SWASHES: a compilation of Shallow Water Analytic Solutions for Hydraulic and Environmental Studies
Olivier Delestre, Carine Lucas, Pierre-Antoine Ksinant, Frédéric Darboux, Christian Laguerre, Thi Ngoc Tuoi Vo, Francois James, Stephane Cordier
aa r X i v : . [ m a t h . NA ] J a n SWASHES: a compilation of Shallow WaterAnalytic Solutions for Hydraulic andEnvironmental Studies
O. Delestre ∗ , ‡ , † , C. Lucas ‡ , P.-A. Ksinant ‡ , § , F. Darboux § ,C. Laguerre ‡ , T.N.T. Vo ‡ , ¶ , F. James ‡ and S. Cordier ‡ January 22, 2016
Abstract
Numerous codes are being developed to solve Shallow Water equa-tions. Because there are used in hydraulic and environmental studies,their capability to simulate properly flow dynamics is critical to guaran-tee infrastructure and human safety. While validating these codes is animportant issue, code validations are currently restricted because analyticsolutions to the Shallow Water equations are rare and have been publishedon an individual basis over a period of more than five decades. This articleaims at making analytic solutions to the Shallow Water equations easilyavailable to code developers and users. It compiles a significant number ofanalytic solutions to the Shallow Water equations that are currently scat-tered through the literature of various scientific disciplines. The analyticsolutions are described in a unified formalism to make a consistent setof test cases. These analytic solutions encompass a wide variety of flowconditions (supercritical, subcritical, shock, etc.), in 1 or 2 space dimen-sions, with or without rain and soil friction, for transitory flow or steadystate. The corresponding source codes are made available to the commu-nity ( ), so that usersof Shallow Water-based models can easily find an adaptable benchmarklibrary to validate their numerical methods.
Keywords
Shallow-Water equation; Analytic solutions; Benchmarking; Vali-dation of numerical methods; Steady-state flow; Transitory flow; Source terms ∗ Corresponding author: [email protected], presently at: Laboratoire de Math´ematiquesJ.A. Dieudonn´e – Polytech Nice-Sophia , Universit´e de Nice – Sophia Antipolis, Parc Valrose,06108 Nice cedex 02, France † Institut Jean le Rond d’Alembert, CNRS & UPMC Universit´e Paris 06, UMR 7190, 4place Jussieu, Boˆıte 162, F-75005 Paris, France ‡ MAPMO UMR CNRS 7349, Universit´e d’Orl´eans, UFR Sciences, Bˆatiment demath´ematiques, B.P. 6759 – F-45067 Orl´eans cedex 2, France § INRA, UR 0272 Science du sol, Centre de recherche d’Orl´eans, CS 40001 Ardon, F-45075Orl´eans cedex 2, France ¶ Presently at: Department of Applied Mathematics, National University of Ireland, Gal-way, Republic of Ireland Introduction
Shallow-Water equations have been proposed by Adh´emar Barr´e de Saint-Venantin 1871 to model flows in a channel [4]. Nowadays, they are widely used to modelflows in various contexts, such as: overland flow [22, 47], rivers [25, 9], flood-ing [10, 17], dam breaks [1, 51], nearshore [6, 37], tsunami [23, 33, 41]. Theseequations consist in a nonlinear system of partial differential equations (PDE-s), more precisely conservation laws describing the evolution of the height andmean velocity of the fluid.In real situations (realistic geometry, sharp spatial or temporal variationsof the parameters in the model, etc.), there is no hope to solve explicitly thissystem of PDE-s, i.e. to produce analytic formulæ for the solutions. It is there-fore necessary to develop specific numerical methods to compute approximatesolutions of such PDE-s, see for instance [49, 34, 7]. Implementation of any ofsuch methods raises the question of the validation of the code.Validation is an essential step to check if a model (that is the equations, thenumerical methods and their implementation) suitably describes the consideredphenomena. There exists at least three complementary types of numerical teststo ensure a numerical code is relevant for the considered systems of equations.First, one can produce convergence or stability results ( e.g. by refining themesh). This validates only the numerical method and its implementation. Sec-ond, approximate solutions can be matched to analytic solutions available forsome simplified or specific cases. Finally, numerical results can be comparedwith experimental data, produced indoor or outdoor. This step should be doneafter the previous two; it is the most difficult one and must be validated by aspecialist of the domain. This paper focuses on the second approach.Analytic solutions seem underused in the validation of numerical codes, pos-sibly for the following reasons. First, each analytic solution has a limited scopein terms of flow conditions. Second, they are currently scattered through theliterature and, thus, are difficult to find. However, there exists a significantnumber of published analytic solutions that encompasses a wide range of flowconditions. Hence, this gives a large potential to analytic solutions for validata-tion of numerical codes.This work aims at overcoming these issues, on the one hand by gatheringa significant set of analytic solutions, on the other hand by providing the cor-responding source codes. The present paper describes the analytic solutionstogether with some comments about their interest and use. The source codesare made freely available to the community through the SWASHES (Shallow-Water Analytic Solutions for Hydraulic and Environmental Studies) software.The SWASHES library does not pretend to list all available analytic solutions.On the contrary, it is open for extension and we take here the opportunity to askusers to contribute to the project by sending other analytic solutions togetherwith the dedicated code.The paper is organized as follows: in Section 2, we briefly present the nota-tions we use and the main properties of Shallow-Water equations. In the nexttwo sections, we briefly outline each analytic solution. A short description of theSWASHES software can be find in Section 5. The final section is an illustrationusing the results of the Shallow-Water code developed by our team for a subsetof analytic solutions. 2igure 1: Notations for 2D Shallow-Water equations
First we describe the rather general settings of viscous Shallow-Water equationsin two space dimensions, with topography, rain, infiltration and soil friction.In the second paragraph, we give the simplified system arising in one spacedimension and recall several classical properties of the equations.
The unknowns of the equations are the water height ( h ( t, x, y ) [L]) and u ( t, x, y ), v ( t, x, y ) the horizontal components of the vertically averaged velocity [L/T](Figure 1).The equations take the following form of balance laws, where g = 9 .
81 m/s is the gravity constant: ∂ t h + ∂ x ( hu ) + ∂ y ( hv ) = R − I∂ t ( hu ) + ∂ x (cid:18) hu + gh (cid:19) + ∂ y ( huv ) = gh ( S x − S f x ) + µS dx ∂ t ( hv ) + ∂ x ( huv ) + ∂ y (cid:18) hv + gh (cid:19) = gh ( S y − S f y ) + µS dy (1)The first equation is actually a mass balance. The fluid density can be replacedby the height because of incompressibility. The other two equations are mo-mentum balances, and involve forces such as gravity and friction. We give nowa short description of all the terms involved, recalling the physical dimensions. • z is the topography [L], since we consider no erosion here, it is a fixedfunction of space, z ( x, y ), and we classically denote by S x (resp. S y ) theopposite of the slope in the x (resp. y ) direction, S x = − ∂ x z ( x, y ) (resp. S y = − ∂ y z ( x, y )); • R ≥ R ( t, x, y ) ≥
0. Inthis paper, it is considered uniform in space;3 I is the infiltration rate [L/T], mentioned for the sake of completeness. Itis given by another model (such as Green-Ampt, Richards, etc.) and isnot taken into account in the following; • S f = (cid:16) S f x , S f y (cid:17) is the friction force. The friction law S f may take severalforms, depending on both soil and flow properties. In the formulæ below, U is the velocity vector U = ( u, v ) with | U | = √ u + v and Q is thedischarge Q = ( hu, hv ). In hydrological models, two families of frictionlaws are encountered, based on empirical considerations. On the one hand,we have the family of Manning-Strickler’s friction laws S f = C f U | U | h / = C f Q | Q | h / C f = n , where n is the Manning’s coefficient [L -1/3 T].On the other hand, the laws of Darcy-Weisbach’s and Ch´ezy’s familywrites S f = C f U | U | h = C f Q | Q | h . With C f = f / (8 g ), f dimensionless coefficient, (resp. C f = 1 /C , C [L /T]) we get the Darcy-Weisbach’s (resp. Ch´ezy’s) friction law. Noticethat the friction may depend on the space variable, especially on largeparcels. In the sequel this will not be the case. • finally, µS d = (cid:0) µS dx , µS dy (cid:1) is the viscous term with µ ≥ /T]. In this section, we recall several properties of the Shallow-Water model that areuseful in the flow description. For the sake of simplicity, we place ourselves in theone-dimensional case, extensions to the general setting being straightforward.The two-dimensional equations (1) rewrite ∂ t h + ∂ x ( hu ) = R − I∂ t ( hu ) + ∂ x (cid:18) hu + gh (cid:19) = gh ( S x − S f ) + µ∂ x ( h∂ x u ) (2)The left-hand side of this system is the transport operator, corresponding tothe flow of an ideal fluid in a flat channel, without friction, rain or infiltration.This is actually the model introduced by Saint-Venant in [4], and it containsseveral important properties of the flow. In order to emphasize these properties,we first rewrite the one-dimensional equations using vectors form: ∂ t W + ∂ x F ( W ) = 0 , where W = (cid:18) hhu (cid:19) , F ( W ) = huhu + gh , (3)with F ( W ) the flux of the equation. The transport is more clearly evidencedin the following nonconservative form, where A ( W ) = F ′ ( W ) is the matrix oftransport coefficients: ∂ t W + A ( W ) ∂ x W = 0 , A ( W ) = F ′ ( W ) = (cid:18) − u + gh u (cid:19) . (4)4ore precisely, when h >
0, the matrix A ( W ) turns out to be diagonalizable,with eigenvalues λ ( W ) = u − p gh < u + p gh = λ ( W ) . This important property is called strict hyperbolicity (see for instance [24] andreferences therein for more complete information). The eigenvalues are indeedvelocities, namely the ones of surface waves on the fluid, which are basic char-acteristics of the flow. Notice here that the eigenvalues coincide if h = 0 m, thatis for dry zones. In that case, the system is no longer hyperbolic, and this in-duces difficulties at both theoretical and numerical levels. Designing numericalschemes that preserve positivity for h is very important in this context.From these formulæ we recover a useful classification of flows, based on therelative values of the velocities of the fluid, u , and of the waves, √ gh . Indeedif | u | < √ gh the characteristic velocities have opposite signs, and informationpropagate upward as well as downward the flow, which is then said subcriticalor fluvial. On the other hand, when | u | > √ gh , the flow is supercritical, ortorrential, all the information go downwards. A transcritical regime exists whensome parts of a flow are subcritical, other supercritical.Since we have two unknowns h and u (or equivalently h and q = hu ), asubcritical flow is therefore determined by one upstream and one downstreamvalue, whereas a supercritical flow is completely determined by the two upstreamvalues. Thus for numerical simulations, we have to impose one variable forsubcritical inflow/outflow. We impose both variables for supercritical inflowand for supercritical outflow, free boundary conditions are considered (see forexample [8]).In this context, two quantities are useful. The first one is a dimensionlessparameter called the Froude numberFr = | u |√ gh . (5)It is the analogue of the Mach number in gas dynamics, and the flow is subcritical(resp. supercritical) if Fr < > h c which writes h c = (cid:18) q √ g (cid:19) / , (6)for a given discharge q = hu . It is a very readable criterion for criticality: theflow is subcritical (resp. supercritical) if h > h c (resp. h < h c ).When additional terms are present, other properties have to be considered,for instance the occurrence of steady states (or equilibrium) solutions. Thesespecific flows are defined and discussed in Section 3. In this section, we focus on a family of steady state solutions, that is solutionsthat satisfy: ∂ t h = ∂ t u = 0 . ∂ x ( hu ) = R or hu = q = Rx + q , where q = q ( t, x = 0).Similarly the momentum equation writes ∂ x (cid:18) q h + gh (cid:19) = − gh∂ x z − ghS f ( h, q ) + µ∂ x (cid:16) h∂ x qh (cid:17) . Thus for h = 0, we have the following system q = Rx + q ,∂ x z = 1 gh (cid:18) q h − gh (cid:19) ∂ x h − S f ( h, q ) + µgh ∂ x (cid:16) h∂ x qh (cid:17) . (7)System (7) is the key point of the following series of analytic solutions. Forthese solutions, the strategy consists in choosing either a topography and get-ting the associated water height or a water height and deducing the associatedtopography.Since [5], it is well known that the source term treatment is a crucial pointin preserving steady states. With the following steady states solutions, one cancheck if the steady state at rest and dynamic steady states are satisfied by theconsidered schemes using various flow conditions (fluvial, torrential, transcriti-cal, with shock, etc.). Moreover, the variety of inflow and outflow configurations(flat bottom/varying topography, with/without friction, etc.) gives a validationof boundary conditions treatment. One must note that, as different source terms(topography, friction, rain and diffusion) are taken into account, these solutionscan also validate the source terms treatment.The last remark deals with initial conditions: if initial conditions are taken equalto the solution at the steady state, one can only conclude on the ability of thenumerical scheme to preserve steady states. In order to prove the capacity to catch these states, initial conditions should be different from the steady state.This is the reason why the initial conditions, as well as the boundary conditions,are described in each case.Table 1 lists all steady-state solutions available in SWASHES and outlinestheir main features. Here we present a series of steady state cases proposed in [27, p.14-17] based onan idea introduced in [32], with a flat topography at the boundaries, no rain,no friction and no diffusion ( R = 0 m/s, S f = 0 and µ = 0 m / s). Thus system(7) reduces to q = q ,∂ x z = 1 gh (cid:18) q h − gh (cid:19) ∂ x h. In the case of a regular solution, we get the Bernoulli relation q gh ( x ) + h ( x ) + z ( x ) = Cst (8)which gives us the link between the topography and the water height.6 teady-state solutions
Flow criticality FrictionType Description § Reference Sub. Sup. Sub. → Sup. Jump Man. D.-W. Other Null CommentsBumps Lake at rest with immersed bump 3.1.1 [16] X Hydrostatic equilibriaLake at rest with emerged bump 3.1.2 [16] X Hydrostatic equilibria andwet-dry transitionSubcritical flow 3.1.3 [27] X X Initially steady state at restTranscritical flow without shock 3.1.4 [27] X X Initially steady state at restTranscritical flow with shock 3.1.5 [27] X X Initially steady state at restFlumes(MacDonald’sbased) Long channelwith subcritical flow 3.2.1 [53] X X X Initially dry channel. 1000 m longLong channelwith supercritical flow 3.2.1 [16] X X X Initially dry channel. 1000 m longLong channelwith sub- to super-critical flow 3.2.1 [53] X X X Initially dry channel. 1000 m longLong channelwith super- to sub-critical flow 3.2.1 [53] X X X Initially dry channel. 1000 m longShort channelwith smooth transition and shock 3.2.2 [53] X X X At t=0, lake downstream.100 m longShort channelwith supercritical flow 3.2.2 [16] X X Initially dry. 100 m longShort channelwith sub- to super-critical flow 3.2.2 [53] X X At t=0, lake downstream.100 m longVery long, undulating and periodic channelwith subcritical flow 3.2.3 [53] X X At t=0, lake downstream.5000 m longRain on a long channelwith subcritical flow 3.3.1 [53] X X X Initially dry. 1000 m longRain on a long channelwith supercritical flow 3.3.2 [53] X X X Initially dry. 1000 m longLong channelwith subcritical flow and diffusion 3.4.1 [19] X X Initially dry. 1000 m longLong channelwith supercritical flow and diffusion 3.4.2 [19] X X Initially dry. 1000 m longPseudo-2D short channelwith subcritical flow 3.5.1 [35] X X Rectangular cross section.Initially dry. 200 m longPseudo-2D short channelwith supercritical flow 3.5.2 [35] X X Rectangular cross section.Initially wet. 200 m longPseudo-2D short channelwith smooth transition 3.5.3 [35] X X Rectangular cross section.Initially partly-wet. 200 m longPseudo-2D short channelwith shock 3.5.4 [35] X X Rectangular cross section.Initially dry. 200 m longPseudo-2D long channelwith subcritical flow 3.5.5 [35] X X Isoscele trapezoidal cross section.Initially dry. 400 m longPseudo-2D long channelwith smooth transition and shock 3.5.6 [35] X X X Isoscele trapezoidal cross section.Initially dry. 400 m long
Slope is always variable. Sub.: Subcritical; Sup.: Supercritical; Man.: Manning; D.-W.: Darcy-Weisbach
Table 1: Analytic solutions for shallow flow equations and their main features— Steady-state cases 7nitial conditions satisfy the hydrostatic equilibrium h + z = Cst and q = 0 m / s . (9)These solutions test the preservation of steady states and the boundary condi-tions treatment.In the following cases, we choose a domain of length L = 25 m with atopography given by: z ( x ) = (cid:26) . − . x − if 8 m < x <
12 m , . In the case of a lake at rest with an immersed bump, the water height is such thatthe topography is totally immersed [16]. In such a configuration, starting fromthe steady state, the velocity must be null and the water surface should stay flat.In SWASHES we have the following initial conditions: h + z = 0 . q = 0 m /sand the boundary conditions (cid:26) h = 0 . ,q = 0 m /s . The case of a lake at rest with an emerged bump is the same as in the previ-ous section except that the water height is smaller in order to have emergenceof some parts of the topography [16]. Here again, we initialize the solution atsteady state and the solution is null velocity and flat water surface.In SWASHES we consider the following initial conditions: h + z = max(0 . , z ) m and q = 0 m /sand the boundary conditions (cid:26) h = 0 . ,q = 0 m /s . After testing the two steady states at rest, the user can increase the difficultywith dynamical steady states. In the case of a subcritical flow, using (8), thewater height is given by the resolution of h ( x ) + (cid:18) z ( x ) − q gh L − h L (cid:19) h ( x ) + q g = 0 , ∀ x ∈ [0 , L ] , where h L = h ( x = L ). 8hanks to the Bernoulli relation (8), we can notice that the water heightis constant when the topography is constant, decreases (respectively increases)when the bed slope increases (resp. decreases). The water height reaches itsminimum at the top of the bump [27].In SWASHES, the initial conditions are h + z = 2 m and q = 0 m /sand the boundary conditions are chosen as (cid:26) upstream: q = 4 .
42 m /s , downstream: h = 2 m . In this part, we consider the case of a transcritical flow, without shock [27].Again thanks to (8), we can express the water height as the solution of h ( x ) + (cid:18) z ( x ) − q gh c − h c − z M (cid:19) h ( x ) + q g = 0 , ∀ x ∈ [0 , L ] , where z M = max x ∈ [0 ,L ] z and h c is the corresponding water height.The flow is fluvial upstream and becomes torrential at the top of the bump.Initial conditions can be taken equal to h + z = 0 .
66 m and q = 0 m /sand for the boundary conditions (cid:26) upstream: q = 1 .
53 m /s , downstream: h = 0 .
66 m while the flow is subcritical . If there is a shock in the solution [27], using (8), the water height is given bythe resolution of h ( x ) + (cid:18) z ( x ) − q gh c − h c − z M (cid:19) h ( x ) + q g = 0 for x < x shock ,h ( x ) + (cid:18) z ( x ) − q gh L − h L (cid:19) h ( x ) + q g = 0 for x > x shock ,q (cid:18) h − h (cid:19) + g (cid:0) h − h (cid:1) = 0 . (10)In these equalities, z M = max x ∈ [0 ,L ] z , h c is the corresponding water height, h L = h ( x = L ) and h = h ( x − shock ), h = h ( x + shock ) are the water heights up-stream and downstream respectively. The shock is located thanks to the thirdrelation in system (10), which is a Rankine-Hugoniot’s relation.As for the previous case, the flow becomes supercritical at the top of thebump but it becomes again fluvial after a hydraulic jump.9ne can choose for initial conditions h + z = 0 .
33 m and q = 0 m /sand the following boundary conditions (cid:26) upstream: q = 0 .
18 m /s , downstream: h = 0 .
33 m . We can find a generalisation of this case with a friction term in Hervouet’swork [31].
Following the lines of [35, 36], we give here some steady state solutions of sys-tem (2) with varying topography and friction term (from [53, 16]). Rain anddiffusion are not considered ( R = 0 m / s, µ = 0 m / s), so the steady statessystem (7) reduces to ∂ x z = (cid:18) q gh − (cid:19) ∂ x h − S f . (11)From this relation, one can make as many solutions as required. In this section,we present some of them that are obtained for specific values of the length L of the domain, and for fixed parameters (such as the friction law and its coef-ficient). The water height profile and the discharge are given, and we computethe corresponding topographies solving equation (11). We have to mention thatthere exists another approach, classical in hydraulics. It consists in consideringa given topography and a discharge. From these, the steady-state water heightis deduced thanks to equation (11) and to the classification of water surface pro-files (see among others [14] and [30]). Solutions obtained using this approachmay be found in [56] for example. Finally, note that some simple choices for thefree surface may lead to exact analytic solutions.The solutions given in this section are more intricate than the ones of theprevious section, as the topography can vary near the boundary. Consequentlythey give a better validation of the boundary conditions. If S f = 0 (we havefriction at the bottom), the following solutions can prove if the friction termsare coded in order to satisfy the steady states. Remark 1
All these solutions are given by the numerical resolution of an equa-tion. So, the space step should be small enough to have a sufficiently precise so-lution. It means that the space step used to get these solutions should be smallerthan the space step of the code to be validated.
We consider a 1000 m long channel with a discharge of q = 2 m / s [53]. The flow is constant at inflow and the water height is prescribedat outflow, with the following values: (cid:26) upstream: q = 2 m / s , downstream: h = h ex (1000) . i.e. initial conditions are h = 0 m and q = 0 m / s . The water height is given by h ex ( x ) = (cid:18) g (cid:19) / − (cid:18) x − (cid:19) !! . (12)We remind the reader that q = 2 m / s on the domain and that the topographyis calculated iteratively thanks to (11). We can consider the two friction lawsexplained in the introduction, with the coefficients n = 0 .
033 m -1/3 s for Man-ning’s and f = 0 .
093 for Darcy-Weisbach’s.Under such conditions, we get a subcritical steady flow.
Supercritical case
We still consider a 1000 m long channel, but with a con-stant discharge q = 2 . / s on the whole domain [16]. The flow is supercriticalboth at inflow and at outflow, thus we consider the following boundary condi-tions: (cid:26) upstream: q = 2 . / s and h = h ex (0) , downstream: free.The initial conditions are a dry channel h = 0 m and q = 0 m / s . With the water height given by h ex ( x ) = (cid:18) g (cid:19) / −
15 exp − (cid:18) x − (cid:19) !! (13)and the friction coefficients equal to n = 0 .
04 m -1/3 s for Manning’s and f =0 .
065 for Darcy-Weisbach’s friction law, the flow is supercritical.
Subcritical-to-supercritical case
The channel is 1000 m long and the dis-charge at equilibrium is q = 2 m / s [53]. The flow is subcritical upstream andsupercritical downstream, thus we consider the following boundary conditions: (cid:26) upstream: q = 2 m / s , downstream: free.As initial conditions, we consider a dry channel h = 0 m and q = 0 m / s . In this configuration, the water height is h ex ( x ) = (cid:18) g (cid:19) / (cid:18) −
13 tanh (cid:18) (cid:18) x − (cid:19)(cid:19)(cid:19) for 0 m ≤ x ≤
500 m , (cid:18) g (cid:19) / (cid:18) −
16 tanh (cid:18) (cid:18) x − (cid:19)(cid:19)(cid:19) for 500 m < x ≤ , with a friction coefficient n = 0 . -1/3 s (resp. f = 0 . upercritical-to-subcritical case As in the previous cases, the domain is1000 m long and the discharge is q = 2 m / s [53]. The boundary conditions area torrential inflow and a fluvial outflow: (cid:26) upstream: q = 2 m / s and h = h ex (0) , downstream: h = h ex (1000) . At time t = 0 s, the channel is initially dry h = 0 m and q = 0 m / s . The water height is defined by the following discontinuous function h ex ( x ) = (cid:18) g (cid:19) / (cid:18) −
16 exp (cid:16) − x (cid:17)(cid:19) for 0 m ≤ x ≤
500 m , (cid:18) g (cid:19) / X k =1 a k exp (cid:18) − k (cid:18) x − (cid:19)(cid:19) + 45 exp (cid:16) x − (cid:17)! for 500 m ≤ x ≤ , with a = − . a = 0 . a = − . n = 0 . -1/3 s for the Manning’s law and f =0 . x =500 m. In this part, the friction law we consider is the Manning’s law. Generalizationto other classical friction laws is straightforward.
Case with smooth transition and shock
The length of the channel is 100m and the discharge at steady states is q = 2 m / s [53]. The flow is fluvial bothupstream and downstream, the boundary conditions are fixed as follows (cid:26) upstream: q = 2 m / s , downstream: h = h ex (100) . To have a case including two kinds of flow (subcritical and supercritical) andtwo kinds of transition (transonic and shock), we consider a channel filled withwater, i.e. h ( x ) = max( h ex (100) + z (100) − z ( x ) ,
0) and q = 0 m / s . h ex ( x ) = (cid:18) g (cid:19) / (cid:18) − x (cid:19) − x (cid:18) x − (cid:19) for 0 m ≤ x ≤ ≈ .
67 m , (cid:18) g (cid:19) / a (cid:18) x − (cid:19) + a (cid:18) x − (cid:19) − a (cid:18) x − (cid:19) + a (cid:18) x − (cid:19) + a ! for 2003 ≈ .
67 m ≤ x ≤
100 m , with a = 0 . a = 21 . a = 14 .
492 et a = 1 . n = 0 . -1/3 s, the inflowis subcritical, becomes supercritical via a sonic point, and, through a shock(located at x = 200 / ≈ .
67 m), becomes subcritical again.
Supercritical case
The channel we consider is still 100 m long and the equi-librium discharge is q = 2 m / s [16]. The flow is torrential at the bounds of thechannel, thus the boundary conditions are (cid:26) upstream: q = 2 m / s and h = h ex (0) , downstream: free.As initial conditions, we consider an empty channel which writes h = 0 m and q = 0 m / s . The water height is given by h ex ( x ) = (cid:18) g (cid:19) / −
14 exp − (cid:18) x − (cid:19) !! and the friction coefficient is n = 0 .
03 m -1/3 s (for the Manning’s law). The flowis entirely torrential.
Subcritical-to-supercritical case
A 100 m long channel has a discharge of q = 2 m / s [53]. The flow is fluvial at inflow and torrential at outflow withfollowing boundary conditions (cid:26) upstream: q = 2 m / s , downstream: free.As in the subcritical case, the initial condition is an empty channel with a puddledownstream h ( x ) = max( h ex (100) + z (100) − z ( x ) ,
0) and q = 0 m / s , h ex ( x ) = (cid:18) g (cid:19) / − ( x − x − ! . The Manning’s friction coefficient for the channel is n = 0 . -1/3 s. We geta transcritical flow: subcritical upstream and supercritical downstream. For this case, the channel is much longer than for the previous ones: 5000 m,but the discharge at equilibrium is still q = 2 m / s [53]. Inflow and outflow areboth subcritical. The boundary conditions are taken as: (cid:26) upstream: q = 2 m / s , downstream: h = h ex (5000) . We consider a dry channel with a little lake at rest downstream as initial con-ditions: h ( x ) = max( h ex (5000) + z (5000) − z ( x ) ,
0) and q = 0 m / s . We take the water height at equilibrium as a periodic function in space, namely h ex ( x ) = 98 + 14 sin (cid:16) πx (cid:17) and the Manning’s constant is n = 0 .
03 m -1/3 s. We get a subcritical flow. Asthe water height is periodic, the associated topography (solution of Equation(11)) is periodic as well: we get a periodic configuration closed to the ridges-and-furrows configuration. Thus this case is interesting for the validation ofnumerical methods for overland flow simulations on agricultural fields.
In this section, we consider the Shallow Water system (2) with rain (but withoutviscosity: µ = 0 m / s) at steady states [53]. The rain intensity is constant, equalto R . The rain is uniform on the domain [0 , L ]. Under these conditions, if wedenote by q the discharge value at inflow q ( t,
0) = q , we have: q ( x ) = q + xR , for 0 ≤ x ≤ L. (14)The solutions are the same as in section 3.2, except that the discharge is givenby (14). But the rain term modifies the expression of the topography througha new rain term as written in (7). More precisely, Equation (11) is replaced by ∂ x z = (cid:18) q gh − (cid:19) ∂ x h − qR gh − S f . These solutions allow the validation of the numerical treatment of the rain.Remark 1, mentioned in the previous section, applies to these solutions too.14 .3.1 Subcritical case for a long channel
For a 1000 m long channel, we consider a flow which is fluvial on the wholedomain. Thus we impose the following boundary conditions: (cid:26) upstream: q = q , downstream: h = h ex (1000) , with the initial conditions h = 0 m and q = 0 m / s , where h ex is the water height at steady state given by (12).In SWASHES, for the friction term, we can choose either Manning’s lawwith n = 0 .
033 m -1/3 s or Darcy-Weisbach’s law with f = 0 . q is fixed at 1 m / s and the rain intensity is R = 0 .
001 m/s.
The channel length remains unchanged (1000 m), but, as the flow is supercriti-cal, the boundary conditions are (cid:26) upstream: q = q and h = h ex (0) , downstream: free.At initial time, the channel is dry h = 0 m and q = 0 m / s . At steady state, the formula for the water height is (13).From a numerical point of view, for this case, we recommend that the raindoes not start at the initial time. The general form of the recommended rainfallevent is R ( t ) = (cid:26) t < t R ,R else , with t R = 1500 s. Indeed, this allows to get two successive steady states: forthe first one the discharge is constant in space q and for the second one thedischarge is (14) with the chosen height profile (13).In SWASHES, we have a friction coefficient n = 0 .
04 m -1/3 s for Manning’slaw, f = 0 .
065 for Darcy-Weisbach’s law. Inflow discharge is q = 2 . / s and R = 0 .
001 m/s.
Following the lines of [35, 36], Delestre and Marche, in [19], proposed newanalytic solutions with a diffusion term (with R = 0 m/s). To our knowledge,these are the only analytic solutions available in the litterature with a diffusionsource term. In [19], the authors considered system (2) with the source termsderived in [38], i.e. : S f = 1 g (cid:18) α ( h ) h u + α ( h ) | u | u (cid:19) and µ = 4 µ h α ( h ) = k l k l h µ v and α ( h ) = k t (cid:18) k l h µ v (cid:19) where µ v [T] (respectively µ h [L /T]) is the vertical (resp. the horizontal) eddyviscosity and k l [T/L] (resp. k t [1/L]) the laminar (resp. the turbulent) frictioncoefficient. At steady states we recover (7) with R = 0 m / s, or again q = q ,∂ x z = 1 gh (cid:18) q h − gh (cid:19) ∂ x h − S f ( h, q ) + µgh (cid:16) − q∂ xx h + qh ( ∂ x h ) (cid:17) . (15)As for the previous cases, the topography is evaluated thanks to the momentumequation of (15).These cases allow for the validation of the diffusion source term treatment.These solutions may be easily adapted to Manning’s and Darcy-Weisbach’s fric-tion terms. Remark 1 applies to these solutions too.In [19], the effect of µ h , µ v , k t and k l is studied by using several values. Inwhat follows, we present only two of these solutions: a subcritical flow and asupercritical flow. A 1000 m long channel has a discharge of q = 1 . / s. The flow is fluvial atboth channel boundaries, thus the boundary conditions are: (cid:26) upstream: q = 1 . / s , downstream: h = h ex (1000) . The channel is initially dry h = 0 m and q = 0 m / sand the water height at steady state ( h ex ) is the same as in section 3.2.1.In SWASHES, the parameters are: k t = 0 . k l = 0 . µ v = 0 .
01 and µ h = 0 . We still consider a 1000 m long channel, with a constant discharge q = 2 . / son the whole domain. Inflow and outflow are both torrential, thus we choosethe following boundary conditions: (cid:26) upstream: q = 2 . / s and h = h ex (0) , downstream: free.We consider a dry channel as initial condition: h = 0 m and q = 0 m / s . The water height h ex at steady state is given by function (13).In SWASHES, we have: k t = 0 . k l = 0 . µ v = 0 .
01 and µ h = 0 . .5 Mac Donald pseudo-2D solutions In this section, we give several analytic solutions for the pseudo-2D Shallow-Water system. This system can be considered as an intermediate between theone-dimensional and the two-dimensional models. More precisely, these equa-tions model a flow in a rectilinear three-dimensional channel with the quantitiesaveraged not only on the vertical direction but also on the width of the channel.For the derivation, see for example [26]. Remark 1, mentioned for Mac Donald’stype 1D solutions, applies to these pseudo-2D solutions too.We consider six cases for non-prismatic channels introduced in [35]. Thesechannels have a variable slope and their width is also variable in space. Moreprecisely, each channel is determined through the definition of the bottom width B (as a function of the space variable x ) and the slope of the boundary Z (Figure 2). The bed slope is an explicit function of the water height, detailedin the following. x yOz Bh Z B + 2 hZ Figure 2: Notations for the Mac Donald pseudo-2D solutionsSolution B (m) Z (m) L (m) h in (m) h out (m)Subcritical flow in a shortdomain B ( x ) 0 200 0.902921Supercritical flow in a shortdomain B ( x ) 0 200 0.503369Smooth transition in a shortdomain B ( x ) 0 200Hydraulic jump in a shortdomain B ( x ) 0 200 0.7 1.215485Subcritical flow in a longdomain B ( x ) 2 400 0.904094Smooth transition followedby a hydraulic jump in along domain B ( x ) 2 400 1.2Table 2: Main features of the cases of pseudo-2D channelsThe features of these cases are summarized in table 2. In this table, the functions17 y x B(x)/2 and -B(x)/2 for B=B1 -5-3-1 1 3 5 0 100 200 300 400 y x B(x)/2 and -B(x)/2 for B=B2
Figure 3: Functions B and B for the shape of the channelfor the bed shape are: B ( x ) = 10 − − (cid:18) x − (cid:19) ! for 0 m ≤ x ≤ L = 200 m ,B ( x ) = 10 − − (cid:18) x − (cid:19) ! − − (cid:18) x − (cid:19) ! for 0 m ≤ x ≤ L = 400 m , (Figure 3) and h in (resp. h out ) is the water height at the inflow (resp. outflow).In each case, the Manning’s friction coefficient is n = 0 .
03 m -1/3 s, the discharge q is taken equal to 20 m s − , the slope of the topography is given by: S ( x ) = (cid:18) − q ( B ( x ) + 2 Zh ( x )) g h ( x ) ( B ( x ) + Zh ( x )) (cid:19) h ′ ( x )+ q n (cid:0) B ( x ) + 2 h ( x ) √ Z (cid:1) / h ( x ) / ( B ( x ) + Zh ( x )) / − q B ′ ( x ) g h ( x ) ( B ( x ) + Zh ( x )) where h is the water height and the topography is defined as z ( x ) = Z Lx S ( X ) dX . Remark 2
When programming these formulae, we noted a few typos in [35],in the expression of S ( x ) , of φ in the hydraulic jump case ( § h in [0; 120] in the solution for the smooth transition followed bya hydraulic jump ( § Remark 3
We recall that the following analytic solutions are solutions of thepseudo-2D Shallow-Water system. This is the reason why, in this section, h ex does not depend on y . In the case of a subcritical flow in a short domain, as in the three that follow, thecross section of the channel is rectangular, the bottom is given by the function B and the length L = 200 m. 18n this current case, the flow is fixed at inflow and the water height is pre-scribed at outflow. We have the following boundary conditions: (cid:26) upstream: q = 20 m s − , downstream: h = h out . The channel is initially dry, with a little puddle downstream, i.e. initialconditions are: h ( x, y ) = max(0 , h out + z (200 , y ) − z ( x, y )) and q = 0 m / s . If we take the mean water height h ex ( x ) = 0 . . − (cid:18) x − (cid:19) ! , the flow stays subcritical in the whole domain of length L = 200 m. In this case, the flow and the water height are fixed at inflow. We have thefollowing boundary conditions: (cid:26) upstream: q = 20 m s − and h = h in , downstream: free.The channel is initially dry, i.e. initial conditions are: h = 0 m , and q = 0 m / s . If we consider the mean water height h ex ( x ) = 0 . . − (cid:18) x − (cid:19) ! , in a channel of length L = 200 m with the B shape and vertical boundary, theflow is supercritical. In the case of smooth transition in a short domain, the flow is fixed at the inflow.We have the following boundary conditions: (cid:26) upstream: q = 20 m s − , downstream: free.The channel is initially dry, i.e. initial conditions are: h = 0 m , and q = 0 m / s . The channel is the same as in the previous cases, with a mean water heightgiven by h ex ( x ) = 1 − . (cid:18) (cid:18) x − (cid:19)(cid:19) . Under these conditions, the flow is first subcritical and becomes supercritical.19 .5.4 Hydraulic jump in a short domain
In the case of a hydraulic jump in a short domain, the flow discharge is fixedat the inflow and the water height is prescribed at both inflow and outflow. Wehave the following boundary conditions: (cid:26) upstream: q = 20 m s − and h = h in , downstream: h = h out . The channel is initially dry, with a little puddle downstream, i.e. initialconditions are: h ( x, y ) = max(0 , h out + z (200 , y ) − z ( x, y )) and q = 0 m / s . We choose the following expression for the mean water height: h ex ( x ) = 0 . . (cid:16) exp (cid:16) x (cid:17) − (cid:17) for 0 m ≤ x ≤
120 m , and h ex ( x ) = exp ( − p ( x − x ⋆ )) M X i =0 k i (cid:18) x − x ⋆ x ⋆⋆ − x ⋆ (cid:19) i + φ ( x ) for 120 m ≤ x ≤
200 m , with: • x ⋆ = 120 m, • x ⋆⋆ = 200 m, • M = 2, • k = − . • k = − . • k = − . • p = 0 . • φ ( x ) = 1 . (cid:16) . (cid:16) x − (cid:17)(cid:17) .We obtain a supercritical flow that turns into a subcritical flow through a hy-draulic jump. From now on, the length of the domain is L = 400 m, the boundaries of thechannel are given by B and the cross sections are isoscele trapezoids.In this case, the flow is fixed at the inflow and the water height is prescribedat the outflow. We have the following boundary conditions: (cid:26) upstream: q = 20 m s − , downstream: h = h out . The channel is initially dry, with a little puddle downstream, i.e. initialconditions are: h ( x, y ) = max(0 , h out + z (400 , y ) − z ( x, y )) and q = 0 m / s . Considering the mean water height h ex ( x ) = 0 . . − (cid:18) x − (cid:19) ! + 0 . − (cid:18) x − (cid:19) ! , the flow is subcritical along the whole channel.20 .5.6 Smooth transition followed by a hydraulic jump in a long do-main In this case, the flow is fixed at the inflow and the water height is prescribed atthe outflow. We have the following boundary conditions: (cid:26) upstream: q = 20 m s − , downstream: h = h out . The channel is initially dry, with a little puddle downstream, i.e. initialconditions are: h ( x, y ) = max(0 , h out + z (400 , y ) − z ( x, y )) and q = 0 m / s . With the second channel, we define the mean water height by h ex ( x ) = 0 . . (cid:16) exp (cid:16) − x (cid:17) − (cid:17) + 0 .
25 exp (cid:18) (cid:18) x − (cid:19)(cid:19) for 0 m ≤ x ≤
120 m , and h ex ( x ) = exp ( − p ( x − x ⋆ )) M X i =0 k i (cid:18) x − x ⋆ x ⋆⋆ − x ⋆ (cid:19) i + φ ( x ) for 120 m ≤ x ≤
200 m , with: • x ⋆ = 120 m, • x ⋆⋆ = 400 m, • M = 2, • k = − . • k = 1 . • k = − . • p = 0 . • φ ( x ) = 1 . (cid:16) . (cid:16) x − (cid:17)(cid:17) − . (cid:16) (cid:16) x − (cid:17)(cid:17) .Starting with a subcritical flow, we get a smooth transition to a supercriti-cal zone, and, through a hydraulic jump, the flow becomes subcritical in theremaining of the domain. In section 3, we gave steady-state solutions of increasing difficulties. Thesesolutions can be used to check if the numerical methods are able to keep/catchsteady-state flows. But even if the initial condition differs from the expectedsteady state, we do not have information about the transitory behavior. Thus,in this section, we list transitory solutions that may improve the validation of thenumerical methods. Moreover, as most of these cases have wet/dry transitions,one can check the ability of the schemes to capture the evolution of these fronts( e.g. some methods may fail and give negative water height). At last, we givesome periodic transitory solutions in order to check whether the schemes arenumerically diffusive or not.Table 3 lists all transitory solutions available in SWASHES and outlines theirmain features. 21 ransitory solutions
Slope FrictionType Description § Reference Null Const. Var. Man. D.-W. Other Null CommentsDam breaks Dam breakon wet domain without friction 4.1.1 [46] X X Moving shock. 1DDam breakon dry domain without friction 4.1.2 [42] X X Wet-dry transition. 1DDam breakon dry domain with friction 4.1.3 [20] X X Wet-dry transition. 1DOscillationswithout damping Planar surfacein a parabola 4.2.1 [48] X X Wet-dry transition. 1DRadially-symmetricalparaboloid 4.2.2 [48] X X Wet-dry transition. 2DPlanar surfacein a paraboloid 4.2.2 [48] X X Wet-dry transition. 2DOscillationswith damping Planar surfacein a parabola with friction 4.2.3 [43] X X Wet-dry transition. 1D
Const.: Constant; Var.: Variable; Man.: Manning; D.-W.: Darcy-Weisbach
Table 3: Analytic solutions for shallow flow equations and their main features— Transitory cases
In this section, we are interested in dam break solutions of increasing complexityon a flat topography namely Stoker’s, Ritter’s and Dressler’s solutions. Theanalysis of dam break flow is part of dam design and safety analysis: dambreaks can release an enormous amount of water in a short time. This could bea threat to human life and to the infrastructures. To quantify the associatedrisk, a detailed description of the dam break flood wave is required. Research ondam break started more than a century ago. In 1892, Ritter was the first whostudied the problem, deriving an analytic solution based on the characteristicsmethod (all the following solutions are generalizations of his method). He gavethe solution for a dam break on a dry bed without friction (in particular, heconsidered an ideal fluid flow at the wavefront): it gives a parabolic water heightprofile connecting the upstream undisturbed region to the wet/dry transitionpoint. In the 1950’s, Dressler (see [20]) and Whitham (see [54]) derived analyticexpressions for dam breaks on a dry bed including the effect of bed resistancewith Ch´ezy friction law. They both proved that the solution is equal to Ritter’ssolution behind the wave tip. But Dressler neglects the tip region (so his solutiongives the location of the tip but not its shape) whereas Whitham’s approach,by treating the tip thanks to an integral method, is more complete. Dresslercompared these two solutions on experimental data [21]. A few years later,Stoker generalized Ritter’s solution for a wet bed downstream the dam to avoidwet/dry transition.Let us mention some other dam break solutions but we do not detail theirexpressions here. Ritter’s solution has been generalized to a trapezoidal crosssection channel thanks to Taylor’s series in [55]. Dam break flows are also exam-ined for problems in hydraulic or coastal engineering for example to discuss thebehavior of a strong bore, caused by a tsunami, over a uniformly sloping beach.Thus in 1985 Matsutomi gave a solution of a dam break on a uniformly slopingbottom (as mentioned in [39]). Another contribution is the one of Chanson,who generalized Dressler’s and Whitham’s dam break solutions to turbulentand laminar flows with horizontal and sloping bottom [12, 13].
In the shallow water community, Stoker’s solution or dam break on a wet22omain is a classical case (introduced first in [46, p. 333–341]). This is a classicalRiemann problem: its analog in compressible gas dynamics is the Sod tube [45]and in blood flow dynamics the ideal tourniquet [18]. In this section, we con-sider an ideal dam break on a wet domain, i.e. the dam break is instanteneous,the bottom is flat and there is no friction. We obtain an analytic solution ofthis case thanks to the characteristics method.The initial condition for this configuration is the following Riemann problem h ( x ) = ( h l for 0 m ≤ x ≤ x ,h r for x < x ≤ L, with h l ≥ h r and u ( x ) = 0 m/s.At time t ≥
0, we have a left-going rarefaction wave (or a part of parabolabetween x A ( t ) and x B ( t )) that reduces the initial depth h l into h m , and a right-going shock (located in x C ( t )) that increases the intial height h r into h m . Foreach time t ≥
0, the analytic solution is given by h ( t, x ) = h l g (cid:18) √ gh l − x − x t (cid:19) c m gh r u ( t, x ) = x ≤ x A ( t ) , (cid:18) x − x t + √ gh l (cid:19) if x A ( t ) ≤ x ≤ x B ( t ) , (cid:0) √ gh l − c m (cid:1) if x B ( t ) ≤ x ≤ x C ( t ) , x C ( t ) ≤ x, where x A ( t ) = x − t p gh l , x B ( t ) = x + t (cid:16) p gh l − c m (cid:17) and x C ( t ) = x + t c m (cid:0) √ gh l − c m (cid:1) c m − gh r , with c m = √ gh m solution of − gh r c m (cid:0) √ gh l − c m (cid:1) + (cid:0) c m − gh r (cid:1) (cid:0) c m + gh r (cid:1) =0. This solution tests whether the code gives the location of the moving shockproperly.In SWASHES, we take the following parameters for the dam: h l = 0 .
005 m, h r = 0 .
001 m, x = 5 m, L = 10 m and T = 6 s. Let us now look at Ritter’s solution [42]: this is an ideal dam break (witha reservoir of constant height h l ) on a dry domain, i.e. as for the Stoker’ssolution, the dam break is instantaneous, the bottom is flat and there is nofriction. The initial condition (Riemann problem) is modified and reads: h ( x ) = (cid:26) h l > ≤ x ≤ x ,h r = 0 m for x < x ≤ L, and u ( x ) = 0 m/s. At time t >
0, the free surface is the constant water height( h l ) at rest connected to a dry zone ( h r ) by a parabola. This parabola is limited23pstream (resp. downstream) by the abscissa x A ( t ) (resp. x B ( t )). The analyticsolution is given by h ( t, x ) = h l g (cid:18) √ gh l − x − x t (cid:19) u ( t, x ) = x ≤ x A ( t ) , (cid:18) x − x t + √ gh l (cid:19) if x A ( t ) ≤ x ≤ x B ( t ) , x B ( t ) ≤ x, where x A ( t ) = x − t p gh l and x B ( t ) = x + 2 t p gh l . This solution shows if the scheme is able to locate and treat correctly thewet/dry transition. It also emphasizes whether the scheme preserves the posi-tivity of the water height, as this property is usually violated near the wettingfront.In SWASHES, we consider the numerical values: h l = 0 .
005 m , x = 5 m , L =10 m and T = 6 s. In this section, we consider a dam break on a dry domain with a friction term[20]. In the literature we may find several approaches for this case. Althoughit is not complete in the wave tip (behind the wet-dry transition), we presenthere Dressler’s approach. Dressler considered Ch´ezy friction law and used aperturbation method in Ritter’s method, i.e. u and h are expanded as powerseries in the friction coefficient C f = 1 /C .The initial condition is h ( x ) = (cid:26) h l > ≤ x ≤ x ,h r = 0 m for x < x ≤ L, and u ( x ) = 0 m/s. Dressler’s first order developments for the flow resistancegive the following corrected water height and velocity h co ( x, t ) = 1 g (cid:18) √ gh l − x − x t + g C α t (cid:19) ,u co ( x, t ) = 2 √ gh l x − x )3 t + g C α t, (16)where α = 65 (cid:18) − x − x t √ gh l (cid:19) −
23 + 4 √ (cid:18) − x − x t √ gh l (cid:19) / and α = 122 − x − x t √ gh l −
83 + 8 √ (cid:18) − x − x t √ gh l (cid:19) / − (cid:18) − x − x t √ gh l (cid:19) . With this approach, four regions are considered: from upstream to downstream,a steady state region (( h l ,
0) for x ≤ x A ( t )), a corrected region (( h co , u co ) for24 A ( t ) ≤ x ≤ x T ( t )), the tip region (for x T ( t ) ≤ x ≤ x B ( t )) and the dry region((0 ,
0) for x B ( t ) ≤ x ). In the tip region, friction term is preponderant thus (16)is no more valid. In the corrected region, the velocity increases with x . Dresslerassumed that at x T ( t ) the velocity reaches the maximum of u co and that the ve-locity is constant in space in the tip region u tip ( t ) = max x ∈ [ x A ( t ) ,x B ( t )] u co ( x, t ).The analytic solution is then given by h ( t, x ) = h l h co ( x, t ) h co ( x, t )0 m u ( t, x ) = x ≤ x A ( t ) ,u co ( x, t ) if x A ( t ) ≤ x ≤ x T ( t ) ,u tip ( t ) if x T ( t ) ≤ x ≤ x B ( t ) , x B ( t ) ≤ x, and with x A ( t ) = x − t p gh l and x B ( t ) = x + 2 t p gh l . We should remark that with this approach the water height is not modified inthe tip zone. This is a limit of Dressler’s approach. Thus we coded the secondorder interpolation used in [50, 51] (not detailed here) and recommanded byValerio Caleffi .Even if we have no information concerning the shape of the wave tip, this caseshows if the scheme is able to locate and treat correctly the wet/dry transition.In SWASHES, we have h l = 6 m , x = 1000 m, C = 40 m / /s (Chezycoefficient), L = 2000 m and T = 40 s. In this section, we are interested in Thacker’s and Sampson’s solutions. Theseare analytic solutions with a variable slope (in space) for which the wet/drytransitions are moving. Such moving-boundary solutions are of great interestin communities interested in tsunami run-up and ocean flow simulations (seeamong others [37], [33] and [41]). A prime motivation for these solutions isto provide tests for numerical techniques and codes in wet/dry transitions onvarying topographies. The first moving-boundary solutions of Shallow-Waterequations for a water wave climbing a linearly sloping beach is obtained in [11](thanks to a hodograph transformation). Using Shallow-Water equations in La-grangian form, Miles and Ball [40] and [3] mentioned exact moving-boundary so-lutions in a parabolic trough and in a paraboloid of revolution. In [48], Thackershows exact moving boundary solutions using Eulerian equations. His approachwas first to make assumptions about the nature of the motion and then to solvethe basin shape in which that motion is possible. His solutions are analyticperiodic solutions (there is no damping) with Coriolis effect. Some of these so-lutions were generalised by Sampson et al. [43, 44] by adding damping due toa linear friction term.Thacker’s solutions described here do not take into account Coriolis effect. Inthe 1D case, the topography is a parabola and in the 2D case it is a paraboloid.These solutions test the ability of schemes to simulate flows with comings and Personal communication
Each solution written by Thacker has two dimensions in space [48]. The exactsolution described here is a simplification to 1D of an artificially 2D Thacker’ssolution. Indeed, for this solution, Thacker considered an infinite channel witha parabolic cross section but the velocity has only one nonzero component (or-thogonal to the axis of the channel). This case provides us with a relevant testin 1D for shallow water model because it deals with a sloping bed as well aswith wetting and drying. The topography is a parabolic bowl given by z ( x ) = h a (cid:18) x − L (cid:19) − ! , and the initial condition on the water height is h ( x ) = − h (cid:18) a (cid:18) x − L (cid:19) + B √ gh (cid:19) − ! for x (0) ≤ x ≤ x (0) , B = √ gh / (2 a ) and for the velocity u ( x ) = 0 m/s. Thacker’s solutionis a periodic solution (without friction) and the free surface remains planar intime. The analytic solution is h ( t, x ) = − h (cid:18) a (cid:18) x − L (cid:19) + B √ gh cos (cid:18) √ gh a t (cid:19)(cid:19) − ! for x ( t ) ≤ x ≤ x ( t ) , u ( t, x ) = B sin (cid:18) √ gh a t (cid:19) for x ( t ) ≤ x ≤ x ( t ) , x ( t ) and x ( t ) are the locations of wet/dry interfaces at time tx ( t ) = −
12 cos (cid:18) √ gh a t (cid:19) − a + L ,x ( t ) = −
12 cos (cid:18) √ gh a t (cid:19) + a + L . In SWASHES, we consider a = 1 m, h = 0 . L = 4 m and T = 10 . .2.2 Two dimensional cases Several two dimensional exact solutions with moving boundaries were developedby Thacker. Most of them include the Coriolis force that we do not considerhere (for further information, see [48]). These solutions are periodic in timewith moving wet/dry transitions. They provide perfect tests for shallow wateras they deal with bed slope and wetting/drying with two dimensional effects.Moreover, as the solution is exact without discontinuity, it is very appropriateto verify the accuracy of a numerical method.
Radially-symmetrical paraboloid
The two dimensional case presented hereis a radially symmetrical oscillating paraboloid [48]. The solution is periodicwithout damping ( i.e. no friction). The topography is a paraboloid of revolu-tion defined by z ( r ) = − h (cid:18) − r a (cid:19) (17)with r = p ( x − L/ + ( y − L/ for each ( x, y ) in [0; L ] × [0; L ], where h isthe water depth at the central point of the domain for a zero elevation and a isthe distance from this central point to the zero elevation of the shoreline. Thesolution is given by: h ( r, t ) = h √ − A − A cos( ωt ) − − r a − A (1 − A cos( ωt )) − !! − z ( r ) u ( x, y, t ) = 11 − A cos( ωt ) (cid:18) ω (cid:18) x − L (cid:19) A sin( ωt ) (cid:19) v ( x, y, t ) = 11 − A cos( ωt ) (cid:18) ω (cid:18) y − L (cid:19) A sin( ωt ) (cid:19) where the frequency ω is defined as ω = √ gh /a , r is the distance fromthe central point to the point where the shoreline is initially located and A =( a − r ) / ( a + r ) (Figure 4).The analytic solution at t = 0 s is taken as initial condition.In SWASHES, we consider a = 1 m, r = 0 . h = 0 . L = 4 m and T = 3 πω . Planar surface in a paraboloid
For this second Thacker’s 2D case, themoving shoreline is a circle and the topography is again given by (17). The freesurface has a periodic motion and remains planar in time [48]. To visualize thiscase, one can think of a glass with some liquid in rotation inside.The exact periodic solution is given by: h ( x, y, t ) = ηh a (cid:18) (cid:18) x − L (cid:19) cos( ωt ) + 2 (cid:18) y − L (cid:19) sin( ωt ) − η (cid:19) − z ( x, y ) u ( x, y, t ) = − ηω sin( ωt ) v ( x, y, t ) = ηω cos( ωt )for each ( x, y ) in [0; L ] × [0; L ], where the frequency ω is defined as ω = √ gh /a and η is a parameter.Here again, the analytic solution at t = 0 s is taken as initial condition.In SWASHES, we consider a = 1 m, h = 0 . η = 0 . L = 4 m and T = 3 πω . 27 x L r a Figure 4: Notations for Thacker’s axisymmetrical solution
Considering a linear friction term ( i.e. S f = τ u/g ) in system (2) (with R =0 m/s and µ = 0 m / s) with Thacker’s approach, Sampson et al. got movingboundaries solutions with damping [43, 44]. These solutions provide a set of 1Dbenchmarks for numerical techniques in wet/dry transitions on varying topogra-phies (as Thacker’s solutions) and with a friction term. One of these solutionsis presented here. The topography is a parabolic bowl given by z ( x ) = h (cid:18) x − L (cid:19) a , where h and a are two parameters and x ∈ [0 , L ]. The initial free surface is( z + h )( x ) = h + a B g h (cid:18) τ − s (cid:19) − B g − g Bs (cid:18) x − L (cid:19) for x (0) ≤ x ≤ x (0) , , with B a constant, s = p p − τ / p = p gh /a . The free surfaceremains planar in time( z + h )( t, x ) = h + a B e − τt g h (cid:18) − sτ sin(2 st ) + (cid:18) τ − s (cid:19) cos(2 st ) (cid:19) − B e − τt g − e − τt/ g (cid:18) Bs cos( st ) + τ B st ) (cid:19) (cid:18) x − L (cid:19) for x ( t ) ≤ x ≤ x ( t ) ,z ( x ) elseand the velocity is given by u ( t, x ) = (cid:26) Be − τt/ sin( st ) for x ( t ) ≤ x ≤ x ( t ) , . x ( t ) and x ( t ) x ( t ) = a e − τt/ gh (cid:18) − Bs cos( st ) − τ B st ) (cid:19) − a + L ,x ( t ) = a e − τt/ gh (cid:18) − Bs cos( st ) − τ B st ) (cid:19) + a + L . In SWASHES, we consider a = 3 ,
000 m, h = 10 m, τ = 0 .
001 s − , B =5 m/s, L = 10 ,
000 m and T = 6 ,
000 s.
In this section, we describe the Shallow Water Analytic Solutions for Hydraulicand Environmental Studies (SWASHES) software. At the moment, SWASHESincludes all the analytic solutions given in this paper. The source code isfreely available to the community through the SWASHES repository hosted at . It is distributed underCeCILL-V2 (GPL-compatible) free software license.When running the program, the user must specify in the command line thechoice of the solution (namely the dimension, the type, the domain and thenumber of the solution) as well as the number of cells for the discretizationof the analytic solution. The solution is computed and can be redirected in agnuplot-compatible ASCII file.SWASHES is written in object-oriented ISO C++. The program is struc-tured as follows: each type of solutions, such as bump , dam break , Thacker ,etc., is written in a specific class. This structure gives the opportunity to easilyimplement a new solution, whether in a class that already exists (for examplea new Mac Donald type solution), or in a new class. Each analytic solution iscoded with specific parameters (most of them taken from [16]). In fact, all theparameters are written in the code, except the number of cells.We claim that such a library can be useful for developers of Shallow-Watercodes to evaluate the performances and properties of their own code, each ana-lytic solution being a potential piece of benchmark. Depending on the targetedapplications and considering the wide range of flow conditions available amongthe analytic solutions, developers may select a subset of the analytic solutionsavailable in SWASHES. We recommend not to change the values of the pa-rameters to ease the comparison of numerical methods among different codes.However, it may be legitimate to modify these parameters to adapt the case toother specific requirements (friction coefficient, dam break height, rain intensity,etc.). This can be easily done in SWASHES but, in that case, the code must berenamed to avoid confusions.
SWASHES was created because we have been developing a software for theresolution of Shallow-Water equations, namely FullSWOF, and we wanted to29alidate it against analytic solutions. To illustrate the use of SWASHES ina practical case, we first give a short description of the 1D and 2D codes ofFullSWOF and then we compare the results of FullSWOF with the analytic so-lutions. The comparisons between FullSWOF results and the analytic solutionsis based on the relative error in percentage of the water height, using, of course,the analytic solution as a reference. This percentage is positive when FullSWOFoverestimates the water height and negative when it underestimates it.
FullSWOF (Full Shallow-Water equations for Overland Flow) is an object-oriented C++ code (free software and GPL-compatible license CeCILL-V2 .Source code available at )developed in the framework of the multidisciplinary project METHODE (see[15, 16] and ). We briefly de-scribe here the principles of the numerical methods used in FullSWOF 2D. Themain strategy consists in a finite volume method on a structured mesh in twospace dimensions. Structured meshes have been chosen because on the onehand digital topographic maps are often provided on such grids, and, on theother hand, it allows to develop numerical schemes in one space dimension(implemented in FullSWOF 1D), the extension to dimension two being thenstraightforward. Finite volume method ensures by construction the conserva-tion of the water mass, and is coupled with the hydrostatic reconstruction [2, 7]to deal with the topography source term. This reconstruction preserves thepositivity of the water height and provides a well-balanced scheme (notion in-troduced in [28]) i.e. it preserves at least hydrostatic equilibrium (9) (typicallypuddles and lakes). Several numerical fluxes and second order reconstructionsare implemented. Currently, we recommend, based on [16], to use the secondorder scheme with MUSCL reconstruction [52] and HLL flux [29]. FullSWOF isstructured in order to ease the implementation of new numerical methods. In this part, we give the results obtained with FullSWOF 1D for three one-dimensional cases. For these examples, we have 500 cells on the FullSWOF 1Ddomain but 2500 for semi-analytic solutions (see Remark 1). In the first twofigures, we also plotted the critical height, in order to show directly whether theflow is fluvial or torrential.
On Figure 5, we plotted the solution for a transcritical flow with shock (Sec-tion 3.1.5) for a time large enough to attain the steady state, namely t = 100 s.Overall, the numerical result of FullSWOF 1D and the analytic solution arein very good agreement. By looking carefully at the differences in water height,it appears that the difference is extremely low before the bump ( − × − m, i.e. -0.001%). Right at the beginning of the bump, this difference increasesto +0.06%, but on a single cell. Over the bump, differences are alternatively w a t e r he i gh t ( i n m e t e r s ) x (in meters) water height - 500 cells - steady state FullSWOF_1DAnalytic solutionCritical heightTopography
Figure 5: Results of FullSWOF 1D for a transcritical flow with shockpositive and negative which leads, overall, to a good estimate. A maximum(+1.2%) is reached exactly at the top of the bump ( i.e. at the transition fromsubcritical flow to supercritical flow). Just before the shock, the difference isonly +0.03%. The largest difference in height (+100%) is achieved at the shockand affects a single cell. While for the analytic solution the shock is extremelysharp, in FullSWOF 1D it spans over four cells. After the shock and up to theoutlet, the heights computed by FullSWOF 1D remain lower than the heights ofthe analytic solution, going from -1% (after the shock) to -0.01% (at the outflowboundary condition).
Figure 6 shows the case of a smooth transition and a shock in a short domain(Section 3.2.2). The final time is t = 150 s.FullSWOF 1D result is very close to the analytic solution. From x = 0 m tothe subcritical-to-supercritical transition, FullSWOF 1D underestimes slightlythe water height and the difference grows smoothly with x (to -0.1%). Around x = 45 m ( i.e. close to the subcritical-to-supercritical transition), the differenceremains negative but oscillates, reaching a maximum difference of -0.22%. Afterthis transition, FullSWOF 1D continues to underestimate the water height andthis difference grows smoothly up to -0.5% right before the shock. As for thecomparison with the transcritical flow with shock (see Section 6.2.1 ), the max-imum difference is reached exactly at the shock ( x = 66 . The last figure for the one-dimensional case is Figure 7, with the solution ofa dam break on a dry domain (Section 4.1.2). We chose the final time equalto t = 6 s. Here too, the numerical result of FullSWOF 1D matches with theanalytic solution well. The height differences on the plateau ( h = 0 .
005 m) are31 w a t e r he i gh t ( i n m e t e r s ) x (in meters) water height - 500 cells - t=1500s FullSWOF_1DAnalytic solutionCritical heightTopography
Figure 6: Results of FullSWOF 1D for the Mac Donald’s type solution witha smooth transition and a shock in a short domain, with Manning’s frictioncoefficient w a t e r he i gh t ( i n m e t e r s ) x (in meters) water height - 500 cells - t=6s FullSWOF_1DAnalytic solution
Figure 7: Results of FullSWOF 1D for a dam break on a dry domain withoutfriction 32ull. On this domain, the water is not flowing yet. This shows FullSWOF 1Dpreserves steady-states at rest properly. The analytic solution predicts a kinkat x = 3 . x = 3 . x = 4 m. Between x = 4 m and x = 6 . x = 6 . x = 7 . x = 7 . x > . x > . We now consider the results given by the two-dimensional software FullSWOF 2D.
In the case of Thacker’s planar surface in a paraboloid (see Section 4.2.2),FullSWOF 2D was run for three periods on a domain of [0 m; 4 m] × [0 m; 4 m]with 500 ×
500 cells. To analyse the performances of FullSWOF 2D, we considera cross section along x (Figure 8).Overall, FullSWOF 2D produces a good approximation of the analytic so-lution: while the maximum water height is 0.1 m, errors are in the domain[ − . × − m; +7 . × − m]. At the wet-dry transition located at x =1 . × − m and,on a single cell, FullSWOF 2D predicts water while the analytic solution givesa dry surface. Close to x = 1 . x > .
57 m.The overestimation persists up to x = 2 .
65 m and then becomes an underes-timation. The underestimation tends to grow up to the wet-dry transition at x = 3 . x = 3 .
38 m). Starting exactly at x = 3 . . × − m). We consider two Mac Donald pseudo-2D solutions. Since FullSWOF 2D doesnot solve the pseudo-2D Shallow-Water system but the full Shallow-Water sys-tem in 2D (1), more significant differences are expected. In both cases, FullSWOF 2Dwas run long enough to reach steady-state.33 w a t e r he i gh t ( i n m e t e r s ) x (in meters) water height - 500 cells - t=3T FullSWOF_2DAnalytic solutionTopography
Figure 8: Cross section along x of FullSWOF 2D result for Thacker’s planarsurface in a paraboloid Supercritical flow in a short domain
The case of a supercritical flow ina short domain (Section 3.5.2) is computed with 400 ×
201 cells on the domain[0 m; 200 m] × [0 m; 9 . B . The y -averaged resultof FullSWOF 2D differs from the analytic solution mainly around x = 100 m(Figure 9), with an underestimation of the water height of up to -0.018 m (-11.85%). This underestimation occurs on the whole domain and gets closer tozero near both the upper and lower boundaries. Subcritical flow in a long domain
The case of a subcritical flow in along domain (Section 3.5.5) is computed with 800 ×
201 cells on the domain[0 m; 400 m] × [0 m; 9 . B . Comparison betweenthe y -averaged FullSWOF 2D result and the analytic solution shows clear dif-ferences (Figure 10), even if the overall shape of the free surface given byFullSWOF 2D matches the analytic solution. FullSWOF 2D underestimatesthe water height on most of the domain. This difference can be up to -0.088 m(-8.8%) at x = 66 m. FullSWOF 2D overestimates water height for x >
297 mand this overestimation can reach up to +0.04 m (+4.1%) at x = 334 m. Acknowledgments
The authors wish to thank Valerio Caleffi and Anne-C´eline Boulanger for theircollaboration. This study is part of the ANR METHODE granted by the FrenchNational Agency for Research ANR-07-BLAN-0232.
References [1] Francisco Alcrudo and Elena Gil. The Malpasset dam break case study. In
Proceedings of the th Concerted Action on Dambreak Modelling Workshop ,pages 95–109, Zaragoza, Spain, November 1999.34 w a t e r h e i gh t ( i n m e t e r s ) x (in meters) water height - 400*201 cells - steady state FullSWOF_2D meanAnalytic solutionTopography
Figure 9: Results of FullSWOF 2D for the Mac Donald pseudo-2D supercriticalsolution in a short domain w a t e r h e i gh t ( i n m e t e r s ) x (in meters) water height - 800*201 cells - steady state FullSWOF_2D meanAnalytic solutionTopography
Figure 10: Result of FullSWOF 2D for the Mac Donald pseudo-2D subcriticalsolution in a long domain 352] E. Audusse, F. Bouchut, M.-O. Bristeau, R. Klein, and B. Perthame. Afast and stable well-balanced scheme with hydrostatic reconstruction forshallow water flows.
SIAM Journal on Scientific Computing , 25(6):2050–2065, 2004.[3] F. K. Ball. Some general theorems concerning the finite motion of a shal-low rotating liquid lying on a paraboloid.
Journal of Fluid Mechanics ,17(2):240–256, 1963.[4] Adh´emar Jean-Claude Barr´e de Saint Venant. Th´eorie du mouvementnon-permanent des eaux, avec application aux crues des rivi`eres et `al’introduction des mar´ees dans leur lit.
Comptes Rendus de l’Acad´emiedes Sciences , 73:147–154, 1871.[5] Alfredo Berm´udez and Ma Elena V´azquez. Upwind methods for hyperbolicconservation laws with source terms.
Computers and Fluids , 23(8):1049–1071, 1994.[6] A. G. L. Borthwick, S. Cruz Le´on, and J. J´ozsa. The shallow flow equationssolved on adaptative quadtree grids.
International Journal for NumericalMethods in Fluids , 37(6):691–719, November 2001.[7] F. Bouchut.
Nonlinear stability of finite volume methods for hyperbolic con-servation laws, and well-balanced schemes for sources , volume 4 of
Frontiersin Mathematics . Birkh¨auser Verlag, Basel - Boston - Berlin, 2004.[8] M.-O. Bristeau and B. Coussin. Boundary conditions for the shallow waterequations solved by kinetic schemes. Technical Report RR-4282, INRIA,October 2001.[9] J. Burguete and P. Garc´ıa-Navarro. Implicit schemes with large time stepfor non-linear equations: application to river flow hydraulics.
InternationalJournal for Numerical Methods in Fluids , 46(6):607–636, October 2004.[10] V. Caleffi, A. Valiani, and A. Zanni. Finite volume method for simulatingextreme flood events in natural channels.
Journal of Hydraulic Research ,41(2):167–177, 2003.[11] G. F. Carrier and H. P. Greenspan. Water waves of finite amplitude on asloping beach.
Journal of Fluid Mechanics , 4(1):97–109, 1958.[12] Hubert Chanson. Applications of the Saint-Venant equations and methodof characteristics to the dam break wave problem. Technical Report CH55-05, Department of Civil Engineering, The University of Queensland, Bris-bane, Australia, May 2005.[13] Hubert Chanson. Solutions analytiques de l’onde de rupture de barrage surplan horizontal et inclin´e.
La Houille Blanche , 3:76–86, 2006.[14] V. T. Chow.
Open-channel hydraulics . McGraw-Hill, New-York, USA,1959.[15] O. Delestre. Ecriture d’un code C++ pour la simulation en hydrologie.Master’s thesis, Universit´e d’Orl´eans, Orl´eans, France, September 2008.3616] Olivier Delestre.
Simulation du ruissellement d’eau de pluie sur des surfacesagricoles . PhD thesis, Universit´e d’Orl´eans, Orl´eans, France, July 2010.[17] Olivier Delestre, St´ephane Cordier, Fran¸cois James, and Fr´ed´eric Dar-boux. Simulation of rain-water overland-flow. In E. Tadmor, J.-G. Liu,and A. Tzavaras, editors,
Proceedings of the th International Conferenceon Hyperbolic Problems , volume 67 of
Proceedings of Symposia in AppliedMathematics , pages 537–546, University of Maryland, College Park (USA),2009. Amer. Math. Soc.[18] Olivier Delestre and Pierre-Yves Lagr´ee. A “well balanced” finite volumescheme for blood flow simulation. submitted, 2011.[19] Olivier Delestre and Fabien Marche. A numerical scheme for a viscousshallow water model with friction.
Journal of Scientific Computing , 48(1–3):41–51, 2011.[20] Robert F. Dressler. Hydraulic resistance effect upon the dam-break func-tions.
Journal of Research of the National Bureau of Standards , 49(3):217–225, September 1952.[21] Robert F. Dressler. Comparison of theories and experiments for the hy-draulic dam-break wave. In
Assembl´ee g´en´erale de Rome — Comptes-Rendus et Rapports de la Commission des Eaux de Surface , volume 38,pages 319–328. Union G´eod´esique et G´eophysique Internationale - Associ-ation Internationale d’Hydrologie Scientifique, Association Internationaled’Hydrologie, 1954.[22] M. Esteves, X. Faucher, S. Galle, and M. Vauclin. Overland flow and in-filtration modelling for small plots during unsteady rain: numerical resultsversus observed values.
Journal of Hydrology , 228(3–4):265–282, March2000.[23] D. L. George.
Finite volume methods and adaptive refinement for tsunamipropagation and inundation . PhD thesis, University of Washington, USA,2006.[24] Edwige Godlewski and Pierre-Arnaud Raviart.
Numerical approximationsof hyperbolic systems of conservation laws , volume 118 of
Applied Mathe-matical Sciences . Springer-Verlag, New York, 1996.[25] N. Goutal and F. Maurel. A finite volume solver for 1D shallow-waterequations applied to an actual river.
International Journal for NumericalMethods in Fluids , 38(1):1–19, January 2002.[26] N. Goutal and J. Sainte-Marie. A kinetic interpretation of the section-averaged Saint-Venant system for natural river hydraulics.
InternationalJournal for Numerical Methods in Fluids , 67(7):914–938, 2011.[27] Nicole Goutal and Fran¸cois Maurel. Proceedings of the 2 nd workshop ondam-break wave simulation. Technical Report HE-43/97/016/B, Electricit´ede France, Direction des ´etudes et recherches, 1997.3728] J. M. Greenberg and A. Y. LeRoux. A well-balanced scheme for the nu-merical processing of source terms in hyperbolic equations. SIAM Journalon Numerical Analysis , 33(1):1–16, 1996.[29] Amiram Harten, Peter D. Lax, and Bram van Leer. On upstream differ-encing and Godunov-type schemes for hyperbolic conservation laws.
SIAMReview , 25(1):35–61, 1983.[30] F. M. Henderson.
Open channel flow . Civil engineering. MacMillan, NewYork, nordby, g. edition, 1966.[31] J.-M. Hervouet.
Hydrodynamics of Free Surface Flows: Modelling with theFinite Element Method . Wiley, April 2007.[32] David D. Houghton and Akira Kasahara. Nonlinear shallow fluid flow overan isolated ridge.
Communications on Pure and Applied Mathematics , 21:1–23, January 1968.[33] D.-H. Kim, Y.-S. Cho, and Y.-K. Yi. Propagation and run-up of nearshoretsunamis with HLLC approximate Riemann solver.
Ocean Engineering ,34(8–9):1164–1173, June 2007.[34] R. J. LeVeque.
Finite volume methods for hyperbolic problems . Number 31in Cambridge Texts in Applied Mathematics. Cambridge University Press,August 2002.[35] I. MacDonald.
Analysis and computation of steady open channel flow . PhDthesis, University of Reading — Department of Mathematics, September1996.[36] I. MacDonald, M. J. Baines, N. K. Nichols, and P. G. Samuels. Analyticbenchmark solutions for open-channel flows.
Journal of Hydraulic Engi-neering , 123(11):1041–1045, November 1997.[37] F. Marche.
Theoretical and numerical study of shallow water models.Applications to nearshore hydrodynamics . PhD thesis, Laboratoire demath´ematiques appliqu´ees - Universit´e de Bordeaux 1, Bordeaux, France,December 2005.[38] F. Marche. Derivation of a new two-dimensional viscous shallow watermodel with varying topography, bottom friction and capillary effects.
Eu-ropean Journal of Mechanics B/Fluids , 26(1):49–63, 2007.[39] H. Matsutomi. Dam-break flow over a uniformly sloping bottom.
Journalof Hydraulics, Coastal and Environmental Engineering. Japan Society ofCivil Engineers , 726/II-62:151–156, February 2003.[40] J. W. Miles and F. K. Ball. On free-surface oscillations in a rotatingparaboloid.
Journal of Fluid Mechanics , 17(2):257–266, 1963.[41] St´ephane Popinet. Quadtree-adaptative tsunami modelling.
Ocean Dy-namics , 61(9):1261–1285, 2011.[42] A. Ritter. Die Fortpflanzung der Wasserwellen.
Zeitschrift des VereinesDeuscher Ingenieure , 36(33):947–954, August 1892.3843] Joe Sampson, Alan Easton, and Manmohan Singh. Moving boundary shal-low water flow above parabolic bottom topography. In Andrew Stacey, BillBlyth, John Shepherd, and A. J. Roberts, editors,
Proceedings of the th Biennial Engineering Mathematics and Applications Conference, EMAC-2005 , volume 47 of
ANZIAM Journal , pages C373–C387. Australian Math-ematical Society, October 2006.[44] Joe Sampson, Alan Easton, and Manmohan Singh. Moving boundaryshallow water flow in a region with quadratic bathymetry. In Geoffry N.Mercer and A. J. Roberts, editors,
Proceedings of the th Biennial Engi-neering Mathematics and Applications Conference, EMAC-2007 , volume 49of
ANZIAM Journal , pages C666–C680. Australian Mathematical Society,July 2008.[45] Gary A. Sod. A survey of several finite difference methods for systems ofnonlinear hyperbolic conservation laws.
Journal of Computational Physics ,27(1):1–31, April 1978.[46] J. J. Stoker.
Water Waves: The Mathematical Theory with Applications ,volume 4 of
Pure and Applied Mathematics . Interscience Publishers, NewYork, USA, 1957.[47] L. Tatard, O. Planchon, J. Wainwright, G. Nord, D. Favis-Mortlock, N. Sil-vera, O. Ribolzi, M. Esteves, and Chi-hua Huang. Measurement and mod-elling of high-resolution flow-velocity data under simulated rainfall on alow-slope sandy soil.
Journal of Hydrology , 348(1–2):1–12, January 2008.[48] William Carlisle Thacker. Some exact solutions to the nonlinear shallow-water wave equations.
Journal of Fluid Mechanics , 107:499–508, 1981.[49] E. F. Toro.
Riemann solvers and numerical methods for fluid dynamics. Apractical introduction . Springer-Verlag, Berlin and New York), 3 edition,2009.[50] A. Valiani, V. Caleffi, and A. Zanni. Finite volume scheme for 2D shallow-water equations. Application to the Malpasset dam-break. In
Proceedings ofthe th Concerted Action on Dambreak Modelling Workshop , pages 63–94,Zaragoza, Spain, November 1999.[51] A. Valiani, V. Caleffi, and A. Zanni. Case study: Malpasset dam-breaksimulation using a two-dimensional finite volume method.
Journal of Hy-draulic Engineering , 128(5):460–472, May 2002.[52] Bram van Leer. Towards the ultimate conservative difference scheme. V.A second-order sequel to Godunov’s method.
Journal of ComputationalPhysics , 32(1):101–136, July 1979.[53] Thi Ngoc Tuoi Vo. One dimensional Saint-Venant system. Master’s thesis,Universit´e d’Orl´eans, France, June 2008.[54] G. B. Whitham. The effects of hydraulic resistance in the dam-break prob-lem.
Proceedings of the Royal Society of London, Ser. A. , 227(1170):399–407, January 1955. 3955] Chao Wu. Ritter’s solution of dam-breaking wave in trapezoid cross-sectionchannel.
Advances in Hydrodynamics , 1(2):82–88, December 1986.[56] J. G. Zhou and P. K. Stansby. 2D shallow water flow model for the hydraulicjump.