It’s a relatively simple algebraic model that produces acceptable results in flows where the turbulence is locally generated mainly from bed friction. It considers the turbulence caused by horizontal velocity gradients, but it does not consider convective transport or turbulence dissipation. Using the model in flows with strong recirculation zones, the results have large errors.
3.6. The Rastogi and Rodi K‐E model (Rastogi & Rodi, 1978) In this model k represents the turbulent kinetic energy and ε the rate of dissipation of the turbulent energy. The model considers turbulence caused by bed friction, velocity gradients, dissipation and convective transport. It solves the following shallow water equations for k‐ε model:
k U x k U y k x y x j t ε U x ε U y ε y x j t x
ν t cμ
k2 ε
ν ν t k σ k x j
ν ν t σε
c k c f1/2
ε x j
3 2ν t S ij S ij c k u f ε h
4 2 c ε1 ε 2ν t S ij S ij c ε u f c ε2 ε k k h2
1/2 c ε 3.6c 3/2 k c ε2 c μ
cf
τb 1 ρ U2
With the following constants:
c μ 0.09
c ε1 1.44
c ε2 1.92
σ k 1.0
σ ε 1.31
Where k is the kinetic turbulent energy, ε is the turbulence dissipation rate, Sij is the deformation sensor. The terms which include the bed friction velocity uf are responsible for modeling the forming of turbulence due to bed friction. The k‐ε model is relatively sophisticated. The results for shallow water turbulent flows are relatively good, being one of the most used models, whenever turbulence levels are important. Nevertheless, its Page 24 of 56
Iber. Turbulent Flow in Shallow Waters
level of complexity does not guarantee correct results for all types of flows. Like any other turbulence model, the results from the k‐ε model must be critically analyzed, for this, user experience in turbulent flow is essential.
3.7. Dimensional Analysis of the turbulent terms in the shallow water equations If the shallow water equations are turned dimensionless, the following non‐dimensional numbers are obtained:
F
U gh
T
1 H Cf L
Rl
UL ν
Rt
UL νt
Which respectively refer to the relationship between water mass inertia (convective forces) and pressure forces (F), bed friction force (T), laminar tangential stresses (Rl) and turbulent tangential stresses (Rt). The relative importance of these associated processes to each dimensionless number is inversely proportional to the value of each number. For example, the larger a dimensionless number is, the less important it will be for the process it represents. That way, for a large laminar Reynolds value, the flow is turbulent and thus the laminar forces loose importance in the flow development. Likewise, the importance of turbulent stresses in the mean velocity will depend on the value of the turbulent Reynolds number, which depends on the turbulent viscosity. A parabolic model can be used to estimate the order of magnitude of the turbulent viscosity, with the following expression:
νt
1 1 κ u f h κ g n h 5/6 U 0.21 n h 5/6 U 6 6
The Manning formula has been used to estimate the bed friction velocity uf. This estimation will be more precise in cases where the turbulence is generated mainly by bed friction, as is the case in rivers, and will provide some errors in cases where the turbulence is originated from horizontal shear stress, for example in cases of recirculation zones. In any case, using this approximation, the turbulent Reynolds number can be expressed as:
Rt
UL 4.8 L νt n h 5/6
This expression can be used as a first evaluation of the importance of turbulent stress in the velocity and depth. For example, if we model a river section with a depth of 10m, a width of 400m, a manning coefficient of 0.025 and a mean velocity of 0.5m/s, the approximate turbulent viscosity is 0.02m2/s and
Page 25 of 56
Iber. Turbulent Flow in Shallow Waters
the turbulent Reynolds number is close to Rt ~ 11000, which is a pretty large value, so the turbulent stresses are expected to have a small effect on the mean flow. According to the expression for Tb, the importance of bed friction grows in shallow water flows, and looses importance as the relationship between depth and the horizontal dimension grows.
Page 26 of 56
Iber. Turbulent Flow in Shallow Waters
IBER Turbulent Flow in Shallow Waters
SEDIMENT TRANSPORT MODULE
Page 27 of 56
Iber. Turbulent Flow in Shallow Waters
4. NON‐STATIONARY SEDIMENT TRANSPORT MODULE The sediment transport module solves the non‐cohesive sediment non‐stationary transport equations. The equations include the bedload transport equations and the suspended sediment transport equations, coupling the bedload and the suspended load through a sedimentation‐rise term. The sediment transport module uses the velocity, depth and turbulence fields from the hydrodynamic and turbulence modules. The bedload is calculated using an empirical formula, chosen from the Meyer‐Peter Muller or the Van Rijn Methods. The suspended load transport is modeled from a depth averaged turbulent transport equation. Hydrodynamic + Turbulence
Bed-Load
Suspended Load
Sediment Conservation
Bed variations
Figure 7. Non‐Stationary sediment transport scheme.
4.1. Sediment Conservation Equations The bed level variation is calculated with the Exner sediment conservation equation:
1 p
Z b q sb,x q sb,y D -E t x y
Where p is the porosity of the sediment that forms the bed layer, Zb is the bed level, qsb,x and qsb,y are the two solid flow components. The difference D‐E is the balance between the bedload and suspended load.
Page 28 of 56
Iber. Turbulent Flow in Shallow Waters
4.2. Bed Transport 4.2.1. Stress Partition The total bedload stress in the river bed is generated by the grain’s roughness (which is proportional to the sediment diameter) and the shape of the bed layer (ripples, dunes, antidunes). Only the grain stress contributes the bedload movement. Thus, before calculating the bed sediment flow it is necessary to estimate this grain stress. To do this, Iber uses the Einstein formula based on the total stress: 1.5
n τ = τ s n * bs
* b
ns
K1/6 s (m)
K s 2 3 D s
25
Where n is the total Manning coefficient, ns is the equivalent Manning coefficient due to the grain, Ds is the sediment’s diameter, Ks the grain’s roughness’s height (measured from the grain’s diameter), τ b is * * the bed total stress, τ bs is the bed stress due to the grain, τ b , τ bs are dimensionless total and grain
stress, calculated with:
τ*b =
τb ρs ρ g Ds
τ*bs =
τ bs ρs ρ g Ds
Where ρs is the sediment’s density, and ρ is the water density. Iber uses a value of Ks 2.5 Ds
4.2.2. Bedload Discharge The Bedload discharge is calculated based on empirical formulas. In the latest version of the model, two well regarded and well known formulas are used: Meyer‐Peter Müller Van Rijn Meyer‐Peter Müller (Meyer‐Peter & Muller, 1948) The original Meyer‐Peter & Müller equation, deduced from gravel beds of diameters up to 30mm, calculates the Bedload discharge with the following expression:
q*sb = 8 τ*bs - τ*c = 8 τ*bs - τ*c 3/2
3/2
Where the dimensionless discharge is calculated as: Page 29 of 56
Iber. Turbulent Flow in Shallow Waters
q*sb
q sb
ρs 3 ρ 1 g Ds
In the case of flat beds, a dimensionless critical bed stress is considered of c
0.047 . On the
contrary, it is necessary to make a bed slope correction. Detailed in section 4.2.3 Other corrections have been made to this formula, (Wong, M, 2003), (Wong & Parker, 2006), resulting in:
q*sb = 3.97 τ*bs - τ*c
3/2
If the bed is flat, c 0.0495 , is considered. On the contrary, it is necessary to make a bed slope
adjustment, detailed in section 4.2.3. This corrected version is included in Iber. Van Rijn (van Rijn, L.C., 1984) In the van Rijn Formula the bedload discharge is calculated from the following expressions:
T 2.1 T 0.3 q 0.053 0.3 D* * sb
T1.5 T 0.3 q 0.100 0.3 D*
* sb
Where T is a dimensionless parameter that measures the excess of bed friction above a critical value which defines the movement threshold.
τ*bs - τ*c T τ*c
The dimensionless diameter is defined as: 1/3
gR D* Ds 2 ν
con R
γs -γ γ
Page 30 of 56
Iber. Turbulent Flow in Shallow Waters
4.2.3. Bed slope correction Whenever the bed is not flat, the equations must be corrected to include the effects of gravity, either to increase the bedload discharge in a positive slope, or to reduce it in an adverse slope. The correcting factor is applied over the critical stress of movement threshold, as detailed by Apsley and Stansby (Apsley & Stansby, 2008), where they present and generalize the works of many other authors, citing at least (Dey, 2003) and (Wu, 2004). To consider the bed slope both in the movement threshold and in the bedload discharge, the sediment’s weight, due to the bed’s slope, is combined in a vector form with the bed stress to get an effective stress. Si b is a unit vector in the direction of the maximum slope; the dimensionless effective stress is defined as:
τ*bs,eff = τ*bs + Do sinβ b Where is the maximum slope, with the horizon, and D0 is a parameter depending on the particle shape. So that the movement begins in the absence of flow, whenever is equal to the internal friction angle of the material ( ), the parameter D0 is defined as:
Do
τ*c,0 tan
*
Where τ c,0 is the dimensionless critical stress for the flat bed. On the other side, the effective critical stress is reduced proportionally to the gravity component normal to the bed slope.
τ*eff,crit = τ*c,0 cosβ Where
τ*c,0 is the same as before. Starting at this point, the bedload discharge equations presented
above are used, substituting the (critical and bed) stresses by effective stresses, and getting the sediment discharge, which is a function of the fluid’s stress and the bed slope in each of the x, y directions. The formulas used are entirely in vector form for the bedload discharge, and thus able to consider any flow direction with respect to the maximum slope
Page 31 of 56
Iber. Turbulent Flow in Shallow Waters
4.2.4. Avalanche model Apsley and Stansby (Apsley & Stansby, 2008) also proposed including an avalanche sliding model to avoid those slopes higher than the bed material’s friction angle. For that, if an angle between two finite volumes is higher than the friction angle , then a unit bedload discharge appears from the upstream to the downstream element equal to:
0.5 L2 (tan β tan ) q aval =(1 p) cos β t Where L is the maximum horizontal dimension of the adjacent finite volumes.
4.2.5. Considerations on a non‐erodible layer While calculating bedload dragging and changes in the bed layer section, the possibility to include a rock layer, or non‐erodible surface, below which the bed erosion cannot go further.
4.3. 2D Turbulent Suspended Load Module 4.3.1. Turbulent Suspended Load Equation The suspended load transport equation uses the velocity, depth and turbulence fields calculated in the hydrodynamic and turbulence modules. The suspended load is modelled using a depth‐averaged equation. The equation used in the Iber is:
ν hC hU x C hU y C t t x y x j Sc,t
C Dsx Dsy E D h y x j x
Where C is the depth‐averaged concentration of suspended solids, Ux, Uy are the horizontal depth‐ averaged velocity components, ν t is the turbulent viscosity, Г is the molecular diffusion coefficient for suspended solids, and Sc,t is the Schmidt number, which relates the moment turbulent diffusion coefficient with the suspended turbulent diffusion coefficient. The terms Dsx, Dsy model the suspended sediment dispersion due to the non‐homogeneous vertical velocity profile and sediment concentration. Normally its effect is undermined in 2D shallow water models, despite its importance may be relevant when the concentrations and velocities vary along the depth of an element, as is the case in open channels with slam curvature radii. Page 32 of 56
Iber. Turbulent Flow in Shallow Waters
The terms E and D respectively model those bedload grains which become suspended (entrainment) and deposition of suspended sediments to the bed layer. The difference represents a balance, and thus a coupling of the bedload and suspended load.
4.3.2. Entrainment/Deposition (E‐D) term Calculation Three formulas are implemented for calculating the entrainment/deposition term (E‐D):Van Rijn (van Rijn, 1987), Smith (Smith, J.D., 1977) and Ariathurai and Arulanandan (Ariathurai & Arulanandan, 1978). The first two are valid on sand bed layers, while the Ariathurai is valid for cohesive bed layers. The three formulas are especially recommended in the latest ASCE Manual of Sediment Transport, among them, the most widely used is the Van Rijn Formula. Van Rijn In the Van Rijn formula (van Rijn, 1987), the term E‐D is evaluated from the following expression
E D Ws c*a ca α Ws C* C Where α is a coefficient that relates the mean suspended particle concentration and the river bedload concentration, which value comes from the Rouse profile for the depth sediment concentration distribution, Ws is the sediment grain velocity, C is the depth‐averaged suspended load concentration,
C* is the depth‐averaged suspended load concentration at equilibrium condition (suspended load transport capacity),
c a and c*a are respectively the instantaneous concentration and the equilibrium
concentration at a height z=a above the river’s bed layer, where a is the height of the layer in which the transport is produced (theoretical separation limit between the bedload and suspended load). This height must be approximated from the grain diameter. The coefficient α is calculated from the vertical concentration distribution (Rouse profile), from the following integral:
h-a
α=
h
a
h-z a z h-a
ws
k u *
a = 3 D50 dz
Where κ=0.41 is the von Karman constant. The equilibrium concentration near the bed layer proposed by Van Rijn (van Rijn, 1987) is:
Page 33 of 56
Iber. Turbulent Flow in Shallow Waters
c*a 0.015
a = ks
D50 T1.5 a D*0.3
ks 3 Ds
1/3
gR D* D 2 ν
Smith This formula is similar to the Van Rijn, its only difference lies in the formula to calculate the equilibrium concentration, for which Smith (Smith, J.D., 1977) proposes the following formula:
1.56 103 T c 1 + 2.4 103 T * a
a = 26.3 τ*s - τ*c Ds + k s
k s 3 Ds
Ariathurai y Arulanandan For cohesive soils the formula proposed by Ariathurai and Arulanandan (Ariathurai & Arulanandan, 1978) is used, in it the erosion depends on the difference between the tangential stress and the critical tangential stress to start the erosion ce , it’s also dependant of a value M representative of the erosion rate (equivalent to an erosion rate when τ b =
2 τ ce ):
τ E = M b 1 τ ce In cohesive soils a modification is introduced for calculating D, which considers the deposition critical tangential stress cd . In this case:
D P α Ws C With:
τ P 1 b if τ b < τ cd and P 0 otherwise τ cd
Page 34 of 56
Iber. Turbulent Flow in Shallow Waters
4.3.3. Sedimentation Velocity The sedimentation velocity is calculated based on its diameter (van Rijn, 1987): 2 R g D50 18 υ 10 υ Ws = 1+0.01 D*3 -1 D50
Ws =
Ws =1.1 R g D50
D50 < 10-4 m 10-4 m < D50 0
Page 41 of 56
Iber. Turbulent Flow in Shallow Waters
5.2.2. Discretization of the bed slope source term Iber uses a centred discretization of all the source terms except of the bed slope term. The main cause of using an upwind discretization of the bed slope instead of a centred discretization is that it calculates exactly the hydrostatic solution in an irregular bathymetry, this way avoiding spurious oscillations in the free surface waters and in the velocities. These oscillations are in general small, but they can become larger in magnitude in irregular bathymetry problems, such is the common case in river and coastal hydraulics. Whenever the upwind numerical schemes are used to discretize convective flow, the upwind discretization of the bed slope provides has better properties and provides more precise results than the classic centred schemes. The used discretization of the bed slope source term in a finite volume Ci can be expressed as:
Si = S dA = Ci
S jKi
ij
Where Sij is an upwind discretization of the bed slope source term in each edge of the considered finite volume and is calculated as:
0 | nij | hL hR -1 -1 S ij = -g zb , R zb , L I X | D | D X n x ,ij 2 2 n y ,ij The same as in the convective flux, to implement this discretization in Iber, it is decomposed using the vectors em as: 3
Sij = mem m 1
1 2
1C cij zij | n ij | 2C 0
1 2
3C cij zij | n ij |
Page 42 of 56
Iber. Turbulent Flow in Shallow Waters
5.3. Discretization of the Transport Equations in the K‐E Turbulence Model, and in the Suspended load Transport Model The conservation equations used in the k‐ε turbulence model and in the suspended load transport model can be expressed as:
Fj S t x j Where = C h is the conserved variable, C is the non‐conserved variable, corresponding to the considered model (sediment concentration, kinetic turbulent energy, turbulent dissipation rate), S are the source terms which model generation or destruction processes of the conserved variable, and F is the convective and diffusive flux, which may be expressed as:
Fj C h u j e h
C x j
Where Γe is the effective diffusivity coefficient, which includes the molecular and turbulent diffusion. In general, turbulent diffusion is many times larger than the molecular diffusion, meaning that this term can be omitted. In the k‐ε turbulence model it is necessary to solve 2 transport equations; one due to the turbulent kinetic energy (k) and the other one for the turbulence dissipation rate (ε).
5.3.1. Depth‐averaged Transport Equation To spatially discretize the transport equation using the finite volume method the differential equation is integrated in each cell of the computational mesh. This process is especially advantageous for solving the conservation equations, since they are being solved in an integral form, which allows the formulation of natural conservative methods. Integrating the convection‐diffusion equation in a two‐ dimensional cell we obtain:
C h i
n 1
C h i
Δt
n
A i C h u dA e h C dA Ai
Ai
Ai
E - D
dA
Where Φ=C.h is the average value at the cell and S=E‐D is the source term due to generation/destruction processes of the considered variable. Applying the divergence theorem to the second and third term we obtain:
Page 43 of 56
Iber. Turbulent Flow in Shallow Waters
C h i
n 1
C h i
n
Δt
A i C h u n dL e h C n dL Li
Li
Ai
E - D
dA
Where the integrals around the cell’s area are extended to the cell’s edges. This equation is the conservation equation expressed in an integral form. The integrals that appear in the integral form conservation equation are solved discretely, this way getting:
C h i
n 1
C h i
Δt
n
Ai
C h u n L j K i
ij
ij
h C n L E - D A j K i
ij
ij
i
i
Where the summations are extended to all the edges that form the boundary of the cell. The sub‐index ij identifies the common face of the cells i and j. Each term of the summations represent a flux of the considered variable that enters the cell through the corresponding face, that way the sum of all the fluxes through all the faces that form the contour of the cess are equal to the balance of what goes out minus what goes in, for example; the net flux outside the cell. In function of the used interpolation to calculate the flux through the cell’s contours, especially the convective flux, different numerical schemes can be found. The diffusive flux between cells is calculated using a centred discretization of second order, without numerical stability problems. The convective flux discretization is more problematic from the numerical stability point of view. In the next paragraphs some implemented numerical schemes are presented in the code to discretize these terms. First Order Upwind scheme In the upwind schemes the direction of the information flow is considered to make the discretization of the convective flux. The diffusive flux is discretized using a centred scheme of second order. In the case of the convective flux, the information travels in the same direction as the water velocity, for example upstream to downstream. The simplest upwind method is the first order upwind scheme, calculated using:
C h u n ij u n ij C h ij un,ij C h ij The velocity is discretized in a centred manner, while the value of C.h is discretized in an upwind manner, considering the upstream node value:
Page 44 of 56
Iber. Turbulent Flow in Shallow Waters
u n ,ij α u n ,i (1 α) u n ,j
C h ij C h i C h ij C h j
si u n ,ij 0 si u n ,ij 0
Where α is a linear interpolation coefficient. For an equidistant discretization the value of α is 0.5. The fact of upwinding the convective term in the first order scheme is equivalent from the mathematical point of view as adding a diffusive term (usually called numerical diffusion) with a diffusivity (numerical) coefficient Γn proportional to the mesh size. The numerical diffusion is a consequence of the used technique for numerical stabilization. The desired situation in a numerical scheme is for it to be stable with the introduction of as little as possible numerical diffusion. This is why it’s so convenient to use fine meshes to reduce the error introduced by numerical diffusion Second Order Upwind scheme The previously mentioned scheme is of first order because the numerical diffusion introduced in the discretization of the convective flux. Despite this it is widely used in Computational Fluid Dynamic (CFD) codes as the default scheme, due to its numerical stability. When more precision is required in a regular size mesh, it is necessary to appeal to higher order schemes. To do this a second order scheme of type MUSCL (Monotonic Upstream Scheme for Conservative Laws) has been implemented Iber, which makes a linear reconstruction of each mesh element of non‐conserved variable. Once this linear reconstruction of the non‐conserved variable is made, the convective flux is calculated at each edge of the computational mesh with:
C h u n ij u n ij C h ij un,ij C h ij u n ,ij α u n ,i (1 α) u n ,j
C h ij C h Ij C h ij C h iJ
si u n ,ij 0
si u n ,ij 0
Where C h I j and C h i J the values of the variable at edge Lij taken from the linear reconstruction of cells Ci y Cj respectively.
5.4. Discretization of the Exner Conservation of Sediment Equation The Exner equation of conservation of sediment can be as: Page 45 of 56
Iber. Turbulent Flow in Shallow Waters
(1 p)
z b q sb,x q sb,y DE t x y
The Exner equation is discretized in a similar way as the suspended load sediment conservation equation. The integral of the Exner equation in each two‐dimensional cell can be written as:
(1 p)
n 1 n zb,i zb,i
Δt
q sb,y q A i sb,x + Ai y x
dA D - E i A i
Applying the divergence theorem to the second term of the equation, we get:
(1 p)
n 1 n zb,i zb,i
Δt
Ai
q n L D - E A jK i
* sb
ij
ij
i
i
The value of the bedload in each edge of the mesh q *sb is calculated in an upwind form as: ij
q q
* sb ij
q sb,i
si q*sb n 0
* sb ij
q sb,j
si q n 0
ij
* sb
ij
The erosion/deposition source term is discretized in a centred form in each mesh cell.
5.4.1. Non‐erodible bed layer There is the possibility to include a rock layer, meaning a non‐erodible bed layer. The implementation in the code of this non‐erodible bed layer has been done using a 2 stage calculation (predictor‐corrector) of the numerical bedload fluxes through the element’s contours. Predictor: searches a new bed layer elevation due mainly to the outflow in an element.
z ip = z in +
Δt (1 p) A i
max 0, q jK i
* sb
n Lij ij
Corrector: if at the predictor the element has eroded below the rock layer, the bedload outflow is corrected en each of the edges, and afterwards it is updated to the new bed layer elevation. p
Si (zi
< z roca ) y ( q*sbn > 0) : ij
Page 46 of 56
Iber. Turbulent Flow in Shallow Waters
q = q *,c sb ij
q
* sb ij
*,c sb ji
(1 p)
n 1 n z b,i zb,i
Δt
Ai
z in -z roca z in -z ip
= - q*sb
ij
q n L D - E A jK i
*,c sb
ij
ij
i
i
5.5. Managing Wet‐Dry Fronts Modeling flooding zones, as is the movement of wave fronts in estuaries and coastal zones is fundamental in environmental hydraulics. In Iber wet‐dry fronts, both stationary and non‐stationary, that may appear in the domain are modelled with a fixed finite volume mesh, allowing each element to have water or not depending on the flow conditions. Among the volumes which do not have water and those which do have, there is a dry‐wet front which is necessary to treat appropriately from a numerical point of view to avoid instabilities and non‐physical oscillations in the results. For this wet‐dry front, either in a flood front or a wave front, there is a wet‐dry tolerance εwd, so if the water depth is lower than εwd it is considered that the cell is dry and is not included in the calculations. The wet‐dry tolerance can be changed by the user to a value close to zero, even though in irregular bathymetry problems, as is very usual in river an coastal hydraulics, it is advisable to use values in the order of 1cm or 1mm to increase the stability of the results without spoiling their precision. In any case, the water height is never forced to be equal to zero, with the purpose of avoiding mass losses in the calculation domain. The numerical scheme used to solve the wet‐dry front is stable and non‐diffusive. The management the wet‐dry fronts in Iber is stable, conservative and non‐diffusive, meaning that the fronts are adequately solved without instabilities of the numerical type, even when these occur in strong bed slopes. Every finite element has an associated bed level. Schematically shown in Figure 9.
Page 47 of 56
Iber. Turbulent Flow in Shallow Waters
Figure 9. Schematic representation of the wet‐dry bed level management.
Between two volumes with different bed level one of the following situations may occur, as presented in Figure 10:
Figure 10. Several situations of the water levels between two adjacent cells.
In the first figure both volumes have water, so no front is produced and thus no special treatment is necessary. In the other two cases there is a wet‐dry front. The difference is that in the second case, the free water level in the wet cell is higher than the bed level of the dry cell, while in the third case it is lower. Only in the third case it is necessary to use a special treatment, which consists of redefining the bed slope to impose a reflexive condition in the front. In this case the bed slope is redefined as:
h i - h j Δz b,ij = z b,j - z b,i
si
h j z b,j - z b,i
si
h j > z b,j - z b,i
The reflexive condition is imposed as:
qn,ij = q x,ij n x,ij + q y,ij n y,ij 0 The use of the mentioned conditions provides can reproduce the hydrostatic solution for any bathymetry, without streatching the front and without generating any spurious oscillations in the free surface. This type of wet‐dry front treatment has been successfully used in the modeling of both stationary and non‐stationary processes, being particularly useful in the simulation of flood zones in rivers and coastal zones, and in calculating the evolution of wave fronts. A more detailed description of the implementation in Iber can be found in the references provided in section 6.
Page 48 of 56
Iber. Turbulent Flow in Shallow Waters
IBER Shallow waters turbulent flow
REFERENCES
Page 49 of 56
Iber. Turbulent Flow in Shallow Waters
6. REFERENCES Apsley, D., & Stansby, P. (2008). Bed‐Load Sediment Transport on large slopes: Model Formulation and Implementation within a RANS solver. Journal of Hydraulic Engineering, ASCE , vol 134 (10). Ariathurai, R., & Arulanandan, K. (1978). Erosion rate of cohesive soils. Journal of the Hydraulics Division. ASCE , 104 (HY2), 279‐283. Bermudez, A., Dervieux, A., Desideri, J., & Vazquez‐Cendon, M. (1998). Upwind schemes for the two dimensional shallow water equations with variable depth using unstructured meshes. Computational Methods Applied for Mechanical Engineering , Vol 155. Bladé, E., & Gómez‐Valentín, M. (2006). Modelación del flujo en lámina libre sobre cauces naturales. Análisis integrado en una y dos dimensiones. Monografía. Barcelona: CIMNE. Bladé, E., Gómez‐Valentín, M., Sánchez‐Juny, M., & Dolz, J. (2008). Preserving steady state in Finite Volume Computations of River Flow. Journal of Hydraulic Engineering, ASCE , Vol 134 (9). Cea, L. (2005). Unstructured finite volume model for unsteady turbulent shallow water flow with wet‐dry flows: numerical solver ans experimental validation. Phd thesis. A Coruña: Universidad de A Coruña. Cea, L., French, J., & Vázquez‐Cendón, M. (2006). Numerical modelling of tidal flows in complex estuaries including turbulence: An unstructured finite volume solver and experimental validation. International Journal for Numerical Methods in Engineering , vol 67 (13). Cea, L., Puertas, J., & Vázquez‐Cendón, M. (2007). Depth averaged modelling of turbulent shallow water flow with wet‐dry fronts. Archives of Computational Methods in Engineering, State of the art reviews , vol 14 (3). Chiew, Y., & Parker, G. (1994). Incipient Sediment Motion on Non‐Horizontal Slopes. Journal of Hydraulics Research , 649‐660. Chow, V., Maidment, D., & Mays, L. (1988). Applied Hydrology. McGraw‐Hill. CIMNE. (2009). GiD The personal pre and post‐processor. Barcelona: www.gidhome.com. Corestein, G., Bladé, E., Gómez‐Valentín, M., Dolz, J., Oñate, E., & Piazzese, J. (2004). New GiD interface for Ramflood‐DSS project hydraulic simulation code. Proceeding of the 2nd Conference on advances and applications of GiD. Barcelona: CIMNE. Dey. (2003). Threshold of sediment motion on combined transverse and longitudinal sloping beds. Journal of Hydraulic Research , #4 405‐415. Page 50 of 56
Iber. Turbulent Flow in Shallow Waters
García, M., & Parker, G. (1991). Entrainment of Bed Sediment into suspension. ASCE Journal of Hydraulic Engineering , #4 414‐435. Meyer‐Peter, E., & Muller, R. (1948). Formulas for bedload transport. Proceeding of the 2nd Conference (pp. 39‐64). Stockholm: IAHR. Rastogi, A., & Rodi, W. (1978). Predictions of heat and mass transfer in open channels. Journal of Hydraulics Division , 104 (3) 397‐420. Smith, J., & McLean, S. (1977). Spatially averaged flow over a wavy surface. Journal of Geophysical Research , 82(12) 1735‐1746. Smith, J.D. (1977). Modelling of sediment transport on continental shelves. The Sea: ideas and observations on progress in the study of the seas. New York: John Wiley and Sons. Van Dorn, W. (1953). Wind Stress on an artificial pond. Journal of Marine Research , Vol 12. van Rijn, L. (1987). Mathematical modelling of morpholigical processes in the case of suspended sediment transport. Delft: Delft Hydraulics Communications #382. van Rijn, L.C. (1984). Sediment Transport, part I: Bedload transport. Journal of Hydraulic Engineering , vol 110 (10). Wong, M. (2003). Does the bedload equation of Meyer‐Peter and Muller fit its own data? 30th Congress (pp. 73‐80). Thessaloniki: International Association of Hydraulic Research. Wong, M., & Parker, G. (2006). Reanalysis and Correction of Bedload relation of Meyer‐Peter and Muller using their own database. Journal of Hydraulic Engineering , 132(11) 1159‐1168. Wu, W. (2004). Depth‐averaged two‐dimensional numerical modeling of unsteady flow and nonuniform sediment transport in Open Channels. Journal of Hydraulic Engineering , 130(10) 1013‐1024.
7.
Page 51 of 56
Iber. Turbulent Flow in Shallow Waters
NOMENCLATURE a width of the layer in which bedload transport occurs
c a , c*a instant concentration and equilibrium concentration of suspended solids at a height z=a above the river bed level C depth‐averaged suspended load
C* Depth‐averaged suspended load in equilibrium conditions Cd discharge coefficient at a gate or a weir Cf bed friction coefficient Cm adjust coefficient for the turbulent viscosity in a parabolic model dwall wall distance Dij lateral dispersion terms Ds sediment diameter D50 mean sediment diameter
D* Non‐dimensional sediment diameter Dsx, Dsy Suspended load dispersion due to the non‐homogeneity of the velocity profiles and of the vertical direction sediment concentration D‐E balance between the bedload and suspended load f potential infiltration rate g gravity acceleration h depth i real infiltration rate k turbulent kinetic energy ks soil’s saturated permeability Ks grain’s roughness height kvd wind drag coefficient
Page 52 of 56
Iber. Turbulent Flow in Shallow Waters
L0 Initial depth of the soil’s saturated region L Depth of the soil’s saturated region M2 parameter that measures the erosion rate because of re‐suspension Ms mass source/drain term Mx, My momentum source/drain terms n Manning coefficient ns equivalent Manning coefficient due to the grain p porosity of the bedload sediments qn unit normal flow at the inlet contour qsb,x, qsb,y components of the bedload discharge Q discharge/Flow Rh hydraulic radius R non‐dimensional submerged specific weight Sc,t Schmidt number Se Effective initial soil saturation Sij deformation tensor T non‐dimensional parameter which measures the excess of friction at the bed layer above the critical value that defines the start of movement uf friction velocity due to bed roughness
u'i u' j Turbulent stresses or Reynolds stresses Ux, Uy depth‐averaged horizontal velocities |U| depth‐averaged mean velocity V10 Wind velocity at 10 meters height Ws sedimentation velocity of solid particles Zs Free layer level Zb Bed layer level
Page 53 of 56
Iber. Turbulent Flow in Shallow Waters
α coefficient that relates the mean concentration of suspended solids and the concentration near the river bed β Bed slope Г Molecular diffusion of suspended solids coefficient δij Kronecker delta ε turbulence dissipation rate Δθ changes in the soil’s humid content as the saturation advances θi initial soil’s humidity content θe soil’s effective porosity (drainable) θr retention capacity (irreducible humidity or non‐drainable) κ=0.41 von Karman constant λ latitude of the considered point μc bed material’s internal friction angle ν fluid’s cinematic viscosity νt turbulent viscosity ρ water density
ρs Sediment density τ b Total stress due to bed friction τ bs Bed stress due to grain τ c Critical bed stress τ*b , τ*bs Total grain stresses (non‐dimensional) τ*c Critical bed stress (non‐adimensional) τs free surface friction due to wind induced roughness τexx, τexy, τeyy tangential effective horizontal stresses
τijv Viscous stresses Page 54 of 56
Iber. Turbulent Flow in Shallow Waters
Soil total porosity ψ soil non‐saturated región suction Ω earth’s rotation angular velocity Ws sediment velocity
Page 55 of 56