Hamiltonian models for the propagation of irrotational surface gravity waves over a variable bottom

A single incompressible, inviscid, irrotational fluid medium bounded by a free surface and varying bottom is considered. The Hamiltonian of the system is expressed in terms of the so-called Dirichlet-Neumann operators. The equations for the surface waves are presented in Hamiltonian form. Specific scaling of the variables is selected which leads to approximations of Boussinesq and KdV types taking into account the effect of the slowly varying bottom. The arising KdV equation with variable coefficients is studied numerically when the initial condition is in the form of the one soliton solution for the initial depth.

Hamiltonian models for the propagation of irrotational surface gravity waves over a variable bottom

Alan Compelli, Rossen I. Ivanov and

Michail D. Todorov

School of Mathematical Sciences, Dublin Institute of Technology,
Kevin Street, Dublin 8, Ireland
Erwin Schrödinger Int. Institute for Mathematics and Physics,
University of Vienna, Vienna, Austria,
Department of Differential Equations,
Faculty of Applied Mathematics and Informatics,
Technical University of Sofia, 8 Kl. Ohridski Blvd., 1000 Sofia, Bulgaria
e-mail: alan.compelli@dit.ie
e-mail: rossen.ivanov@dit.ie
e-mail: mtod@tu-sofia.bg

Mathematics Subject Classification (2010): 35Q53, 35Q35, 37K05, 37K40

Keywords: Dirichlet-Neumann Operators, KdV equation, Water waves, Solitons

1 Introduction

In 1968 V. E. Zakharov in his work [33] demonstrated that the equations for the surface waves of a deep inviscid irrotational water have a canonical Hamiltonian formulation. The result has been extended to models with finite depth and flat bottom [9, 13], for internal waves between layers of different density [10] as well as waves with added shear for constant vorticity [7, 5, 6, 3, 4]. A multilayer model based on the Green-Naghdi approximation has been proposed in [8].

The Hamiltonian approach to the wave motion has been extended by inclusion of variations of the bottom surface in several papers [11, 12, 2]. The Hamiltonian framework allows for approximations taking into account different scales when considering shallow water and long-wave regimes. As a result the arising model equations are of the type of well known shallow water equations like the Korteweg-de Vries (KdV) and Boussinesq equations.

The problem of waves with a variable bottom has a long history. In the pioneering work of Johnson [18] a perturbed KdV equation is derived as a model for surface waves from Euler’s governing equations for irrotational inviscid fluid (cf. also with [21]). The problem has been studied further by Johnson [18, 19] and several other authors, e.g. in [23, 24, 25, 26].

Boussinesq-type models with variable bottom have been derived and analysed in [28, 14].

In this review paper we consider the scaling regime adopted in [18] within the Hamiltonian framework of waves under the variable bottom of Craig et al [11, 12, 2]. We use numerical solutions of the Johnson equation in order to analyse the effects of the propagation of solitary waves over a variable bottom.

2 Setup

An inviscid, incompressible system is presented with a surface wave and variable bottom as shown in Fig. 1.

Figure 1: The system under study.

The average water surface is at and the wave elevation is given by the function Therefore we have


The average bottom depth is at and the bottom elevation from the average depth is given by the function Hence the function that represents the local depth is


The body of the fluid which occupies the domain is defined as


The subscript notation will be used to refer to valuation on the surface and will refer to valuation on the variable bottom .

3 Governing equations

Let us introduce the velocity field The incompressibility and the irrotationality of the flow allow the introduction of a stream function and velocity potential as follows:


In addition,

in This leads to

The Euler equations written in terms of the velocity potential produce the Bernoulli condition on the surface


where is the acceleration due to gravity.

There is a kinematic boundary condition on the wave surface


and on the bottom


We make the assumption that all considered functions , , are in the Schwartz class for the variable, that is declining fast enough when (for all values of the other variables). In other words we describe the propagation of solitary waves.

4 Hamiltonian formulation

The Hamiltonian of the system will be represented as the total energy of the fluid


It can be written in terms of the variables as


We introduce as usual the variable proportional to the potential evaluated on the surface


and the Dirichlet-Neumann operator given by


where is the outward-pointing unit normal vector (with respect to ) to the wave surface.

With Green’s Theorem (Divergence Theorem) the Hamiltonian can be written as


On the bottom the outward-pointing unit normal vector is


thus and no bottom-related terms are present. Noting that the term is a constant and will not contribute to , we renormalise the Hamiltonian to


The variation of the Hamiltonian can be evaluated as follows. Green’s Theorem transforms the following expression to contributions from the surface and the bottom:




Noting that the variation of the potential on the wave surface is given as




we write


It is noted from (7) that


which represents the bottom condition in variational form. It also explains the fact that does not depend on . Evaluating we remember that and therefore


due to (6).

Next we compute


Noting that, using the kinematic boundary condition (6),


we write


Recall that



by the virtue of the Bernoulli equation (5).

Thus we have canonical equations of motion


where the Hamiltonian is given by (12). Introducing the variable we can represent (25) in the form


5 Taylor expansion of the Dirichlet-Neumann operator

We introduce some basic properties of the Dirichlet-Neumann operator. The details can be found in [11, 2]. The operator can be expanded as


where The surface waves are assumed small, i.e. and one can expand with respect to as follows:


The operator has an eigenvalue for a given wavelength , when acting on monochromatic plane wave solutions proportional to In the long-wave case has an eigenvalue

thus when is small, one can formally expand in powers of .

For the operator the expansion is


where . It is assumed that can be of the order of provided that the bed stays away from the surface, that is of with This way one can expand in powers of Using the recursive formulae in [11],

The Dirichlet-Neumann operator is


Keeping expansions up to and in we need

and hence, the truncated expansion is


In the case when with of order (or smaller) the commutator of and is proportional to which is of order (or smaller) and therefore small with respect to or . Therefore we take the truncated expansion


where is the local depth and indicates that the bottom depth varies slowly with .

6 Boussinesq and KdV approximations

We introduce, as usual, the small scale parameters

where is the wave amplitude and is the wavelength, and consider the long-wave and shallow water scaling regime. We consider the wave propagation regime with of order ; with its eigenvalue of order The quantity has the magnitude of a velocity (multiplied by ) and therefore is of order ; is of order 1 and most importantly, which usually leads to the Boussinesq and KdV propagation regimes.

The magnitudes of the quantities can be made explicit by the change


where now all quantities like and are . We write the Hamiltonian (13) in terms of the scaled variables, using the expansion for the Dirichlet-Neumann operator given in (32), and keeping only terms of order :


As is an overall scale factor we can work with the rescaled Hamiltonian


The Boussinesq-type equations of motion, truncating at , obtained from (35) and (26) and the commutation of and in the terms of orders and are


In the case of constant depth =const ( , , and can be scaled out) the system becomes

This system is integrable, known as the Kaup - Boussinesq system. Indeed, the transformation eliminates the term. A further transformation with brings the system to the form

which has a scalar Lax pair representation (due to D.J. Kaup, [22]) with a spectral parameter :

The solitons of the KB system have been studied by several authors, see for example [22, 17, 16] and the references therein.

Returning back to (36), it is noted that the leading order terms satisfy the system of equations



which shows that satisfies the wave equation for a wavespeed , that is the wavespeed is nearly constant, depending on the slowly varying variable . Thus, in the leading order, the initial disturbance propagates (to the right) with a slowly varying speed :


In the leading order approximation also


The observations from the leading order approximation indicate that the so-called slow variables, like the characteristic might be more adequate for the analysis of the model. Following Johnson [18, 20] we select a slow variable of the form


reminiscent of the far field variable . The function will be determined in what follows so that will be the characteristic for the right running wave. As a next step, the equations will be written in terms of the slow variables . The derivatives are related as follows:


The operator is hence


Note that the leading order approximation in the new variables from (36) is


giving or so we define


Hence we write the equations (36) in the new variables


We substitute from the second equation in (45) into the first one. In the terms of orders and we substitute with its leading order approximation (39). The equation for is


We observe that all terms are of smaller orders and . Rescaling appropriately the variables to remove the scale parameters and and introducing the variable we write the equation in (36) in the form of a KdV equation with variable coefficients:


This is exactly the equation obtained by Johnson [18] via appropriate expansions from the governing equations of the fluid motion, (see also the derivation in [20]) so we can refer to it as Johnson’s equation.

In the propagation regime where clearly the -term in (46) has to be neglected and one obtains the dispersionless Burgers equation


This equation does not have globally smooth solutions, its solutions always form a vertical slope and break.

7 The Johnson equation and soliton propagation

The Johnson model (46) has been studied previously in [19, 24, 25, 26, 23]. It has been noted that a change of variables

and rescaling of transforms the equation to the form of a perturbed KdV equation


where the perturbation is on the right hand side with

The perturbed KdV equation can be treated within the framework of the inverse scattering approach for the KdV equation. The basics of this approach are outlined for example in [27, 23, 15]. If the initial condition is a pure soliton solution for the KdV equation, due to the perturbation, waves of radiation will appear and will decrease the energy of the initial soliton. The soliton perturbation theory however is unwieldy and in the case when the perturbation gives birth to new solitons is even more problematic. In our study, as we shall see below, new solitons are born when the initial soliton travels over an uneven bottom.

The model equation (47) can be written in terms of the local depth , that is (taking for simplicity ):


We observe that plays the role of the time-like variable in the usual KdV setting and - the space-like variable. In order to keep this analogy in what follows we replace with i.e. while is a space-like variable. This surprising outcome is due to the fact that the original variables are both of order (while combinations like the characteristics are of order 1) and at leading order both and can measure time, see also the explanation in [20]. Thus we consider now the KdV with “time”-dependent coefficients


According to the classification of the KdV equations with variable coefficients, (51) does not appear to be transformable to the KdV equation for any choice of , see for example [32]. We proceed by specifying the -dependent function


where and are appropriate constants. This function represents a ramp at between two constant values (two depths), describes the steepness of the ramp at . Thus, the initial condition will be taken at before the ramp is switched on at . The initial profile then is taken as the KdV 1-soliton when is constant: () and


where is a constant amplitude. For this of course is not the 1- soliton solution for the new depth, and we are interested in investigating the changes in the behaviour triggered at . With rescaling

(and appropriate rescaling of ) the KdV-type equation (50) can be written as a KdV in a canonical form


with initial condition taken as the one soliton solution for depth at

Let us introduce the constant

then the initial condition acquires the form


Recall that the initial condition for the one-soliton is therefore this initial condition can produce several solitons, waves of radiation, or, in any case can change the shape of the incoming soliton.

For the 1-soliton solution is of the form


Let us denote the constant

It is important to know how many solitons () will be born with the initial condition (55). Clearly, will be a function of the threshold . (In the case we have obviously and .)

For (54) the associated spectral problem is [27, 29]

where is a spectral parameter, which however is time-independent. Therefore it is sufficient to study the spectral problem at

where has been given in (55). Change of variables leads to the associated Legendre equation for

It has solutions on the interval if

where is an integer, the number of the discrete eigenvalues of the spectral problem and thus the number of the solitons. is another integer, labelling the negative discrete eigenvalues, . The solutions to this equation are called the associated Legendre polynomials [1]. Therefore there are special depths (eigendepths) leading to the appearance of solitons for :


a result that appears in [30, 18]. The corresponding values for are given in Table 1.

We study numerically the evolution of an incoming soliton


which enters a new depth at and and can transform into a multi-soliton solution for the new depth (if respectively ). To this end we consider the fully implicit finite-difference implementation of (51) complemented by an inner iteration with respect to the nonlinear term (for more details see, for example [31]). In Fig. 2, where , and above the value necessary for the emergence of the two soliton solution (Table 1) the birth of the second soliton (of a much smaller amplitude) is visible. The difference between and increases with and so does the amplitude and the velocity of the second soliton, e.g. when for the propagation from Fig. 3. The increase in the amplitude of the incoming soliton only increases the reflected waves which are waves of radiation. They are highly unstable and decay rapidly with , which can be seen from Fig. 4. Note that if the incoming soliton moves from shallow to deep region ( and ) then new solitons do not appear, due to (56). Then only reflected waves of radiation reduce the energy of the incoming soliton. This is illustrated in Fig. 5. Note that the dispersive radiation waves of small amplitude move to the left. This is because their phase velocity determined from is and becomes significant for the short waves where is not small. This effect is unphysical since the KdV model does not work as a water-wave model for short waves. The soliton velocity is positive as it can be seen e.g. from (57).

As expected from (56) and Table 1 the increase of the threshold (the increase of ) increases the number of the emerging solitons, Fig. 6. Although qualitatively the numerical results are in an agreement with the theory, the exact values of from (56) are not matched. The possible reason is that strictly speaking (56) is valid for a rapid jump from to at while our assumption is for slow (and smooth) bottom variations which we model via the profile (52).

Figure 2: with two emerging solitons ( ). A small amplitude (and velocity) soliton is visible behind the bigger one.The horizontal axis is and the snapshots are taken for different in relative units.

Figure 3: with two solitons. The amplitude of the slower soliton is bigger in comparison to the situation in Fig. 2. ( ).

Figure 4: with higher initial amplitude . The number of solitons in comparison to the situation on Fig. 3 is unchanged, however the reflected waves (waves of radiation) are visible on the left,

Figure 5: with initial amplitude . This corresponds to soliton propagation from shallow water to deep water. There are no new solitons emerging - the energy of the incoming soliton is transformed to energy of the reflected wave only (which is a rapidly decaying wave of radiation),

Figure 6: with initial amplitude . There are several solitons emerging due to the bigger threshold,
1 2 3 4 5 6
0 0.24 0.38 0.47 0.54 0.59
Table 1: The dependence of on

8 Discussion

The motion of the wave surface is determined by two functions – the Hamiltonian variables and the potential on the surface . However the fluid motion in the entire domain of the fluid can only be recovered from the entire boundary of which includes the bottom. An expression for the continuation of the potential from the surface into the bulk of the fluid is provided in [11] in terms of the Dirichlet-Neumann operator. This at least formally determines the dynamic of the fluid in .

There are of course many other possibilities for the nature of the bottom variation, e.g. random topography studied in [2, 12], as well as for the propagation regimes which will be studied in forthcoming publications. The wave dynamics in the presence of shear currents (vorticity) with a variable bottom is another very important and interesting possibility for future research.


AC and RI acknowledge funding from the Erwin Schrödinger International Institute for Mathematics and Physics (ESI), Vienna (Austria) as participants in the Research in Teams Project Hamiltonian approach to modelling geophysical waves and currents with impact on natural hazards, where a big part of this work has been done. AC is funded by a Fiosraigh fellowship at Dublin Institute of Technology (Ireland). MT acknowledges financial support from the Bulgarian Science Fund under grant DFNI I-02/9.


The authors are thankful to Prof. Adrian Constantin, Prof. Robin Johnson, Dr Calin I. Martin and Prof. André Nachbin for many valuable discussions. The authors are also thankful to two anonymous referees for their very constructive remarks and suggestions.


  • [1] Abramowitz, M. & Stegun, I. A. (Eds.) 1972 Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, 9th printing. New York: Dover, p. 332.
  • [2] De Bouard, A., Craig, W., Díaz-Espinosa, O., Guyenne, P. & Sulem, C. 2008 Long wave expansions for water waves over random topography. Nonlinearity 21, 2143–2178. (DOI:10.1088/0951-7715/21/9/014)
  • [3] Compelli, A & Ivanov, R. 2015 On the dynamics of internal waves interacting with the Equatorial Undercurrent. J. Nonlinear Math. Phys. 22, 531-539. (DOI: 10.1080/14029251.2015.1113052) arXiv:1510.04096 [math-ph].
  • [4] Compelli, A. & Ivanov R. 2016 The dynamics of flat surface internal geophysical waves with currents. J. Math. Fluid Mech. (DOI: 10.1007/s00021-016-0283-4) arXiv:1611.06581 [physics.flu-dyn]
  • [5] Constantin, A. & Ivanov, R. 2015 A Hamiltonian approach to wave-current interactions in two-layer fluids. Phys. Fluids 27, 086603. (DOI: 10.1063/1.4929457)
  • [6] Constantin, A., Ivanov, R. I. & Martin, C.-I. 2016 Hamiltonian formulation for wave-current interactions in stratified rotational flows. Arch. Rational Mech. Anal. 221, 1417–1447. (DOI: 10.1007/s00205-016-0990-2).
  • [7] Constantin, A., Ivanov, R. & Prodanov, E. 2008 Nearly-Hamiltonian structure for water waves with constant vorticity. J. Math. Fluid Mech. 10, 224–237. (DOI 10.1007/s00021-006-0230-x) arXiv:math-ph/0610014.
  • [8] Cotter, C. J., Holm, D. D. & Percival, J. R. 2010 The square root depth wave equations, Proc. R. Soc. A 466, 3621–3633. (DOI: 10.1098/rspa.2010.0124) arXiv:0912.2194 [physics.flu-dyn].
  • [9] Craig, W., Groves, M. 1994 Hamiltonian long-wave approximations to the water-wave problem. Wave Motion 19, 367–389. (DOI: 10.1016/0165-2125(94)90003-5).
  • [10] Craig, W., Guyenne, P., Kalisch, H. 2005 Hamiltonian long wave expansions for free surfaces and interfaces. Comm. Pure Appl. Math. 58(12), 1587–1641. (DOI: 10.1002/cpa.20098).
  • [11] Craig, W., Guyenne, P., Nicholls, D. P. & Sulem C. 2005 Hamiltonian long-wave expansions for water waves over a rough bottom. Proc. R. Soc. A 461, 839–873. (DOI: 10.1098/rspa.2004.1367).
  • [12] Craig, W., Guyenne, P., Sulem, C. 2009 Water waves over a random bottom. J. Fluid Mech. 640, pp. 79–107. (DOI:10.1017/S0022112009991248).
  • [13] Craig, W. & Sulem, C. 1993 Numerical simulation of gravity waves. J. Computat. Phys. 108, 73–83. (DOI: 10.1006/jcph.1993.1164).
  • [14] Grajales, J.C.M. & Nachbin, A. 2004 Dispersive wave attenuation due to orographic forcing. SIAM J. Appl. Math. 64, 977–1001. (DOI: 10.1137/S0036139902412769).
  • [15] Grahovski, G. & Ivanov, R. 2009 Generalised Fourier transform and perturbations to soliton equations. Discrete Contin. Dyn. Syst. Ser. B 12 (3), 579–595. (DOI: 10.3934/dcdsb.2009.12.579) arXiv:0907.2062 [nlin.SI].
  • [16] Haberlin, J. & Lyons, T. 2017 Solitons of shallow-water models from energy-dependent spectral problems. arXiv:1705.04989 [math-ph]
  • [17] Ivanov, R. & Lyons, T. 2012 Integrable models for shallow water with energy dependent spectral problems. J. Nonlin. Math. Phys., 19 (Suppl. 1), 124008 (17 pages).(DOI: 10.1142/S1402925112400086) arXiv:1211.5567 [nlin.SI]
  • [18] Johnson, R. S. 1973 On the development of a solitary wave moving over an uneven bottom. Proc. Camb. Phil. Soc. 73, 183–203. (DOI: 10.1017/S0305004100047605)
  • [19] Johnson, R. S. 1973 On an asymptotic solution of the Korteweg-de Vries equation with slowly varying coefficients, J. Fluid Mech. 60, pp. 813–824, DOI: 10.1017/S0022112073000492
  • [20] Johnson, R. S. 1997 A Modern Introduction to the Mathematical Theory of Water Waves. Cambridge University Press.
  • [21] Kakutani, T. 1971 Effects of an uneven bottom on gravity waves. J. Phys. Soc. Japan 30, 272–276. (DOI: 10.1143/JPSJ.30.272).
  • [22] Kaup, D. J. 1975 A Higher-order water-wave equation and the method for solving it. Progr. Theor. Phys. 54, 396–408. (DOI: 10.1143/PTP.54.396)
  • [23] Kaup, D. J. & Newell, A. C. 1978 Solitons as particles, oscillators, and in slowly changing media: a singular perturbation theory. Proc. R. Soc. Lond. A 361, 413–446. (DOI: 10.1098/rspa.1978.0110).
  • [24] Knickerbocker, C. J. & Newell, A. C. 1980 Shelves and the Korteweg-de Vries equation. J. Fluid Mech. 98(4), 803–818. (DOI: 10.1017/S0022112080000407).
  • [25] Knickerbocker, C. J. & Newell, A. C. 1985 Propagation of solitary waves in channels of decreasing depth. J. Stat. Phys. 39, pp 653–674. (DOI: 10.1007/BF01008358).
  • [26] Knickerbocker, C. J. & Newell, A. C. 1985 Reflections from solitary waves in channels of decreasing depth. J. Fluid Mech. 153, pp. 1–16; (DOI: 10.1017/S0022112085001112).
  • [27] Lamb, G. L., Jr. 1980 Elements of Soliton Theory, Wiley.
  • [28] Nachbin, A. 2003 A Terrain-Following Boussinesq System, SIAM J. Appl. Math. 63(3), 905–922. (DOI:10.1137/S0036139901397583)
  • [29] Novikov, S.P., Manakov, S.V., Pitaevsky, L.P. & Zakharov, V.E.. 1984 Theory of Solitons: the Inverse Scattering Method. New York: Plenum.
  • [30] Tappert, F. & Zabusky N. J. 1971 Gradient-induced fission of solitons. Phys. Rev. Lett. 27, 1774–1776. (DOI: 10.1103/PhysRevLett.27.1774)
  • [31] Todorov, M.D., Christov, C.I. 2007 Conservative numerical scheme in complex arithmetic for Coupled Nonlinear Schrodinger Equations. Discrete Contin. Dyn. Syst. Suppl. 2007, 982–992.
  • [32] Vaneeva, O. 2013 Group Classification of Variable Coefficient KdV-like Equations, In: Dobrev V. (eds) Lie Theory and Its Applications in Physics. Springer Proceedings in Mathematics & Statistics, vol 36. Springer, Tokyo. (DOI: 10.1007/978-4-431-54270-4_32) arXiv:1204.4875 [nlin.SI].
  • [33] Zakharov, V. E. 1968 Stability of periodic waves of finite amplitude on the surface of a deep fluid. J. Appl. Mech. Tech. Phys. 9, 86–89. (DOI: 10.1007/BF00913182).
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
Add comment
Loading ...
This is a comment super asjknd jkasnjk adsnkj
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test description